Introduction to

Tagmemic Analysis
 Introduction to
Tagmemic Analysis

Walter A. Cook, S.J.

Georgetown University Press, Washington, D.C. 20057

Library of Congress Cataloging in Publication Data

Cook, Walter Anthony, 1922-

Introduction to tagmemic analysis.

Originally published in 1969.

Bibliography: p.
Includes index.
1. Tagmemics. I. Title.
P160.C6 1978 415 78-1268
ISBN 0-87840-171-7

Copyright © 1969 by Georgetown University

All rights reserved
Printed in the United States of America

International Standard Book Number: 0-87840-171-7

 Preface

Introduction to Tagmemic Analysis is an application of the

methods of linguistic science to practical language problems at
the level of grammar. It presupposes a general introduction to
the science of linguistics and a knowledge of phonetics and
phonemics. This introduction deals with inductive methods for
analyzing real languages beginning with explicit sets of data.
In the application of these methods, the beginner is encouraged
to use his knowledge of traditional grammar, within the limita-
tions imposed by a strictly formal approach to analysis.
The tagmemic system is used as the formal method for pre-
senting the results of this inductive analysis. It is maximally
taxonomic; every function and form is given a name. This system
seems best suited for students beginning language analysis and
for more advanced students who are analyzing an unknown
language for the first time. At a later stage, more sophisticated
methods for presenting the results of analysis might be at-
tempted,
Tagmemic analysis cuts across the boundary between mor-
phology and syntax, and presents the structures of grammar by
a single method. In the present work, after an introduction deal-
ing with method, models, and practice, the system is presented
in Chapter 1, Tagmemic Analysis. Structures at the various
levels of grammar are presented in the following chapters: 2
Sentence Level; 3, Clause Level; 4, Phrase Level; and 5 Word
Level. Methods for determining the number of sentences gene-
rated and their underlying structural descriptions are given m
Chapter 6, Morpheme Level. Finally, the basic units of the
phonological, lexical, and syntactic components of the grammar
are described in terms of their feature, manifestation, and dis-
tribution modes in Chapter 7, Language Description. Suggested
problems and a list of tagmemic symbols for use in problem
work are included at the end of the text.
At each stage of development, an attempt has been made to
be as explicit as possible with regard to the sources used and
each chapter is supplemented by suggested readings with anno-
tations. Tagmemic theory is developed from the works of Ken-
VI PREFACE

neth L. Pike and Robert E. Longacre. Practical tagmemic formu-

lations are based on the work of Benjamin Elson and Velma
Pickett and the laboratory manual of William R. Merrifield. To
a lesser degree, use is made of the writings of other tagmemic
authors, particularly those of the Summer Institute of Linguis-
tics, who in facing practical language problems have contributed
so much to the development of the theory.
Grateful acknowledgment is made to Kenneth L. Pike Rob-
ert E. Longacre, and William R. Merrifield for their constructive
criticism of early tagmemic writings, and for materials supplied
from the Summer Institute of Linguistics; also to Ruth M.
Brend for comments on an earlier version of the manuscript. A
debt of gratitude is also due my former teachers at the School
of Languages and Linguistics and its dean, Robert Lado for
their encouragement. And then there are students, who always
manage to teach us while we teach them.
WALTER A. COOK S.J.

Washington, D.C.
May 1969
 Contents

Preface v

INTRODUCTION: METHOD MODELS

AND PRACTICE 1
The Scientific Method 2
Models of Grammar 5
Theory and Practice 9
1 TAGMEMIC ANALYSIS 13
The Unit: The Tagmeme 15
The Construction: The Syntagmeme 21
The System: Levels of Grammar 27

2 SENTENCE LEVEL 39
Sentence Level Sorting 41
Major Sentences 47
Minor Sentences 54
3 CLAUSE LEVEL 65
Independent Clauses 67
Dependent Clauses 73
Clause Level Analysis 79
4 PHRASE LEVEL 91
Relater-Axis Phrases 93
Multiple Head Phrases 99
Head-Modifier Phrases 106

5 WORD LEVEL 117

Inflections: Outer Formations 119
Derivations: Inner Formations 125
Compounds: Core Formations 131
6 MORPHEME LEVEL 143
Grammatical Patterns 145
Structural Descriptions 151
Generative Possibilities of the Model 159
Vlli CONTENTS

LANGUAGE DESCRIPTION 169

The Phonological Component 171
The Lexical Component 177
The Syntactic Component 184

Supplementary Exercises 195

Tagmemic Symbols 197

Selected Bibliography 200

Index 205
 In human speech, different sounds
have different meanings. To study this
coordination of certain sounds with
certain meanings is to study
language.
Leonard Bloomfield, Language, 1933
INTRODUCTION: METHOD MODELS AND PRACTICE

Linguistics is the scientific study of language. The object of this

study is language, "a purely human and noninstinctive method
of communicating ideas, emotions and desires by means of a
system of voluntarily produced symbols." (Sapir, 1921:8V The
particular viewpoint adopted by the linguist in this study of
language is the search for language structure. The methods
he uses to discover and describe language structure are scientific.
The central problem in language analysis is the problem of
the inaccessible machinery in the mind of man which produces
the utterances of language in a regular and systematic way. It
is the problem of fundamental grammar. In the early history of
linguistic science, Ferdinand DeSaussure (1916:14) distinguished
between language and speaking. Speaking is willful individual
and accidental; language is systematic, conventional and belongs
to the psychological order. It is language, not speaking, that is
the object of linguistic study. More recently Noam Chomsky
(1965:4) distinguished competence from performance. Com-
petence is "the speaker-hearer's knowledge of his language,
and performance is "the actual use of language in concrete
situations." It is competence, not performance, that is the object
of our study. The problem of language analysis is to discover and
describe the language competence of the speaker.
Grammar in this fundamental state the competence of the
speaker, is largely unconscious. People use language uncon-
sciously, without being aware of the shift in grammatical struc-
tures as they use the language for communication. Whether

'This refers to the author, date, and page number of the reading in the Selected Bibliog-
raphy at the end of the text.
 2 INTRODUCTION TO TAGMEMIC ANALYSIS

language is learned in early childhood, or at an advanced age,

the processes rapidly become habitual. In order to analyze lan-
guage, one must work backward from performance, which can
be observed to a competence that is not accessible by any direct
approach. The analyst then attempts to construct a model of
grammar, which is a formal statement of competence, based
upon the objective evidence that is presented to him in the actual
performance of the speaker.

The Scientific Method

As in most physical and some behavioral sciences, the method
used in the analysis of languages is primarily an inductive one.
This method is not a matter of free choice; it is forced upon us by
the nature of the language problem.
The Little Black Box. To illustrate how inductive science
works John G. Kemeny (1959:131) gives the example of the box
with inaccessible machinery. Suppose you have a locked box,
which cannot be opened without destroying its contents. One
can observe how the box works, and predict how it will act. We
can form a theory as to what kind of machinery is in the box, and
we will maintain that theory as long as our predictions concern-
ing the box are verified.
The problem of the box is similar to the problem which faces
the linguistic analyst. He can note the performance of the native
speaker of a language, but the competence of the speaker is not
accessible to him. Therefore he constructs a formal grammar, a
statement of competence, and maintains his theory as long as
it continues to predict accurately the performance of native
speakers. But this formal grammar has only a relation of equiva-
lence, not a relation of identity, with the fundamental grammar
in the mind of the speaker. The grammatical model generates
language utterances as if it were a native speaker.
The Search for Competence. The linguistic analyst has
competence as the object of his study, yet this is not directly ac-
cessible. He can only approach competence through: (1) the input,
the material which goes into the mind of man to form a grammar;
(2) reflection on his own language competence; or (3) the output,
the performance of the speaker.
No one has ever completely measured the material which
goes into the mind of the child to form a grammar, nor do we
know how this material is assimilated and organized. Reflection
also seems to be a poor tool, since language is largely unconscious.
Although the analyst is clearly capable of editing his own Ian-
 INTRODUCTION: METHOD, MODELS, AND PRACTICE 3

guage to make it grammatical, he is wary of subjectivism; and

if he analyzes his own language at all, he prefers to work from
some objective evidence of actual performance.
The result is that the only really objective evidence available
to the linguistic analyst is the performance of native speakers
of the language. In the search for competence, the goal of the
analyst is to catch the unconscious competence of the native
speaker in the actual performance, to work backward from what
is factual and evident to the structures underlying this system-
atic behavior.

The Inductive Cycle. The inductive method is a cyclic method

which goes from facts to theory to new facts. It consists of four
steps: (1) observation of the data, (2) insight into the structure
of the data, (3) formulation of a hypothesis, and (4) verification
of the hypothesis. In the first two steps we observe and guess,
a regressive argument from consequent to antecedent. In the
next two steps we formulate and check, a progressive argument
from antecedent to consequent. The process is described by
Bochenski (1965:92) and Kemeny (1959:86).
Applying this theory to language, we first observe the per-
formance of the native speaker and guess at the underlying
language competence. We then formulate a statement of com-
petence and check this against the known performance of the
native speaker. The observe and guess steps are called discovery;
the formulate and check steps are called verification. The four
steps may be represented in the following schematic diagram:

(INSIGHT)^ Language Competence ,(THEORY)

Discovery f 2> GUESS - 3 . FORMULATE^ Verification
Process I Process
(Regressive)! 1. OBSERVE 4. CHECK J(Progressive)
Language Performance
The Four Steps. The steps of the inductive method character-
ize this scientific method as empirical, intelligent, exact and
objective.
Step 1: Observation of the data is an empirical step in which
the linguist records language performance in pro-
tocol statements —statements in a form recogniz-
able and readable by other analysts. In problems
at the grammatical level, the given data is assumed
to be in phonemic script.
Step 2: Insight into the structure of the data is an intelli-
gent step and prevents the process from being
 4 INTRODUCTION TO TAGMEMIC ANALYSIS

mechanical. It is intuitive, based upon the data,

but aided by the background, training and intelli-
gence of the analyst, who must be able to general-
ize, to compare, to see regularity.
Step 3: Formulation of the hypothesis is an exact step in
which the results of discovery are formulated in
an exact way according to some given theory. The
theory provides new terms, general ways of speak-
ing about language, and must be used in a consistent
way by the analyst.
Step U: Verification of the hypothesis is an objective step in
which the formulated results are mechanically veri-
fied. Assuming the investigation has been empirical
(Step 1) and exact (Step 3), the results should be
independent of the subjective attitudes of the
analyst, and publicly verifiable by any other analyst
using the same data and methods.
Discovery Process. The tagmemic theory of language analy-
sis has traditionally emphasized the discovery process. Accord-
ing to the author of the theory, Kenneth L. Pike, the discovery
procedures are not mechanical, but contain elements of intuition
(1967:225, fn. 6). In any discovery process there are "analytic
leaps" (1967:493) which, in the study of language, are based on
the facts of proportion and analogy, substitution and expansion.
The intuitive, nonmechanical nature of the discovery process,
however, does not prevent the formal presentation of results
from being exact. Analytic procedures require reference to
meaning and intuitive judgment, but presentation does not
(1967:493).
Robert E. Longacre openly calls the procedures guess-and-
check but states that they "need not be mechanical to be use-
ful" (1964:12). The guesses contain "elements of creative think-
ing," and are based on the facts of "closure" and "choice."
Speakers are aware when a construction is closed, and often
backtrack to reword a construction. The element of choice is
evident in the hesitation of a speaker, as he attempts to find the
right word or expression (1964:14). The guesses made by the
analyst are then formulated in an exact way and subjected to a
series of checks which call for revision of the formal statement
and allow the analyst to come up with some better guesses along
the way.
Verification Process. With the advent of transformational
models, attention has been focused upon mechanical verification
procedures. Although traditionally the tagmemic theory has
been more concerned with the discovery process, the language
 INTRODUCTION: METHOD, MODELS, AND PRACTICE S
generated from tagmemic models has been tested with native
speakers in extensive translations. The impact of this new
emphasis on verification has, however, caused a reevaluation
of the theory in terms of explicit generative power.
Longacre (1964:32) develops t h e power of a tagmemic gram-
mar through four steps: (1) readings (combinations), (2) permuta-
tions, (3) manifestations, and (4) substitution of lexical and
phonological material. By these steps, tagmemic grammars
can be formalized as devices which generate the grammatical
sentences of the language. "Provided the grammar has been
carefully constructed," says Longacre (1964:31), the analyst will
"end up with t h e grammatically possible, and never with the
grammatically impossible."
Following the suggestion of Longacre, it can be demonstrated
t h a t the tagmemic model is an exact model which is capable of
(1) generating t h e original data, (2) generating trial utterances
at random which are acceptable to native speakers, and (3) gen-
erating, given a finite grammar and lexicon, the entire gram-
matical output of the grammar together with the underlying
structural description of each sentence.

Models of Grammar

The result of the discovery process in language analysis is

a model of grammar ready for verification. Most language models
consider both the forms of language and the function of these
forms in the language. Models which emphasize the functional
aspect are functional models.
Functional Models. In the analysis of language, three ques-
tions must be asked about each linguistic form: (1) What is it?
(2) What does it do? and (3) Where does it occur? The answer to
the first question identifies the form as a member of a definite
form class. The answer to the second question identifies how this
form functions within the language. The answer to the third
question gives the distribution of the functioning form unit.
The concept of function is not new in the study of language.
As Chomsky has pointed out (1965:63), even traditional gram-
mars gave us information about grammatical categories (which
we call "form") and grammatical relationships (which we call
"function"), and grammatical models must account for both in
language description. In practice, however he advocates a
strictly formal model of grammar, in which grammatical func-
tions, or relationships, are implicit (1965:69).
Functions by Levels. Function is present at all levels of the
grammar. Function may be distinguished from form in phono-
 6 INTRODUCTION TO TAGMEMIC ANALYSIS

logy, in lexicon and in grammar. In phonology, the phonemes of

a language have a differential function. They do not carry mean-
ing; they are sound units which differentiate the meaningful
forms of language. They are "points in the pattern" to which the
native speaker reacts (Sapir, 1921:56). In the lexicon, the mor-
phemes, or form units of the language, have a referential func-
tion. Within the communication system of language, the function
of the morpheme is simply to refer to an area of meaning. This
association is purely conventional, and independent of the shape
of the forms which are used. For example, vowel replacement is
used in such forms as goose/geese to represent plural and in
such forms as take/took to represent past. "Wherever we go,
says Sapir (1921:59), "we are impressed by the fact that pattern
is one thing, the utilization of pattern quite another." Further
research in languages other than the English language, he sug-
gests, might further emphasize "this relative independence of
function and form" (1921:59). In grammar, the units have a
grammatical function, and add a grammatical meaning to forms,
over and above the meaning they have as lexical items (Fries,
1962:70). Total linguistic meaning equals lexical plus grammati-
cal meaning.

Word Functions. The form of words in isolation is distinct

from the function of words in syntax. The form, or category, of
a word is denned by the inflectional paradigm of the word. A
noun is a word that takes noun inflections. The function of a word
in syntax is denned by the usage of the word; a nominal is a word
that "acts like" (hence—AL) a noun in syntax. In The Structure of
English (1952), Fries presented a model of English structure
based upon functional usage. The groups that he isolates are
called Class 1 (nominal), Class 2 (verbal), Class 3 (adjectival) and
Class 4 (adverbial). In addition, the groups of functional words
A through O, fifteen groups in all, are isolated by the same pro-
cedures.
The results of the slot and filler technique employed by Fries
are really function, not form, classes. This analysis could be fol-
lowed by subanalysis, a technique in which the forms which act
like nouns and fill noun slots in the structure are sorted. Among
these forms will be found nouns —forms that take noun inflec-
tions—and other forms that appear in cross-functional usage
with nouns. For example, in the sentence, Barking is noisy,
the form barking is used as a noun. But this is a verb form,
marked by the verbal inflection -ing. The same form may appear
in adjectival use in The barking dog, and in verbal use as a main
verb in sentences such as The dog is barking.
Group Functions. Groups may be named by the head word
of the construction as noun phrase, verb phrase, and so on.
 INTRODUCTION: METHOD, MODELS, AND PRACTICE 7
These are categorical labels. The same groups, as they are used
in syntax, may be given functional labels, such as nominal
phrase, to indicate what function they have in syntactic usage.
Extending Fries' technique to include word groups is a type of
superanalysis t h a t labels group functions. Listing word form
word function, and group function shows three levels of struc-
ture. Applied to the sentence given by Longacre (I960), these
levels are:
Group Function 1(S) < > 2(P) 1(0)
Word Function (D + 3 + 3 + 3 + 1) (2) (D + 1)
Word Form (D + 3 + 2-ing + 2-d + 1) 2-d (D + 1)
Sentence The slow, lumbering covered wagon pulled
the family.
Phrase structures emerge at the word function level, and clause
patterns at the group level. Function words tend to disappear at
higher levels, and the resulting grammatical patterns can be
expressed in terms of four basic functions as nominal, verbal,
adjectival, and adverbial groups.

Function-Form Units. Subsequent to the development of the

model using functional word classes, Kenneth L. Pike introduced
the tagmemic model which uses a function-form unit. While
working with artificial languages, Pike found t h a t the basic units
of grammar could neither be expressed in terms of function
alone, in such strings as S + P 4- O, nor in terms of form alone in
such strings as N + V + N, but demanded t h a t both function and
form be expressed, in such strings as S:N + P:V + 0:N which is
read as "subject slot filled by a noun phrase, predicate slot filled
by a verb phrase, and object slot filled by a noun phrase." He then
concluded t h a t the basic unit of grammar must be a correlation
of function and form the correlation of a grammatical function
or slot with the list of mutually substitutable fillers that fill t h a t
slot.
This unit was labeled the tagmeme, from t h e Greek word
tagma meaning "arrangement," and posited as the fundamental
unit of grammatical arrangement, corresponding to the units
of sound in phonology, and the units of meaningful form in mor-
phology. The tagmeme unit is a slot-class correlative. The func-
tional slot gives the grammatical relation, the filler class gives
the grammatical categories involved, but neither exists without
the other. Function is manifested by forms and forms do not
occur in grammar without having identifiable function.

String Constituents. The emphasis that the tagmemic model

placed upon t h e functional aspect of grammar caused a radical
change in the analysis of language. If function is primary, then
the analysis of any construction is going to result in as many
 8 INTRODUCTION TO TAGMEMIC ANALYSIS

functional parts as can be identified within the construction.

This results in the cutting of the construction string into many
functional parts simultaneously. Other grammatical models
current at the time used the principle of immediate constituent
analysis, which required that the construction string, with rare
exceptions, always be cut into two parts. This binary cutting
operation was repeated successively until the morphemes or
ultimate constituents of the string, were reached.
In tagmemic string-type analysis, the construction is viewed
as a set of multiple relations. The tagmeme units are points in
the pattern, and are related to each other within that pattern.
The functional points are not rigidly denned. They are largely
intuitive and correspond to traditional notions of subject and
predicate, head and modifier. But this laxness of definition in the
early stages of analysis helps to provide flexibility in separating
functions within the construction string. Further refinements
of this analysis permit a pinpointing of more exact grammatical
meanings attached to each functional slot, such as subject-as-
actor in active sentences and subject-as-goal in passive sen-
tences.

Grammatical Levels. The concept of the tagmemic model was

further clarified by the introduction of grammatical levels. The
function-form unit, or tagmeme, occurred in constructions of the
string type, which, in turn, occurred as well-defined levels within
the grammar. These levels are the natural levels of sentence,
clause, phrase, word, and morpheme. The idea of tagmemes at
various levels is attributed by Pike to Longacre (1967:432, fn. 1).
Elsewhere Pike says that the concept of levels is now widely
accepted as part of tagmemic theory (1967:232, fn. 9).
The introduction of grammatical levels that are well defined
allows the analyst to begin his analysis at any level of the gram-
mar. He need not begin with the sentence; he may begin by ana-
lyzing words or phrases, and can continue analysis to levels
above the sentence. But the final grammar will present a well-
ordered set of rules dealing with the structures of sentences,
clauses, phrases, and words down to the ultimate constituents
or morphemes, of the language.
In generating sentences from this grammatical model the
form units which are listed in the lexicon are programmed into
word level structures, words into phrases, phrases into clauses,
and clauses into sentences. For those who wish to carry the anal-
ysis further, sentences can be programmed into higher and more
complex structures.

Tagmemic Model. The resulting tagmemic model contains

a grammar, a lexicon, and a phonological component. The gram-
mar is a series of syntactic statements concerning sentence
 INTRODUCTION: METHOD, MODELS, AND PRACTICE 9

clause, phrase, and word level structures. Each construction at

each level is formulated in terms of tagmeme units, which explic-
itly give both the function and the form of each element in the
construction. The lexicon lists the form units of language, to-
gether with their classification and gloss, and the morphophone-
mic rules to account for the varying forms of morphemes. Finally,
the phonological component gives the phonemic sentence a
phonetic realization in the language.
The model generates all and only the grammatical sentences
of the language; it assigns a structural description to each sen-
tence generated and an explicit function to each part of each
structural description. The "all and only" provision can be re-
duced to an exact mathematical figure for all sentences gener-
ated by a finite grammar with a limited lexical inventory. This
is the generative potential of the grammar. The structural de-
scriptions can be constructed as tree diagrams, if the grammar is
properly constructed in levels from higher to lower. And in the
structural descriptions, every branch of every underlying phrase
structure diagram may be explicitly labeled as to its function
within its own construction, thus making the branching process
explicit.

Theory and Practice

The test of a linguistic model is how it stands up in practice.
It is only by coming face to face with real analytic problems that
one is able to judge the value, as well as the limitations, of the
model. The present text is designed to be used in practical anal-
ysis.
Supplementary Exercises. On page 195, a list of supplemen-
tary exercises is suggested for use with the present text. The
best laboratory manual available is William R. Merrifield's
Laboratory Manual for Morphology and Syntax, revised 1967
Summer Institute of Linguistics. In this revision, problems #1
to #91 deal with morphology, and problems #92 to #188 deal
with syntax. The list of exercises suggested here includes the
first twenty problems dealing with syntax, #92 to #111 but
regular use of the manual will suggest other combinations just
as effective.
The syntax problems of the manual are series of sentences
given in phonemic script, together with the translation of each
sentence. In solving the problems, the set of sentences is ac-
cepted as the language, and solutions are based upon the data
given in the problem. Where assumptions must be made about
the language in question, they must be explicitly stated. All
of the data is taken from real languages, so the problems en-
countered are real not hypothetical problems.
 10 INTRODUCTION TO TAGMEMIC ANALYSIS

Tagmemic Notation. A list of eighty tagmemic symbols, with

their definitions and use is given on page 197. It is a basic list,
intended to be sufficient for simple declarative sentences, al-
though new symbols might be needed for other sentence types.
Symbols constitute a language about language. Working with
a limited list of symbols, analysts are able to communicate funda-
mental insights concerning language structure with the same
basic vocabulary.
The solution based on a limited symbol inventory expresses
a commitment to a particular concrete solution to a problem. If
more than one solution presents itself, it is up to the analyst,
with his prior knowledge of the theory, to present his own chosen
alternative in the most rigorous fashion possible within the
limits of that theory. Of many solutions that are adequate to rep-
resent the structure in a given set of data one solution will
usually turn out to be simpler than the others.
Solutions that are rigorously formulated are easily under-
stood by others familiar with the same notation. If, in practice,
it becomes necessary to introduce new symbols by analogy with
the old, each new symbol should be explained as it is introduced.
The resulting description can then be interpreted without am-
biguity.

Use of Meaning. In using a functional model for the descrip-

tion of language, no apology is given for the use of meaning. The
recognition of the lexical meaning of forms, and the grammatical
meanings which these forms have in syntax, is essential to the
system. In general, the tagmemic system follows the principle
enunciated by Fries (1952:8) that the use of meaning is unscien-
tific only when "the fact of our knowing the meaning leads us to
stop short of finding the precise formal signals that operate to
convey that meaning.
A translation is given with the data, often supplemented
by a word for word gloss of the forms. First the individual words
and word parts must be isolated, together with their meanings,
by comparing recurring partials in the language. When there is
a change in meaning parallel to a change in form, the meaning
change is attributed to the form change. The forms and mean-
ings isolated are listed in a lexicon. In the second step, these
forms are programmed into the patterns of the language. Formu-
las are constructed that describe how morphemes combine to
form words words to form phrases, phrases to form clauses and
clauses to form sentences. This set of formulas constitutes the
grammar. The points in the pattern are named with functional
labels, and the forms that occur at these points are labeled with
form or category labels from the lexicon.
 INTRODUCTION: METHOD, MODELS, AND PRACTICE 11

Verification of the Model. With the formulation of the tag-

memic model in terms of grammar and lexicon, the problem has
been solved but it is useful to follow this work with verification
procedures. One method of doing this is to calculate, for each
finite problem, the exact number of sentences that can be gener-
ated by a given solution.
If the solution to a problem is a commitment, the analyst
should be aware of the full implications of this commitment. He
should be aware that the solution he proposes, whether he makes
the calculation or not, does in fact generate an exact number of
sentences. The attempt to determine this number leads to a
fuller realization of the generative power of the grammar, and
focuses attention upon restrictions which should be imposed on
the grammar, and included in the solution.
With each grammar produced, there is an underlying struc-
tural description assigned to each sentence generated. The ana-
lyst should be aware, not only of the number of sentences gener-
ated by the model he has created, but of the number of underlying
structural patterns, and the number of sentences that belong to
each of the structural patterns. The number of structural pat-
terns generated is a function of the number of optional elements
occurring at any level from sentence to word; the number of
sentences generated is a function of the number of morphemes
that happen to belong to each morpheme class isolated.
Conclusion. The tagmemic model is only one of many models
used in linguistic analysis. Because of its functional orientation,
and its well-developed discovery procedures, it is useful in teach-
ing language analysis to beginners, and ideally suited even for
the more advanced in a field methods situation. With the new
emphasis upon verification procedure, it can be demonstrated
that the model has generative capacity, and that it generates a
well-defined set of sentences that can be calculated exactly and
generated by computer. The model also generates a set of phrase
structure descriptions which represent the "deep structure" of
the sentences generated. Finally, it assigns explicit function to
all points of the branching patterns in that underlying structure.
The science of language analysis profits from the existence of
conflicting theories; each can learn from the other. The tag-
memic system has been applied to many languages of the world
with considerable success, and many of the current language
descriptions of unknown languages are done in some form of
tagmemic description. Anyone interested in the science should be
acquainted with this functional approach to language, simply in
order to understand what others are doing. Those using this
approach should also be familiar with other current grammati-
cal models.
TABLE 1: THE SYSTEM OF GRAMMATICAL LEVELS

SENTENCE LEVEL
Base + Intonation

CLAUSE LEVEL
Subject + Predicate + Object + Adjuncts

PHRASE LEVEL
Relater + Axis, and Endocentric Word Groups

WORD LEVEL
Stems + Derivations + Inflections

MORPHEME LEVKL
The ultimate level of analysis
1 TAGMEMIC ANALYSIS

Tagmemic analysis is a set of procedures for the description of

language, with a basic grammatical unit called the tagmeme
mapped into string-type constructions located at specific levels
in the grammar. The system and theory were developed by Ken-
neth L. Pike and used by the Summer Institute of Linguistics
for the training of language analysts. The system was designed
to meet concrete field problems. Because of the large number of
linguists who have come to use the system and the frequency of
their publications, the system of tagmemic analysis is now one of
the major systems of analysis in modern linguistic science.
Early Tagmemic Theory. The notion of the tagmeme as a
basic unit of grammar, consisting of a functional slot and a list
of the mutually substitutable items that fill the slot, first ap-
peared in Kenneth L. Pike's Language in Relation to a Unified
Theory of the Structure of Human Behaviour, Part I, 1954, and
Part ill, 1955. The term "tagmeme" was substituted for the origi-
nal name, ''grameme," in a later article by Pike, "On Tagmemes
nee Gramemes" (IJAL, 24:273-278,1958). In this article, Pike de-
scribes how he arrived at the notion of the tagmeme by working
with artificial languages with a constant lexical inventory but
varying grammatical structures, and contrasts his use of the
word "tagmeme" with that of Bloomfield. The tagmeme is a unit,
parallel to the phoneme and the morpheme, in a triple grammati-
cal hierarchy of phonology, lexicon, and grammar. All three basic
units are shown to be trimodally structured in a later article:
"Language as Particle, Wave and Field" (The Texas Quarterly,
1959, 2:37-54). Finally, the last volume of Pike's original work
was published as Part III (1960) to complete the theoretical

13
 14 INTRODUCTION TO TAGMEMIC ANALYSIS

framework. All three volumes have recently been republished in

a single volume (The Hague, Mouton and Co., 1967). Bibliog-
raphies have been updated and footnotes added to reflect re-
cent discoveries in linguistics.
Later Developments. With the tagmeme clearly defined as a
unit, Robert E. Longacre, in his article, "String Constituent
Analysis" (Language, 1960, 36:63-88), shows how these units
combine in a string type analysis which is radically different
from the binary type constructions used in the older immediate
constituent model. The cuts in tagmemics are simultaneous at
each level of the grammar, with multiple branching in a single
formula whereas the immediate constituent type of analysis is
based on a system of successive binary cuts. This system is re-
duced to practice in Longacre's Grammar Discovery Procedures
(The Hague, Mouton and Co., 1964), which explains analytic pro-
cedures at the sentence, clause, phrase, and word levels. Finally,
the main points of the theory are summarized by Longacre in a
later article, "Some Fundamental Insights of Tagmemics"
(Language, 41:65-66, 1965).
With unit and constructions now clearly denned, Pike next
turned his attention to the grammatical system, and developed
the idea of the grammatical matrix as a field system in two arti-
cles: "Dimensions of Grammatical Structure" (Language, 38:221-
244, 1962), an exposition of matrix theory, and "A Syntactic
Paradigm" (Language, 39:216-230, 1963), a practical application
of the matrix theory. In these articles, the grammatical matrix
is described as an array, similar to the systematic arrangements
found in word paradigms and phonemic charts. While recogniz-
ing the usefulness of matrices to show relations between con-
structions, Longacre, in Grammar Discovery Procedures, also
recommends the use of grammatical transformations. The use of
both methods in the same description is not contradictory; the
two ways are "not immiscible" (1964:16).

Pedagogical Materials. The application of the tagmemic

system has been greatly aided by the practical work of the teach-
ers of the Summer Institute of Linguistics and the pedagogical
textbooks prepared for this teaching program. Beginning with
Velma Pickett, An Introduction to the Study of Grammatical
Structure, and Benjamin Elson, Beginning Morphology and Syn-
tax a set of revised texts was published in 1962, including Ben-
jamin Elson and Velma Pickett, An Introduction to Morphology
and Syntax, and the Laboratory Manual for Morphology and
Syntax, by William R. Merrifield, Constance M. Naish, Calvin R.
Rensch, and Gillian Story. These works provide a clear and sim-
ple method of applying tagmemics in practice, together with
problems needed to exercise these skills.
 TAGMEMIC ANALYSIS 15

Although this brief survey has been confined to the principal

texts available in tagmemic theory, many other analysts have
contributed to the system by solving problems of particular lan-
guages using tagmemic methods. References to these tagmemic
publications, with a critical appraisal of each, may be found in
"A Guide to Publications in Tagmemic Theory," by Kenneth L.
Pike, published in Current Trends in Linguistics, edited by
Thomas A. Sebeok (The Hague, Mouton and Co., 1966, III 365-
394).

THE UNIT: THE TAGMEME

The unit in tagmemic analysis is the tagmeme, the correlation of

a functional slot with the class of items that fill that slot. This
unit is not merely a form unit, as in other grammatical models
but a composite of function and form. For a complete understand-
ing of this basic unit, we will consider: (1) the definition of the
tagmeme, (2) the kinds of tagmemes that may occur in construc-
tion, and (3) tagmemes as essential units of the grammar.

The Tagmeme Defined

The tagmeme is defined as "the correlation of a grammatical
function, or slot, with the class of mutually substitutable items
that fill that slot" (Elson and Piekett, 1962:57). It is a slot-class
correlative in which both function and form are explicitly
named. This concept of a grammatical function, correlated with a
set of manifesting items, is the first of four fundamental insights
of tagmemics (Longacre, 1965:65).

The Functional Slot. A slot is a position in a construction

frame. Functional slots are positions in construction frames
which define the role of linguistic forms in the construction,
relative to other parts of the same construction. Functions are
grammatical relationships. They answer the question of what
the form does in the construction, and are labeled as subject,
predicate, head, modifier, and the like. Functional slots may be
identified by (1) position, (2) proportion, and (3) meaning, as de-
scribed by Pike (1967:218-219).
The slot is defined by position in the construction frame. How-
ever* the function is structural rather than linear. Although
normally functional slots will be found in a fixed position, the
system provides for units that are movable and occur in more
than,one position. Further, cases of intonation and stress, which
occur simultaneously with other elements, are by convention
placed to the right of the accompanying elements. The linear
order of the slot is not to be interpreted too rigidly.
 16 INTRODUCTION TO TAGMEMIC ANALYSIS

The slot is also defined by structural meaning, that is the

meaning which is added to the lexical items filling the slot, over
and above the lexical meaning as it appears in the dictionary.
The structural meaning of the slot is indicated by the choice of
slot label, for example, "Subject." Structural meaning, in turn
sets up a proportional relationship between items filling the dif-
ferent slots of a construction. One subject is to its predicate as
any other subject is to any other predicate within the same con-
struction. Items change, but the proportion remains constant.
The Filler Class. The filler class is the list of all the items that
can fill the functional slot. These items are mutually substitutable
within the slot. This filler class by definition is a distribution
class which in many cases is heterogeneous. Whatever fills the
slot belongs to the distribution class, no matter what its form.
These fillers must be sorted into form classes, and the symbol for
each form class that occurs must be listed as part of the tag-
meme. In the subject slot, for example, the fillers may include
pronouns, proper nouns, noun phrases, clauses, infinitives, and
participles.
Within the filler class one class of forms may be the exclu-
sive, or the typical filler. Particularly in case-marked systems,
the slot may be identified by form class marked with a particular
case. The subject fillers, for example, may be marked by the
nominative case. When listing the filler classes, an effort must be
made to find the typical filler of the slot, and identify it by formal
markings where possible. Subject, in English, for example, is a
slot filled by forms which can be replaced by the series /, you,
he she, it, they and is opposed to the object slot, which can be re-
placed by the series me, her, him, them.
Slot-Class Correlative. The tagmeme is neither a functional
slot nor a filler class but a correlation of both slot and class. If
the forms of language are held constant, a new language struc-
ture can be created by changing the function of these forms. If
the functions are held constant a new language can be created
by changing the forms. Both function and form are necessary
in order to identify the units of grammar in any particular
language.
Any linguistic sign is fully defined by its meaning, form and
distribution. The meaning of the tagmeme is carried by its func-
tional slot. The form of the tagmeme consists of the form classes
which are found to manifest this function in the sentences of
the language. Finally, this form-meaning unit has a distribution
in the constructions of the language.
The correlation of the functional slot with the filler class is
a correlation of function and form. The tagmemic system is
 TAGMEMIC ANALYSIS 17

maximally explicit in the naming process. Function and form are

both named in a notation, such as S:N, read as: "subject slot,
filled by a noun phrase." We simultaneously name the form, by
listing the forms that fill the slot to the right of the ratio mark,
and name the function by giving a label to the grammatical
meaning carried by the forms to the left of the ratio sign. When
this tagmemic unit is placed in a construction string, we can
answer all these pertinent questions about the linguistic unit:
What is it? What does it do? Where does it occur?

Kinds of Tagmemes in Constructions

Tagmemes are function-form correlatives which are dis-
tributed in the constructions of the language. Relative to the
constructions in which they occur, tagmemes are of various
kinds. They may be: (1) obligatory or optional to the construction;
(2) nuclear or peripheral to the structure; and (3) distributed in
fixed or movable positions.
Obligatory versus Optional Tagmemes. Tagmemes are either
obligatory or optional to the construction in which they occur.
An obligatory tagmeme is a tagmeme that occurs in every mani-
festation of the structure in the given data. It is marked with a
plus (+) sign to indicate that it must occur whenever the con-
struction occurs. An optional tagmeme is a tagmeme that occurs
in some but not all of the manifestations of the construction. It
is marked with a plus/minus (±) sign to indicate that it may
occur but need not necessarily occur, whenever the construction
occurs. Every tagmeme in the construction string, including the
very first tagmeme, must be marked as obligatory or optional.
Suprasegmentals may be marked with a minus (—) sign. This
sign is interpreted as an obligatory tagmeme, but one which is
not in linear order. It is used for the intonation patterns of sen-
tences, and may be extended to include the suprasegmentals that
affect the structure of the word, such as the features of pitch,
tone and stress.
The symbols that mark the tagmeme as obligatory or optional
are also the concatenation symbols of the construction string.
They may occur in the following combinations to express concat-
enation relations:
-f A +B Tagmemes A and B are both obligatory
+A ±B Tagmeme A is obligatory, B is optional
±A +B Either A or B must occur, but not both.
By the use of parentheses, further relations of concatenation
also can be expressed, with an obligatory or optional marker out-
side the parentheses.
 18 INTRODUCTION TO TAGMEMIC ANALYSIS

-K+A ±B) for combinations A, AB (B requires A)

±(+A ±B) for combinations A, AB, 0 (B requires A)
-K±A ±B) for combinations A, B AB
±(±A ±B) for combinations A, B, AB, 0
Particularly in cases where the affected tagmemes are separated
in the linear string, the algebraic sign outside the parentheses
may be placed upon a tie bar, connecting the tagmemes which
show mutual dependence.

+(+A ±B) may be written +A ±B

Further combinations of obligatory and optional tagmemes,
when more than two tagmemes are involved, are handled in
similar fashion.
Nuclear versus Peripheral Tagmemes. Tagmemes are either
nuclear or peripheral to the constructions in which they occur.
This is not the same distinction as obligatory and optional. A
nuclear tagmeme is a tagmeme that is diagnostic of the con-
struction in which it occurs. It may be either obligatory or
optional. A peripheral tagmeme is a tagmeme that is not diag-
nostic of the construction in which it occurs. It is always optional.
Peripheral tagmemes are sometimes called satellite or marginal
tagmemes.
All obligatory tagmemes are nuclear, but all nuclear tag-
memes are not obligatory; all optional tagmemes are not periph-
eral but all peripheral tagmemes are optional. The application
of these two rules results in a three-way classification of tag-
memes (Longacre, 1964:19): (1) nuclear and obligatory; (2) nu-
clear and optional; and (3) peripheral (and optional).
In clause structures for example, the nuclear tagmemes are
those tagmemes that help us to diagnose the clause structure,
and generally include the subject, predicate, and object. Yet
these tagmemes are, in different languages or in different situa-
tions in the same language, optional tagmemes. Thus, the subject
is often optional at the clause level, the object is optional in
transitive clauses and even the predicate may be optional in
predicate attribute constructions. Yet all of these tagmemes
are nuclear, and are useful in separating clause types. A tag-
meme may be both nuclear and optional.
Movable versus Fixed Order Tagmemes. Tagmemes are
generally assumed to occur m the position in which they are
represented in a fixed linear sequence. In the case of languages
with movable word order, the statistically prominent order is
 TAGMEMIC ANALYSIS 19

represented in the string, but the movability of the tagmeme

must be indicated.
If the tagmeme has two and only two fixed positions in the
string, the alternate positions may be represented by writing the
tagmeme in each position, in an either/or notation such as
±A . . . . . +A. If the tagmeme is freely movable within the string,
one device which has been suggested is an arrow written above
the movable tagmeme. If, in the same situation, the tagmemes
are freely movable, but may not interrupt the nucleus, or SPO
complex, then these nuclear elements may be enclosed in paren-
theses to limit the movability of tagmemes.
1.1 tCl = (+S:N +P:tV +O:N) ±L:loc ±T:tem
The above formulation allows such constructions as (in regular
order): SPO, SPOL, SPOLT, SPOT; but either L or T or both may
occur before the SPO nucleus, and the order TL is permitted
after the nucleus.

Etic and Emic Units of Grammar

Tagmemes are the essential units of grammar; but parallel
to phonology and lexicon, the grammar has both essential units
and nonessential units. The nonessential unit is called an etic
unit and is the first approximation of the analyst to the unit
from the point of view of an outsider. The essential unit is
called the emic unit, and is the unit of language from the point
of view of a native speaker of the language. In grammar, the etic
unit is called the tagma, and these tagmas are grouped as
allotagmas of essential units, called tagmemes. Thus, in a paral-
lelism suggested by Elson and Pickett (1962:131):
Tagmas (and allotagmas) are to tagmemes, what
Morphs (and allomorphs) are to morphemes, and what
Phones (and allophones) are to phonemes.
The first approximation of the analyst is the tagma, a correlation
of function and form. But after initial investigation, tagmas that
are the same are grouped as allotagmas of one tagmeme. Tagmas
that are different are set up as belonging to different tagmemes.
Norms must be laid down to determine when tagmas are the
same or different.
Tagmas as Etic Units. The isolation and classification of
tagmas is called the science of tagmatics. It is essentially a
cutting process, in which the slot-class correlatives discovered
are isolated from each other in the linear string. The words and
 20 INTRODUCTION TO TAGMEMIC ANALYSIS

word sequences are grouped into units, and the string is cut
simultaneously into its functional parts at a given level in
the grammar.
In practice, every tagma is considered, in the first analysis,
as if it were an invariant unit. One works as if every tagma were
a tagmeme, that is, as if every tagmeme had one and only one
allotagma. This is parallel to the science of morphetics where
the units isolated are morphs, but initially considered as if they
were the only manifestations of the morpheme. Form and mean-
ing are clearly isolated.
Once the slot-class correlatives have been initially formu-
lated it is possible to institute a grouping process and group
together those tagmas which seem to have the same functional
meaning, even though the fillers of the slots may be different.
Thus, given such structures as S:N, a subject slot filled by a noun
phrase, and S:pn, a subject slot filled by a pronoun, these may be
combined as S:N/pn, a subject slot filled by either a noun phrase
or a pronoun. Alternate filler classes are represented by the slash
symbol (/) and repeated as often as needed. When variants of a
tagmeme seem to have different grammatical meaning, the
grouping process becomes more difficult.
Allotagmas of a Tagmeme. The tagmeme is the essential unit
of grammar, but may be represented by one or more allotagmas.
Since the unit is a function-form unit, the variants of a tagmeme
may differ in function, or in form, or in both. Since the function-
form unit has a distribution in the language, there may also be
differences in distribution. The linguistic sign (LS) is fully de-
scribed in terms of meaning, form, and distribution (MFD); the
variants of the tagmeme will be variants: (1) of meaning, (2) of
form, (3) of distribution or some combination of these features.
If tagmas are totally different in form and meaning, and also
in distribution, they belong to different tagmemes. If tagmas are
partially different, that is, if they differ in one of these features
the following norms may be useful:
1. Tagmas differing in form alone are tagmas with the same
functional meaning and the same position in the string.
These are easily grouped as allotagmas of the same
tagmeme, by listing the alternate fillers in the one
tagmemic slot.
2. Tagmas differing in meaning alone may belong to the
same tagmeme. Although the structural meaning of the
tagmeme is the principal identifying feature, it is pos-
sible to carry over meanings from one's own native
language into the target language. Therefore, unless
the meaning difference is correlated with a parallel
 TAGMEMIC ANALYSIS 21

difference in filler class or position in the string, group

the elements as allotagmas of the same tagmeme.
3. Tagmas differing in position alone may belong to the
same tagmeme, and would then be merely positional
variants of the same unit. Unless change of position is
correlated with change of meaning or with change of
form consider the units as variants of one tagmeme.
Tagmemes as Essential Units. The tagmeme is the essential
unit of grammar which may be manifested by one or many tag-
mas. The unifying element in the tagmeme is the grammatical
meaning attached to the slot, just as the unifying element in
the morpheme is the meaning attached to the form or set of
forms. The tagmeme may have, as its variants: (1) form variants,
made up of different filler classes; (2) positional variants if the
tagmeme is movable without change of meaning; and (3) varia-
tion in structural meaning, provided form and distribution are
constant. The sum of these meaning variations would then be
the meaning of the slot, as interpreted by native speakers in
the target language. The area of meaning of the slot often does
not coincide with the analyst's initial guesses in setting up the
etic tagmas, prior to the tagmemic grouping process.

THE CONSTRUCTION: THE SYNTAGMEME

In tagmemic analysis, the unit is the tagmeme—a correlation of
a functional slot with a filler class. These units are strung to-
gether in constructions. Once the unit of a grammatical descrip-
tion has been clearly denned, the next step is to consider how
these units are put together in constructions. In tagmemic
analysis, the construction is called the syntagmeme, reminiscent
of DeSaussure's syntagm, "a combination supported by lin-
earity, . . . always composed of two or more consecutive units"
(1916:123). To understand the nature of the syntagmeme in tag-
memic analysis, we consider, in order: (1) the definition of the
construction, or syntagmeme; (2) the kinds of construction that
occur; and (3) etic and emic constructions.

The Syntagmeme Defined

The construction, or syntagmeme, is denned as "a potential
string of tagmemes, whose manifesting sequence of morphemes
fills ^grammatical slot" (Elson and Pickett, 1962:59). This defini-
tion includes the notions of (1) string constituency, (2) potential
string, and (3)rthe internal and external unity of the construc-
tion. The concept of syntagmeme, as a "functionally contrastive
 22 INTRODUCTION TO TAGMEMIC ANALYSIS

string on a given level," is the second fundamental insight of

tagmemics (Longacre, 1965:70).
String Constituents. There are two principal systems for the
description of constructions, no matter what type of unit is
used. The first is a string constituent type of analysis; the second
is an immediate constituent (IC)type of analysis. In a string type
analysis, the utterance is cut simultaneously into all of its func-
tional parts. In making these cuts, the analyst is guided by his
knowledge of function. In an immediate constituent analysis,
the utterance is cut successively into binary constituents. In
making these cuts, the analyst is partly guided by a knowledge
of function as well as by the theory that all constructions, with
rare exceptions, consist of two parts.

String Analysis cuts this sentence into 5 parts at once:

The slow, lumbering covered wagon / pulled / the pioneer's
family / across the prairie / just yesterday.
IC Analysis cuts this sentence in 4 successive 4
cuts: The
slow, lumbering covered wagon, /' pulled / the pioneer's
family I3 across the prairie /2 just yesterday.
Potential Strings. Given a string type of analysis, how many
units are required for the string? Following DeSaussure, one
would think that at least two consecutive units are required.
However in the tagmemic use of string analysis, there need not
be two obligatory elements to make a construction. Construc-
tions are also possible with one obligatory and one optional
element. If the string is potentially expandable —and in any
concrete set of data, if it is expanded in fact at least once—then
the string is considered to be a construction.
Constructions in tagmemics are not obligatorily complex, but
they must be at least potential strings. For example, in discuss-
ing the phrase, Pike says it is "either composed of a sequence
of two or more words or is one word which is optionally ex-
pandable in that same slot into a sequence of two or more words,"
(1967:439) and he formulates this as: a phrase is+(+word+word),
-K+word ±word), but not +(+word).
The principle of the potential string allows for conciseness
in the formulation of single words and phrases. The notation
S:N, means that the subject slot is filled by a noun phrase. But
the term "noun phrase" includes both single nouns and nouns
taken together with modifiers. The notation S:N/n would be re-
dundant since N (noun phrase) includes n (noun). In practice,
the symbol N (noun phrase) would be used only if the noun did
in fact in the set of data, occur at least once with modifiers.
 TAGMEMIC ANALYSIS 23

External Unity and Internal Unity. The external unity of

the construction is determined by its occurrence in a higher
level slot. To depend upon this occurrence for the unity of the
construction is to start an infinite series. Somewhere there is a
highest level construction, and this does not occur in a higher
slot. In practice, in a grammar limited to the analysis of sen-
tences, these sentences are bounded by their intonational pat-
terns and are not considered as occurring in higher slots.
Although external unity is useful for lower level construc-
tions each construction has its own internal unity. This internal
unity is based upon the correlativity of the tagmeme unit and
the construction in which it occurs. The tagmemic functions are
defined in terms of the construction in which they occur. For
example, subject is meaningless without its relation to the other
tagmemes of the string, such as predicate and object.
The tagmeme, then, is not only a unit, as Pike insists, but it
also expresses grammatical relationships within the context of
the construction. The construction cannot exist without the tag-
memes as component parts, and the tagmemes cannot exist with-
out reference to the syntagmeme, or construction. It is this
built-in set of relations, of the whole to its parts, that gives
the constructioa its internal unity, and this feature of internal
unity is independent of whether or not the whole occurs in a
higher slot.

Kinds of Syntagmemes
The syntagmeme, or construction, is a potential string of
tagmemes. All of the strings are not of the same kind. Strings
of various kinds occur at all levels of grammatical analysis.
These construction types include: (1) exocentric and endocentric
constructions, (2) closed and open-ended constructions and
(3) recursive and nonrecursive layering.
Exocentric versus Endocentric Constructions. Constructions
may be of the exocentric or endocentric type. An endocentric
construction is centered about one or more head tagmemes. An
exocentiie construction is not centered. In endocentric construc-
tionsi the whole construction may be replaced by a form similar
m form class to the head of the construction; in exocentric con-
structions, where there is no head tagmeme, the whole construc-
tion does not fill the same slot as one of its parts.
In string-type analysis, endocentric constructions include
multiple head constructions, such as coordinate phrase and
appositive phrase, as well as single headed phrases, composed of
nouns verbs, adjectives, and adverbs with their modifiers; they
 24 INTRODUCTION TO TAGMEMIC ANALYSIS

are proper to phrase level. Exocentric constructions occur at

phrase level only in the relater-axis phrase (prepositional
phrase), but include most constructions at the clause and sen-
tence levels, where the whole construction is not centered about
any particular head tagmeme.
Closed and Open-Ended Constructions. Some constructions
are closed while others seem to have no limit to their expansion
possibilities. When words are combined in a phrase, or phrases
into clauses these seemingly can be strung out in endless fash-
ion with no practical limit.
Coordinate constructions are open-ended at almost all levels
except the word level, where morphemes are combined into
words. At the phrase level, for example:
He went to the store and bought books, pencils, pens . .
In this English example, the object slot, filled by a coordinate
noun phrase, indicates a phrase level construction that can be
extended without limit. The conjunction generally occurs before
the last item recorded. The noun phrase in question might be
formulated as follows:
1.2 Nco = 4-Hjin ±H2:n ±H 3 :n.. . +C:c +Hn:n
The discontinuity of the construction is indicated by the three
dots (.. .), and the final expression indicates how the construc-
tion has closure (see also formula 4.9).
Modification structures may be open-ended; and limitless
numbers of modifiers can, in certain cases be added at the will
of the speaker. Placing these modifiers in predicate attribute
position, as an expansion of the adjective coordinate phrase
acting as attribute, we might have:
We want students who are bright, honest, diligent.. .
1.3 Ajc0 = +H,:aj +H2:aj +H3:aj .
Closure may be indicated in the same way as m the preceding
noun phrase. In these examples, provision is made for construc-
tions which may be infinitely long. With practical data, however
these potentially infinite constructions are, in fact limited to a
concrete length.
Recursive versus Nonrecursive Layering. As Chomsky has
pointed out (1957:24), grammatical systems must have methods
for handling recursive constructions. Tagmemics handles this
recursiveness with layering formulas, but not all cases of layer-
ing are recursive.
Recursive layering demands a rule of the type X -* X + Y
 TAGMEMIC ANALYSIS 25
where the same symbol occurs on both sides of the arrow or
"rewrite" sign. In tagmemics, this requirement is fulfilled for
example, by allowing whole clauses to fill clause level slots. In
cases where the identical clause level symbol occurs on both sides
of the equals sign, such a construction is infinitely recursive and
may be repeated indefinitely.
The police said the assassin shouted he hoped he had killed him.
1.4 tCl = +S:N +P:TV +O:tCl/pn
The tagmemic formulation is a simple clause formula in which
the symbol tCl occurs on both sides of the equation, showing
recursiveness. This formula, applied four times (as is evidenced
by the repetition of four subjects, four verbs, and four objects),
describes the sentence.
The police said / the assassin shouted .
The assassin shouted / he hoped . .
He hoped / he had killed him.
He had killed / him.
Nonrecursive layering is an inclusion of phrase within phrase
or clause within clause, and need not be recursive. For example
an adjective phrase might be included within a noun phrase.
Thus, in the phrase, a very old man, the word very modifies old
in an adjective phrase. The whole phrase, very old is then one
of the modifiers of the word man.

Etic and Emic Constructions

Just as the tagmeme unit has etic variants, there are also
both etic and emic constructions. The etic construction is the first
guess of the analyst; the emic construction is the construction
as used by a native speaker. Constructions are patterns; in order
for patterns to be distinctive as patterns, they must differ in
at least two points in the construction. In order to determine
the emic constructions of a language, we must: (1) describe the
etic constructions, (2) apply the rule of two differences, and (3)
find the essential constructions of the language.
Etic Constructions. The first step in the discovery procedure
is to write the strings representing constructions according to
what is found in the data. In the beginning, it is better to list
too many rather than too few construction types. Each element
in the construction is marked as obligatory or optional. Even in
preliminary analysis, some of these constructions will show
similarities. At this point, one must consider whether two
similar constructions are to be combined in one formula.
 26 INTRODUCTION TO TAGMEMIC ANALYSIS

The presence or absence of an optional element is not suffi-

cient reason for setting up a new construction type unless the
optional element is also nuclear. Differences in obligatory ele-
ments should be carefully noted, as they probably indicate dif-
ferent construction patterns.
The tagmemic slot may have many different form classes
as fillers. If these fillers are mutually exclusive, they are con-
sidered as filling the same slot in the structure, pending fur-
ther investigation.
The most practical procedure, after identifying the morphs
in the construction is to set up a temporary formula for the
longest utterances in the data, in order to get all of the slot-class
units in the correct order. Many shorter sentences are included
in the maximum formula.

Longacre's Rule of Two. In order for two constructions to be

essentially, or emically, different, they must be different in two
ways. Only with two differences can patterns differ as patterns
in language. One of these two differences must affect the nu-
clear tagmemes, according to Longacre (1964:18):
For two syntagmemes to be in contrast they must have more than
one structural difference between them; at least one of these dif-
ferences must involve the nuclei of the syntagmemes.
In applying the rule of two, it should be noted that nuclear is not
the same as obligatory. The rule specifies that one difference
must be nuclear, but this nuclear tagmeme may be either oblig-
atory or optional.

Essential Constructions. After applying the rule of two,

those constructions that are found in contrast—that is those
found to have two differences, one of which involves the nu-
cleus—are called essential. If two differences are not found the
two constructions are listed as etic variants of the same con-
struction and combined in one formula. Since the string is com-
posed of tagmemes, the application of the rule depends upon
knowing when tagmemes are different, and also depends upon
the definition of nuclear tagmeme.
Tagmemes are listed as essential contrastive units if they
differ in at least two of these features: (1) slot, (2) filler class, or
(3) position. As already outlined in the section on the Unit, a dif-
ference in either slot meaning, or filler class, or position would
only indicate that this is a variant of the essential unit, or tag-
meme.
Tagmemes are nuclear if they are diagnostic of the construc-
tion in which they occur. The following are rules of thumb to
 TAGMEMIC ANALYSIS 27

determine if a tagmeme is nuclear, without reference to the

construction as such:
1. If the tagmeme is obligatory, it is nuclear. Therefore a
difference in obligatory tagmemes is always an essen-
tial difference.
2. If the tagmeme is a transform of a nuclear tagmeme, it
is considered nuclear. Transform potential of a nuclear
element thus specifies such elements as Agent (Ag:)
as nuclear in the passive.
3. Tagmemes in concord with nuclear tagmemes are con-
sidered nuclear at least at the clause level. The agree-
ment in endocentric constructions is excluded from this
rule. Thus, subjects tied to the predicate or objects
governed by the predicate are nuclear.
The real difference in applying the rule of two revolves
around the question of whether the external distribution of the
construction as a whole should be considered as one of two dif-
ferences. According to Longacre, external distribution is not a
countable contrastive feature (1964:21). Constructions should
be regarded as contrastive only when they show two differences
in their internal structure.
Elson and Pickett (1962:134) suggest the use of external dis-
tribution as one of the two differences in their rule: "Assume
that two similar constructions are different construction types
if there is one internal difference . . . and a correlated difference
of distribution of the construction as a whole." Pike (1962:232)
agrees that a difference in external distribution like a trans-
form difference, may be counted, provided this distributional
difference is paralleled by a difference in structural meaning.
In practice, this difference in external distribution has been
found to be useful in limiting the output of the grammar by
setting up two different constructions which differ internally by
only one feature but have a different external distribution.

THE SYSTEM: LEVELS OF GRAMMAR

A grammatical model should have clearly defined units which
enter into constructions, and these constructions should be or-
ganized into some kind of grammatical system. In tagmemics,
the unit is the tagmeme, a correlation of function and form; the
construction is a potential string of tagmeme units, the syntag-
meme; and the system is the grammatical hierachy, arranged in
a series of systematic levels. By geometric analogy, the tagmeme
is a point, the construction a line made up of points, and the
grammatical hierarchy lines arranged from higher to lower.
 28 INTRODUCTION TO TAGMEMIC ANALYSIS

The model resembles an abacus, with beads as units, strung on

horizontal wires to form constructions with the wires arranged
from higher to lower within a frame.

Grammatical Level Defined

A grammatical level is described as a relative position in
space—where space is the grammatical hierarchy within which
constructions may occur. Just as physical celestial objects can-
not be located in space in any absolute way but only in reference
to each other, so levels are relative positions in grammatical
space denned in reference to each other. The concept of "struc-
tural levels, arranged in explicit systemic hierarchy," is the
third fundamental insight of tagmemics (Longacre, 1965:72).
Hierarchical Level. According to Pike, language may be de-
scribed in terms of a triple hierarchy of phonology, lexicon and
grammar. Within the grammatical hierarchy, constructions are
arranged at a series of well-defined levels. The development of
the concept of levels within grammar is attributed to Longacre,
who felt that there should be a close analogy with phonology,
where a single phoneme may be a syllable. Similarly, in gram-
mar a single morpheme may be simultaneously a word, a poten-
tial phrase, and a clause, as in the English command Go! The
most common levels in use are the sentence, clause, phrase, and
word levels.
By grouping constructions at a series of natural levels lan-
guage structure is represented as an orderly mapping of lower
level into higher level structures. Morphemes are mapped into
words, words into phrases, phrases into clauses, and clauses into
sentences, in an orderly manner. This regular mapping of
smaller units into larger units is, by using levels, clearly distin-
guished from nesting and layering tendencies within the same
grammatical level. Provision is also made in the system for
atypical mapping of structures, for level-skipping, and for back-
looping.
Tagmemic versus IC Levels. In tagmemic analysis, the ana-
lyst works with a series of levels within the grammatical hier-
archy, normally the sentence, clause, phrase, word, and mor-
pheme levels. Constructions are arranged in the grammar at
one of the four construction levels. In IC analysis, the only levels
represented are included under the terms immediate, mediate
and ultimate constituents. The immediate constituents are the
results of the first cut, the ultimate constituents are morphemes
and mediate constituents are the results of any intervening
binary cut.
 TAGMEMIC ANALYSIS 29

In the analysis of a simple sentence the first cut in both IC

analysis and in tagmemics is to remove the intonation from the
sentence base. In tagmemics, this is a sentence level construc-
tion; in IC the sentence base and its intonation are the immedi-
ate constituents. The final level of the tagmemic grammar is the
level of ultimate constituents, the morphemes. The clause
phrase, and word level constructions of tagmemics correspond
roughly to the mediate cuts of the IC grammar, except that the
tagmemic levels are well denned whereas the mediate constitu-
ents of IC grammar are ad hoc constituents changing with each
sentence analyzed.
IC analysis, according to Longacre, fails to reveal construc-
tion types of "maximum relevance and comparability" (1964:16),
and fails to distinguish mapping of constructions from lower
into higher levels, from layerings on the same level of analysis
(1965:72). Tagmemics, on the other hand, places constructions
within the grammar at well-defined levels, arranging the levels
from higher to lower within the grammar.
Significance of Levels. The concept of levels places the
strings or syntagmemes of the grammar in an ordered set of
rules.'Just as the idea of string allowed multiple branching in
the grammar, so the idea of level orders the rules from higher to
lower in some well-defined way. Within a level, however con-
siderable vaEiety of construction is found, and ordering within
the level is not so rigidly defined as to exclude different orderings
with differentuSiefes of data.
The purpose of the level is to set up a hyperclass of construc-
tions which are similar in some specified way. Once we have de-
fined these contrastive characteristics, we may speak of sentence
level, clause level, phrase level, and word level constructions.
Each of these constructions is defined, again, in terms of its
constituent tagmemes, which are then recognized as sentence
level clause level, phrase level, and word level tagmemes re-
spectively* Consequently, the concept of level not only places
the constructions in hierarchical order, it allows for a more posi-
tive iidentifteation of the tagmeme unit. The tagmeme is fixed at
a definite level or depth, in the grammar, and at the same time
has a position in the linear construction string.

Kinds of Levels in Grammar

Grammatical levels form a telescopic system in which the
constructions at lower levels are typically mapped into construc-
tions at higher levels in the grammar. In tagmemics, levels
above the sentence are of considerable importance; sentences
occur in larger contexts. In this introduction, however we con-
 30 INTRODUCTION TO TAGMEMIC ANALYSIS

sider only the sentence level and levels below the sentence. This
includes: (1) the five typical levels of grammar, (2) atypical
mapping of levels, and (3) the systematic organization of levels
in a field structure of hierarchy.
Typical Levels of Grammar. The five typical levels of gram-
mar are the sentence, clause, phrase, word, and morpheme levels.
These correspond to the sentence, clause, group, word, and mor-
pheme levels of M. A. K. Halliday's scale-and-category grammar.
The first four levels are levels at which constructions occur. The
final, or morpheme level, is not a level of construction, but an
ultimate point of reference.
1. The sentence level is that level of the grammar at which
the major and minor sentences of language are broken
down into dependent and independent clauses, together
with their intonation patterns.
2. The clause level is that level of the grammar at which
the clauses are broken down into their subjects, predi-
cates, objects, and various clause adjuncts such as time,
place, manner, and circumstance.
3. The phrase level is that level of the grammar at which the
structured word groups which are not clauses are
broken down into words.
4. The word level is that level of the grammar at which the
words of the language are broken down into their con-
stituent morphemes, including the analysis of the pro-
cesses of inflection, derivation, and compounding.
5. The morpheme level is that level of the grammar at
which the morphemes are seen as the ultimate mean-
ingful constituents of which the utterances of a lan-
guage are composed. Each morpheme is listed in a
lexicon accompanying the grammar, with its form
class, and gloss.
A tagmemic grammar will therefore consist of a set of for-
mulas ranging from the .sentence through the clause, phrase
and word levels, and this grammar will be accompanied by a
lexicon of constituent morphemes. The five levels as listed are
typical of most languages. It may be possible to establish more
or fewer levels within a given language, but strong points of
contrast between levels would have to be established in order
to set up new levels in the grammar. A priori, the five-level sys-
tem seems to correspond to natural units of language — found in
most grammars—which native speakers feel intuitively are
sentences, clauses, phrases, and words (see Pike, 1967:444).
Atypical Mapping. Although the normal processes of gram-
mar call for a mapping of lower level constructions into higher
 TAGMEMIC ANALYSIS 31

levels, there are cases of atypical mapping, including level

skips, layering, and loopbacks in the grammatical structure:
1. Level skips are omissions of a level in proceeding from
higher to lower. When a level is skipped, a filler from a
lower level construction is used in a higher level con-
struction. The use of a phrase clitic a bound form to
fill a phrase level slot, is a typical example of skipping.
2. Layering is the inclusion of a construction within an-
other construction at the same level (Elson and Pickett,
1962:59). Thus, clause within clause and phrase within
phrase are examples of layering. This layering may be
recursive or nonrecursive, as explained in the section
on Tagmemes in Construction.
3. Loopbacks are inclusions of higher level constructions
within the slots of a lower level construction. A relative
clause filling the identifier slot in phrase structure is a
typical example of a loopback. (See Practice 2 Clauses
within Phrases, for examples.)

Field Structure of Hierarchy, Longacre has pictured the

typical and atypical mappings of structures in a field (1965:75),
as the fourth fundamental insight in tagmemic grammar. The
following chart is adapted from Longacre's chart in Language
(1965:76), eliminating levels above the sentence level. At clause
level, for example, the chart may be read as follows:
1. If tagmemic fillers at clause level are words (W), this is
level-skipping: the phrase level has been skipped and
the clause slot is filled by a lower level form the word.
2. If tagmemic fillers at clause level are phrases (P), this
is normal mapping: clauses are normally filled by
phrases in proper hierarchical structuring.
3. If tagmemic fillers at clause level are clauses (C), this is
layering at a level: a clause is embedded within another
clause. This layering need not be recursive.
4. If tagmemic fillers at clause level are sentences (S),
this is back-looping: a higher level construction the
sentence, fills a slot in a lower level construction the
clause.
Tagmemic Level Normal Layering Back-
Filters Skipping Mapping at a Level looping
Above the Sentence C S
At Sentence Level P c S
At Clause Level W p c S
At Phrase Level M w p cp
At Word Level M w
 32 INTRODUCTION TO TAGMEMIC ANALYSIS

Etic and Emic Levels of Grammar

The tagmemic system of analysis posits well-defined units
called tagmemes, the correlation of functional slot with filler
class. These units combine in syntagmemes, or constructions.
The constructions in turn are grouped at the essential levels of
grammar.
Just as there are etic and emic units, tagmas and tagmemes,
and etic and emic constructions so also there are etic and emic
levels of grammar. Longacre (1964:16) suggests that these es-
sential levels of grammar are "discoverable" for each language.
Pike (1967:437) suggests criteria for setting up levels within a
given language as follows:
1. Constructions at a level must specify the kind of con-
stituents at that level, including their occurrence and
relative order in construction.
2. Constructions at one level should be sharply in contrast
with those at other levels immediately above and below
the given level.
If we accept the natural levels of grammar as the sentence
clause, phrase, and word then within these levels we find there
are three possible identifying characteristics of levels three
geological strata, with the functions of (1) relation, (2) coordina-
tion and (3) subordination. These features may help to define
the essential levels of grammar.
Relational Strata. Within a given level, provision should be
made for taking a whole construction at that level, and relating
it to higher level constructions. In tagmemic analysis, Longacre
has pointed out the parallelism between relater-axis structures
at the clause and phrase levels, with a proportion evident be-
tween relater and axis (1964:38):
Preposition: its object : : subordinator: its clause
Similar relational markers may be found at sentence level
where sequence markers are used to connect sentences in higher
structures, distinguishing sentences in "absolute" and "in-
cluded" position (Fries, 1952:240). At word level also, the process
of inflection is relational and fits the word for syntactic use. So
close is the parallelism between inflectional ending and preposi-
tion that in some languages they are indistinguishable. These
four natural levels show parallelism in the use of relaters:
At Sentence Level, S-Subordinator + Sentence
At Clause Level, C-Subordinator + Clause
At Phrase Level, Preposition + Phrase
At Word Level Inflection + Word
 TAGMEMIC ANALYSIS 33

In each of these structures the relater is added to a whole sen-

tence, clause, phrase, or word and fits it for use in higher level
structures. If this parallelism is correct, it would seem that re-
lational properties are one identifying-contrastive feature of
an emic level.

Coordinate Strata. Within a given level, provision should

also be made for the coordination of two whole constructions to
form a single coordinate construction at that level. In an article
on scale-and-category grammar, R. D. Huddleston suggests that
coordination be included at the various levels of the grammar
("Rank and Depth," Language, 1965:41). The levels of this gram-
mar are sentence clause, group, and word.
In tagmemic analysis, some provision is made for coordina-
tion at phrase level, where noun phrase + noun phrase = coordi-
nate noun phrase. This is paralleled at the word level by com-
pound words, and at the clause level by the coordination of
clauses to form compound sentences. This parallelism, if followed
rigidly, would produce the following structures:
Coordinated Sentence = Sentence + Sentence
Coordinated Clause = Clause + Clause
Coordinated Phrase = Phrase + Phrase
Coordinated Word = Word + Word
The schema is accurate for compound words and coordinate
phrases, but suggests some indeterminacy between clause and
sentence levels, where the coordination of clauses results in a
compound sentence, and raises the question whether clause and
sentence are really distinct levels. The parallelism does suggest,
however that coordination might be a second identifying-con-
trastive feature of the emic level.

Subordinate Strata. The units at a level must be broken down

into constituent parts. This is the last or subordinate layer of
structure at a level and constitutes the notion of level in its nar-
rowest sense. Whatever else happens at a level whole construc-
tions must be broken into parts.
1. Sentence is composed of bases, margins, and intona-
tion.
2. Clause is composed of subjects, predicates, objects, and
adjuncts.
3. Phrase is composed of head and modifiers.
4. Word is composed of roots and derivations.
If this process of subordination is considered as the third iden-
tifying-contrastive feature of an emic level and occurs in con-
junction with the two features of relation and coordination the
 34 INTRODUCTION TO TAGMEMIC ANALYSIS

subordination feature is limited. It does not overlap with the

other two features.
Word level becomes a single level, with layers of inflections
compounding, and derivation. Phrase level is a single level, with
relater-axis phrases, coordinate phrases and modification struc-
tures. Clause level as found in current use, is mainly subordi-
nate structure with the coordinate and relational features used
to form compound and complex sentences at the sentence level.
If levels are well defined as reference points, they may serve as a
basis for future universal grammars, by forewarning us what to
expect in the structures of different languages.

PRACTICE 1: TAGMEMIC FORMULATION

The tagmemic model consists of a series of syntactic statements

at the sentence, clause, phrase, and word levels. In the analysis
of a single sentence those elements which can be omitted from
the structure are marked as optional, and all other elements are
marked as obligatory. Write the tagmemic formulas at each
level for the following sentence from Longacre (1960:63):
The slow, lumbering covered wagon pulled the pioneer's family
across the prairie just yesterday.

At the sentence level, the sentence base is separated from

its final intonation contour. The intonation slot is marked with
a minus sign to indicate it is suprasegmental. The intonation
if unknown, is marked ICF.
Sentence Level Construction:
Sent = +Base:tCl -Into:ICF
Read: A sentence consists of a base slot filled by a transitive
clause and an intonation slot filled by a final intonation
contour.

At the clause level, the clause that constitutes the sentence

base is analyzed as a string made up of subject, predicate, object,
and the sentence adjuncts of time and place. Slot names are
marked with capitals; fillers have capital letters for word groups
and small letters for words.
Clause Level Construction:
tCl = +S:N +P:tv ±O:N ±L:RA ±T:Tem
 TAGMEMIC ANALYSIS 35
Read: A transitive clause consists of a subject slot filled by
a noun phrase, a predicate slot filled by a transitive verb, an
optional object slot filled by a noun phrase, an optional lo-
cation slot filled by a relater-axis phrase, and an optional
temporal slot filled by a temporal phrase.

At the phrase level, the word groups that fill clause level
slots are analyzed into constituent words. Slot names are marked
with capitals; fillers have capital letters for word groups and
small letters for words.
Phrase Level Construction:
RA = +R:rel +Ax:N across + (the prairie)
Read: A relater-axis phrase consists of a relater slot filled
by a relater (preposition), and an axis slot filled by a noun
phrase.
N = +Det:det±Pos:pos +H:n the + pioneer's + family
Read: A noun phrase consists of a determiner slot filled by a
determiner, an optional possessive slot filled by a possessive,
and a head slot filled by a noun.
Tern = ±Int:int +H:tem just + yesterday
Read: A temporal phrase consists of an optional intensifier
slot filled by an intensifier and a head slot filled by a tem-
poral word.

At the word level, words that consist of more than one mor-
pheme are analyzed into constituent morphemes. Slots have
small letters; fillers have small letters if they are either mor-
phemes or words. (For possessive noun phrase, see formula 4.13.)
Word Level Construction:
pos = +nnuc:ns ±num:plm +pos:posm pioneer + -'s
Read: A possessive consists of a noun nuclear slot filled by a
noun stem, an optional number slot filled by a plural marker,
and a possessive slot filled by a possessive marker.
tv = +vnuc:tvs ±t:tm pull + -ed
Read: A transitive verb consists of a verb nuclear slot filled
by a transitive verb stem and an optional tense slot filled
by a tense marker.
aj = +core:ivs/tvs +ajzr:{-ing}/{-ed} cover + -ed
Read: One type of adjective consists of a core slot filled by
a verb stem (transitive or intransitive), and an adjectivizer
slot filled by the (derivational) suffixes, {-ing} and {-ed}.
 36 INTRODUCTION TO TAGMEMIC ANALYSIS

SUPPLEMENTARY READINGS 1
Cook, Walter A., S.J., On Tagmemes and Transforms, Washington, D.C.
Georgetown University Press, 1964. A summary of the tagmemic
system, including the use of transforms and generative potential.
Elson, Benjamin, and Pickett, Velma, An Introduction to Morphology
and Syntax, Santa Ana, Summer Institute of Linguistics, 1964.
Standard text for beginning morphology and syntax in tagmemics.
Longacre, Robert E., "String Constituent Analysis," Language, 36:63-
88, (1960). Compares tagmemic string construction with IC analy-
sis.
, Grammar Discovery Procedures, The Hague, Mouton & Co.,
1964. For procedures of tagmemics, including the use of transforms
and the generative potential in terms of a set of rewrite operations.
, "Some Fundamental Insights of Tagmemics," Language, 41:
65-76. For tagmeme, syntagmeme, hierarchy, and field structure.
Monograph Series on Languages and Linguistics, 18th Annual Round
Table, No. 20 (1967), E. L. Blansitt, Jr, ed. Conference with panels
on tagmemic theory, current research in tagmemic description
and grammatical analysis.
Pike, Kenneth L., Language in Relation to a Unified Theory of the Struc-
ture of Human Behaviour, Glendale, California, Summer Institute
of Linguistics, Part I (1954), includes Chaps. 1-7, on behavior,
morphemes, Part II (1955), includes Chaps. 8-10, phonemes, syl-
lables, Part III (1960), includes Chaps. 11-17, tagmeme, syntag-
meme. Reprinted in one volume with updated bibliographies and
footnotes on current developments. The Hague, Mouton & Co.,
1967.
"On Tagmemes nee Gramemes," UAL, 24:273-278, (1958). For
the concept of tagmeme and its discovery.
, "Language as Particle, Wave and Field." The Texas Quarterly,
2:37-54, (1959). For the development of the trimodal nature of the
unit.
"Dimensions of Grammatical Structure." Language, 38:221-
244, (1962). For the exposition of matrix theory as applied to gram-
mar,
"A Syntactic Paradigm." Language, 39:216-230, (1963). For
matrix theory applied to a practical example.
"A Guide to Publications Related to Tagmemic Theory." Cur-
rent Trends in Linguistics, Thomas A. Sebeok, ed., The Hague,
Mouton & Co., Vol. Ill (1966) 365-394. A survey and critical ap-
praisal of literature on tagmemic theory and practice.
 TABLE 2: SENTENCE LEVEL ANALYSIS
According to the According to the According to the According to the
Type of Intonation Type of Base Type of Clauses Type of Situation
1. Compound Sentence 1. Statement (S)
two independent clauses formed to relay
with conjoining information
Major Types 2. Complex Sentence 2. Question (Q)
with independent and dependent clause formed to elicit
Complete Base with embedding answer response
3. Simple Sentence 3. Command (C)
one independent clause, formed to elicit
Sentences no dependent clause action response
with
Final Intonation
1. Sequential Sentence 1. Addition (A)
from compound with statements
2. Marginal Sentence 2. Response (R)
Minor Types from complex with questions
with 3. Elliptical Sentence 3. Exclamation (E)
Incomplete Base from simple in any context
Nonelliptical Types Vocatives, greetings,
no clause structure calls, titles, mottoes
 SENTENCE LEVEL

The sentence level of grammar is that level at which clauses are

combined into larger units. All sentences occur in some linguistic
or nonlinguistic context, but the structures above the level of
sentence are so varied that it is more practical for the beginner
to focus attention upon sentence structure and the structure of
levels below-the sentence. The sentence is a grammatical unit a
construction in which the constitute is any utterance with final
intonation contour, and the constituents are the clauses con-
necting particles, and intonation patterns.
Sentence Defined. The sentence has been defined by Bloom-
field as "an independent linguistic form, not included by virtue
of any grammatical construction in any larger form" (1933:170).
This definition is paraphrased by Hockett as "a constitute which
is not a constituent; a grammatical form which is not in con-
struction with any other grammatical form" (1958:199). In tag-
memic analysis, this grammatical independence is accepted as a
relative independence, which must be structurally defined for
each language. The possibility of higher level analysis is never
excluded. In the definition of the sentence, the following features
are essential:
1. Sentences are relatively isolatable. Sentences may be
isolated, and any corpus may be reduced to sentences,
with no residue (Longacre, 1964:17, fn. 14). The sentence
is isolatable in its own right (Pike, 1967:442).
2. Sentences have final intonation patterns. In a practical
approach to sentences, we accept them as we find them.
Intonation patterns, with pause and other phonological
features, help to isolate sentences.
 40 INTRODUCTION TO TAGMEMIC ANALYSIS

3. Sentences are composed of clauses. The clauses combine

in some kind of "patterned dependency" (Longacre,
1964:126) involving combinations of clauses which do not
have the overall structure of a single clause. These
clauses may be dependent or independent clauses.
Traditional Sentence. The sentence has traditionally been
the focal point of attention in grammatical work, but its defini-
tion was based upon meaning. The sentence was denned as con-
taining "a complete thought" and was composed of a subject
"about which something is said" and a predicate, "what is said
about the subject." This tradition dates back to Aristotle and
the Greek logicians and is carried over via Priscian's Latin
Grammar into the English grammars of today.
In the linguistic tradition dating from Bloomfield, an attempt
was made to define the sentence, not according to its meaning
but according to its form, as an independent linguistic form.
Tagmemic analysis would add the further dimension of distribu-
tion and claim that the sentence is only one level of the gram-
matical hierarchy. Sentences are distributed in higher units
and ultimately in the context of behavior. Discourse analysis,
however has been little developed. Elson and Pickett treat
levels above the sentence only briefly (1962:127-128), and Long-
acre barely mentions them (1964:17, 125-126). Part of the diffi-
culty is the fact that structures above the sentence occur in a
wide variety of literary genres, including dialogue, poetry, and
narration. Further references to levels above the sentence may
be found in Pike (1966:372).

Sentence Types. The definition of sentence becomes more

concrete once the decision is made as to what is acceptable as a
sentence. The notion of sentence is realized in concrete sentence
types. Classification of sentences may be based on any of the
following features:
1. According to the number and kind of clauses in the base
sentences are classified as simple, complex, and com-
pound. This feature is used by Pike (1967:442-443) and
Longacre (1964:130) to separate sentences into multiple
clause, simple clause, and nonclause types.
2. According to the internal structure of the mam clause
sentences are complete or incomplete. These are also
called major, or favorite, or full sentences, as opposed
to minor sentences (see Bloomfield 1933:171, and Hoc-
kett, 1958:200). Minor types are completive or exclama-
tory.
3. According to the type of response expected, sentences
are classed as statements, questions, and commands.
 SENTENCE LEVEL 41

Although included in traditional grammars, these

sentence types can be denned formally for a language.
4. According to the nature of the actor-action relationship,
sentences are classed as active, middle, or passive. This
classification is based on the voice feature and belongs
more properly to clause level.
5, According to the presence or absence of negatives in the
mam verb phrase, sentences are classed as affirmative
or negative sentences. This classification is best handled
at the level of the verb phrase, at least in English.

SENTENCE LEVEL SORTING

In tagmemic analysis, it is possible to begin work at any of the
well-defined levels of the grammar. The practical procedures
followed here will be to begin at the sentence level and proceed
to the morpheme level, omitting levels above the sentence. The
sentence level is then practically viewed as the starting point
m analysis. Accepting the corpus as a set of sentences the first
task of the analyst is to institute a sort of sentence types and
reduce the continuous narrative or dialogue to sets of homoge-
neously grouped sentences of well-defined types. The task here
is threefold: CD to separate the kernel sentences from derived
sentences; (2) to reduce multiple clause sentences to simple clause
types; and (3) to separate sentence base from sentence intonation.

Kernel Sentences
Within a given language, certain sentences belong to a set
of basic structures, and all other structures may be expressed
as derived from these basic structures. The basic structures
are called kernel sentences. We first identify the distinctive
features of kernel sentences; then consider the derived sen-
tences opposed to these basic structures; and finally show how
the nonkernel sentences are derived from kernel sentences.
Distinctive Features of the Kernel. Traditionally, language
analysis has focused upon the declarative sentence as of primary
importance. With the introduction of transformational grammar,
analysis has once more concentrated upon this basic type. In
response to the question of which set of sentences belongs to the
kernel, Chomsky answers, that for English at least, "the kernel
consists of simple, declarative, active sentences" (1957:80). How-
ever since both negative transformations and deletion trans-
formations are included in the grammar, we assume that the
kernel is also affirmative and nonelliptical.
A kernel sentence is defined as a sentence of the language
 42 INTRODUCTION TO TAGMEMIC ANALYSIS

that is (1) simple, (2) complete, (3) statement, (4) active and
(5) affirmative. Any sentence that simultaneously has these
five distinctive features is a kernel sentence; any sentence that
lacks even one of these five features is a derived sentence. In
our analysis, we first institute a sort of all sentences in the
corpus to separate the kernel set from the nonkernel or de-
rived set. Later on, we attempt to describe the derived sentences
in terms of the kernel set, using either transformational rules
or matrix displays, to show how nonkernel sentences are derived.
In this way we concentrate attention on the basic sentence
structures.
Derived Sentences. The notion of kernel sentences is more
clearly understood when kernel and nonkernel sentences are
compared. The derived sentences lack at least one of the five
distinctive features of kernel sentences. Kernel and derived
sentences may be contrasted as follows:
Kernel Sentences Derived Sentences
Simple versus Complex, Compound
Complete versus Incomplete, Elliptical
Statement versus Question, Command
Active versus Middle, Passive
Affirmative versus Negative
In a sentence level sort, kernel sentences are separated from
derived sentences, and primary attention is focused upon kernel
sentences. The analysis of derived sentences is then related to
the simple structures from which they are derived.
Transformation. The transformational rule is simply a rule
of change. This rule has an input string, a rule of change, and an
output string. With kernel sentences as input, it is possible to
set up a series of optional rules that will produce the output,
the derived sentences:
INPUT STRING T-RULE OUTPUT STRING
Kernel sentence + transformation = derived sentence
Tagmemic grammars formerly described all the sentences of a
language by describing both kernel and derived sentences. The
resulting description was a complete description, but often failed
to show the relationships between similar sentences. With the
introduction of transformational rules or matrix devices to show
the relationships between sentences, it is still necessary to
describe both kernel and derived sentences in order to discover
the differences between structures. However, the final grammar
may be considerably simplified by employing some type of trans-
 SENTENCE LEVEL 43

formational rule or matrix display, together with an analysis of

only kernel sentences.
Early models of transformational grammar produced derived
sentences by a series of optional transformational rules, deriving
the nonkernel passive by an optional passive transformation
for example. In more recent transformational models the kernel
set of sentences is a "proper subset" of sentences generated;
derived sentences are also generated by the grammar by includ-
ing optional; markers in the base. The changes in methodology
in transformational grammars do not destroy the "important
intuitive significance" (Chomsky, 1965:18) of the kernel.

Primary Sorting Procedures

The sorting process at sentence level is a preparation for the

work of analysis at clause level. The first procedure is to reduce
complex and compound sentences to simple clause structures.
The results of this reduction should be the clauses of the corpus
written one to a line, ready for analysis. The first procedure
then is to (1) reduce the corpus to simple clause structures
(2) identify complex and compound sentence types, and (3) iso-
late intonational patterns.

Reduction. In the reduction process, the corpus is considered

to be a body of sentences, made up of simple clauses. These
clauses may be dependent or independent. An independent
clause is a clause that may stand alone in a given language as a
simple major sentence; a dependent clause is a clause which
may not, in this language, stand alone as a simple major sen-
tence type. If we consider independent clauses as type A and
dependent clauses as type B, with sentence boundary marked
as #, we can define the following sentence types:
Simple Sentence # A# one A-clause only
Complex Sentence # A + B # one A at least one
B-clause
Compound Sentence # A + A # more than one
A-clause
In the reduction of complex and compound sentences to
simple sentences, we are performing an operation that is the
reverse of the conjoining process for building compound sen-
tences, and of the embedding process for building complex
sentences. It is important that the formulas for complex and
compound sentences be preserved as the corpus is prepared
for clause level analysis. The formulas for these nonsimple
 44 INTRODUCTION TO TAGMEMIC ANALYSIS

type sentences will then be included in the sentence level. In

this analysis, the following types of clauses may be found:
1. Independent clause types
a. Simple independent type A
b. Sequential independent type (c) + A
2. Dependent clause types
a. Subordinated type (r) + A = B
b. Subordinate type B = A type, with
built in (r)
Whether clauses are dependent or independent in the sentence
structure the clauses may be analyzed together at the clause
level neglecting the connecting particles (c) and relating par-
ticles (r) which pertain rather to sentence structure.
The results of the reduction process will be a series of simple
clauses and a list of sentence formulas for complex and com-
pound sentences complete with intonational patterns. Further
sorting of simple clauses isolates the complete statements ac-
tive and affirmative.
Embedding Process. Dependent clauses are embedded within
the structure of language at the sentence, clause, and phrase
levels. The process of subordinating a clause to another struc-
ture is called the embedding process. Within the tagmemic system
a decision concerning the level of embedding must be made for
each dependent clause.
1. At sentence level, dependent clauses are embedded in
sentence structures and fill marginal slots in that
structure. The resulting structure is a complex sentence,
formulated as margin + base. The structure of a complex
sentence may be indicated by: (a) choice of clause type,
(b) connectors, (c) sequence of tenses, (d) order of clauses
(Longacre, 1964:128). Conditions, indirect discourse,
and so on may be of this type.
2. At clause level dependent clauses are embedded in the
clause structure and fill the same slots as words and
phrases. The resulting structure is a simple sentence
with the overall structure of a single clause (Longacre,
1967:17). Time and place clauses are of this type.
3. At phrase level, dependent clauses are embedded into
the phrase structure as modifiers. They fill identifier
slots in the phrase formula and are an example of loop-
back in the hierarchy of clause within phrase. These
are principally relative clauses, modifying single words
in the phrase structure.
Conjoining Process. The conjoining process deals with the
joining of similar constituents at any level. The elements con-
 SENTENCE LEVEL 45

joined may be clauses, phrases, or words. The conjoining process

at the sentence level deals with the conjoining of clauses, that is,
with the conjoining of strings of tagmemes which contain one
and only one predicate in the string.
According to Alton L. Becker (1967:110), conjoining is (1)
recursive in the linear sense, (2) context sensitive in that con-
stituents must be of the same type, and (3) generally optional.
Conjoining at the clause and phrase levels deals with the joining
of single tagmemes and need not be traced to complete under-
lying sentences. The constituents joined are the same in that
they fill the same slot, but the fillers of the slot may be words or
phrases.
In contrast, the conjoining of clauses to form sentences deals
not with single tagmemes, but with strings of tagmemes. The
strings which are conjoined are not necessarily complete struc-
tures nor are they always dominated by the same node of the
underlying phrase structure. Thus, in She rode and I walked
home S and P are conjoined; but the locational, L, while prop-
erly belonging to both clauses occurs once. One method of han-
dling the problem is to consider both clauses as complete struc-
tures and allow optional deletion of elements in one of the clause
structures thus deriving the compound from two simple under-
lying complete clauses.

Multiple Clause Formulation

In a tagmemic analysis restricted to the sentence and lower
levels of analysis, the sentence level serves chiefly as a clearing
stage to prepare the work of clause level analysis. Sentence level
structures are recorded as simple or multiple clause structures,
with their intonation patterns. The underlying structure of the
sentence is then analyzed at the clause level.

Sentence Level Tagmemes. Sentence level constructions are

those constructions in which the constitute is an isolated utter-
ance, and the constituents are the clauses which constitute the
sentence base, together with connecting particles and intona-
tional or junctural features. The principal tagmemes useful in
sentence level analysis are:
Base: Base tagmeme, filled by independent clauses (IndCl).
Multiple base slots may be subscripted: Base t Base2.
Marg: Margin tagmeme, filled by dependent clauses (DepCl).
Margins occur only in complex and marginal sentences.
C: Connector tagmeme, filled by connecting conjunctions
called connectors (c). These also occur at phrase level.
Into: Intonation tagmeme, filled by intonation contours marked
as final (ICF), as nonfinal (ICN), or by pattern: 231 I.
 46 INTRODUCTION TO TAGMEMIC ANALYSIS

The base and intonation tagmemes are always nuclear to the

structure; the margin tagmeme is generally optional and periph-
eral; and the connector tagmemes may be nuclear if they are
obligatory to the construction.
Sentence Level Constructions. Typical constructions at the
sentence level are the formulations for simple, complex and
compound sentences. Each of these constructions will be accom-
panied by intonation patterns.
2.1 (Simple) Sent = +Base:Ind.Cl -Into:ICF
Read: A (simple) sentence consists of a base slot filled by
an independent clause and an intonation slot filled by a final
intonation contour.
In this simple sentence formula, the + sign means that the
sentence base is obligatory. The — sign also indicates an oblig-
atory tagmeme, but shows that the tagmeme is a suprasegmen-
tal, and not in linear sequence. The intonation pattern, if not
known is simply labeled ICF, for final intonation contour or
ICN for nonfinal intonation contour. If, however, the pattern
is known, it may be marked, for example, as 231 | .
In early stages of the analysis, the fillers of base and margin
slots, respectively, are marked as independent clause (Ind.Cl)
and dependent clause (Dep.Cl). Once the type of clause is known,
however these temporary symbols are replaced by concretely
designated clause types.
Complex sentences consist of one independent and at least one
dependent clause. Their sentence level formulas contain a
margin tagmeme when the dependent clause is analyzed as part
of sentence level structure.
2.2 (Complex) Sent = +Base:Ind.Cl
±Marg:Dep.Cl -Into:ICF
Read: A (complex) sentence consists of a base slot filled by
an independent clause, an optional margin slot filled by a
dependent clause, and an intonation slot filled by a final
intonation contour.
Compound sentences consist of at least two independent clauses,
and may or may not have dependent clauses as well. Such a
structure will have at least two base tagmemes, and often con-
nector tagmemes between clauses.
2.3 (Compound) Sent = +Base,:Ind.Cl ±C:c
+Base2:Ind.Cl -Into:ICF
Read: A (compound) sentence consists of a base slot filled
by an independent clause, an optional connector slot filled
 SENTENCE LEVEL 47

by a connector, a base slot filled by an independent clause,

and an intonation slot filled by a final intonation contour.
Complex and compound sentences may show more than one
intonation contour. If it is necessary to record more than one
intonation pattern in the same sentence a second intonation
slot may be introduced for each clause in the pattern, usually
filled by nonfinal intonations.
2.4 Sent = +Base:Ind.Cl -Into:ICN
+Marg:Dep.Cl -Into:IGF,
2.5 Sent = +Base,:Ind.Cl -Into:ICN ±C:c
+Base2:Ind. Cl -Into.-ICF
Sentence as Initial String. In this introduction to tagmemic
analysis, the symbol #Sent# is accepted as the initial string, or
absolute starting point for analysis. Every analysis begins with
this symbol, and proceeds as far as the ultimate constituents, or
morphemes, in the utterance. In this limited analysis, every
structure is dominated by a sentence symbol. This method is con-
sistent and allows comparison with the structures of the trans-
formational grammars. Secondly, a place is always provided for
intonational patterns in the language, whether these are known
or not known. The procedure is comparable to IC analysis, where
the first cut separates the sentence base from its intonational
pattern. In tagmemic analysis, the intonation pattern has a
place only in sentence level construction, and contrasts of in-
tonation must be handled at this particular level.

MAJOR SENTENCE TYPES

The kernel sentences of the language have been described as
those which are: <1) simple, (2) complete, (3) statements, (4) active,
and (5) affirmative. First, the simple sentences were opposed to
complex and compound sentences. Now we investigate the fea-
ture of "completeness," and distinguish between major sentence
types, which are complete sentences, and minor sentence types,
which are noncomplete.

Major Type Defined

The major sentences of a language are sentences which have
the feature of completeness; minor sentences lack this feature.
The judgment as to whether or not a sentence is complete is
determined by the type of clause base involved. If the base is an
independent clause, the sentence is complete; if the base is only
a dependent clause, or has no clause structure at all then the
 48 INTRODUCTION TO TAGMEMIC ANALYSIS

sentence is incomplete. A major sentence is a sentence which has

at least one independent clause.
Major sentences, as understood here, are the set of sentences
described by Bloomfield (1933:170-172) as full sentences favor-
ite types, with the meaning "complete and novel utterance."
The emphasis here is on the word "complete," which is opposed
to "noncomplete" minor types.
Complete versus Incomplete. Whether or not a sentence is
complete depends upon whether it contains an independent
clause. Since an independent clause is a clause which can stand
alone as a sentence this definition is circular unless criteria are
given for when a clause may stand alone as a sentence. This will
depend upon the structure of the language in question.
English, for example, requires that an independent clause
of the statement type contain both subject and predicate. In
Latin, where the subject is already contained within the verb
form the subject is not obligatory. Sentences such as veni, 'I
came are already complete.
Complete sentences may be simple, complex, or compound,
since all of these structures contain at least one independent
clause. But if the reduction process is completed first then the
single clause structures must be tested for completeness in the
second step. In the early stages of analysis, the "standing alone"
property may be tested with a native speaker; but as the analyst
completes clause level analysis, he should become aware of the
formal features of the particular language that determine when
a clause may stand alone and when it may not. The intuitive
analysis is then reduced to a formal analysis of the structure.

Major Kernel Sentences. A sentence that has the feature of

completeness, and contains at least one independent clause is
a major sentence in the language. It may be simple, complex, or
compound. Of these major sentences, only the simple major types
belong to the kernel; the other major types, and minor sentences
are derived from the kernel. The features of simplicity and
completeness may be used to set up a classification of kernel
and derived sentence types as follows:
1. Well-formed sentences (formata) are both simple and
complete. These sentences have one and only one in-
dependent clause—a clause which, by formal criteria
can stand alone in the language as a major sentence.
This set of well-formed sentences contains the kernel
as a proper subset.
2. Transformed sentences (transformata) are complete, not
simple. These sentences are complex and compound.
They are major sentences because they have at least
 SENTENCE LEVEL 49

one independent clause, but are not part of the kernel

because they are not simple. They may be derived from
simple sentences by applying the processes of embed-
ding and conjoining.
3. Incomplete sentences (deformata) are simple not com-
plete. These sentences include both dependent clause
structures and nonclause structures which occur in
the language as minor type sentences. When the clause
structure is only partial, these sentences may be de-
rived from simple, complete sentences by a process of
deletion.

Simple Complete Sentences. The well-formed sentences that

belong to the kernel next become the focus of attention for the
analyst. The corpus is reduced to simple sentences and the
complex and compound major types are rejected as not belong-
ing to the kernel. The simple sentences are sorted for complete-
ness, and the minor type sentences with incomplete structure
are rejected as derived. What remains is a set of simple complete
sentences; these are then sorted according to the remaining
three features of the kernel:
1. Statements, questions, and commands which are identi-
fied according to the type of response expected. State-
ments (S) expect no particular response; questions (Q)
expect an answer; and commands (C) expect an action
response.
2. Active middle, and passive sentences are identified
according to the voice of the main verb form. Active
sentences have subject-as-actor; passive sentences have
subject-as-goal; and middle sentences have subject as
both actor and goal. Some languages have a fourth
voice called reciprocal, in which the subject and object
act in reciprocal action.
3. Affirmative and negative sentences are identified by
the presence of a negating particle within the mam
verb phrase. The sentence may be said to be negative
when a negative particle occurs with the main verb.
These last three features are used to form a kernel sentence
matrix.
Sentence Level Matrix
Once the complex and compound sentences have been broken
down by reduction and the complete sentences isolated from the
incomplete, a sentence level matrix is set up that deals with
simple, complete sentences. This matrix distinguishes the kinds
 50 INTRODUCTION TO TAGMEMIC ANALYSIS

of sentences described in the other three features of the kernel

namely: (1) statement, as opposed to question and command;
(2) active, as opposed to middle and passive; and (3) affirmative
as opposed to negative.

Parameters of the Matrix. A matrix in mathematics is an

array of rows and columns, in which the elements are the coeffi-
cients for a set of simultaneous equations. In linguistic analysis,
a matrix is an array of rows and columns, an array in the sense
m which this word is used in computer analysis. The elements of
the array may be complete construction symbols or units which
enter into construction.
The notion of matrix was introduced by Pike in a series of
articles entitled "Dimensions of Grammatical Structure" (Lan-
guage, 1962:221-244) and "A Syntactic Paradigm" (Language,
1963:216-230), in which grammatical structures are charted in
much the same way as morphological paradigms. Longacre like-
wise makes use of matrices as well as transformational rules,
to show relationships between constructions. These two methods
of showing structural relationships are "not immiscible" (Long-
acre, 1960:16). Both methods may be used in the same grammati-
cal description.
A grammatical matrix charts constructions in different di-
mensions or parameters. These parameters must be chosen
before the matrix can be set up. For the sentence level matrix
two useful parameters are:
1. Sentence types, such as statement, question, and com-
mand.
2. Basic clause types, such as transitive, intransitive, and
equational.
Together, these parameters give us a 3 x 3, or 9-element matrix
of sentence types. These nine sentence types are useful for early
analysis.

CLAUSE
TYPE STATEMENT QUESTION COMMAND
S-iCl Q-iCl C-iCl
intransitive j o hn went. DidJohn go? Go, John!
S-tCl Q-tCl C-tCl
iransmve John ate it. Did John eat it? Eat it, John!
E uational S-eqCl Q-eqCl C-eqCl
qua lona j o n n is good. Is John good? Be good, John!
Matrix Operators. The analogy of a mathematical matrix
may be carried one step further by the introduction of matrix op-
erators. In matrix algebra, the matrix may be used as a unit in
 SENTENCE LEVEL 51

calculations. The whole matrix is multiplied by a single number,

by multiplying each element in each cell of the matrix by the
number, called an operator.
In tagmemic field theory, the derivation of construction types
is represented as the multiplication of a matrix by a constant
(Pike, 1962:226), Some of the constants which might occur in this
use are the features of interrogation, emphasis, quotation, and
negation. Each of the elements in the matrix is multiplied by
the constant to derive new types (Pike, 1967:473, fn. 9). This
operation is essentially the same as transformational rules
which derive negatives, passives, and so on. In the sentence
matrix given above, the first column is the basic kernel and
the Q and C columns are derived from the first column by use
of Q and G operators respectively. The whole 9-element matrix
may then be operated on by (operator #1) negative.
S-iCl Q-iCl C-iCl
S-tCl Q-tCl C-tCl /Negative/
S-eqCl Q-eqCl C-eqCl
The product will be a 9-element negative matrix, with nine nega-
tive sentence types in addition to the nine affirmative sentence
types in the original matrix. The transitive elements of the
matrix may be operated on by (operator #2) voice.
S-iCl s-tcip s-tcim s-tcir
Q-tCl /Voice/ Q-tClp Q-tCl m Q-tClr
C-tCl C-tGL c-tcim C-tClr
With the use of both operators, negative passive sentences are
derived. Beginning with three clause types, a 9-element matrix
is developed. With the use of both operators, 36 different sen-
tence types may be derived. Of these 36, only three are kernel
types and the other 33 are derived. If voice is only active/passive,
as in English, there are three basic and 21 derived, or 24 types.
MAJOR SENTENCE TYPES
At Sentence Level At Clause Level
Compound Statement Intransitive
Complex Question Transitive Active Affirmative
Simple Equational Passive Negative
Command

Simple Clause Formulations

The sentence level is that Level at which clauses combine to
make sentences. In the case of simple clause structures it is the
level at which single clause bases combine with intonation pat-
terns. In the formulation of these structures only base and in-
 52 INTRODUCTION TO TAGMEMIC ANALYSIS

tonation tagmemes are required, with possible special tagmemes

for sentence level particles. Simple major sentences are state-
ments, questions, or commands; the clauses that fill the base
slot are transitive, intransitive, or equational. In an initial sort
of simple sentences these features of the sentence matrix are
considered first. The voice and negative operators are considered
at the clause level of construction. Applying this type of sort to
a problem in Sierra Popoluca (#165, Laboratory Manual, 1967),
a corpus of 75 sentences was separated into five homogeneous
sets as follows:
s-tci S-iCl S-eqCl C-tCl C-iCl
17 25 14 8 11
Working on the divide-and-conquer principle, the corpus was
reduced to five manageable problems, which could be easily and
clearly charted. No questions occurred in the corpus given.
Statements. Turning our attention first to the statements,
which belong to the kernel, we formulate the sentence type for
statement, assuming all statements have similar intonation and
that this intonation is in contrast with the intonation patterns
of the command sentences. In the formulas, sentences are as-
sumed to be statements unless marked otherwise.
2.6 Sent = +Base:iCl/tCl/eqCl -Into:ICF
Read: A sentence consists of a base slot filled by an intran-
sitive clause a transitive clause, or an equational clause,
and an intonation slot filled by a final intonation contour.
The individual fillers of the base slot, labeled as tCl, iCl, and eqCl
are then expanded at the clause level. The two points of contrast
between this sentence level formula and that for the command
sentences given below are found in the fillers of the base and
intonation slots respectively.
2.7 C-Sent = +Base:C-iCl/C-tCl -Into:C-ICF
Read: A command sentence consists of a base slot filled by
an intransitive or a transitive command clause, and an
intonation slot filled by a final intonation contour for
commands.
The fillers of the base slot labeled as C-iCl and C-tCl, are then
expanded at the clause level, and are demonstrably different
from the corresponding fillers of the statement type sentence.
Commands. The commands of language, marked as C-iCl
and C-tCl, usually differ from statements in their internal clause
structure and may differ as well in their intonation patterns.
 SENTENCE LEVEL 53

Subjects in the second person are omitted or replaced by voca-

tives; verbs are in the imperative mood; and often the negative
particles used with commands are distinctive. Other differences
may occur, including restrictions on peripheral tagmemes. The
differences at the sentence level may be listed as:
1. Intonation Pattern. May be special for commands; dif-
ferences may be represented by a substitution trans-
formation.
2. Base Fillers. May be different clause level strings as
fillers; differences may be represented by a substitution
transformation.
3. Command Markers. May be marked by special particles;
differences may be represented by an addition trans-
formation.
Questions. The questions of language may differ from state-
ments by intonation pattern, by the use of special question mark-
ers, or by different internal clause structure in the fillers of the
base slot. Question words are used in most languages, and
change of order is a question-marking signal in English. Other
changes may occur, such as limitation in the use of peripheral
tagmemes in the clause structure. Differences at the sentence
level may be listed as:
1. Intonation Pattern. Some patterns are used as question
signals; differences may be represented by a substitu-
tion: 231 i;-» 233 f.
2.8 Q-Sent = +Base:iCl/tCl/eqCl -Into:2331
Read: A question sentence consists of a base slot filled by
an intransitive a transitive, or an equational clause, and
an intonation slot, filled by the intonation pattern: 233 f •
2. Base Fillers. Question type clauses occur, as Q-iCl, Q-tCl,
Q-eqCl in which the clause structure which forms the
sentence base is different, while the intonation pat-
tern may be the same as for statements.
2.9 Q-Sent = +Base:Q-iCl/Q-tCl/Q-eqCl -Into:231j
Read: A question sentence consists of a base slot filled by
an intransitive, a transitive, or an equational question
clause, and an intonation slot filled by the intonation
pattern: 2311.
3. Question Marker a. Questions may be signaled by special
question markers, marked QM, used to transform any
statement into a question; for example, Hindi uses un-
stressed {kyaa} as one method of marking questions.
 64 INTRODUCTION TO TAGMEMIC ANALYSIS

2.10 Q-Sent = ±QM:{kyaa} +Base:iCl/tCl/eqCl

-Into:233 f
Bead: A question sentence consists of an optional question
marker slot filled by {kyaa} and a base slot filled by an in-
transitive, transitive, or equational clause, and so on.

MINOR SENTENCE TYPES

An utterance either has final intonation or it does not have final

intonation. Those utterances found with final intonation con-
tours are called sentences; those without final intonation contour
are fragments. It is to be expected that with this definition of a
sentence, many types of sentences will be found that do not have
complete clause structures as a sentence base. These are the
minor sentences of language.

Minor Sentence Defined

A minor sentence is a sentence with an incomplete sentence
base and a final intonation contour. Since the contour is final
the utterance is a sentence; since the base is incomplete the sen-
tence is a minor type of sentence. Some minor sentences have
clause structure while others have no clause structure intended
or implied. Of the minor sentences with clause structure some
are (1) sequential; some are (2) marginal; and some are (3) ellipti-
cal, with only partial clause structure.

Sequential Sentences. Minor sentences of the sequential

type are those sentences which, in fact, contain a complete ind^
pendent clause, but also contain an obligatory sequence-marking
tagmeme. This type of sentence is not used to initiate discourse,
but is found in included position, with recognizable conjunctions
as sequence signals (Fries, 1952:250), for example: also, besides
consequently, however, likewise, thus, yet, and so on. Elson and
Pickett (1962:121), following Waterhouse, "Dependent and Inde-
pendent Sentences" (IJAL-.29,45-54,1963) classify the sequential
sentence as a minor type. The sequence sentence has a connector
and an independent clause:
2.11 Seq-Sent = +C:c +Base:Ind.Cl -Into:ICF
Read: A sequence sentence consists of a connector slot filled
by a connector, a base slot filled by an independent clause,
and an intonation slot filled by a final intonation contour.
This minor sentence type is related to the compound sentence
and the formula is identical with the second half of the compound
sentence formula (see formula 2.3). Without the final intonation
 SENTENCE LEVEL SS

contour, this sentence could be joined to a preceding simple sen-

tence to form a single compound sentence. Without the sequence
marking tagmeme, since the base is filled by an independent
clause, the structure could stand alone as a simple sentence in
the language.
Marginal Sentences. The marginal sentence has a margin
slot filled by a dependent clause and a final intonation contour.
This type is derived^-from the complex sentence (see formula 2.2).
and could be joined to a preceding simple sentence to form a
complex sentence.
2,12 Marg-Sent = +Marg:depCl-Into:ICF
Read: A marginal sentence consists of a margin slot filled
by a dependent clause, and an intonation slot filled by a
final intonation contour.
The filler of the margin slot might be either a dependent clause
of the relater-axis type, with the clause subordinated by a relater
particle; or the filler might be an indefinite clause, in which the
subordinating particle is also a constituent of the clause in which
it stands, for example, indefinite pronoun.
Elliptical Sentences. The elliptical sentence is a sentence in
which the clause structure is not complete due to the deletion
of some of its elements. This sentence type is directly derived
from the simple sentence type (see formula 2.1). These sentences
include* or are interpreted in terms of, elements understood
m traditional grammar, or have a nonfinite verb filling the
predicate slot in the structure.
These structures are similar to sequential or marginal struc-
tures, but differ from them in that they do not contain either a
complete independent clause or a complete dependent clause;
the clause structure here is a partial structure, or a nonfinite
structure.
1. Deletions. After sequence marking tagmemes, subject dele-
tion, is common. The sequence marker connects the minor
sentence with the preceding sentence, and the subject is
understood to be the same.
He finished his work at the office.
And then went home, (subject omitted)
In marginal minor sentences, the subject and auxiliaries
are often deleted with a resulting participial form filling the
predicate slot.
What are you doing this evening?
Watching television, (subject and auxiliary omitted)
 56 INTRODUCTION TO TAGMEMIC ANALYSIS

2. Nonfinite Verb Forma. Partial clause structures also occur

as minor sentences, which in normal use would be nominals
in clause structure. The internal structure of these clauses is
characterized by the form of the verb filling the predicate
slot which might be either (1) an infinitive form (marked
+P:Vlnf) or (2) a participle form (marked +P:Vpart). These forms
occur in combination with other clause level tagmemes, al-
though the subject is generally omitted from the string.

Types of Minor Sentences

The minor sentences of language are not only characterized

by their formal structure as sequential, marginal, or elliptical.
They are also characterized, according to their function in the
context of the situation, as (1) additions, (2) responses, and (3) ex-
clamations. Although the tagmemic structure has been outlined
according to form the label of the construction should coincide
with its use in the language.
Addition Sentences. Addition sentences (marked A-Sent)
are minor sentences that occur in discourse as additions to state-
ments already made. They occur with final intonation and are
sentences. They do not have full clause structure and are minor
sentences. In form, they may be sequential, marginal, or ellip-
tical; or they may have only the structure of a word or a phrase.
They are formulated:
2.13 A-Sent = +Base:depCl/phrase/word -Into:ICF
Read: An addition sentence consists of a base slot filled by
a dependent clause, phrase or word and an intonation slot
filled by a final intonation contour.
These additions qualifications, corrections, or afterthoughts
need not be fully expressed in the language, as they are clearly
understood in the given context as added to the previous sen-
tence. The addition is part of a major sentence, a major sentence
which can be reconstructed into a full sentence using elements
from the preceding sentence. Thus:
He is leaving for summer vacation soon. (Statement)
The day after tomorrow. (Addition sentence)
He is leaving for summer vacation the day after tomorrow.
Response Sentences. The response sentences (marked R-
Sent) are minor sentences which occur in response to questions.
They are sentences with final intonation, but they often are
minor sentences of the marginal or elliptical types, or have the
 SENTENCE LEVEL 57

structure of only a phrase or a word. Short responses are very

frequent. Thejr are formulated:
2.14 R-Sent = +Base:depCl/phrase/word -Into:ICF
Read: A response sentence consists of a base slot filled by
a dependent clause, phrase or word and an intonation slot
filled by a final intonation contour.
The responses to questions need not be complete structures as
the response is fully intelligible in terms of the question asked.
Where necessary, the full response can be reconstructed in terms
of the original question into a full and complete major sentence-
type.
Where are you going this afternoon? (Question)
To the supermarket. (Response sentence)
I am going to the supermarket this afternoon.
In reconstructing the full sentence, the question word where
which fills a location slot in clause structure, is replaced by the
response phrase, to the supermarket, which fills the locational
slot in clause structure. All other elements in the clause remain
constant, except for the change of person, here, from speaker
to nonspeaker.
Exclamatory Sentences. Addition and response sentences
are minor sentences of the completive type; they complete a
statement or a question that has gone before. These are opposed
to the exclamatory type, which is syntactically independent
(see Bloomneld, 1933:176). These exclamatory minor sentences
might combine with any sentence as a peripheral exclamatory
tagmeme, but when used in isolation with their own final intona-
tion contour, they are minor sentences.
Completive minor sentences, including additions and re-
sponses, tend to have an underlying structure which ranges
from simple word structures, such as a yes or no answer, through
phrases, to elliptical, marginal, and sequential dependent clause
structures.
Exclamatory minor sentences tend to be limited to simple
word and phrase groups, with no underlying clause structure at
all. Because of their syntactic independence, no underlying struc-
ture is implied. This lack of clause structure is evidenced by the
lack of a predicate tagmeme. These structures are discussed in
the following section, Nonclause Structures. The structure of
minor sentences, with incomplete clause structure, may be
charted as to sentence and clause level features. Each minor
sentence type may then be characterized by picking one feature
from each column in the diagram of the minor sentence types.
 58 INTRODUCTION TO TAGMEMIC ANALYSIS

MINOR SENTENCE TYPES

At Sentence Level At Clause Level
Sequential Addition Intransitive
Active Affirmative
Marginal Response Transitive
Elliptical Equational Passive Negative
Exclamatory

Nonclause Structures
The minor sentences of language include single words or
short phrases that do not have the underlying structure of a
clause. These structures are characterized, in general, by the
absence of a predicate or predicate-like tagmeme in the string,
which defines a clause. Yet such forms do occur with final intona-
tion and fall within the general class of minor sentences.
Calls, Greetings, Interjections. Minor sentences of this group
have no clause structure and are generally limited to one or two
words. They are functionally of the exclamatory type, syntac-
tically independent.
1. Calls, or vocatives, are generally the names of persons
or titles of address designating persons. Many case-
marked language systems have a vocative case used
with or without exclamatory particles.
John! Waiter! Mr. Secretary!
2. Greetings in most languages are stereotyped expres-
sions, used in a ritual for meeting people, initiating
conversation, or leaving. Where the individual words
had meaning, this original meaning is often lost.
Hello! Goodbye! Be seeing you!
3. Interjections are usually short and expressive, not ex-
pecting a particular response. They are used particu-
larly to express strong emotion, such as pain, surprise,
enthusiasm but not limited to these uses.
Ouch! Oh boy! Doggone it!
Most of these types need no grammatical formulation. They are
simply listed in the lexicon of the language for use in particular
context situations; and they are listed as single lexical entries
without further analysis.
Titles, Mottoes, Inscriptions. This group of minor sentences
is generally of greater length and indicates some of the phrase
structures of the language. This type may overlap with favorite
sentence types.
 SENTENCE LEVEL 59
1. Titles, when they consist of more than one word are
generally a legitimate phrase structure, with the author
included as an agentive. In reading, the structure is
spoken with a single final intonation pattern.
Alice in Wonderland by Lewis Carroll
2. Mottoes consisting of more than one word, show regular
phrase structure. The type of phrase used may differ
from language to language, as each culture has its own
favorite phra.se types in this use.
Latin: In caritate et iustitia (An adverbial phrase)
French: Liberti e*galite,fraternite (A nominal phrase)
3. Inscriptions, when limited to phrase structure, as well
as spoken toasts, often begin, in English, with To mean-
ing "dedicated to," followed by nominals often with
embedded modifying clauses.
Nontypical Structures. There are sentences in language use
which do not conform to any of the major or minor sentence
types. They are special uses of language which may be pitfalls
for the unwary analyst.
Metalanguage is language about language. In this usage,
some of the forms of language become the topic of conversation,
and thereby become nominal in use, losing their original func-
tional class, for example:
The is a definite article. {the = a noun)
Using metalanguage, it is possible to string together lists of
conjunctions, auxiliaries, and other parts of speech, m ways
normally ungrammatical.
Between hotdogs and and and and and hamburgers
there is the same spacing given on the neon sign.
Once the forms in metalinguistic use are recognized, the gram-
matical structures appear to be quite regular.
Abbreviated language occurs in common use in headlines in
the writing of telegrams, and in some types of radio announce-
mentSi The structure is shortened by eliminating many of the
function words, with the result that the message is cryptic and
often becomes ambiguous.
Army camps in the open Is camps noun or verb?
Officers Flying Home Is officers plural, possessive?
Anyone who has tried writing a ten-word telegram has experi-
enced the difficulties of using this abbreviated form of language.
 60 INTRODUCTION TO TAGMEMIC ANALYSIS

Certain types of written and spoken reports also have an abbre-

viated style, for example: the weather bulletins —a recitation
of facts regarding clouds, wind, tides, and temperatures-or
police descriptions of a suspect in a case, with name, height,
weight, and identifying characteristics. These special uses of
language must be interpreted in the light of their stylistic for-
mat.
There are other forms of language which are grammatical
and neither abbreviated nor metalinguistic. These specialized
uses of language are the aphorisms, or sayings, popular in a
language, or the use of language in special fields such as sports
science or business. The limited use of abstract inanimate
nouns as subject, described as "less grammatical" by Chomsky,
(1957:78), in such sentences as Sincerity admires John is contra-
dicted by the aphorisms:
Misery loves company. Power corrupts.
Fortune favors the brave. Love conquers all.
The use of simile and metaphor in both ancient and modern
languages also allows for a wide range of subject-predicate com-
binations that would not be allowed by an overrestricted se-
mantic component. The collocations of individual words and
consequent grammatical patterns, are used with considerable
liberty in the sports field, where each sport has its own spec-
ialized vocabulary. These and other special forms of language
must be interpreted in terms of its nonspecialized uses.

PRACTICE 2: SENTENCE LEVEL SORTING

The corpus is a set of sentences ready for analysis. The sen-
tences, particularly in running text, must be isolated, sorted ac-
cording to type, and then analyzed at the clause level. Procedures
at the sentence level include reduction of the corpus to simple
clause structures and then recording of complex and compound
structures with their intonations. Reduce the following para-
graph and list the type of sentence patterns:
As the sun set, he remembered, to give himself more confi-
dence the time in the tavern at Casablanca when he had played
the hand game with the great negro from Cienfuegos who was the
strongest man on the docks.
Hemingway, The Old Man and the Sea
Reduction of the Corpus. Rewrite the paragraph, one clause
to a line; isolate connectors, relaters, and simple clause patterns
discovered.
 SENTENCE LEVEL 81

+R +S +P ±0
1. as the sun set
2. he remembered the time in the tavern at
Casablanca
3. to give himself more confidence
4. when he had played the hand game with the great
negro from Cienfuegos
5. X who was the strongest man on the docks
Clauses within Sentences. After reduction of the corpus, com-
plete formulas for complex and compound sentences can be re-
corded with intonation pattern. If clause (2) main and clause
(3) purpose, combine:
(Complex) Sent = ±Marg:F-tCl +Base:tCl -Into:ICF
Read: A complex sentence consists of an optional margin
slot, filled by a purpose clause (F), a base slot filled by a
transitive clause, and an intonation slot filled by final in-
tonation contour.
F-tCl = +P:tvmt +IO:refl.pn +D0:N (Purpose)
Read: A purpose clause consists of a predicate slot filled by a
transitive verb infinitive, an indirect object slot filled by a
reflexive pronoun, and a direct object slot filled by a noun
phrase.
Alternate Solution: If the complex sentence formula is not a
useful pattern in the overall structure of the language, the pur-
pose clause may be interpreted as filling a purpose slot within
the main clause:
tCl = +S:pn +P:tv ±F:F-tCl ±0:N id (Main Clause)
Read: The (main) transitive clause consists of a subject
slot filled by a pronoun, a predicate slot filled by a transitive
verb an optional purpose slot filled by a purpose clause
and an optional object slot filled by an identified noun phrase.
Clauses within Clauses. Some dependent clauses will be
found as fillers of clause level slots, with one clause layered
within another clause. The temporal clause (1) fills a clause
level temporal slot.
tCl = ±T:T-iCl +S:pn +P:tv ±0:Nld (Main Clause)
Read: The (main) transitive clause consists of an optional
temporal slot filled by a temporal clause and so on.
T-iCl = +R:rel +Ax:iCl (Temporal)
Read: A temporal clause consists of a relater slot filled by
a relater and an axis slot filled by an intransitive clause.
 62 INTRODUCTION TO TAGMEMIC ANALYSIS

Clauses within Phrases. Some dependent clauses will be

found as fillers of modifier slots at the phrase level. These are
generally relative clauses. Clauses (4) when and (5) who are of
this type. The identifier (Id:) tagmeme is used for these loopback
clauses.
N,d = +Det:det +H:n +Id:R-tCl (Noun Phrase)
Read: An identified noun phrase consists of a determiner
slot filled by a determiner, a head slot filled by a noun and
an identifier slot filled by a relative transitive clause.
R-tCl = +T:rel.av. +S:pn +P:tV ±O:N. . (Relative)
Read: A relative clause consists of a temporal slot filled by a
relative adverb a subject slot filled by a pronoun, a predi-
cate slot filled by a transitive verb phrase, and an optional
object slot filled by a noun phrase and so on.

SUPPLEMENTARY READINGS 2
Bloomfield, Leonard, Language, New York, Holt, Rinehart and Winston,
Inc., 1933. See Chap. 11, "Sentence Types," 170-177, for the classi-
cal definition of sentence. Favorite (= major) types are distin-
guished from minor. The minor are completive or exclamatory.
Chomsky, Noam, Syntactic Structures, The Hague, Mouton & Co., 1957.
Kernel sentences are the sentences which result when we apply
only obligatory and no optional transformations (46). In English,
the kernel sentences are simple declarative active (80) with no
complex noun or verb phrases (107). All other sentences are derived.
Elson, Benjamin, and Pickett, Velma, An Introduction to Morphology
and Syntax, Santa Ana, Summer Institute of Linguistics, 1962.
Chap. 11, "Tagmemes and Constructions at the Sentence Level,"
82-83, and Chap. 18, "Survey of Senten.e Types," 121-128. The
levels above the sentence are treated briefly in Chap. 19,127-128.
Fries, Charles C, The Structure of English, New York, Harcourt, Brace
and World, Inc., 1952. See Chap. 2, "What is a Sentence?" 9-28, for
use of Bloomfield's definition versus traditional definitions. Chap.
3, "Kinds of Sentences," 29-53, based on situation and response.
Gleason, Henry A. Jr, Linguistics and English Grammar, New York,
Holt, Rinehart and Winston, Inc., 1965. Chap. 14, "Joining Clauses,"
329-350. Rejects division of sentences into simple, complex, and
compound, in favor of IC coordination and subordination of
clauses.
Liem, Nguyen Dang, English Grammar, A Combined Tagmemic and
Transformational Approach, Linguistic Circle of Canberra, 1966.
A Contrastive Analysis of Vietnamese and English, vol. I, Chap.
5, "English Sentence Types and Sentence Level Structures," 139-
149. Lists major independent, major dependent, and minor types.
 SENTENCE LEVEL 83
Longacre, Robert E., Grammar Discovery Procedures, The Hague,
Mouton & Co., 1964. Chap. 4, "Sentence Level Procedures," 125-
161. Treats sentence level after the levels of clause, phrase, and
word.
, "The Notion of Sentence," Monograph Series on Languages and
Linguisties, No. 20 (Washington, D.C., Georgetown University
Press, 1967>, 15-25. Defines sentence level as a level of clause com-
bination, and classifies sentences by conjunction, alternation, im-
plication, and negation.
 TABLE 3: CLAUSE LEVEL ANALYSIS

According to the According to the According to Its Characteristics of

Type of Verb Form Distribution of Unit Internal Structure the Clause Type

1. Transitive (tCl) Transitivity recognized

transitive verb, if object occurs with
may take object the verb at least once.
Independent Clause 2. Intransitive (iCl) Intransitivity recognized
which can stand alone intransitive verb, if no object ever occurs
takes no object with the verb
3. Equational (eqCl) Verb links subject with
equational verb, predicate attribute,
predicate attribute noun, adjective, or
Main Clause adverb
with
Finite Verb Form 1. Nominal Clause Recognized dependent
acts like a noun by indefinite pronoun
Dependent Clause 2. Adjectival Clause Recognized dependent
which cannot stand alone acts like adjective by relative pronouns
3. Adverbial Clause Recognized dependent
acts like adverb by clause-type relaters

Nonfinite Verb Forms Recognized as dependent

with partial clause by gerunds, participles,
structure infinitives
3 CLAUSE LEVEL

The clause level of grammar is that level that is below the sen-
tence level and above the phrase level. The clause is composed of
words and phrases and, in turn, fills slots at the sentence level.
The clause is a unit of grammar. It is a construction in which the
constitute is a potential sentence base and in which the con-
stituents are the subjects, predicates, objects, and adjuncts that
combine to form this base. Clause level analysis is central to the
system.
Clause Defined. The clause is "a string of tagmemes that con-
sists of or includes one arid only one predicate, or predicate-like
tagmeme, in the string, and whose manifesting morpheme se-
quence typically fills slots at the sentence level" (Elson and
Pickett, 1962H34). In the wording of this definition, the following
essential features should be noted:
1. Clauses typically fill slots on the sentence level. In typical
mapping of lower constructions into higher levels
clauses combine to form sentences, or combine with in-
tonation to form sentences. However, atypical mapping
of clause within clause and clause within phrase occurs.
2. Clauses consist of or include one and only one predicate.
Accordingly, there are as many clauses as there are
predicate tagmemes. A single clause may, however, have
a compound verb-form in the predicate slot. Some lan-
guages have verb forms which, in themselves, constitute
whatLongacre calls a "clause-in-miniature," containing
both subject and predicate; other languages require
subject and predicate tagmemes.
3. Clauses may have a predicate-like tagmeme. Particularly

65
 66 INTRODUCTION TO TAGMEMIC ANALYSIS

in the equational type structures, the predicate may

be optional. In these cases there is another tagmeme,
the predicate attribute tagmeme, which is obligatory.
This predicate attribute occurs even when the linking
verb does not occur and the clause unit has no predicate
tagmeme.
Traditional Clause. The clause has traditionally been con-
sidered a unit of grammar and has been denned as "a group of
words, containing a subject and a predicate, and used as part of
a sentence," Clause units were subclassified as independent, or
main clauses, and dependent, or subordinate clauses. An in-
dependent clause was defined as a clause that could stand alone
as a sentence and a dependent clause as a clause that could not
stand alone as a sentence (see Warriner, 1965:53).
In tagmemic analysis, the clause is accepted as one of the
fundamental units of grammar, but is denned as a word group
containing one and only one predicate. This allows the inclusion
not only of independent and dependent clauses, which have both
subject and predicate, but also of partial clause structures under
the label of clause. An infinitive or participle, expanded with ob-
jects or modifiers, is also classified as a dependent clause. The
definition of independent clause is also amended. This clause
must be able to stand alone as a major sentence in the language.
The features that determine whether the clause can stand
alone as a major sentence in the language will be unique for each
language investigated.

Clause Types. The clause types traditionally investigated in

the study of grammar must also be investigated in formal lin-
guistic analysis. Whatever the system of analysis, provision must
be made for describing known clause structures, and for stating
the relations of dependency by which clauses are subordinated
within sentence structure.
Independent clauses are classified by these features:
1. Transitivity: intransitive, transitive, or equational.
2. Voice: active middle, passive, or reciprocal.
3. Negation: affirmative or negative.
Dependent clauses are classified by their structure:
1. Complete Clause Structures: Subordinated or subordi-
nate.
2. Partial Clause Structures: Infinitives or participles.
3. Nonclause Structures: Base for minor sentences.
In the analysis of independent clause structures, primary
 CLAUSE LEVEL «7

attention is given to kernel sentences. Simple, complete state-

ments whether transitive, intransitive, or equational, are
analyzed first. Then these basic sentences are related to non-
kernel sentences. Statements are opposed to commands and
questions, active to nonactive, affirmative to negative.
In the analysis of dependent clause structures clause struc-
tures must be analyzed as filling slots at a particular level of
analysis. The clause fills a marginal slot at sentence level, or the
clause occurs filling a clause level slot in a layering structure or
the clause fills a phrase level slot in a loopback structure. Once
the distribution of the clause is known, the clause may be further
classified according to its function as nominal, adjectival, or ad-
verbial, and according to its form as a complete, partial, or non-
clause structure.

INDEPENDENT CLAUSES
The clause level in tagmemics is the heart of the analytic process.
The sentence level provides a sorting ground for reducing sen-
tences to homogeneous sets of clauses, apart from their intona-
tion. After sorting, the underlying structure of the sentence is
analyzed by taking apart its clause base, in terms of subject-
predicate-object-adjunct tagmemes that form the underlying
structure. The work done here is comparable to the upper level
branching rules of a transformational grammar.

Independent Clause Defined

An independent clause is a clause that can stand alone as a
major sentence in the language. This capacity to stand alone
must be judged according to its acceptability by a native speaker
of the language. Each language will have its own features of
both external distribution and internal structure which formally
categorize clause independence.
Nuclear Clause Level Tagmemes. Clauses are potential
strings of tagmemes which include one and only one predicate
in the string. Within that string, some tagmemes are nuclear
some peripheral. The nuclear tagmemes are diagnostic of the
construction in which they occur and may be obligatory or op-
tional; peripheral tagmemes are optional. In clause structure,
subject-predicate-object are generally nuclear.
S: Subject tagmeme manifests the topic, or actor, or thing de-
scribed. Subject may be obligatory or optional in the con-
struction, and is often linked with the predicate by a cross-
reference type of concord tie.
 68 INTRODUCTION TO TAGMEMIC ANALYSIS

P: Predicate tagmeme manifests the type of action or serves

as a link between the subject and its attribute. Predicate
is obligatory in all clause constructions except the equa-
tional type construction.
PA: Predicate attribute manifests an attribute of the subject
by a nominal, adjectival, or adverbial word group. It is
linked to the subject by a linking verb and an agreement
type of concord. It is always obligatory.
O: Object tagmeme completes the meaning of the predicate,
and often is case-governed by the predicate in a govern-
ment type concord tie. Objects may be specified as DO:
direct object; 10: indirect object; or OC: object com-
plement, in structures which take two objects.
Peripheral Clause Level Tagmemes. The peripheral tag-
memes at clause level are not diagnostic of the construction, and
are always optional. They fill in details of time, place and man-
ner, and are mainly adverbs and adverbial phrases or clauses.
Typical peripherals are:
L: Location tagmeme answers the question where and places
the verbal action in a place setting. The fillers of this slot
are adverbs of place and their substitutes. Locative adverbs
are labeled loc.
T: Temporal tagmeme answers the question when and places
the verbal action in a time setting. The fillers of this slot are
adverbs of time and their substitutes. Temporal adverbs
are labeled tern.
M: Manner tagmeme answers the question how and fills in the
details of the circumstances of the action. The fillers of this
slot are adverbs of manner and their substitutes. Manner
adverbs are labeled av. In initial analysis, this slot is used to
differentiate manner from time and place as a catchall for
all adverbs except those of time and place.
/; Introducer tagmeme introduces a clause and occurs almost
always in initial position. In later analysis, introducers may
turn out to be temporals, or sequence signals proper to the
sentence level. Introducers are labeled i.
The Drama Analogy. The grammatical relationships in the
clause structure have been explained by Longacre, using the
analogy of a drama. The dramatis personae, or roles, are the
nominals, subjects, and objects. The plot is the verbal action. The
scenery, or props, are provided by the adverbial elements of
place, time, manner, and so on (1964:35; 1965:65). To this, we
would add the costuming department, manifested by adjectival
words and phrases, used as fillers in the predicate attribute slot.
This four-fold classification parallels the —AL ('acts like') des-
 CLAUSE LEVEL 69

ignation of laager and Smith (1951:74), and the functional classes

of Fries (1952:76). These terms are extended to include both
single words and word groups.
Fries Trager-Smith Longacre
Class 1 Words nominals Roles, dramatis personae, S:, 0:
Class 2 Words verbals Plot, verbal action, P:
Glass 3 Words adjectivals Costuming features PA:
Class 4 Words adverbials Props, scenery, L:, M:, T:, I:
The drama analogy is within the tagmemic tradition and sets
up classes of tagmemes which are similar in function. The role-
plot distinction seems to be a fruitful line of investigation for
discourse analysis (Pike, 1367:246, fn. 14; Gleason, Georgetown
Monograph Series: No. 21,1968:39-63). It is also the basis for the
case grammar of Charles JMllmore (1968), who seeks to establish
deep structures with plot, tense and mode, in one branch, and
role, agent and goal, in another, marked by case.

Clause Level Matrix

At sentence level, a sentence level matrix was set up with the
principal sentence types—statement, question, command—in
one dimension and the principal clause types in the other. At
clause level, a clause level matrix may be set up with the prin-
cipal clause types in one dimension and the nuclear tagmemes
in the other dimension. This type of matrix reveals the differ-
ences m structure between clause types.
Clause Type Subject Predicate Object
1. Intransitive ±S:N +P:iv 0
2. Transitive ±S:N +P:tv ±0:N
3. Equational ±S:N ±P:eqv +PA:N/Aj/Av
In preliminary sorting, it is assumed that these three types of
clauses will be found in a language, but further investigation
may require that these types be subclassined or that new types
of clauses be added.
Intransitive Clause. An intransitive clause is denned as a
clause which contains an intransitive verb. This verb, in turn
is aiverb which can never take an object. According to the rule
of two (see p. 26), this clause must differ from other clauses by
two Jstructural differences, acme of which affects the nuclear tag-
meraes of the construction. The differences here are: (1) absence
of object tagmeme, (2) filler of predicate slot, and (3) inability of
intransitive verbs to transform into passive. All are nuclear. A
typical clause level formula for an intransitive clause would be:
3.1 iCl = +S:pn +P:iv ±L:loc He went home.
 70 INTRODUCTION TO TAGMEMIC ANALYSIS

Read: An intransitive clause consists of a subject slot filled

by a pronoun, a predicate slot filled by an intransitive verb
and an optional location slot filled by a locative.
Transitivity could have been denned in a different way, so
that verbs that occur sometimes with and sometimes without
objects are classified as both transitive and intransitive. How-
ever, transitivity is often marked in the verb form and verbs
so marked for transitivity do occur without objects at times.
Therefore the tagmemic system permits the use of optional
objects and classifies intransitive verbs as verbs that never take
an object, and transitive verbs as verbs that have the capacity to
take an object.
Transitive Clause. A transitive clause is defined as a clause
that contains a transitive verb. This verb, in turn, is a verb that
has the capacity to take one or more objects. This capacity to
take an object is demonstrated by the occurrence of the object
with the verb at least once. According to the rule of two, the tran-
sitive clause differs from other clauses in: (1) the presence of an
optional object, (2) filler of the predicate slot, and (3) the ability
of the transitive clause to transform into the passive. A typical
clause level formula for the transitive clause is as follows:
3.2 tCl = +S:pn +P:tv ±O:n He ate dinner.
Read: A transitive clause consists of a subject slot filled by
a pronoun, a predicate slot filled by a transitive verb, and
an optional object slot filled by a noun.
Every verb is either transitive or intransitive; but transitive
verbs can be further subclassified as: (1) semitransitive with
optional object; (2) transitive, with obligatory object; and (3) di-
transitive with two objects. In English, for example verbs occur
with one object, with both indirect and direct object, or with
object and object complement.
Equational Clause. An equational clause is defined as a
clause that contains an equational, or linking, verb. The verb
connects the subject with the predicate attribute, which may be
nominal, adjectival, or adverbial. In this structure, the PA: slot
is obligatory and in agreement with the subject. In some lan-
guages the subject is omitted, in others, the predicate linking
verb is omitted. In English both are obligatory. A typical clause
level formula for the equational clause is as follows:
3.3 eqCl = +S:pn +P:eqv +PA:N He is the chief.
Read: An equational clause consists of a subject slot filled
by a pronoun, a predicate slot filled by an equational verb,
and a predicate attribute slot filled by a noun phrase.
 CLAUSE LEVEL 71

In investigating equational clauses, it is often useful to test

the clause formula in different tenses. Many languages that allow
a zero link in the present tense have an overt linking verb form
in the past! tense. This clause type is sometimes classified into:
(1) equational, if the predicate attribute is a noun; and (2) stative,
if the predicate attribute is an adjective (Elson and Pickett,
1962:113). In Liem's analysis of English, clauses beginning with
It and There are set up as stative clauses (1966:160-161) distinct
from the equational type clause.

Derived Clause Types

Independent clauses are used as the fillers of the base slot

of both kernel and derived sentences. The basic list of nuclear
and peripheral tagmemes is sufficient for most kernel sentences.
For derived sentences new tagmemes must be established. The
derivation --of sentences from the kernel may be shown by indi-
cating the con trastive patterns. The principal derived sentences
are: (1) commands and questions, opposed to statements; and (2)
passive and negative, opposed to active and affirmative; to these
may be added (3) emphatic clauses.

Questions and Commands. The sentence types opposed to

statements are questions and commands. These differ from
statements by features at the sentence level, at the clause level
or both.
Questions differ from statements at the sentence level by
intonation patterns, or by special question markers, or by the
type of clause used (see p. 53, Questions). The differences in the
structure of the clause are:
1. Use of question words as fillers for clause level slots. In
English, for example, questions are recognized by the
words who, what, when, where, why, how, generally with
a shift of the slot to initial position.
3.4 Q-eqCl = +S:Q-pn +P:eqv +PA:N
Who is the chief?
2. Change of order, in English, is used as a question mark-
ing signal, The first verbal auxiliary, if there is one is
placed before the subject; where there is no auxiliary,
some form of the verb do is used.
3.5 Q-iCl = +Aux:aux +S:pn +P:iv ±L:loc
Is he going home?
Commands differ from statements at the sentence level by
intonation patterns, by command markers, or by a special type of
 72 INTRODUCTION TO TAGMEMIC ANALYSIS

clause base (see p. 52, Commands). The principal differences

in clause structure are:
1. Subject deletion in imperatives of the second person. The
subject is often supplied by a vocative form, which is
syntactically independent.
2. Verb forms are often formally marked by an imperative
mood, which is different from the regular finite verb
forms of the language.
3.6 C-tCl = +P:tv,mp ±O:N (John!) Eat your dinner!
3. In commands, the form of the negative is often different
from that of statements (see Sierra Popoluca, {dya},
'not,' versus {odoy}, 'don't'). Other differences may be
found including limits on peripheral tagmemes.
Passive and Negative. The basic sentence types may change
from active to passive, middle, or reciprocal voice; affirmative
sentences may be changed into negative. These changes are
proper to clause level.
Passive clauses differ from active clauses by changes in the
subject of the active clause and changes in the voice of the verb.
1. Subject of the active clause becomes agent (Ag:) of the
passive. Object of the active clause becomes subject of
the passive clause.
2. Verb forms for active and passive differ in the marking
of the voice feature. In the passive clause, the predicate
slot is filled by a passive verb form.
3.7 tCl = +S:Ni +P:tv +O:N2
Clyde robbed the bank.
3.8 tClp = +S:N2 +P:tvp +Ag:RA
The bank was robbed by Clyde.
Negative clauses differ from the active in the presence of a
negative within the main verb phrase. The negative is best ex-
pressed at the level of the phrase as one optional element in the
phrase level string. In larger problems, it is often convenient to
use a negative T-Rule.
3.9 Topt Negative S + P + O =>S + Neg + P + O
This rule, used in Sierra Popoluca (#165), indicates that the neg-
ative {dya} is inserted in the clause level string, immediately
before the predicate.
Emphatic Clause. Changes of order, combined with change of
meaning or the use of emphatic particles, may indicate that a
special emphatic clause type is required in the language. Since
tagmemes may be movable, simple change of position may place
 CLAUSE LEVEL 73

different elements in focus; but this does not indicate the need
for a new clause type. The emphatic clause must differ from
others by two structural differences. For example in Cashina-
hua (#164), the object is moved and marked with a particle.
3.10 E-tCl = +O:Nm/pnm ±M:RA +S:N/pn +P:tv/tV
The object slot normally occurs immediately before the predi-
cate. In the emphatic type clause, the object slot is moved to
first position, and the noun phrase or pronoun is marked with a
special emphatic particle {-ra}. There is a corresponding change
of meaning to strong emphasis. In the Laboratory Manual by
Merrifield, emphatic stress is indicated by underlining the forms
stressed. Palantla Chinantec (#125), Northern Tepehuan (#126),
Kalagan (#168, #170), and Bukidnon Manobo (#169) are sample
problems dealing with emphatic clause structures.
DEPENDENT CLAUSES
Dependent clauses are clauses that may not stand alone as major
sentences, though they occur, with final intonation, as minor
sentences. To understand the use of dependent clauses within
major sentence structure, we consider the external distribution
of the clause, the functional meaning of the clause and its in-
ternal structure.

External Distribution of Clauses

Dependent clauses fill subordinate positions in major sen-
tences. The process by which a clause is subordinated is called
an embedding process. The structure in which the clause is
embedded is called the matrix; the dependent clause that is
embedded is called a constituent. In a tagmemic analysis, em-
bedding may take place at the sentence level, the clause level
or the phrase level.
Sentence Level Embedding. A dependent clause embedded
m a sentence structure is called margin, and the resulting
sentence is a complex sentence, consisting of base, margin, and
intonation. Glauses embedded at sentence level form the only
true complex sentences and are formulated in a complex sen-
tence formula (see p. 46). Nominal and adjectival dependent
clauses are never embedded at the sentence level but fill noun
and adjective slots at clause and phrase level. Embeddings at
sentence level are called marginal.
The structure must be a true multiple clause structure what
Longacre calls a "patterned dependency," involving more than
one clause but not having the overall structure of a single
 74 INTRODUCTION TO TAGMEMIC ANALYSIS

clause (1964:128). Within this structure, there may be restric-

tions governing the type of clause used, connecting particles,
order of clauses, and sequence of tenses, which assist in deciding
whether a structure is a complex sentence. Some of the types sug-
gested by Longacre are direct and indirect quotation, conditions,
effect-cause, circumstance-event, and antecedent-consequent
(1964:130). A fuller description of sentence types is found in "The
Notion of Sentence" (18th Annual Round Table, Georgetown Uni-
versity Monograph Series No. 20, 1967:15-25).
In English, certain sets of correlative particles seem to indi-
cate that the structure is a multiple clause sentence structure,
and that the sentences that use these particles are truly com-
plex sentences. For example:
Conditional: if... then
Concessive: although . . . nevertheless/still/yet
Causal: since/because . . . therefore

Clause Level Embedding. In clause level embedding, the

structure has the overall structure of a single clause, even
though embedded clauses occur in the structure. The dependent
clauses fill clause level slots and are parallel to constructions
where the same clause level slots are filled by single words or by
phrases. In their external distribution, these clauses may be
nominal, filling subject and object slots, or adverbial, filling
adjunct slots such as time, place or manner.
The embedding of clause within clause is a particular case
of the recursiveness of a tagmemic grammar when the same
symbol occurs to both right and left of the equals sign (see p. 24).
Nonrecursive embedding of clause within clause also may occur
when the clauses do not have the same symbol. Any occurrence
of clause within clause is called a layering in the system.
In English, and other languages as well, introductory clause
particles may signal the type of clause and the kind of slot it
fills in the structure. For example:
Temporal: before, while, after, when until
Locational: whence (from), whither (to), where
Manner: as, like, as if

Phrase Level Embedding. In phrase level embedding, the

structure has the overall structure of a single phrase. If the
phrase is relater-axis, then the clause may be embedded as the
object of a preposition, a nominal. If the phrase is a modification
structure then the clause is embedded as a single modifier re-
lated as an adjectival, to the phrase head.
Nominal clause: Here is a recording of it.
Here is a recording of what was said.
 CLAUSE LEVEL 75

Adjectival clause: The man was my uncle.

The man who came to dinner was my uncle.
In English, adjectival clauses embedded at phrase level are
usually relative clauses whose antecedent is the head noun of
the phrase. The order in English is that single word modifiers
precede, and phrases and clauses follow (see also p. 96):
Single modifiers + Head noun + Phrases + Clauses
The embedded clauses that fill nominal slots at phrase or clause
level are indefinite clauses, in which the relative indefinite pro-
noun is nonanaphoric. It has no antecedent in the sentence struc-
ture. Relative pronouns are who, whose, what, which, whom;
indefinites are whoever, whatever, whichever, and any nonana-
phoric relative pronoun. Any occurrence of clause within phrase
is called a loopback in the system.

Functional Meaning of Clauses

Dependent clauses are embedded into sentence structure. They
are partially defined by their external distribution. The first fact
that must be determined is the level at which the clause is em-
bedded. The second fact that must be determined is the func-
tional use of the clause. At the sentence level clauses fill mar-
ginal slots in the sentence structure, as already treated under
complex sentences. Clauses embedded at clause and phrase levels
can be identified according to their functional meaning.

Nominal Dependent Clauses. Dependent clauses embedded

at the clause and phrase level have nominal functions. Depend-
ent clauses may serve as the subject or object of clauses, or they
may function as the object of a preposition, at the phrase level,
in relater-axis structure.
Nominate at clause level are subjects and objects; they fill
the same slots in the structure as pronouns, nouns, or noun
phrases. For example:
Subject: He practices language analysis.
Whoever takes this course practices language
analysis.
3.1.1 tCl = +S:indef.Cl +P:tv +O:N
3.12 indef.Cl=+S:indef.pn+P:tv+O:N
Nominal dependent clauses are generally introduced by indefi-
nite relatives that have no antecedent, such as: whoever what-
ever, whichever, and more rarely, simple relative pronouns used
without an antecedent. The indefinite nominal dependent clause
may also occur as object.
 76 INTRODUCTION TO TAGMEMIC ANALYSIS

Object: John said it.

John said that he wanted to go home.
Nominate at phrase level are objects of prepositions; they fill
the same slots in relater-axis phrases as pronouns, nouns, or
noun phrases. The restrictions on internal structure are the
same as at clause level. For example:
RA Object: Here is a recording of it.
Here is a recording of what was said.
3.13 RA = +R:rel +Ax:indef.Cl
3.14 indef.Cl =+S:indef.pn +P:tVp
The indefinite clause is discovered within the axis slot of the
relater-axis construction, and then is analyzed according to its
clause structure.
Adjectival Dependent Clauses. Dependent clauses embedded
at the clause and phrase level are found with adjectival function,
filling the same slots in clause or phrase structure as adjectives.
They may act as predicate attribute in clause structure, or as
modifiers in the phrase:
1. Adjectivals at Clause Level. The only occurrence of ad-
jectives at the clause level is in the predicate attribute
slot (marked PA:), which also takes nominal and ad-
verbial clauses. In some languages, the use of an aux-
iliary + participle may be interpreted as a linking
verb + attribute construction with the attribute slot
filled by an adjectival partial clause.
2. Adjectivals at the Phrase Level. The most common use
of dependent clauses in adjectival function is the use of
relative clauses as modifiers. These occur after the head
noun, and fill identifier slots (marked Id:). The phrase
is called an identified noun phrase. For example:
The old man was my uncle.
The man who came to dinner was my uncle.
3.15 N,d = +Det:det +H:n +Id:rel.Cl
3.16 rel.Cl = +S:rel.pn +P:iv +L:RA
The relative clause is said to manifest the identifier tagmeme,
and the use of clause within phrase constitutes a loopback in
the grammar.
Adverbial Dependent Clauses. Dependent clauses embedded
at the clause level fill peripheral slots of time, place, and man-
ner. They are often introduced by relative and indefinite pro-
words of the adverb class.
 CLAUSE LEVEL 77

Locational: He went there.

He went wherever he wished.
3.17 iCl = +S:pn + P:iv ±L:L-iCl
3.18 L-iCl=+R:reH-Ax:iCl
Temporal: John will fix the porch then.
John will fix the porch when he has time.
3.19 tCl =+S:np +P:tV +O:N ±T:T-tCl
3.20 T-tCl =+R:rel +Ax:tCl
In both of these examples, the clauses are subordinated by the
use of a temporal or locational relater. The clauses are depend-
ent and specified as to function by the use of the subordinating
particles.
Internal Structure of Clauses
Dependent clauses are classified according to their external
distribution as embedded at a particular level, and as fulfilling
specific nominal, adjectival, or adverbial function at that level.
Dependent clauses may also be classified according to their in-
ternal structure as clauses which are: (1) subordinated by a re-
later, (2) subordinate because of a built-in relater, or (3) partial
clause structures and therefore dependent. The recognition of
these types helps in setting up the analysis of the dependent
clause, once it has been discovered in the structure.
Subordinated Clause Types. Dependent clauses of the sub-
ordinated type have an overt relater marking dependency. The
test for this type of clause is to remove the relater. If the clause
without the relater is an independent clause, then the clause
with relater is a subordinated type. These clauses are analyzed
as relater-axis structures consisting of two tagmemes, a clause
subordinator and a clause as axis, in a way parallel to the analy-
sis of prepositional phrases. This analogy is established by Long-
acre (1964i3§) in the following set of relations (see also p. 32):
Preposition: its object:: subordinator: its clause
The class of clause subordinators is the class of subordinating
connectors and includes relative and indefinite adverbials as
well. The: subordinated type of clause is typically found in ad-
verbial use at the clause level.
Subordinate Clause Types. Dependent clauses of the sub-
ordinate type have no overt reiater but have an internal rela-
tive or indefinite pronoun which is a portmanteau representation
of relater and pronoun. If the relative is removed, the resulting
 78 INTRODUCTION TO TAGMEMIC ANALYSIS

structure is not a clause; if the relative is replaced by the equiva-

lent pronoun, the resulting clause is an independent clause. The
relative acts both as a constituent of the clause and as a subor-
dinator, making the clause dependent in the sentence. Thus in
the man who came to dinner, the relative who = R + he. It acts
simultaneously as subject of the sentence, and as clause sub-
ordinator.
Relatives are often omitted as Koutsoudas has observed
(1966:292), in object position, but are never omitted in the sub-
ject position. The relative which is object of the preposition is
also often omitted. For example:
the girl whom he likes becomes the girl he likes
the girl whom he gave becomes the girl he gave
the ring to the ring to
the girl who lives but not "the girl, lives
next door, next door.
The relatives and indefinites that are adverbial in function are
never omitted, because they serve as clause subordinators in a
relater-axis type construction, and are obligatory to the con-
struction.
Partial Clause Structure. Dependent clauses that show only
partial clause structure are identified by the presence of a pre-
dicate tagmeme, together with other clause level elements oblig-
atory or optional to the string. The principal dependent clauses
of this type in English are the participles and infinitives used
with other clause level elements in an embedded string. Other
languages show similar use of nonfinite verbs.
Participle Clauses. In English, the two participles, past and
present, are used as part of an embedded string, which is de-
scribed as a partial clause structure. Their use is nominal or
adjectival, and they fill the same structural positions as other
subordinated and subordinate clauses. For example:
Subject: Watching baseball is entertaining.
Object: He enjoyed playing the horses.
RA Phrase: He found a new method for raising money.
The participle clauses fill nominal slots in the clause or RA
phrase structure. They may be labeled as dependent (depCl),
with an internal structure consisting of a predicate slot filled by
participle (tvpart), and other clause level slots in the string. In
adjectival use, for example:
Past Participle: The ship, lost in the storm was recovered.
Present Participle: The dog, barking loudly, ran away.
The participle clauses here fill adjectival slots in the phrase and
 CLAUSE LEVEL 79

are best characterized as identifiers (marked Id:) in the same

way as the equivalent dependent clauses of the subordinated or
subordinate type. They resemble deleted versions of the more
complete dependent structure:
The ship, (which was) lost in the storm was recovered.
The dog, (which was) barking loudly, ran away.
In external distribution, the participle clause types fill the same
slots as other dependent clauses; they differ only in their internal
structure. They are formally different fillers in functionally
identical slots.
Infinitive Clauses. In English, the infinitive is recognized by
the infinitive marker to, together with the base form of the
verb. Infinitives occur either without subjects or with subjects
in the objective case. When a subject occurs with the infinitive
it is often simultaneously the object of the main verb, in a port-
manteau manifestation (Longacre, 1967:326). For example:
Without Subject: To err is human.
With subject: I asked him to go home.
In the occurrence of infinitives without subject, the infinitive
and the other clause level tagmemes which it dominates are
interpreted as filling a single slot, as a clause embedded within
the overall pattern. In the example given, the infinitive fills the
subject slot, +S:ivinf. In portmanteau representation, the analy-
sis gives precedence to one element. The second example may be
interpreted as the string, -t-S +P +10 (him) +D0 (to go home),
or as an SPO string with him internal to the embedded clause
acting as subject of the infinitive.

CLAUSE LEVEL ANALYSIS

The clause level stands below the sentence level and above the
phrase level. Sentences are composed of clauses, which, in turn
are composed of words and phrases. At the sentence level com-
plex and compound sentences are reduced to simple clause struc-
tures by the procedures outlined in the section on reduction. At
the clause level, the simple clauses are separated into homogene-
ous groups, and these clauses are then analyzed into subject,
predicate, objects, and adjuncts.

Clause Level Sorting

Once the corpus has been reduced to a set of simple clauses,
these clauses must be homogeneously sorted. In problems pre-
pared for analysis, the clauses are often presorted and give the
 80 INTRODUCTION TO TAGMEMIC ANALYSIS

appearance of being contrived. Each analyst should be able

to presort clauses himself, so that every problem he does will
in fact have clauses of the same type.
Kernel Sentences. The clause types that underlie kernel sen-
tences should be sorted out and analyzed first. Derived non-
kernel structures can then be analyzed as variations upon the
kernel type structures. Kernel sentences are simple, complete
statements, active and affirmative.
At the sentence level, simple complete statements are isolated
by reduction and sorting, as follows:
1. Simple sentences are isolated by reduction, listing one
clause to a line. Complex and compound sentences are
not discarded but are broken down into simple sen-
tences; relaters are removed and relative pronouns
treated as if they were personal pronouns.
2. Complete sentences are subject to analysis; incomplete
sentences are filed as "residue" until after complete
sentences are analyzed. Then incomplete structures
become intelligible in the light of the complete structure.
3. Statements are separated from questions and com-
mands. Mark each simple complete statement as S, Q,
or C, and begin the analysis with the study of statements
only. Questions and commands are treated later.
At the clause level, the clause bases underlying the simple
complete statements isolated at the sentence level are analyzed.
These are first sorted into basic clause types, such as transitive
intransitive, and equational:
1. Transitivity: Mark each clause as iCl, tCl or eqCl, accord-
ing to whether the verb seems to be a transitive in-
transitive or equational verb.
2. Affirmative clauses should be separated from negative.
The negative type is generally characterized by a nega-
tive particle in the verb phrase.
3. Active clauses must be separated from middle and pas-
sive. The clause structure for nonactive clauses is gen-
erally distinctive.
Charting the Data. Once clause types have been sorted, each
type is analyzed singly. Within that type, any group of clauses
of the same type may be analyzed as a set of structures. Con-
venient quantities range from 12 to 24 clauses, although there
is no theoretical limit to the number. A chart is a list of all the
clauses in the given data, separated into string constituents.
List clause level elements, such as S: subject, P: predicate, 0:
object, PA: predicate attribute, at the top of the page, and under
 CLAUSE LEVEL 81

each heading list the words or phrases that fill these particular
slots. Peripheral elements may initially be labeled as T: tem-
poral, L: Locational, I: introducer, and M: manner. Elements
m the string should be listed, in the chart, roughly in the order
m which they occur in the data. If the order is not fixed use the
statistically predominant order. If only one element is movable,
it may be listed in both positions. A partial chart of the data in
Sierra Popoluca (#165, Laboratory Manual, Merrifield, 1967) for
transitive clauses, showing subject, predicate, object, and loca-
tional slots, is given below.
±S +P ±0 ±L

1. tu-m pA^Sin ipa?tne ikawah

a man' 'has found' 'his horse'
2. he?m Siwan d y a iko?cgakum he?m ikawah
'that John' 'did not hit again' 'his horse'
9. (he?m pA-Sin) iwatpa ikama (X) yA?m
that man' 'will make' 'his cornfield' 'here'
12. he?m kawah iku?tne ha?yanmok
'that horse' 'has eaten' 'much corn
In clause 9, the subject occurs after the verb. In this case the
position occupied by the subject is marked (X), and the subject,
in parentheses, is placed in the regular column marked subject.
Even in a preliminary survey of the transitive clause type it
should become obvious that the regular order is SPO with an
alternate order POS expressed in clause 9. In the data thus far
all three elements are obligatory. However, further analysis of
transitive clauses shows both subject and object are missing
in some clauses, and so these are marked as optional to the con-
struction. Some clause fillers are words; some are phrases.

Tagmemic Grammar

A tagmemic solution is given in terms of a grammar and

lexicon. The grammar is a list of formulas at the sentence, clause,
phrase, and word levels, with whatever restrictions and assump-
tions are required. The lexicon is a list of the morphemes which
occur in the data, together with their form, classification and
gloss. To this are appended whatever morphophonemic rules are
required.
Tagmemic Formulas. Grammar is represented by formulas.
We label this section of the solution "GRAMMAR," and state
the level of construction to which the formula belongs. The
formula consists of a construction symbol to the left, an equals
sign, and a string of tagmemes to the right of the equals sign.
Tagmemes are marked as obligatory (+) or as optional (±) depend-
 82 INTRODUCTION TO TAGMEMIC ANALYSIS

ing upon whether they always occur, or only sometimes occur,

within the construction.
Formulas use capital and small letters. Capital letters refer
to slots or fillers greater than the word; small letters refer to
slots or fillers representing words or morphemes. Thus the slots
at sentence clause, and phrase levels are capital letters. If the
fillers are phrases or clauses, they are marked with capital let-
ters. Fillers which are words or morphemes are marked with
small letters.
GRAMMAR
Sentence Level Construction:
Sent «• +Base:tCl -Into:ICF
Read: A sentence consists of a base slot filled by a transi-
tive clause and an intonation slot filled by a final intona-
tion contour.
Clause Level Construction:
tCl = ±S:N+P:tv±O:N
Read: A transitive clause consists of an optional subject
slot filled by a noun phrase, a predicate slot filled by a tran-
sitive verb, and an optional object slot filled by a noun
phrase.
The labels and readouts of the formulas are not part of the solu-
tion but are added for the sake of clarity. In submitting solu-
tions include the labels, but exclude the readout which only
restates what the formula says.
Restrictions and Assumptions. When restrictions or assump-
tions are necessary to the solution, they are added immediately
after the formulas of the grammar. Besides these essential for-
mal statements, no comments should be necessary. The solution
should be stated in such a way as to express a nonambiguous
commitment.
Restrictions are strictly formal and should pertain to gram-
mar, not to lexicon. They generally deal with the facts of co-oc-
currence and are of the form: Item A excludes Item B or Item A
requires Item B. Many restrictions can be built into the solution,
using available devices. For example:
1. Mutually exclusive fillers in a single slot may be listed
as alternates, using the (/) notation. Thus, N/pn, means
noun phrase or pronoun.
2. Mutually exclusive tagmemes can be listed in an "ei-
ther . . . or" notation, as ±A +B, meaning either A or B,
but not both (see p. 17).
 CLAUSE LEVEL 83

Assumptions are strictly formal, and involve a conclusion by

the analyst which is beyond the given evidence. Those assump-
tions should be stated which have caused a modification of the
grammatical formula.
Maximum Generation Potential. The maximum generation
potential(MGP) of a given solution ts the total number of sen-
tences that may be generated from a given grammar and lexi-
con. This number is finite and calculable whenever the data
isfimlteid,To calculate the MGP:
1. Count the number of morphemes in a given class and
list this number below the symbol of that filler class in
the lowest level formula.
2. If the slot is optional, add +1 (to account for the pos-
sibility of zero occurrence within that slot) to the num-
ber of real morphemes.
3. Multiply the resulting totals to obtain the number of
combinations possible in each single construction. Place
this total under the label of the construction in any
higher level slot in which it occurs.
4. Continue <.lie same procedure, working from lower to
higher levels, until the MGP is determined for the con-
struction marked "#Sent#." This is the total number of
sentences generated by the solution.
Restricted Generation Potential. If there are restrictions
in the solution, the number of sentences will be less. The total
number of sentences, less the restricted sentences, gives the
restricted potential: RGP— MGP — restricted sentences. No gen-
eral rules can be given to cover all possible restrictions, but these
suggestions may help:
1. Classes may be divided into subclasses and calculated
one at a time. Thus, nouns as subject or object may be
called ni and n2.
2. Mutually exclusive items must be added together to
form one set, to which zero is added only once, if the set
is optional.
3. Where concord of items is expressed, the numbers
genders, and persons must be calculated separately,
for example, singular versus plural constructions.

Tagmemic Lexicon
The lexicon is the "total stock of morphemes in a language"
Bloomneld (1933:162). In a particular language problem, the
lexicon is the totallist of morphemes in the particular language,
 84 INTRODUCTION TO TAGMEMIC ANALYSIS

together with the morphophonemic rules which describe mor-

pheme variants. The lexicon is the second half of a tagmemic
solution.
Form, Class, Gloss. The lexicon accompanying a tagmemic
grammar has entries listed in three columns one for the mor-
phemic form, one for the classification, and one for the gloss or
translation:
1. Form is entered in the first column. Where the mor-
pheme is invariant, the phonemic form is entered. If the
morpheme has variant forms, either the morpheire is
listed as a base symbol in braces { }, or all forms of the
morpheme are listed, separated by alternation signs.
2. Class is entered in the second column. This is the form
class label chosen to represent the morpheme class in
the formulas of the grammar. The classification in the
lexicon, must be identical with the form class symbols
used in the filler slots of the grammar, so that the forms
can be automatically programmed into the formula.
3. Gloss is entered in the third column, always in single
quotes. The gloss is the assumed meaning of the mor-
pheme. It is not an exact translation. Forms with con-
trasting meanings must be clearly identified and the
gloss should show where their meaning is contrastive.
Zeros in the Lexicon. Both zero morphs and zero morphemes
may occur in the lexicon. Zero morph is a zero representing one
alternate of a morpheme, in cases where the morpheme has
nonzero alternates. Zero morpheme is a zero representing the
only alternate for a morpheme. To set up a zero morpheme, zero
must be one of a structural series, a filler of a tagmeme which has
overt elements as fillers. Thus, some languages have pronominal
affixes for first and second persons, but the third person is recog-
nized by the absence of any form. In these cases, a clear positive
meaning is assigned to zero in this particular slot.
If zero is listed as meaningful in the lexicon, then the slot
which it fills is marked as obligatory; it must occur but one filler
may be zero. If zero is not listed in the lexicon, then the slot re-
mains optional wherever, in at least one case, the slot remains
unfilled. Care should be taken that zeros do not proliferate in
analysis. Zero morphs are in contrast with overt variants of the
same morpheme; zero morphemes are in contrast with other
fillers of the same tagmeme. Only in these contexts is the zero
notation meaningful.

The Classified Lexicon. The entries of the lexicon should be

in a fixed order. One possible order is alphabetical but this is
 CLAUSE LEVEL 85

not always the most useful. An alternative method is to use a

classified lexicon. The morphemes of the lexicon are classified
according to a tripartite system, into: (1) nouns and noun modi-
fiers; (2) verbs and verb modifiers; and (3) uninflected particles.
Morphemes can be separated into these three groups, and then
listed alphabetically, where required, within these classifica-
tions. This system allows ready reference to all verb forms, all
noun forms, and so on, at a glance. For example:
I. The Noun System: (form) (class) (gloss)
Nouns pA-Sin n. 'man
Adjectives ha?yao aj. 'much'
II. The Verb System:
Verbs pa?t tvs. 'find'
Adverbs yA?m loc. 'here'
III; Particles dya neg. 'not'
Again, the classifying labels are not part of the solution but are
included as a descriptive convenience, enabling the analyst to
locate immediately the set of nouns, verbs, and so on which are
included under each form class label.
Morphophonemics. If all morphemes are invariant, then no
morphophonemic statement is required. If, however, some of the
morphemes have variants, then this is indicated by the notation
{ } in the form column, and a statement is included within the
lexicon giving the form and distribution of the morpheme var-
iants. This may be done in several ways:
1. Solution by Allomorph8. Each variant is listed as an
allomorph of the morpheme, and a distribution state-
ment is added for each allomorph, telling where, under
what conditions, the allomorph will occur.
2. Solution by Morphophoneme8. One base form is giv^n in
morphophonemic notation, in which phonemes wnich
are variable are marked in capital letters, and phonemes
which are constant in the form are marked in small
letters. The distribution statement then lists the
changes that occur in the variable phonemes only.
This is useful in partial changes.
3. Solution by Process Rule. The process rule is a general
statement which explains the changes which occur, in
terms of a process or change. These rules may be ex-
pressed in terms of transformational statements in-
volving either whole forms or parts of forms. Thus, the
summary process rule
A -» B / X—_Y may be read as:
A is rewritten as B in the environment
where either X or Y may be null.
 86 INTRODUCTION TO TAGMEMIC ANALYSIS

The rule is adequate, in that it explains what change

takes place (A becomes B) and explains in what par-
ticular environments this occurs. The environment may
be phonemic, morphemic, or even tagmemic.

PRACTICE 3: CLAUSE LEVEL STRUCTURE

Write the sentence level formula for the following set of sen-
tences, showing the type of clause present. Isolate subjects, pre-
dicates, objects, and adjuncts and write a clause level formula.
Write the lexicon, and calculate the maximum generation poten-
tial of the solution. The material is taken from Thomas J. Egan,
C.SS.R., Tereno Grammar.
TERENO (Mato Grosso, Brazil)
1. kaliwano kuriko yaye kohoyene
'The children played here today.
2. koituke neukacheke
'He worked yesterday.'
3. chaane imoko horokocheno yotike
'The people slept there during the night.'
4. ceeno koyoho
'The woman spoke.'
5. imoko
'He slept.'
6. ceeno koituke yaye
'The women worked here.'
7. hoyeno kuriko horokocheno neukacheke
'The men played there yesterday.'
8. kaliwano koyoho yaye yotike
'The child spoke here during the night.'
Problem 3: TERENO (Brazil)
GRAMMAR (240 Sent)
Sentence Level Construction:
Sent = +Base:iCl -Into:ICF
Read: A sentence consists of a base slot filled by an intran-
sitive clause, and an intonation slot filled by a final intona-
tion contour.
Clause Level Construction:
iCl = ±S:n +P:iv ±L:loc ±T:tem
Read: An intransitive clause consists of an optional subject
slot filled by a noun, a predicate slot filled by an intransitive
 CLAUSE LEVEL 87

verb an optional location slot filled by a locational adverb

and an optional temporal slot filled by a temporal adverb.

LEXICON (13 morphs)

I. Noun System:
Nouns (4) chaane n. 'people'
ceeno n. 'woman/women'
hoyeno n. 'man/men'
kaliwano n. 'child/children'
II. Verb System:
Verbs (4) imoko IV. 'slept'
koyoho iv. 'spoke'
koituke iv. 'worked'
kuriko iv. 'played'
Adverbs (4) horokocheno loc. 'there'
yaye loc. 'here'
kohoyene tern. 'today'
neukacheke tern. 'yesterday'
yotike tern. 'during the night'

SUPPLEMENTARY READINGS 3
Elson, Benjamin, and Pickett, Velma, An Introduction to Morphology
and Syntax, Santa Ana, Summer Institute of Linguistics, 1962.
Chap. 7, "Tagmemes and Constructions at the Clause Level,"
64-72; Chap. 17, "Survey of Clause Types," 108-121, active, reflex-
ive, passive, indefinite, stative, imperative, optative, interrogative,
emphatic, and various types of dependent clauses.
Fries, Charles C, The Structure of English, New York, Harcourt, Brace
and World, Inc., 1952. Chap. 9, "Structural Meanings, Subjects
and Objects," 173-201; Chap. 12, "Immediate Constituents, Layers
of Structure," 256-273, with branching rules given 271-272. For
word classes, Class 1, 76; Class 2, 80; Class 3, 82; Class 4 83.
Gleason, Henry A., Jr., Linguistics and English Grammar, New York,
Holt, Rinehart and Winston, Inc., 1965. Chap. 7, "Syntactic rela-
tions," 138-167, with traditional and IC diagrams; Chap. 13,
"Clause Patterns," 299-328, with transformations, 304-306.
Greenberg, Joseph H., Univeraals of Language, Cambridge, Mass., The
M.I.T. Press 1963. Chap. 5, "Some Universals of Grammar with
Particular Reference to the Order of Meaningful Elements,"
73-113, treats SPO order in language and prepositional/postposi-
tional contrast.
Koutsoudas, Andre, Writing Transformational Grammars, New York,
McGraw-Hill, Inc., 1966. Chap. 7, "Conjoining," 231-232; Chap.
8, "Embedding," 269-270, in transformational grammars.
Liem, Nguyen Dang, English Grammar, A Combined Tagmemtc and
Transformational Approach, Linguistic Circle of Canberra, 1966.
 88 INTRODUCTION TO TAGMEMIC ANALYSIS
Chap. 1, "Independent Declarative Clause Types," 1-20, with sum-
mary chart 1.2, listing 10 independent declarative types. Minimum
and expanded formulas are given, with examples.
Longacre, Robert E., Grammar Discovery Procedures, The Hague,
Mouton & Co., 1964. Chap. 1, "Clause Level," 35-73, including
drama analogy, 35; rule of two, 47; and clause matrix, 70.
, and Williams, Ann F., "Popoluca Clause Types," Ada Linguis-
tica Hafniensia, vol. X, no. 2, 161-186 (Copenhagen, 1962). Dis-
tinguishes 6 kernel (statement) clause types and 12 derived (in-
terrogative) clause types, based on formal criteria.
 TABLE 4: PHRASE LEVEL ANALYSIS
According1 to the According to the According to the Characteristics
Type of Grouping Type of Structure Internal Structure of Phrase Type
Exocentric Relater-Axis (RA) Recognized by
Noncentered relater and phrase relater class
1. Coordinate Phrase Recognized as
similar phrases, same type phrases
Phrase different referents with conjunction
Endocentric
as a 2. Item-Appositive (IA) Recognized as
Multiple Head
Structured similar phrases, same type phrases
Word Group same referent without conjunction

1. Noun Phrase (N) With possessives,

with noun head determiner, adjective
2. Verb Phrase (V) With auxiliaries,
Endocentric with verb head negative, adverbs
Modifier-Head 3. Adjective Phrase (Aj) With intensifiers,
with adjective head degree markers
4. Adverb Phrase (Av) With intensifiers
with adverb head for adverb class
 PHRASE LEVEL

The phrase level of grammar is that level that is below the

clause level and above the word level. The phrase is composed
of words and typically fills slots at the clause level. It is a con-
struction in which the constitute is a close-knit morpheme se-
quence which functions as a typical unit at the clause level and
whose constituents are words. Phrases are word groups that fill
the same slots at clause level as are filled by single words. This
level is comparable to the "group level" of Halliday's scale-and-
category; grammar.
Phrase Defined. The phrase is denned as "a unit composed
of two orinore words potentially, which does not have the charac-
teristics of a clause, and typically, but not always, fills slots on
the clause level" (Elson and Pickett, 1962:73). In this definition
the following features must be noted, in order to understand
phrase level construction:
1. Phrases typically fill slots on the clause level. In typical
mapping of lower level constructions into higher words
combine to form phrases, and phrases combine to fill
clauses. The phrase is discovered at the clause level as
a functioning unit, and is analyzed into constituent
parts at the phrase level. The clause level slot is filled
either by words or by word groups; if the group is not an
embedded clause (see p. 74), then the word group is a
phrase.
2. Phrases do not have the characteristics of a clause. Func-
tioning word groups may be clauses or phrases. Clauses
are recognized as having one and only one predicate tag-
meme in the string (see p. 65). All other word groups are

91
 92 INTRODUCTION TO TAGMEMIC ANALYSIS

phrases. Even the verb phrase, which fills the predicate

slot, is distinguished from clause, and has its own inter-
nal unity, although the same word group may be phrase,
clause, and, with intonation sentence.
3. Phrases consist potentially of two or more words. The
phrase is a potential word group. It is not obligatorily
complex. The phrase symbol signifies both the head word
in- isolation and the head word with modifiers.
Traditional Phrase. The phrase has traditionally been con-
sidered a unit, but the various types of phrases have often lacked
systematic treatments. The phrase has been defined as a word
group not containing subject and predicate, which functions as
a single part of speech. Some of the phrase types treated include:
(1) prepositional phrases, (2) coordinate and appositive phrases,
and (3) modification phrases. Verbal phrases are often extended
to include infinitives and participles.
In tagmemic analysis, the definition of clause is extended to
any word group containing one and only one predicate, so that
word groups which contain a predicate are listed as clause struc-
tures. Infinitive or participle constructions that contain, po-
tentially, other clause level tagmemes are listed as embedded
clause structures. The phrase, on the other hand, is extended
to include "potential phrases," that is, single words with op-
tional modifiers are listed as phrases. Thus, according to Pike
(1967:439), a phrase is +(+word +word), or +(+word ±word), but
not +(+word). A phrase is a unit which is composed of either two
or more words or is one word which is optionally expandable.
Phrase Types. The phrase types of traditional grammar may
be organized systematically according to formal features as
exocentric and endocentric structures. The latter may have one
head or many. An exocentric construction is a noncentered con-
struction. Endocentric constructions are centered constructions,
in which the whole construction fills the same clause level slots
as the head of construction. Endocentric constructions may be
multiple head or single head constructions. In multiple head
construction, the two or more heads may be coordinated or if
they have the same external referent, may be in apposition.
These basic construction types are listed by Hockett (1958:184-
185).
Phrase types may be outlined as follows:
1. Exocentric Phrase: prepositional phrase.
2. Endocentric Phrase: fills same slots as head words:
A. Multiple Head Phrase: more than one head word:
(1.) Coordinate: heads have different referents.
(2.) Appositive: heads have same referent.
 PHRASE LEVEL 93

B. Modifier-Head Phrase: only one head word.

Noun verb, adjective, adverb phrases.
From this outline of phrase types, it may be seen that the phrase
level is essentially the level of endocentric construction. All
phrases are centered about head words except the prepositional
phrase, in which the relater, or preposition, is attached to the
word or word group and fits this word or group for a particular
syntactic use. Elsewhere, phrases are structured word groups
filling the same slots as the head of construction.

RELATER-AXIS PHRASES
The phrase level of grammar deals with structured word groups
which are not clauses. Since these word groups are of various
kinds we find within the phrase level, various geological strata
or layers. One such layer is the layer of relation. Words or word
groups, together with a phrase relater, act as functioning units.
These are called relater-axis phrases.
Relater-Axis Phrase Defined
A relater-axis phrase is a structured word group with two
immediate constituents, one of which is a phrase relater and the
other a word or word group governed by the relater and called
the axis. This construction type is exocentric because neither the
relater alone, nor the axis alone, may fill the same clause level
slots as the relater-axis group.
Relational Structure. In our discussion of the features which
might be found in emic levels of structure within a language
(see p. 32), we suggested that levels may be characterized by
three geological strata: the layer of relation, the layer of co-
ordination, and the layer of subordination. At the phrase level,
the layer of relation is exemplified by the relater-axis type con-
struction which is a phrase, but of a type different from other
phrases.
The function of the relater, in this type of structure, is to
relate the whole constitute of the axis whether word or phrase,
to the grammatical structure of which it is a part. The relater
acts as a sky hook, which takes the word or word group that is
its axis and ties it into the structure. The function of the relater
here is analogous to the clause level relater which subordinates
clauses to sentence structure and to inflectional endings at the
word level which relate the word form to syntactic use.
This structure should always be treated as an immediate
constituent structure with two and only two constituents the
relater and its axis. The reason for this is that the axis in turn
 94 INTRODUCTION TO TAGMEMIC ANALYSIS

will often prove to be another phrase, of an endocentric type,

similar in structure to the other endocentric phrases of the lan-
guage. In an ordered analysis, the relater-axis phrase must al-
ways be analyzed before the phrases that might constitute the
axis. First, remove the relater, then analyze the axis together
with other endocentric structures. This will generally lead to
economy of description, although it is possible, with multiple
nesting structures, to have a relater-axis phrase within a sim-
pler phrase, containing, in turn, a simpler phrase.
The axis is often "governed" by the relater. Government is
one type of concord in which the case is determined by the other
words in the structure. Government occurs in relater-axis
phrases, and also in the relation between verbs and objects in
many languages.
Phrase Relaters. The relater-axis construction may be recog-
nized by the set of phrase relaters of the language. Phrase re-
laters are a group of function words, opposed to clause level
relaters (subordinating conjunctions) on the one hand, and to
inflectional endings on the other. In languages which have no
written literature it is often difficult to determine whether re-
laters are bound or free. If free, they are phrase relaters; if
bound they are inflectional endings. One clue to their identity
is whether they occur in construction only with single words, or
with word groups. Those relaters which govern word groups tend
to be free relaters.
Relater is preferred to the term "preposition," because all
relaters in all languages are not preposed. The relaters of Eng-
lish are called prepositions; the relaters of Hindi are called post-
positions, and always occur after the word or words they govern.
Greenberg (1963:77) has attempted to correlate the order of
Subject-Predicate-Object, with the use of prepositions and post-
positions in language structure. He concludes:
Type I: PSO always prepositional (Univ. 3)
Type II: SPO mainly prepositional
Type III: SOP mainly postpositional (Univ. 4)
Relaters need not be single words, whether prepositional or post-
positional. Most languages contain complex, as well as simple
relaters. The best procedure for these complex relaters is simply
to include them within the closed list which contains single word
relaters.
Relater Class. The most useful way to define classes of func-
tion words is either to list the set exhaustively or to give such a
representative list of the function word class that other members
may be recognized in similar positions. The function word class
is then defined as this typical list plus all other words which fill
 PHRASE LEVEL 95

similar slots in syntactic structure. For English, a typical list of

prepositional relaters is as follows:
about in
above into
aeross of
after off
against on
among out
around over
at through
before to
behind under
between until
by up
down upon
during with
for within
from without
The prepositions of this list are included within the 1000 most
frequently used words in English, according to the Thorndike
count (1944). Similar lists may be found in Fries, (1952:95), and in
Lado (1964:123).
Types of Relater-Axis Phrase
Relater-axis phrases are similar in their internal structure;
they are all composed of two obligatory tagmemes, a relater and
an axis. But they may be distinguished according to their exter-
nal distribution. The relater-axis phrases fill different slots in
clause and phrase level structures, and on this basis may be dis-
tinguished as nominal, adjectival, and adverbial types. This basis
for contrast is used in many traditional English grammars and
corresponds, at clause level, to the drama analogy (see p. 68).
Adverbial Relater-Axis Phrases. Those relater-axis phrases
that fill the same clause level slots as single adverbs are adver-
bial relater-axis phrases. These occur at the clause level of analy-
sis filling peripheral slots of time, place, manner, as props or
scenery for the central action.
Temporal relater axis phrases answer the question when and
fill slots at the clause level normally filled by adverbs of time.
Temporal relater-axis phrases may be recognized by their exter-
nal distribution; they fill temporal slots in clause structure. They
may also often be recognized by their internal structure. The
temporal relaters may be listed as a separate class, for example
before, until, during, since, after, are typical temporal relaters.
Other relaters in this class include: in, at, on, about used with
 96 INTRODUCTION TO TAGMEMIC ANALYSIS

time phrases. Principal relaters in a set can often be systemati-

cally arranged in a chart, as suggested by Lado (1964:123):
before during after
until at since
Locational relater-axis phrases answer the question where and
fill slots at the clause level normally filled by adverbs of place.
Locational relater-axis phrases may be recognized by their ex-
ternal distribution; they fill locational slots in clause structure.
They may also be recognized by internal structure. The loca-
tional relaters may be listed as a separate subclass of relaters
for example, into, in, out of, through, around, over, under, to, from,
up, down, on, off, are typical locational relaters (see chart, Lado
1964:123).
into at out of
to in from
Manner relater-axis phrases answer the question how and fill
slots at the clause level normally filled by adverbs of manner. In
early stages of the analysis, it is useful to group all adverbial
phrases except time and place under the manner label and then
establish more particular classes as the analysis proceeds. The
manner phrases show distinct subclasses, such as instrument,
accompaniment, purpose, cause, benefactive, and so on. Typical
relaters in this function are: for, against, by, by means of, with
without like. Excluded from this grouping would be such nuclear
slots as agent, marked with by, and indirect object, marked with
to or for.
Adjectival Relater-Axis Phrases. Those relater-axis phrases
that fill noun modification slots at phrase level, or the predicate
attribute slot at the clause level, are adjectival relater-axis
phrases. They fill the same slots as single adjectives in clause and
phrase structure. Phrases which fill modifier slots at the phrase
level are nested phrases and manifest a layering (see p. 31) of
phrase within phrase. In English, nested relater-axis phrases
occur immediately after the head noun and before clauses:
Single modifier + Head noun + Phrases + Clauses
The descriptive function of the nested relater-axis adjectival
phrase may often be demonstrated by transforming the relater-
axis phrase into a preposed modifier. For example:
the mayor o/Boston becomes Boston's mayor
the girl with the red hair becomes the red-haired girl
the man in the gray becomes the gray-flannel-suited
flannel suit man.
 PHRASE LEVEL »7

The transformed phrase illustrates another type of adjectival

relater-axis phrase, in which a bound relater, or phrase clitic,
governs the phrase. The use of bound relaters, such as {—Z2},
possessive,' and {-ed}, 'having,' is an example of level skip-
ping (see p. 31); a bound form, belonging to the word level is used
in a higher level construction (here a phrase construction).
Nominal Relater-Axis Phrases. When phrases occur in nomi-
nal nuclear slots at the clause level and are marked by specific
particles rather than case endings, they resemble, in a superficial
way, the relater-axis phrase. But relater-axis phrases more
typically fill peripheral slots at the clause level, or modifier slots
at the phrase level. Therefore it is often advisable not to use the
relater-axis terminology for the nuclear slots in analysis. One
alternative is to consider these phrases as "marked" nominal
phrases, and the particles involved as "markers" for nominal
slots.
1. Indirect object, in English, is often marked with to or
for. But this case is also marked by position, and the
marker here is optional. For example:
I gave the money to him. I gave him the money.
2. Infinitive forms, in English, are marked with to. But
this marking belongs to the infinitive form, and does
not mark the whole embedded clause. In the predicate
slot of the embedded clause, the verb form can be listed
as Vlnf, which is then analyzed as marker 4- verb base.
I allow him to come. (iVlnf = to + iv)
Marked nominal phrases are treated in the same way as the
cases of a case inflected language. Latin has subjects in the
nominative case and objects in the accusative case; Japanese
marks these same slots with particles {o} and {wa}. Latin has
indirect objects in the dative case; English marks these cases
with {to} and {for}, or by position. If these marked nominals
with markers that are free forms, are treated in the same way as
case-marked nominals in which the markers are bound forms
then the fillers of nuclear clause level slots need never be for-
mulated as relater-axis phrases.

Formulating Relater-Axis Phrases

The relater-axis phrase is a phrase that always consists of
two obligatory tagmemes, a relater and an axis (Elson and
Pickett, 1962:75,106). The relater is a preposition, a postposition,
a noun in special use, or a bound particle. The axis may be a word
another phrase, or an embedded clause. The phrase is called RA
 98 INTRODUCTION TO TAGMEMIC ANALYSIS

and the obligatory tagmemes are marked as +R: (relater) and

+Ax: (Axis) and are placed in the order in which they occur in the
given construction.

Typical Mapping. Relater-axis phrases that fill typical clause

level slots of time, place, and manner are examples of typical
mapping of phrase into clause. When relater-axis phrases are
discovered at the clause level, they are analyzed before those
phrases that constitute the axis alone. For example:
During the afternoon he worked at the university.
4.1 iCl = ±T:RA,+S:pn+P:iv ±L:RA
4.2 RA = +R:rel] +Ax:N during +(the afternoon)
Read: A (temporal) relater-axis phrase consists of a relater
slot filled by a relater (class 1), and an axis slot filled by a
noun phrase. (Class 1 = before, during, after, and so on.)
4.3 RA = +R:rel2 +Ax:N at +(the university)
Read: A (locational) relater-axis phrase consists of a relater
slot filled by a relater (class 2), and an axis slot filled by a
noun phrase. (Class 2 = into, in, out of, to, at, from, and
so on.)
The noun phrase which fills the axis slot is subsequently ana-
lyzed as a head-modifier type of phrase, and is generally similar
to other noun phrases found at the clause level in subject and
object slots.

Atypical Mapping. Relater-axis phrases that fill phrase level

slots are examples of layering of phrase within phrase. Relater-
axis phrases that have bound forms as relaters are examples of
level-skipping. Both are cases of atypical mapping, for they are
outside of the normal mapping of the lower level constructions
directly into higher level constructions. Both are adjectival in
function and both are illustrated in normal adjective type con-
structions as well as in special possessive constructions.
Layering of clause within phrase is handled by means of the
identifier tagmeme (see p. 76). This tagmeme is always mani-
fested by embedded clauses and never by embedded phrases.
The phrase that contains the embedded clause is called an "iden-
tified" noun phrase (Nld).
Layering of phrase within phrase is handled by means of the
modifier tagmeme, filled by an (adjectival) relater-axis phrase.
In English, this modifier tagmeme follows the head noun when
the relater is a free form. For example:
 PHRASE LEVEL 99

The girl with the red hair

4.4 N = +Det:det +H:n ±Mod:RA
4.5 RA3 = +R:rel3 +Ax:N with +(the red hair)
Read: An (adjectival) relater-axis phrase consists of a re-
later slot filled by a relater (class 3), and an axis slot filled
by a noun phrase. (Class 3 = of, in, with, and so on.)
Level-skipping, in which a bound relater is used to govern a
phrase, is best analyzed at the phrase level before the phrase
that constitutes the axis. The alternate solution, of treating the
phrase as a word stem and performing the analysis at the word
level, fails to demonstrate the use of bound particles (generally
restricted) as phrase clitics.
The red-haired girl
4.6 N = +Det:det±Mod:RA +H:n
4.7 RA, = +Ax:N +R:{-ed} (red hair) + -ed
Read: An (adjectival) relater-axis phrase consists of an axis
slot filled by a noun phrase, and a relater slot filled by the
bound relater {-ed}.
Government is a type of concord that may occur m relater-
axis phrases. If the choice of case, for the axis, is determined by
the choice of relater then the axis is said to be "governed by the
relater." To program into the grammar a subroutine for selecting
the proper case with each relater, the R: and Ax: slots are con-
c
nected by a concord tie: +R: +Ax: Such a device automati-
cally selects the correct case. In Latin, for example, it would
select urbem, after ad, in ad urbem, 'to the city,' and would select
urbe after ab, in the phrase, ab urbe, 'from the city.
MULTIPLE HEAD PHRASES
With the exception of the relater-axis phrases, the phrase level
of structure is endocentric. Endocentric phrases are coordinate
or subordinate. The coordinate endocentric phrases are multiple
head phrases, in which the multiple head phrase fills the same
slot as either of its constituent heads. The layer of coordination,
or conjoining, is the second layer of phrase level construction.
Multiple Head Phrase Defined
Multiple head phrases are defined as structured word groups
which contain more than one head. The heads of constructions
may be single words, or they may be subordinate phrases. If the
 100 INTRODUCTION TO TAGMEMIC ANALYSIS

two heads have the same referent, the construction is appositive;

if the two heads have not the same referent, the construction is
strictly coordinate.
Coordinate Structure. Every level of the grammar should
admit the possibility of conjoining elements at that level, accord-
ing to Huddleston (1965, 41), cited on page 33. The conjoining
process has been defined by Noam Chomsky (1957:36) in the fol-
lowing rule:
If S and S2 are grammatical sentences, and Si differs from S2 only
in that X appears in St where Y appears in S2, and X and Y are
constituents of the same type m Sj and S2 respectively, then S
is a sentence, where S3 is the result of replacing X by X+ and +Y
inS..
In an article entitled "Conjoining in a Tagmemic Grammar of
English" (18th Annual Round Table, Georgetown University,
No. 20 1967,109-121), Alton L. Becker sets up rules for conjoin-
ing for tagmemics, with emphasis on the decision procedure
involved in choosing constituents of the same type. The simi-
larity must be based on function rather than on form. Restrict-
ing conjoining to the phrase level sets up the following rules:
1. Conjoining is independent of form. That is, phrases may
be joined to words, or phrase to phrase, or word to word.
However, constituents will be of the same functional
type, such as nominal, adjectival, and so on.
2. Conjoining is dependent on function. Elements con-
joined must fill the same functional slot, that is, they
must belong to the same tagmeme. Elements joined are
both subjects, predicates, temporals, locationals and
so on.
3. Conjoining at phrase level tends to be open-ended, pro-
viding the grammar with what Becker calls "linear
recursiveness" (see p. 45), which is opposed to the recur-
siveness due to layering or embedding.
Phrase Connectors. Multiple head phrases that are coordi-
nated may be recognized by a limited set of function words called
connectors. The coordinate phrase may be joined by zero con-
nectors, by single connectors, or by double connectors. The ele-
ments connected in the coordinate phrase have the same general
function, but have different external referents:
Parataxis: Read Ibsen, Tolstoy
One connector: Read Ibsen and Tolstoy
Two connectors: Read both Ibsen and Tolstoy
The functional meaning of the phrase may change according
to the type of connectors used. These connectors are: (1) additive,
 PHRASE LEVEL 101

such as: and; (2) disjunctive, but not; (3) alternative, either/or;
(4) comparative, rather/than.
In contrast, the appositive multiple head phrase is char-
acterized by zero connector and the heads of construction have
both the same grammatical function and the same referent in
the extra-linguistic world:
Apposition: Read Tolstoy, the author of War and Peace
If connectors are introduced, the phrase is interpreted as co-
ordinate and the referents of the two heads of construction are
presumed different:
here in Washington Apposition: here = in Washington
here and in Washington Coordinate: here 5* in Washington
Connector class. The connector class used in conjoining at
the phrase level, is also used at the sentence level in combining
clauses to form compound sentences (see p. 45). The following
list of connectors is common in English; it is listed by Fries
(1952:94-95) as Group E:
and both .. .. and
but, but not not ..but
not neither. ..nor
or either .. . .or
rather than rather . ..than
In the listing above, Fries notes that the connector, but used
with words and phrases, is more common with adjectivals, poor
but honest or with adverbials, slowly but carefully. With nomin-
als this connector is common only with indefinite pronouns,
in such expressions as everything but, nothing but. It is not com-
mon elsewhere with nominals or verbals. In tagmemic formula-
tions, the first list is the list of single connectors and the second
list is the list of double connectors which fill the first (cj and
second (Cj.) connector slots respectively. The connector slot as
part of phrase level, is marked with a capital letter, as C:; the
fillers, being words, are represented by small letters, as c or as
Cj . . . C 2 ,

Types of Multiple Head Phrases

Multiple head phrases are divided into subtypes according
to both formal and semantic norms. The two main types are
coordinate phrase and item-appositive phrase. The coordinate
phrase may be further subdivided as coordinate (1) noun, (2)
verb, (3) adjective, and (4) adverb phrases. The list is not exhaus-
tive; other parts of speech, including the function words, may
often be coordinated by connectors.
 102 INTRODUCTION TO TAGMEMIC ANALYSIS

Coordinate Phrases. These are multiple head phrases, gen-

erally joined by a connector in which the heads of construction
have different external referents. Heads of construction belong
to the same function class and individually fill the same slots as
the whole construction.
1. Coordinated Noun Phrase(NC0) The conjoining of two or
more phrases of the nominal type. These include nouns,
noun substitutes, and phrases with nouns as head,
which fill the same tagmemic slot. For example:
He and I
Jack and Jill
The old man and the sea
2. Coordinate Verb Phrase(Vco). The conjoining of two or
more verb forms whether these be main verbs, partici-
ples, or auxiliaries. The forms that are conjoined belong
to the same functioning subclass:

They were singing and dancing.

He could and should pay the bill.
3. Coordinated Adjective Phrase(AjC0). The conjoining of
two or more phrases or words of the adjectival type.
These include single adjectives, as well as intensified
adjective phrases, composed of intensifier + adjective.
The adjectives joined must be functionally the same
and fill one slot.
She was young and very pretty.
There were three or four red flowers.
4. Coordinate Adverb Phrase(AvC0). The conjoining of two
or more phrases or words of the adverbial type. These
include both single adverbs and intensified or other
relater-axis adverbial phrases. The adverbials must
fill the same function as temporal, locational, manner
instrumental, and so on.
He drove the car slowly and with caution.
The snow fell softly and very silently.
Besides the function classes based upon the four major form
classes the function words also may often be connected. In this
case the group of words so connected fills the same slot as a
single word would fill. For example:
The bats flew in, out, over, around, and through the belfry.
Come on up out from down in under there.
In the second example the relaters are not coordinated. This
 PHRASE LEVEL 103

example is rather a recursive application of a relater-axis con-

struction.
Itera-Appositive Phrase (IA). These are multiple head phrases
in which the two heads have the same extralinguistic referent
and are not joined by members of the connector class. Like the
relater-axis phrase, the item-appositive phrase is best analyzed
as consisting of two immediate constituents, that is of two and
only two obligatory tagmemes, the item tagmeme and the ap-
positive tagmeme. As with coordinate phrases the two heads
of construction must fill the same tagmemic slot in the structure.
Concord in item-appositive phrases is more generalized, since
the heads have the same referent, and we expect concord of
gender, number, and case. But the grammatical forms used in
apposition in language do not always neatly conform to this
pattern. Item-appositive phrases are most commonly nominals;
but other types of apposition may occur with parenthetical
phrases, such as that is and for example, introducing apposi-
tives. Typical examples of item-appositive phrases include:
Tony, the barber We, the people
He the man in the street You the defendent
Item-appositive phrases are analyzed as multiple fillers of a
single tagmeme and not as repetitions of the same tagmeme.
Thus:
I live here in Washington
is first analyzed as Subject-Predicate-Locational, at the clause
level. The single locational slot is filled by an item-appositive
phrase (L:IA), and the item-appositive phrase is then analyzed
as consisting of an item slot filled by here, and an appositive slot
filled by the phrase in Washington. This analysis belongs to
phrase level.
Repetition of Tagmemes. If like tagmemes are either coor-
dinated or in apposition, how is it possible to have a repetition
of the same tagmeme in the string? Either the tagmemes are the
same or different; if the same, they may be analyzed as coordi-
nates; if different, they are not repetitions of the same tagmeme
but belong to two different tagmemes.
Items and appositives are analyzed as single fillers in the
slot, for example, as a locational slot filled by an item-appositive
phrase. Elements in apposition are certainly not repetitions
of the same tagmeme but are complex manifestations of a single
tagmeme.
Coordinate phrases with overt connectors are analyzed as
a single coordinate filler of a single slot. Because of the connector
such phrases cannot be analyzed as repetitions of the same
 104 INTRODUCTION TO TAGMEMIC ANALYSIS

tagmeme but are complex manifestations of a single tagmeme.

Coordinate phrases with no overt connector constitute the
difficult case. If they are functionally the same, they may be
considered a case of a coordinate phrase with zero connector
If not, they should belong to different tagmemes. Repetitions
should be regarded with some suspicion, and tagmemes should
be tested for formal differences.

Formulating Multiple Head Phrases

The multiple head phrase is a phrase which consists of more

than one obligatory head slot, with or without intervening con-
nector slots. The two types of phrases are coordinate and ap-
positive (Elson and Pickett, 1962:106). The same types are listed
by Longacre as double-centered, coordinate, linking construc-
tions (1964:74-75), as opposed to relater-axis constructions on
the one hand, and modification structures on the other.
Coordinate Phrases. The coordinate phrase consists of more
than one head tagmeme, generally all obligatory, and one or
more connector tagmemes, which may be optional. Slots are
marked with capitals. Fillers, if they are phrases, are marked
with capitals; if they are words, with small letters. A typical
formula for a coordinate noun phrase is:
4.8 N C 0 =+H 1 :N+C:c+H 2 :N the old man and the sea
Read: A coordinate noun phrase consists of a head slot
filled by a noun phrase, a connector slot filled by a connector,
and a head slot filled by a noun phrase.
Alternate fillers for the head slot include pronouns (pn), proper
nouns or names (np), and possibly nested appositive or coordi-
nate phrases. Since serial, or open-ended, coordination often
occurs in language at this level, a discontinuous notation may
be used to indicate these open-ended constructions in which
new items may always be added (see also formulas 1.2 and 1.3).
4.9 Nco = +HX:N ±H 2 :N . . . +C:c +H n :N
Read: A coordinate noun phrase consists of a head slot filled
by a noun phrase, an optional indefinite series of head slots
filled by noun phrases, a connector slot filled by a connector,
and a final head slot filled by a noun phrase.
Coordinate verb phrase (Vc0), adjective phrase (Ajeo) and adverb
phrase (Avco) are formulated in much the same way. Problems
arise not in the formulation of the coordinate phrase but in
recognizing coordination.
 PHRASE LEVEL 105

He drove the car slowly and with caution.

This transitive clause shows a manner slot filled by a coordinate
adverb phrase (M:Avco) in which the first head slot is filled by
an adverb (+Hi;av) and the second by a relater-axis phrase
(+Hs!:RA). Coordination must be recognized, and formulated
where it is discovered.
Item-Appositive Phrase. The item-appositive phrase consists
of two and only two obligatory slots, the item (marked It:) and
the appositive (marked App:). As with relater-axis phrases, the
double capital (IA) designation is used to stress the fact that this
construction is composed of two obligatory parts. A typical
formula for an item-appositive phrase is as follows:
4.10 IA = +It:np +App:N Tony, the barber
Read: An item-appositive phrase consists of an item slot
filled by a proper noun and an appositive slot filled by a
noun phrase.
Item and appositive are functionally identical, so it does not
make mueh difference which element is called item which
appositive; both have the same external referent and either
can substitute for the whole construction.
Pronouns and noun phrases are regular fillers for the item-
appositive phrases; clauses occur more rarely. The test for
clauses in apposition is whether the clause could stand alone in
place of the item-appositive phrase. According to this test, rela-
tive clauses which modify nouns are not in apposition, but fill
identifier slots modifying the head noun in the construction.
In true apposition, the item can be deleted, and the result is
still a sentence.
Concord in Multiple Head Phrase. The concord problems
raised by multiple head structure concern both internal struc-
ture and external distribution. The internal structure of phrases
involves agreement; external distribution of these phrases
involves problems of cross-reference.
Cross-reference is concord of subject and predicate. If the
subject is a coordinate noun phrase, cross-reference concerns
the combined subject in both head slots; if the subject is an item-
appositive phrase, the cross-reference is that of one of the head
slots only. For example:
Jack and Jill are plural subject, plural cross-reference
Tony, the barber, is singular subject and cross-reference
Agreement is concord between parts of the same construc-
tion. In coordinate endocentric phrases this concord is between
 106 INTRODUCTION TO TAGMEMIC ANALYSIS

the heads of construction. Since coordinate heads are con-

stituents of the same type," there will be at least case concord.
In item-appositive phrases, there is also agreement; but since
the two heads have the same referent, there is agreement of
gender, number, and case, but not of person. For example:
We the people, are . . .
You the president, are .
In an item-appositive phrase with pronouns, the pronoun tends
to dominate the construction and is in cross-reference concord
with the verb. The appositive elements here belong to a dif-
ferent person category, but agree with the item in gender, num-
ber and case.

HEAD-MODIFIER PHRASES
The first layer at phrase level concerns the relater-axis phrase,
the second layer the multiple head phrases. The third level deals
with the single headed phrases and the structure of modifica-
tion. This latter layer is also endocentric, but is endocentric and
subordinate. All of the elements of the phrase are subordinate
to the single head tagmeme.

Head-Modifier Phrase Defined

The head-modifier phrase is a phrase in which one of the
tagmemes is an obligatory head tagmeme, and the other tagme-
mes are modifiers which are subordinate to that head tagmeme.
Most modification slots are optional; but in a particular lan-
guage, as for the determiner slot in English one or more modi-
fiers may be obligatory, at least for some subclasses of head
words.
Modification Structure. The modification structure is a
subordinate endocentric structure in which there is one head
slot and a series of (optional) modifier slots. The whole construc-
tion fills the same slot as the fillers of the head slot, or the equiva-
lent substitute form. This layer of structure deals with the
costuming in Longacre's drama analogy and seems to be a uni-
versal feature of languages.
Head-modifier constructions are further subclassified ac-
cording to the form class of the words filling the head slot in the
construction. Constructions for all the major form classes occur:
1. A noun phrase (N) is a head-modifier phrase with a noun
as the head. The modifiers are adjectivals and include
determiners, quantifiers, possessives, and descriptive
adjectives, which may modify the noun.
 PHRASE LEVEL 107

2. A verb phrase (V) is a head-modifier phrase with a verb

as head. The modifiers are adverbials used in a close-
knit sequence with the verb, verbal auxiliaries or mod-
als, and negatives.
3. An adjective phrase (Aj) is a head-modifier phrase with
an adjective as head. The modifiers include intensi-
fiers used to modify adjectives and the markers for com-
parison when these markers are free forms.
4. An adverb -phrase (Av) is a head-modifier phrase with
an adverb as the head. The modifiers include intensi-
fiers that modify adverbs and the markers for compari-
son when these are free forms.
The designations of the head-modifier construction are
strictly formal and based on the form of the head word. These
designations are not to be confused with functional designa-
tions such as nominal and verbal phrases, which deal with the
function of word groups in syntax.

Noun Modifiers. The internal structure of head-modifier

noun phrases is predictable in terms of the noun modifiers
expected within the structure. These modifiers, generally but
not always optional, are combined with a single head slot (marked
H:) filled by the head noun.
Det: Determiner tagmeme, filled by determiners (det) includ-
ing the articles: a/aw, the; and the demonstratives: this/
that, theselthose.
Pos: Possessive tagmeme, filled by possessive pronouns (pos)
such as: my/mine, your/yours, his, her/hers, its, our/ours,
their/theirs. Nouns or noun phrases marked for posses-
sion also occur.
Qn: Quantifier tagmeme, filled by numerals (num) such as:
one two, first, second; or quantitative adjectives (qn):
all, many, more and so on.
Mod: Modifier tagmeme, filled by descriptive adjectives (aj)
which may be broken into subclasses of color, size, quality,
such as: red/white, big/little, thick/thin, old/new, dry/wet.

Verb Modifiers. The internal structure of verb phrases is

also predictable; and elements in the string include modals and
auxiliaries, together with negatives. More rarely, adverbs may
be used in modifier or intensifier slots when they form a close-
knit verb sequence. This is particularly true of adverbs of man-
ner, rarely of time and place. Adverbs generally fill peripheral
clause level slots (Elson and Pickett, 1962:104).
Aux: Auxiliary tagmeme, filled by modals or auxiliaries distinct
from the main verb form. Modals (mo) include: can, may,
 108 INTRODUCTION TO TAGMEMIC ANALYSIS

must, shall, will. Auxiliaries (aux) include forms such as:

have be do.
Neg: Negative tagmeme, filled by negatives (neg) such as: not
don't. Negatives may be different for statements and
commands for example: Hindi statements use nahli 'not';
Hindi commands use mat 'don't.' Sierra Popoluca has a
contrast between dua 'not' and odoy, 'don't.'

Adjective and Adverb Modifiers. The internal structure of

adverb and adjective phrases is generally limited to a closed
set of intensifiers, together with markers for comparative and
superlative degree.
Int: Intensifier tagmeme, filled by intensifiers (int) and markers
for comparison, in English: more and most, where these
are free forms. Intensifiers include: pretty, rather, very,
really, truly, and so on.
In English, there is growing use of a wide range of adverbs in
intensifier function. Thus, such phrases as remarkably intelli-
gent and intensely interesting manifest the intensifier + adjec-
tive construction.

Types of Head-Modifier Phrases

Head-modifier phrases are divided into subtypes according

to the word which fills the head slot of the construction. The
principal types parallel the major form classes, and are: (1) noun
phrase (N), (2) verb phrase (V), (3) adjective phrase (Aj), and
(4) adverb phrase (Av). Other word groups may occur in lan-
guage, but they generally represent patterns that are not widely
productive in the language.

Noun Phrase Subtypes. The noun phrase is a phrase in which

the head word is a noun. This phrase type may be divided into
different subtypes, according to the ways in which nouns are
classified in the language. Nouns are classified according to
distinctive features, as:
Common Count Animate Singular
Proper Mass Inanimate Plural
These features must be grammatically defined for the language,
and may be inflective or selective categories (Hockett, 1958:
230). Inflective categories are grammatical meanings carried
by inflectional endings; selective categories are grammatical
features recognized by the choice of items in constructions. Some
categories may show considerable overlap:
 PHRASE LEVEL 109

1. Proper nouns are "names," which can be preceded by

neither a definite nor an indefinite article in English;
common nouns may take articles. Note: Since names are
unique, a definite article is often part of the name. For
example: the whole phrase the Amazon is a name; Ama-
zon does not occur.
2. Count nouns are items which are countable. Mass nouns
refer to items which cannot be counted; they never occur
with the indefinite article. For example: a course, two
courses; but not *a music *two musics.
3. Animate nouns in English, are those nouns which may
be replaced by he/she/they; inanimate nouns may be
replaced by the pronoun, it. These gender classes may
overlap; for example: babies, household pets, which
may be used functionally now as animate now as inani-
mate nouns.
4. Singular/plural, in English, is in most cases an inflective
category, which is marked for plural by the inflection
{=Z,}, 'plural.' However, in a limited number of cases,
plural is selective, for example: cattle (plural).
Tagmemics formulates inflective categories at the word
level by separating inflection from stem. Selective categories,
such as gender (in English) and the count and common features,
may require separate formulas. All categories are covered at
the phrase level by concord ties, (1) of agreement within the
endocentric phrase, (2) of cross-reference between subject and
predicate, and (3) of government by verbs or relaters.
Concord Marking. Agreement is concord between head and
modifiers in a modifier-head phrase. In tagmemics, concord is
marked with a tie bar connecting fillers of the slots and marked
with a "c" for grammatical concord. This concord must be speci-
fied as to features, such as gender, number and case.

4.11 N = +Det:det +H:n ±Mod:aj la casa blanca

Read: A noun phrase consists of a determiner slot filled by
a determiner a head slot filled by a noun, and an optional
modifier slot filled by an adjective (in Spanish).
Read V ; Concord of gender and number.
The concord tie is a feature of selectivity, and calls for a sub-
routine that selects the proper fillers for each slot. This subrou-
tine is analagous to the square brackets of transformational
grammar, as in Koutsoudas (1968:13). In writing a tagmatic
manifestation of any one construction, items on the same line
 110 INTRODUCTION TO TAGMEMIC ANALYSIS

within the brackets are chosen. These are "etic variants of the
emic construction" (Pike, 1962:236). The formulation in 4.11 may
be made explicit in the following formulation:
rel -j "m.sg. "—o
4.12 N=+Det: la +H:n f.sg. ±Mod:aj —a
los m.pl. —OS
.las_ Lf.pi. J .—as.
T h e m e c h a n i c s of concord a r e outlined by Elson a n d P i c k e t t
(1962:87-91), developed by Pike as " t a g m a s in reciprocally-con-
ditioned v a r i a t i o n , " (1962:236), a n d specified by D a n M. M a t s o n
in " T a g m e m i c Description of A g r e e m e n t " (18th A n n u a l R o u n d
Table, Georgetown, 1967:103-108).
In a g e n e r a t i v e t a g m e m i c model we view t h e concord m a r k -
ing as calling up a subroutine that will select the proper modi-
fiers according to the head word selected. The rule, as Matson
has pointed out, is context-restrictive, and exponentiation (cate-
sian multiplication) is not uniformly possible. The output must
be broken down into four types of output as in the above example
with no cross-multiplication. Thus the output may be calculated
separately for the noun phrases:
NnNfsic N mpl Nfl
The subroutine would identify the head noun with the selective
feature for example, casa, n.f.; (note that it can be marked for
plural) and generate the phrases, la casa blanca and las casas
blancas while excluding the masculine articles and the adjec-
tives with masculine endings. The output generated contains
all grammatical and no ungrammatical sequences.

Formulating Head-Modifier Phrases

The head-modifier phrase consists of an obligatory head slot

with one or more modifier slots which may be obligatory or
optional. Except for nested constructions, the subordinate endo-
centric type of phrase is generally analyzed as the last stage of
phrase construction.

Noun Phrases. The noun phrase is a construction with a

noun filling the head slot and with various noun modifiers in the
string, for example: determiners, quantifiers, adjectives. Pos-
sessives are also noun type modifiers; but it is tactically often
more useful to consider the item + possessor phrase as consist-
ing of two and only two constituents.
In item-possessor phrases (IP), both item and possessor are
obligatory and nuclear (Longacre, 1964:84). Either item, or pos-
 PHRASE LEVEL 111

sessor, or both may be marked. The string consists of -Htem

+Possessor. Attributes in the string belong to either item or
possessor:
4.13 IP = +Pos:RA + It:n (the pioneer's) + family
Read: An item-possessor phrase consists of a possessor slot
filled by a relater-axis phrase, and an item slot filled by a
noun.
4.14 RA = +Ax:N +R:{Z2} (the pioneer) + -'s
In cases where the possessive is a single word, such as a
possessive pronoun that alternates with the determiner slot,
it is more useful to analyze the possessive as part of the string
modifying the head noun.
4.15 N = +Det:det/pos ±Qn:num ±Mod:aj +H:n
the/his three young children
Read: A noun phrase consists of a determiner slot filled by
a determiner or possessive, an optional quantifier slot filled
by a numeral, an optional modifier slot filled by an adjective,
and a head slot filled by a noun.
Nonpossessed noun phrases, aside from the problems of pos-
session and the marking of concord, tend to fall into a fixed pat-
tern with the elements in fixed order, and only the head nuclear
and obligatory. In internal structure, the one formula has many
readings, depending upon whether the optional elements occur
or do not occur in a given structure. In external distribution, it
is often convenient to label noun phrases as N1( N2, and so on to
limit their distribution in higher level structures.
Verb Phrases. The verb phrase is a construction with a verb
as the head word and with various auxiliaries, negatives, and
adverbs in the string. The problem of formulating verb phrases
is often concentrated upon the special problem of the English
verb phrase. In other languages, the verb phrase is often simpler
and can follow a strict word level analysis. For example, in Hindi,
verbs are either words, or participle plus auxiliary:
mai Jamshedpur nahii jaataa huu.
'I am not going to Jamshedpur.'
4.16 iV = ±Neg:neg +H:ivpart +Aux:aux
Read: An intransitive verb phrase consists of an optional
negative slot filled by a negative, a head slot filled by a
participle, intransitive verb, and an auxiliary slot filled by
an auxiliary verb.
English verb phrases consist of discontinuous elements for the
 112 INTRODUCTION TO TAGMEMIC ANALYSIS

aspect perfective have + en, and progressive be + vng, as in the

formula:
Aux = Tense (Modal) (Perfect) (Progressive)
formulated by Chomsky in Aspects of the Theory of Syntax (1965:
43), where modal is denned as the set: can, may, must, shall, will.
To produce a tagmemic formula which is the equivalent of this
analysis would require the mixing of phrase and word level
constructions, for example:
4.17 iV/tV = +T:tm ±Mo:mo ±Asp,:aspi ±Aspz:asp2
+H:iv/tv
Further rules would be required for permuting elements and
for the combination of morphemes into words to form the verb
phrase. The solution of Elson and Pickett (1962:106) seems to
mix levels of analysis.
One possible solution is to place the auxiliaries in three posi-
tion classes, with Aux3 = modals, Aux2 = have, and Aux, = be, and
correlate the occurrence of these with forms of the verb in -ing,
•en. Thus
1. Verb Phrase = +Aux3 ±Neg +Aux 2 +Aux, +Head verb
(modal) Modal have been V-ing
Modal 0 be V-ing
Modal have 0 V-en
2. Verb Phrase = +Aux2 ±Neg +Aux, +Head Verb
(have) have been V-ing
have V-en
3. Verb Phrase = +Auxi ±Neg +Head Verb
(be) be V-ing
The above forms are limited to the active voice. The negative is
optional; forms that precede the negative are tense-marked in
the construction.

PRACTICE 4: PHRASE LEVEL GROUPING

Write the sentence and clause level formulas for each of the
sentences below. For every word group that functions as a unit
at clause level, write a phrase level formula. Write the lexicon
and calculate the maximum generation potential of the solution.
The material was supplied by Teodore A. Llamzon, S.J., Philip-
pines.
TAGALOG (Philippines)
1. kumain arj isda
'The fish ate.'
 PHRASE LEVEL 113

2. nagluto arj tao nag matabag baboy

'The man cooked a fat pig.
3. kumain ag malakir) tao nag manok
*The big man ate a chicken.'
4. nagluto ag babae
The woman cooked.'
5. kumain ag baboy nag malakig isda
'The pig ate a big fish.1
6. kumain ag manok nag isda
'The chicken ate a fish.'
7. kumain ag matabag babae nag maliit na baboy
'The fat woman ate a little pig-
8. nagluto ag maliit na babae nag malakig manok
'The little woman cooked a big chicken.'

Problem 4: TAGALOG (Philippines)

GRAMMAR (520 Sent)
Sentence Level Construction:
Sent = +Base:tCl -Into:ICF
Clause Level Construction:
tCl = +P:tv +S:N ±O:N
Phrase Level Construction:
N = +Det:det! ±Mod:aj +H:n
N2 = +Det:det2 ±Mod:aj +H:n

LEXICON (12 morphs)

I. Noun System:
Nouns (5) babae n.l. 'woman
tao n.l. 'man
baboy n.1/2 'Pig
isda n.1/2 'fish'
manok n.1/2 'chicken'
Determiners (2) ag det.l 'the (subj)'
nag det.2 'a/an (obj)'
Adjectives (3) malakig aj. 'big'
matabag aj. 'fat'
maliit na aj. 'little'
II. Verb System:
Verbs (2) kumain tv. 'ate'
nagluto tv. 'cooked'
 114 INTRODUCTION TO TAGMEMIC ANALYSIS

SUPPLEMENTARY READINGS 4
Becker, Alton L., "Conjoining in a Tagmemic Grammar of English,"
Monograph Series on Languages and Linguistics, No. 20, 109-121.
Washington, D.C., Georgetown University Press, 1967.
Elson, Benjamin, and Pickett, Velma, An Introduction to Morphology
and Syntax, Santa Ana, Summer Institute of Linguistics, 1962.
Chap. 8, "Tagmemes and Constructions at the Phrase Level "
73-75; Chap. 16, "Survey of Phrase Types," 103-108, with head-
modifier, coordinate, appositional, relater-axis, and close-knit
verb phrase.
Fries, Charles C, The Structure of English, New York, Harcourt, Brace
and World, Inc., 1952. Chap. 10, "Structural Meanings, Modifiers,"
202-239, and Chap. 6, "Function Words," 87-109.
Gleason, Henry A., Jr., Linguistics and English Grammar, New York,
Holt, Rinehart and Winston, Inc., 1965. Chap. 8, "Structure Sig-
nals," see "Parts of Speech and Function Words," 186-194, use of
meaning.
Hockett, Charles F., A Course in Modern Linguistics, New York, Mac-
millan Company, 1958. Chap. 21, "Endocentric Constructions,"
183-190 and Chap. 22, "Exocentric Constructions," 191-198.
Huddleston, R. D., "Rank and Depth," Language, 41: 574-586 (1965).
Proposes a coordinate layer for all levels of the grammar, in a
scale-and-category grammar of five (natural) levels.
Law, Howard W., "The Use of Function-Set in English Adverbial Classi-
fication, Monograph Series on Languages and Linguistics, No. 20
93-102. Washington, D.C., Georgetown University Press, 1967.
Sorting adverbials by form, function, and position.
Matson Dan M., "Tagmemic Description of Agreement," Monograph
Series on Languages and Linguistics, 18th Annual Round Table
No. 20, 103-108. Washington, D.C., Georgetown University Press
1967. Tagmemic agreement reduced to subroutine using square
bracket notation.
Warriner, John E., English Grammar and Composition, New York,
Harcourt, Brace and World, Inc., 1951, revised 1965. Chap. 2 "The
Phrase," 35-51, including traditional definition of phrase (36)
and explanation of prepositional, verbal, and appositive phrases.
 TABLE 5: WORD LEVEL ANALYSIS
According to Its According to the According to Its Characteristics
Freedom in Use Formation Process Internal Structure of the Word Type

1. Nouns (n), with Gender, number,

noun inflections case, possessive
Inflection 2. Verbs (v), with Tense, aspect,
Outer Formation verb inflections mood, voice
3. Adjectives (aj) Agreement with
adjective inflections nouns, comparison

1. Restrictive type Nominalizer (nom)

Word Derivation form class unchanged Verbalizer (vbzr)
as a Inner Formation 2. Governing type Adjectivizer (ajzr)
Minimum Free Form form class changed Adverbializer (avzr)

1. Endocentric Roots related as

compound same as multiple heads, or
Compounds one of the roots head and modifier
Core Formation 2. Exocentric Roots unrelated, or
compound not same one subordinated
as one of the roots to the other
 WORD LEVEL

The word level of grammar is that level which is below the phrase
level and above the morpheme level. The word is composed of
morphemes and typically fills slots at the phrase level. It is a con-
struction in which the constitute is a minimally free form in the
language and whlose constituents are morphemes. Words are
composed of one or more morphemes; the morphemes are the
ultimate grammatical constituents, the minimum meaningful
forms in the language.
Word Defined; The word has been defined by Bloomfield
(1933:178) as "a minimum free form; a free form which does not
consist entirely of two or more lesser free forms." The phrase
minimum free form" is to be understood as a minimally free
form, that is, a free form which does not consist entirely of other
free forms. The word "entirely" must be emphasized. If free
forms are compounded, then something else enters into the
composition of the! compound to form the word.
In tagmemic analysis, word level construction is defined as a
construction which consists, potentially, of two or more word level
tagmemes, filled by morphemes. Word level includes the layers
of (1) inflection, (2) derivation, and (3) compounding, and parallels
the phrase level in which there are layers of relation coordina-
tion, and subordination.
Elson and Pickett distinguish between word level construc-
tions, which consists of a stem (or root) and an (inflectional) affix
(1962:76), and stem; level constructions, composed of two or more
tagmemes which form the stem and fill the nuclear slot in words
(1962:79). The disadvantage of separating a word and stem level
is that the parallelism with the phrase level is destroyed, and

117
 118 INTRODUCTION TO TAGMEMIC ANALYSIS

the stem is given preeminence among the linguistic forms of lan-

guage, as if it had the emic status of such levels as the sentence
clause, phrase, and word in a language.

Linguistic Form. Forms that carry meaning are called lin-

guistic forms. These forms may be simple, monomorphic forms,
or they may be complex, polymorphic forms. Forms may also be
classified as free or bound, and as stems or affixes.
Free forms are forms that can occur alone; bound forms are
forms which cannot occur alone. If a form is found in one occur-
rence as free, it is listed as free. A form, once free, is always
free.
Sterna are forms that carry the basic meaning of the word;
affixes add meanings to the stem. If a stem consists of a single
morpheme, it is also called a root. Stems and roots may be bound
or free, but affixes are always bound. The affixes are further
classified as prefix, infix, suffix, or suprafix, depending upon
whether they occur before, within, after, or simultaneously with
the stem.
These are the forms that enter into the composition of words
at the word level. If (A) is a free stem, (a) a bound stem, and (b) an
affix, then the basic formulas for classifying word types are as
follows (see Sapir, 1921:29):
1. Simple Word consists of a single morpheme and
is free. A
2.a. Derived Word consists of bound stem and affix. a+b
b. Derived Word consists of free stem and affix. A+b
3. Compound Word consists of two free stems. A+A
Bloomfield classified the first two types as primary words, sim-
ple or derived, and the last two types as secondary words, which
contain a free form plus some other element in their composition
(see Bloomfield, 1933:209).

Word Level Analysis. At the word level in tagmemics, as

understood here words are broken down into constituent mor-
phemes. Thus:
1. Simple words, consisting of one morpheme are not con-
structions and need not be represented at the word level.
They are simply entered into the lexicon as free morphs.
Words that show composition, and that contain more
than one morpheme, are analyzed at this level.
2. Complex words are polymorphic; they may be broken
down into stem and affix. The affixes may be inflectional
or derivational. Inflectional affixes are removed first
then derivational affixes.
 WORD LEVEL 119

3. Compound words are polymorphic and consist of more

than one stem or root. If the combining stems are free
forms, then some other device is used in the language to
combine these forms into a single word.
Exceptional cases at the word level include compound-com-
pound words, such as wind+shield+wiper, complex-compound
words, such as gentle-man-ly. These are analyzed as nested
constructions at word level. The problem of phrase derivatives,
such as old maid-ish, are best handled as loopbacks in the gram-
matical hierarchy (see p. 31). The phrase fills a word level slot,
and the derivative is an adjective.

INFLECTIONS: OUTER FORMATIONS

Word level constructions are those constructions in which the
constitute is the word; the constituents are the inflectional,
derivational, and root tagmemes that enter into word structure.
The outer layer of formation of the word is the layer of inflection.
Inflectional affixes are first stripped from the word, then the
stems of the words are analyzed into roots and derivational
affixes.
Inflection Denned
Most languages of the world are characterized by some
type of inflectional system. There is a contrast between inflected
and noninflected forms. Where this contrast exists, inflections
are recognized by the following set of distinctive features:
1. They are outer formations.
2. They are used as defining norms for major form classes.
3. They are relational in function, fitting words for use
in syntax.
Outer Formations. Inflectional affixes, according to their
form, are outer formations. The word has an onion-like structure,
with inflections as the outer layer, derivations as inner layer,
and roots at the core. When words are inflected, the inflectional
affixes are found in the outer positions, furthest from the root.
This does not mean that inflections may only occur at the outer-
most position. There may be series of inflections, but they are
outer formations with respect to derivational affixes.
1. The word may be uninflected. If so, derivational affixes
may occur in word final position. Therefore it is not true
to say that any affix which is found in final position in a
word is an inflectional affix.
 120 INTRODUCTION TO TAGMEMIC ANALYSIS

2. The word may have more than one inflection. If so, one
inflectional affix will appear in final position, and an-
other inflectional affix in nonfinal position. Therefore
it is not true to say that inflectional affixes occur only
in word final position.
3. Considering word formation as a whole the set of in-
flectional affixes tends to be outer and the set of deriva-
tional affixes tends to be inner. In one analysis of Latin
(O'Brien, 1965:29), the Latin word is characterized by
a stem, followed by three derivational affixes and three
inflectional affixes, in that order. The three inflectional
affixes are outer and the three derivational affixes are
inside the inflections.
It is generally true for English that no derivational affix ever
occurs outside of an inflectional affix. Forms like spoonfuls,
noun+derivation+inflection, are preferred over spoonsful,
noun+inflection+derivation. Forms like betterment are regular
verb+derivation formations, not adjective-(-inflection+deri-
vation.

Wide Distribution. Inflectional affixes are general through-

out a form class and therefore are both limited in number and
have a very high frequency of occurrence. Derivational affixes
tend to be more numerous, but each derivation has very limited
use in the language.
1. Inflections are general throughout a word class; they
are used as the defining norms for the major form
classes, such as noun, adjective, verb and adverb. These
word types are defined by their word paradigms. A noun
is a word which takes the inflectional affixes of a noun.
2. Inflections are limited in number. In English, the noun
has only two inflections, the verb has four, and the adjec-
tive two for a total of eight. All other affixes in the
language are classed as derivational. Even in highly
inflected languages, such as Latin or Sanskrit, the in-
flectional systems are well defined and general through-
out the word class.
3. Inflections have a high frequency of occurrence. Com-
pared with the derivational affixes which are limited to
a few members of a class, inflections generally occur
throughout the class and appear more often.
Relational Function. Inflections are relational markers,
which fit the members of the major form class for use in syntax,
rather than change the form class of words. This does not mean
that uninflected forms do not occur in syntax; it means that once
 WORD LEVEL 121

the inflection, or relational marker, is added to a form, that form

is limited by that marker to certain positions in the syntactic
construction. This relational marking takes place without
changing the form class of the word. The changes made in the
form of the word! are syntactically relevant.
English, for example, has only two inflections for the noun
stem a plural and a possessive. The addition of a possessive
suffix fits the noun for use in syntax as an adjectival; the noun
with the possessive marking, can only be used in slots normally
filled by adjectives, and never in slots normally filled by (un-
marked) nouns. Thus, John + {-Z2} 'possessive,' forms John's,
in such constructions as John's hat. Yet the form class of the
word has not changed. To say that the form John's is now an
adjective, is to imply that this form can undergo comparison in
the same way as the simplest members of the adjective class.
This form is not an adjective, because it cannot take the affixes
{-er}, 'comparative,' and {-est}, 'superlative,' proper to that
class.
The relational function of inflectional affixes may also be
shown in the case-marked forms of a language. In Latin nouns
marked as nominative, genitive, dative, accusative, and ablative
all have their own specific syntactic positions in constructions.
They remain nouns, but are marked for special use in syntax as
subjects, possessors, indirect objects, direct objects, or objects
m construction with particular prepositions.

Types of Inflectional Systems

In inflected languages, inflected forms are opposed to non-
inflected forms. Within the inflected forms nouns and their
modifiers can usually be distinguished from verbs and their modi-
fiers. Noun systems, with nouns and adjectives, are generally
characterized by gender, number, and case; verbal systems,
including verbs and auxiliaries, by tense, mood and voice. Un-
inflected forms are defined by their syntactic use.
Inflectional Systems. In inflected languages, it is useful to
establish a morphblogical sieve which will sort out the form
classes of the language according to their inflectional paradigms.
The uninflected forms are first sorted according to their syn-
tactic use, and then defined by listing. A typical sieve of this
type may be set up; for English as follows:
Inflected words, defined by inflection
1. Words not inflected for tense
but inflected for number and case = NOUNS
{&,}, plural
{%i}, possessive
 122 INTRODUCTION TO TAGMEMIC ANALYSIS

but inflected for comparison = ADJECTIVES

{-er}, comparative
{-est}, superlative
2. Words inflected for tense = VERBS
{Z3}, 3d singular present
{DJ, simple past
{D2}, past participle
{-ig}, present participle
Uninflected words, defined by syntactic use
Similar morphological sieves have been set up, for Latin, by
Hockett (1958:221) and by O'Brien (1965:41). The purpose of the
sieve is to sort the word forms of language and group them into
major form classes which are clearly defined by the inflectional
system of that particular language.
Noun systems are characterized by gender, number, and
case. English has only plural and possessive inflections, but
most Indo-European languages inflect nouns for number and
case and adjectives for gender, number and case. Other useful
categories include person, in pronouns. Verbal systems are
characterized by tense, mood, and voice. Other useful categories
are transitivity, aspect, person, and number of subject. The cate-
gories marked will vary from one language to another.

Inflectional Paradigms. The paradigm of a major form class

consists of a single stem of that class, with the inflectional
affixes which the stem may take. The paradigm serves as a defi-
nition of the form class; the forms need not take every inflec-
tional affix but to belong to the class the form must take some
affixes of the set as opposed to affixes of other paradigms.
1. Noun (n) is defined as a word that takes noun inflections.
For example, in English, nouns show the following in-
flectional contrasts:
Base Form Stem + Z Stem + Z2 Stem + Z, + Z
man men man s men s
2. Adjective (aj) is defined as a word that takes adjective in-
flections. For example, in English, adjectives show the
following inflectional contrasts:
Base Form Stem + {-er} Stem + {-est}
cold colder coldest
3. Verb (v) is defined as a word that takes verb inflections.
For example, in English, verbs (except the verb be) show
the following inflectional contrasts:
Base Form Stem+Z3 Stem+D, Stem+D2 Stem+{-ir)}
sing sings sang sung singing
 WORD LEVEL 123

Once the inflectional paradigms of a language have

been established and the major form classes of the
language defined, it is often possible to establish sub-
classes and classes of substitutes by the same paradigm.
4. Pronoun (pn), in its form, is a subclass of the noun. It
does not add affixes which are inflections but its forms
fit the noun inflectional paradigm:
man men man s men s
f me we, us my, mine our ours
you you your, yours your, yours
he him they, them his their theirs
she, her her, hers
The alternates listed in each cell of the paradigm are
positional: variants, and the forms may be considered as
allomorphs of the same morpheme. Thus, the morpheme
{I}, has two forms, /ay/, 'I,' and /miy/, 'me.' The first
occurs before verbs, the latter after verbs and as object
of a preposition.
5. Auxiliary (aux) in its form, is a subclass of the verb
which fillsipart of the verbal paradigm, but not the whole
paradigm. In English, the verb be has eight forms, most
verbs have five or four and the modals only two.
sing sings sang sung singing
be/am/are is was/were been being
can could
In some verbs, the five part paradigm only has four
parts, because the Di and D2 morphemes are identical
m farm. However, they may be recognized as different
morphemes which happen to have identical shape. Like-
wise the auxiliary must has one form whether it is
used in the past or the present tense, but it is still classed
as a modal.

Word Level Formulas: Inflections

In combining the word and stem levels into a single word
level in tagmemics, there is no basic change in writing tagmemic
formulas. The word level formulas of earlier tagmemic gram-
mars are simply the first layer of analysis; the stem level of
these tagmemic grammars constitutes the two subsequent
layers. The three layers, taken together, form a single word level.
Inflectional Constructions. An inflectional construction (for-
merly word level construction) consists potentially of two or
 124 INTRODUCTION TO TAGMEMIC ANALYSIS

more tagmemes, one of which is manifested by a stem or root

and the other by an (inflectional) affix (see Elson and Pickett,
1962:76). There are two kinds of tagmemes in this construction:
(1) nuclear tagmemes filled by stems and (2) a series of affix tag-
memes filled by inflectional affixes. Word level inflection formu-
las may be symbolized in the formula:
Woi'd = Stem + Inflections
When dealing with inflectional constructions, it is immaterial
whether the stem is simple (a root), or complex (affix derived),
or compound (formed of more than one root). At this layer of
construction, the stem is separated from all of its inflectional
affixes. If the stem is not simple, it is analyzed later as a complex
or a compound stem.

Word Level Tagmemes. The tagmemes in the first layer of

word level construction are (1) nuclear slots filled by stems and
(2) inflectional slots filled by inflections. The whole constitute is
called a word. Some typical slot-class correlations at this stage
of analysis are as follows:
Word = Stem 4- Inflections
n (noun) nnuc:ns pos:posm num:numm
v (verb) vnuc:vs t:tm radiradm
aj (adjective) #jnuc:ajs comp:{-er} supl:{-est}
av (adverb) avnuc:avs
At the word level all slots and fillers are represented by small
letters. Nuclear slots are filled by the corresponding stems for
example, nnuc: noun nuclear slot; ns, noun stem. For the inflec-
tions the slot name is the name of the inflectional function, and
the filler is called a marker. Thus t:tm is read as "tense slot filled
by a tense marker"; asp.-aspm is read as "aspect slot filled by an
aspect marker." Nuclear slots are obligatory; inflectional slots
may be obligatory or optional.

Typical Inflectional Formulas. The tagmemes which repre-

sent the inflectional layer of the word level are combined in for-
mulas which may either represent the full inflectional paradigm
in the language, or may be written, ad hoc, for particular lan-
guage problems.
1. Noun Formulas: A typical formula for nouns, in English,
would be as follows, assuming that number is a category
which must be expressed:
5.1 n = +nnuc:ns +num:numm ±pos:posm
 WORD LEVEL 125
Read: A noun consists of a noun nuclear slot filled by a
noun stem,;a number slot filled by a number marker, and
an optional^ possessive slot filled by a possessive marker.
An alternate formulation would list number as optional
(±)i>ut t^iis obscures the fact that every noun in English
expresses number, either singular or plural. Possession
on the contrary, is an optional category and constitutes
the onlyj case marking outside of the pronoun system.
2. Verb Formulas: A typical formula for verbs, in English,
would be as follows, with either tense or aspect obliga-
tory to the construction:
5.2 iv/tv = +vnue:ivs/tvs ±t:tm +asp:aspm
Redd: An intransitive or transitive verb consists of a verb
nuclear Slot,filled by an intransitive or a transitive verb
stem and either a tense slot filled by a tense marker or
an aspect slot filled by an aspect marker (but not both).
This formula requires one inflection, and allows only
one, with a verb stem. If tense occurs, it is present,
marked biy — <l> or {Z3}, or past, marked by {DJ; if aspect
occurs, the verb form is marked as a present participle
with {—i^l, or as a past participle, with {D2}.
3. Adjective j Formulas: A tyipical formula for adjectives,
in English, would be as follows, with comparison markers
optional:
5*3 aji = +ajnuc:ajs ±comp: {—er}/{-est}
Read: An adjective consists of an adjective nucleus filled
by an adjective stem, and an optional comparison slot filled
by {-er}/{-est}.
In this formulation, the adjective may show no compari-
son jtposit^ve degree) or it; may take the optional com-
parative {i-er} or superlative {-est} marker.

DERIVATIONS: INNER FORMATIONS

The top layer of word level construction deals with inflectional

affixes; once these! are stripped away, what remains is a word
stemj, Thisjftem, which takes affixes, and other uninflected words
of the language aiie then grouped together. Any stem or word
which is not simple, that is, which consists of more than one
morpheme, must be analyzed into its constituent parts. The next
laye* of cottstrueta^n deals with roots and derivational affixes.
 126 INTRODUCTION TO TAGMEMIC ANALYSIS

Derivation Defined
Derivational affixes are affixes which are not inflections. If
the inflections of language are denned as affixes which do not
change the form class of the word, but fit the word for use in syn-
tax in the sense explained above, then derivational affixes are
affixes which can change the form class and merely establish
words as members of the various form classes.

Inner Formations. According to their form derivational

affixes are inner formations. They are inner with respect to in-
flections so that if derivations and inflections both occur, deriva-
tions are inner, closer to the stem, and inflections are outer
furthest from the stem.
Derivational affixes may occur in word final position, when
the word is not inflected or the inflection has a zero manifesta-
tion. More than one derivational affix may appear in the same
word. The position classes of the derivations must then be deter-
mined. The set of derivational affixes tends to be inner, while the
set of inflectional affixes tends to be outer.

Restricted Distribution. Derivational affixes have a distribu-

tion that is very restricted; they are not general through a word
class. They tend to be statistically more numerous than inflec-
tions, but each single derivation has a very limited work load.
Derivations establish a form in a particular class; by the deri-
vational ending, one can recognize the word as belonging to a
certain class. Derivations are practically unlimited in number;
there is no theoretical limit to the number of derivations. English
has over 60 common ones. Derivations have a low functional load.
Each single derivation occurs rarely and is limited to a few set
combinations with particular stems.
The range of productivity of a derivational pattern may vary
from one or two isolated forms to a fairly general use with a
particular form class, but rarely does a derivational affix combine
with all the forms of a class.

Subordinating Function. Derivational affixes are markers

which establish a form within a particular form class. They are
of two kinds (Hockett, 1958:243):
1. Governing derivational affixes change the form class of
the word to which they are added. Thus, constitute a
verb, + {-tion} = constitution a noun.
2. Restrictive derivational affixes do not change the form
class of the word, but change the meaning of the deriva-
tive. Thus, state, a noun, + {-hood} = statehood, a noun.
 WORD LEVEL 127

Whether or not the derivational affix changes the form class of

noun, the constitute, once the derivation is added, is recognizable
as a member of'a particular class. Constitution is recognized
as a noun and statehood is recognized as a noun. Once formed
the constitute is treated as the simplest member of the same
form class, both in syntax and in the addition of inflectional
affixes. The principal differences between inflection and deriva-
tion may then be summarized as follows (see Nida, 1949:99):
Inflections Derivations
1. Tend to be outer formations, 1. Tend to be inner formations,
occur further frofn the stem occur closer to the stem than
than any derivational affix. any inflectional affix.
2. Tend to be less numerous, but 2. Tend to be more numerous but
with wide distribution. Each with limited distribution. Each
affix is used through the class. affix limited to small subclass.
3. Used to fit words for use in 3. Used to establish words in a
syntax, but never change the class, and generally change the
form class of the word. form class of the word.
4. Derived words do belong to the
4. Inflected words do not belong same distribution class as the
to the same distribution class underived members of the
as uninflected members of the class. The change affected by
same class. The inflection is derivation is morphologically
syntactically relevant. relevant.
5. Inflectional paradigms tend to 5. Derivational paradigms tend to
be well denned, homogeneous, be ill defined, heterogeneous,
and define major form classes. and only define single words.
Because of the difficulties involved in characterizing inflection
and derivation across languages, no clear-cut differences can be
established. In each language, a judgment must be made, using
the above general norms, as to what constitutes inflection; other
affixes are derivational.
Types of Derivational Systems
Derivations are affixes which are not inflections; they are
bound forms and nt^y be prefixes, suffixes, infixes, or suprafixes.
These affixes may be combined in derivational systems, based on
the following differences: (1) derivations are governing or re-
strictive; (2) derivations establish a word in a form class; and
(3) derivations differ in form, as one of the four types of affixes.
Governing Derivations. Derivations which change the form
class of the stein are' called governing derivations. They govern
or determine the fortn class of the derivative. The newly formed
derivative is marked by the formative, as a noun, verb, adjective,
or
 128 INTRODUCTION TO TAGMEMIC ANALYSIS

Noun formatives are derivational affixes that, when added

to a given stem form a noun. These derivations are called nomi-
nalizers (nom). Each language has its favorite patterns of nomi-
nalization. In English, nouns are generally formed from verbs
sometimes from adjectives.
1. Favorite Pattern: Verb + D.2-1 -» Noun
break-AGE resign-ATION decis-ION deliver-Y
arriv-AL employ-EE pay-MENT CONvert (stress)
admitt-ANCE catch-ER defen-SE bragg-ART
assist-ANT freez-ER depart-URE associate /ey -» A /
2. Secondary Pattern: Adjective + D.3-1 —* Noun
accur-ACY social-ISM good-NESS activ-ITY
free-DOM social-IST social-ITE tru-TH
Verb formatives are derivational affixes that when added
to a given stem, form a verb. These derivations are called verbal-
izers(vbzr). In English, verb formatives are comparatively rare.
Verbs are the most basic form, in the sense they are used to de-
rive other forms. The verbalizers that occur are causatives a
frequent verb formative in language.
1. Favorite Pattern: Noun + D.l-2 -> Verb
fright-EN glory-FY idol-IZE (EN-joy, BE-friend)
2. Secondary Pattern: Adjective 4- D.3-2 -» Verb
cheap-EN equal-IZE (EN-able)
Adjective formatives are derivational affixes which, when
added to a given stem, form an adjective. They are called adjec-
tivizers (ajzr). In English, adjectives are generally formed from
nouns rarely from verbs.
1. Favorite Pattern: Noun 4- D.l-3 -* Adjective
season-AL fortun-ATE cub-IC beast-LY
suburb-AN ragg-ED book-ISH fam-OUS
circul-AR wood-EN child-LESS cream-Y
vision-ARY peace-FUL life-LIKE
2. Secondary Pattern: Verb + D.2-3 -» Adjective
pay-ABLE watch-FUL creat-IVE tire-SOME
confid-ENT sens-ORY
Adverb formatives are derivational affixes which when added
to a given stem, form an adverb. They are called adverbializers
(avzr). Adverbs, in English, are generally formed from adjec-
tives. Once it is formed the adverb is a closed construction and
is not used to form words of other classes, such as nouns, verbs,
or adjectives.
1. Favorite Pattern: Adjective -I- D.3-4 -» Adverb
glad-LY (the most productive of the derivatives)
 WORD LEVEL 129

2. Secondary Pattern: Noun + D.l-4 -*• Adverb

dai-LY home-WARDS rule-WISE (A-shore)
Governing derivations always change the form class; in English
they are mainly suffixes. Prefixes are included in parentheses.
Restrictive Derivations. Derivations which do not change the
form class of the stem to which they are added, but merely
change the meaning, are called restrictive. In English, they are
mainly prefixes. Suffixes are included in parentheses.
1. Noun Patterns: Noun + D. 1-1 -* Noun
ANTI-body UN-truth (king-DOM) (lord-SHIP)
EX-wife (child-HOOD) (duch-Y)
2. Verb Patterns: Verb + D.2-2 -* Verb
AD-join DE-brief PRE-form RE-dress
OOL-locate DIS-agree PRO-rate UN-do
3. Adjective Patterns: Adjective + D.3-3 -* Adjective
ANTI-social UN-real (kind-LY) (hard-LY)
IM-possible (green-ISH)
There are no restrictive adverbializers. Adverbs do not form the
base either for other form classes, or for further formation of
adverbs.
Chart of English Derivational Affixes. The derivational af-
fixes are marked &n input-output; D.l-2, means noun becomes a
verb. The restrictive derivations are found along the diagonal.
Stem Glass Noun •Verb Adj Adv
1. Noun D.1U D.l-2 D.l-3 D.l-4
kingdom lorify lifeless rulewise
2. Verb TX241 .2-2 D.2-3 B.2-4
payment disagree ayable No case
3. Adjective 15.3-il
goodness
D.3-2 -fi.3-3 D.3-4
equalize kindly gladly

Word Level Formulas: Derivations

After words have been separated from the outer layer of in-
flection the stems of words must be analyzed at that layer of
grammar within the word level, sometimes called stem level,
which includes the pirocesses of derivation and compounding. Der-
ivation deals with the addition of bound forms which are not
inflections to roots. iThis is the second, or subordinate layer, of
the word level.
Derivational Constructions. A derivational construction
(formerly stem level construction) consists potentially of two or
 130 INTRODUCTION TO TAGMEMIC ANALYSIS

more tagmemes, one of which is manifested by a stem or root,

and the other by some derivational affix. The stem or root is said
to manifest a core tagmeme, and the derivation manifests a der-
ivational tagmeme. The formula is:
Complex Stem = Core + Derivations
The term complex stem is sometimes replaced by affix-derived
stem or simply derived stem. The first tagmeme is a core slot
filled by a root; and complex stems must be reduced to the ulti-
mate roots of the word. However, in practice, it is more often the
case that the formula must be written in terms of stems, not
roots to show the productivity of the formula. Hence the func-
tional term "core" is used in the formula.

Derivational Layer Tagmemes. The tagmemes in the second

layer of word formation are: (1) core slots filled by stems, includ-
ing roots, and (2) derivational slots filled by derivational affixes.
With governing derivations, both slots are obligatory, and the
derivations are called nominalizers (nom), verbalizers (vbzr),
adjectivizers (ajzr), and adverbializers (avzr). Typical slot-class
correlations at this layer of analysis are as follows:
Complex Stem = Core + Derivations
ns (noun stem) core:vs/ajs nom:D.2-l/D.3-l
vs (verb stem) core:ns/ajs vbzr:D.l-2/D.3-2
ajs (adjective stem) core:ns/vs ajzr:D.l-3/D.2-3
avs (adverb stem) core:ajs/ns avzr:D.3-4/D.l-4
For restrictive derivations where the form class is not changed,
the slot with the affix is optional. This slot may be labeled accord-
ing to function, such as nom: m formula 5.7. All formulas apply
not only to the stems of inflected words, but apply to uninflected
words as well. The process of derivation is independent of the
process of inflection.

Typical Derivational Formulas. Although the formulas

given for inflections are best analyzed as strings of inflections
with a single stem, the formulas for derivations, for full produc-
tivity, are best represented in terms of two and only two con-
stituents, especially in English. The general order of word deri-
vation in English may be represented as:
Verb —* Noun —* Adjective —» Adverb
constitute + -tion + -al + -ly
This one formula represents the three favorite patterns, in
English, for deriving nouns, adjectives, and adverbs. Verbs are
 WORD LEVEL 131

generally basic forms. These favorite patterns may be repre-

sented in tagmemics, as follows:
5.4 hs = +core:tvs +nom:{-tipn} constitute + -tion
Read: One type of noun stem consists of a core slot filled by
a transitive verb stem and a nominalizer slot filled by
{-tion},
5.5 ajs = +core:ns -f ajzr:{-al} constitution + -al
Bead:'One type of adjective stem consists of a core slot filled
by a noun stem and an adjectivizer slot filled by {-al}.
5.6 avs = +core:ajs +avzr:{-ly} constitutional + -ly
Read: One type of adverb stem consists of a core slot filled
by an adjective stem, and an adverbializer slot filled by
{-iy}.
In the readouts of these formulas, we specify one type of noun
stem not the whole class of noun stems, as composed of verb
stem and derivation.

Restrictive Derivations. For those derivations which do not

change the form cliass, the derivational slot is marked as optional
andi may be labeled according to function or meaning.
5.7 ns = +core:ns ±nom:{-hood} child + hood
Read: One type of noun stem consists of a core slot filled by
a noun stem, and an optional nominalizer filled by {-hood}.
5.8 ajs = ±neg:{*m-} +core:ajs lm- + possible
Read: One type of adjective stem consists of an optional
negative slot filled by {tm-}, and a core slot filled by an ad-
jective stem.
Where the derived noun forms a separate subclass in the lan-
guage which is syntactically relevant, the derived stem may be
marked as stem class 2 and the core stem as stem class 1, with
the derivational affix as obligatory.

COMPOUNDS: CORE FORMATIONS

Inflectional constructions consist of stem and affix. The stem

bound or free, carries the basic meaning of the word. Affixes are
bound forms, which add to the meaning of the stem, and are
either inflectional or derivational. The stems which remain after
the inflections are stripped away are: (1) simple, of one root;
 132 INTRODUCTION TO TAGMEMIC ANALYSIS

(2) complex, with root and derivation; or (3) compound, consisting

of root plus root. Roots form the innermost core of the word and
constitute the final layer of construction.

Compounding Denned
Compound stems are stems consisting of more than one root.
Roots are monomorphic forms, which carry the basic meaning of
words. Monomorphic stems are roots; polymorphic stems may
consist of more than one root. Stems of more than one root are
reduced to single roots at the compounding or coordinate layer
of word formation.
Compound Recognition. When compounds are discovered
composed of two roots, it is not difficult to analyze them as having
two parts, or two roots. What is difficult is to recognize the dis-
tinctive features of compounds so that compounds are distin-
guished from phrases. Some features are:
1. Phonological Features. This includes such features as the
patterns for consonants and vowels in the phonology of
the language, and suprasegmental features like stress.
In English, words are characterized by a single primary
stress, so that compounds are often recognized by stress
pattern, and lack of juncture. For example: black bird
has primary stress on each word, and a juncture. Black-
bird, the compound, has one primary stress and no
juncture.
2. Syntactic Features. Compounds are distinguished from
phrases, in that they have asyntactic features, which
are contrary to phrase patterns:
a. Word Order. In compounds, unusual orders may be
found and usual orders are not regular even for a
particular compounded pattern. For example: sea-
sick noun followed by adjective, is not a usual
phrase pattern. Verb + particle, as in splash-down, is
regular for verbs, not for nouns.
b. Interruptibility. The parts of a compound are not
interrupted, but form a rigid noninterruptible pat-
tern; the form is inseparable. For example: dare-
devil cannot be used as dare-the-devil which is a
phrase.
c. Modification. Elements of the compound cannot be
separately expanded with modifiers, although the
whole compound may be modified. For example:
sea-sick may not occur as deep seasick, with deep
modifying sea.
 WORD LEVEL 133

d. Infitctibility. Elements of the compound may not be

separately inflected, although the whole compound
may; be inflected in its class. For example: ash-tray,
may not occur as ashes-tray, with ask inflected for
plural.
3. Semantic Features. Compounds tend to take on special-
ized meanings, and so attain idiomatic status. The
meaning of the compound may be quite specialized, so
that the meaning of its members is obscured. For ex-
ample: a blackboard may be green, and may be made of
slate, not wood. The features that set compounds apart
may differ from language to language, but some initial
investigation for each should be made on the basis of
phonological, syntactic, and semantic features. The com-
pound needs some binding force, so that it is not com-
posed entirely of free forms, and can be differentiated
from phrase structures.
Parts of a Compound. The simplest level of compounding is
the joining of tw0 simple roots; and this type of compound pre-
dominates in many languages. In English, the range of root-root
compounds may be illustrated by listing some of the initial roots
end verifying, from any dictionary, the vast number of com-
pounds formed.
1. Nouns as (Initial Element
arm, eye, ear, foot, hair, hand, heart, lip, mouth
wrist,! bird, bull, cat, cow, duck, ash, rail, rose, suit,
steam, air, beach, rain, snow, time, light, moon, sun,
star, ivind, book, car, door, farm, house, lamp, shoe,
tea, Hble, wall.
2. Verbs as Initial Element
blast, jbreak, cast, crack, drive, drop, count, carry,
feed, line, lean, loop, kick, pull, push, pick, play,
print, stick, shoot, splash, show, throw, take, turn
swim, wash, write, read, set, come, fall, go, run, sit,
stand.
3. Adjeetivesas Initial Element
black, I blue, gold, green, grey, red, white, yellow,
north, east, south, west, right, left, straight, big,
small, hot, cold, high, low, long, short, far, near, brief,
fair tcip, sweet, tight, quick, slow.
4. Adverbs ad Initial Element
about, iafter, back, by, cross, down, fore, front, hind
in, off, ton, out, over, in, under, up.
The second element of the compound is occasionally complex, and
carelmust be taken to make the proper immediate constituent
 134 INTRODUCTION TO TAGMEMIC ANALYSIS

cuts, in order to determine the pattern for the compound and its
underlying forms.
1. Forma in -ER, used as second root, fill the same function
in the compound as simple forms. The construction is
Noun + Noun (-ER). For example: house-keeper and
line-backer are simple noun + noun compounds. The
second noun is a derived form with verb root + deriva-
tion.
2. Forms in -ED used as second root, are generally phrase
derivatives with the -ED governing the whole phrase.
(Noun + Noun) + -ED. Thus, for example, red-haired in
which red modifies hair, and the whole phrase fills
the core slot in a derivational formula, governed by -ED.

Types of Compounding Systems

Compounds are stems or uninflected words consisting of
more than one root. They may be bound or free, and the elements
entering into compounds may be other stems. Compounds may
be classified according to form class, according to the syntactic
relationship between the two roots, or according to both.
Form Class of Compounds. Compound forms belong to some
form class and function syntactically like simple or derived
members of the same form class. Compounds may be classified
by specifying (1) the form class of the constitute and (2) the form
class and order of occurrence of the two roots entering into the
compound. Classification by form is independent of syntactic
relations between the two roots and resembles an algebraic
equation.
Noun Compounds. Any major form + noun = noun compound.
The second root is a noun and the first root may be a noun, a verb,
an adjective, or an adverb. Examples of noun compounds are as
follows:
Noun -t- Noun (modifier-head) ash-tray, arm-chair
Verb + Noun (verb-object) dare-devil, pick-pocket
Adjective + Noun (modifier-head) black-bird red-coat
Adverb 4-Noun (a-syntactic) after-thought back-talk
Verb Compounds. Any major form + verb = verb compound.
The second root is a verb and the first root may be a noun, a verb
an adjective, or an adverb.
Noun + Verb (object-verb) house-keep, baby-sit
Verb + Verb (coordinate) dive-bomb, drop-kick
Adjective + Verb (a-syntactic) white-wash sweet-talk
Adverb + Verb (modifier-head) down-grade, over-do
 WORD LEVEL 135

Adjective Compounds. Any major form (except verb) + adjec-

tive = adjective compound. The second root is an adjective and
the first root may be a noun, an adjective, or an adverb. Verbs do
not combine with adjectives.
Noun + Adjective (a-syntactic) sea-sick, ox-eyed
Adjective 4- Adj> (coordinate) blue-green, south-west
Adverb + Adj. (modifier-head) off-white, near-sighted
Adverb Compound. Adverb + adverb = adverb compound.
Adverb + Adverb (coordinate) through-out in-to
Special Noun Compound. Verb + adverb = noun compound.
The first root is a verb and the second root is an adverb. The
compound differs from the verb + particle = verb construction
by a change of stress.
Verb + Adverb (derived) drive-in blast-off
This last type of compound, in English, is the only compound com-
posed of two roots in which the compound is not the same class
as the second root.
Syntactic Relationships. The relationship between the two
roots of a compound is generally the same as between those
form classes elsewhere in the grammar, but is in highly con-
densed form. Where the grammatical relations are obscure, the
compound is called asyntactic; the syntactic relations can be
further described I as endocentric and exocentric.
1. Aayntactic Compounds
Noun 4- Adjective, sea-sick, = sick because of the sea.
Adverb 4- Noun back-talk, = derived from talk back
2. Syntactic Compounds
a. Endocentric: at least one head root
(1) Coordinate: two head roots
girl-friend = girl who is a friend
(2) Subordinate: only one head root
black-bird = a modification structure
b. Exocentric: neither root is the head
pick-pocket = one who picks pockets
splash-down = used as a noun or verb
The functional meaning of compounds may be further clarified
by tracing them to their underlying sentence patterns (Lees,
196d:48). In noun compounds the type of phrase indicated by
noun + noun, in terms of relaters, for example: ash-tray- tray
for ashes; arm-chair — chair with arms; sea-shore = shore of the
sea; sun-light = light from the sun.
Chart of English Compounds. The following chart is based
on the form class of the roots and derived stem. The derived stem
 136 INTRODUCTION TO TAGMEMIC ANALYSIS

is always the same as the second root, except for the verb + adverb
construction, which is listed here as a compound noun with
single primary stress. The chart is based on form alone in a
manner analogous to the charting of consonant clusters in pho-
nology. Ones form classes have been sorted, they may be further
differentiated according to functional norms.
Compound + Noun -t-Verb + Adj. + Adv.
1. Noun N+N N+V N+Aj N + Av
text-book baby-sit home-sick No case
2. Verb •V+ N v +v V + Aj V + Av
turn-key No case = Noun
3. Adjective Aj+N dive-bomb
Aj+V Aj+Aj Aj +Av
blue-bird white-wash blue-green No case
4. Adverb Av + N Av + V Av + Aj Av +Av
back-talk out-shine off-white through-out

Word Level Formulas: Compounds

After inflectional and derivational affixes have been removed
from words and stems, what remains are either single roots or
compounds of more than one root. The analysis of compound
stems into component roots is the third, or coordinate, layer of
the word level.
Compound Stem Constructions. A compound construction
(formerly part of stem level) consists potentially of two or more
core tagmemes, manifested by roots. Both core slots may be
obligatory, or one of the core slots may be optional. The general
formula for compounds is:
Compound Stem = Corei + Core2
The functional term core is used in the general formula be-
cause the full productivity of the formula often requires that the
core slots be filled by stems. However, any analysis proceeds to
the ultimate roots of the words. In English, the most common
compounds are root + root.
Compound Layer Tagmemes. The tagmemes in the third
layer of word formation are core slots filled by roots (or stems).
In asyntactic constructions, exocentric and coordinate construc-
tions both core slots are obligatory. In endocentric subordinate
construction, of the head-modifier type, the head core slot is
obligatory, and the modifier optional.
Compound Stem = Coret + Core2
ns (noun stem) -t-core^tvr/avr/nr +core2:nr
vs (verb stem) +core,:nr/vr +core2:vr
 WORD LEVEL 137

ajs (adjective stem) +corel:nr/ajr +core2:ajr

avs (adverb stem) +corej:avr +core2:avr
ns (noun stem) -f-coref.vr +core2:avr
Compound Stem = Core, (mod) + Core2 (head)
ns (noun stem) ±corei:nr/ajr +core2:nr
vs (verb stem) ±corei:avr +core2:vr
ajs (adjective stem) ±core!:avr +core2:ajr
In the above formulation, all of the core slots are unqualified as
to their specific function. In more detailed analyses, the core
slots are labeled according to function, obc: (object core) for
the verb .+• object compound, and qc: (qualifying core) for modi-
fier-head compounds. In an analysis that concentrates primarily
on form what is significant is the type of roots that enter the
combination, not the specific function.

Typical Compound Formulas. The formulas for derivation

and for compounding are best expressed in terms of two constitu-
ents. Where the stress pattern or other suprasegmentals are
essential to the compound, these may be expressed using the
(-) concatenation symbol, as was suggested for the intonation slot
at the sentence level.
1. In asyntactic or exocentric structures, both roots are
obligatory, and the construction is recognized by the
form class of the roots.
5.9 ns •= +coret:tvr +core2:nr pick + pocket
Read: One type of noun stem consists of a core slot filled by
a transitive verb root, and a second core slot filled by a
noun root.
2. In endocentric coordinate structures, both roots are ob-
ligatory, and the two roots are in a coordinate relation-
ship with each other.
5.10 ajs = -Hjpreirajr +core2:ajr blue + green
Read: One type of adjective stem consists of a core slot
filled by an adjective root, and a second core slot filled by
an adjective root.
3. In endocentric subordinate structures the first root, in
English, is a modifier, and the second root is the head of
the construction.
5i11 ns = ±corei:nr +core2:nr gate + house
Read: One type of noun stem consists of an optional (quali-
fying) cere slot filled by a noun root, and a second core slot
filled by a noun root.
In any of the formulas where the stress pattern is significant,
 138 INTRODUCTION TO TAGMEMIC ANALYSIS

the stress pattern may be included in a slot, marked with a (-)

concatenation symbol.
5.12 ns = ±corei:ajr -fcorejinr -stress: / _ \
black + bird
Read: One type of noun stem consists of an optional core
slot filled by an adjective root, a second core slot filled by a
noun root and a stress pattern consisting of primary-
secondary stresses, with no intervening (+) juncture.

PRACTICE 5: WORD LEVEL COMPOSITION

Write the sentence, clause, and phrase level formulas for the
following sentences. For every word which shows composi-
tion, write a word level formula. List all the morphs in a lexicon
and calculate the maximum generation potential of the given
solution. The material was supplied by Cesar A. Hidalgo.
IVATAN (Philippines)
1. maymuhasa u mahahakay
'The men are planting.'
2. manivakami su paray
'We are harvesting the rice.
3. naymuhakami su uvi kakuyab
'We planted yams yesterday.
4. nanivasa u mavavakis sidaraw
'The women harvested today.'
5. naymuhasa
'They planted.'
6. manivasia sicaraw
'He is harvesting today.'
7. naniva u mahakay su uvi kakuyab
'The man harvested the yams yesterday.'
8. maymuha u mavakis su paray sicaraw
'The woman is planting rice today.
IVATAN (Philippines)
GRAMMAR (540 Sent)
Sentence Level Construction:
Sent = +Base:tCl -Into:ICF
Clause Level Construction:
tCl = +P:tv ±S:N, ±O:N2 ±T:tem
 WORD LEVEL 139

Phrase Level Construction:

Ni =+Det«iet.l. +H:n,/ji = +Det:det.2. +H:n
Word Level Construction:
tv = -H:tm H-vnuc:tvs +subj:pn
J ±num:{-R-}
:— +nnucj:ns:
LEXICON (16 morphs)
mahakay ns. 'man' may- ~ ma- tm. 'present'
mavakis ns. 'woman' nay- ~ na- tm. 'past'
{-R-} numim. 'plural' -muha- tvs. 'plant'
paray n.2. 'rice' -niva- tvs. 'harvest'
UVl n.2. 'yams' -kami pn. 'we
u det.1; 'the(subj)' -sa pn. 'they'
su det.2. 'the(obj)' -sia ~ -0 pn. 'he'
sidaraw tern. 'today' kakuyab tern. 'yesterday'
Morphophonemics:
1. {may}, 'present' = /ma-/before/n/., and /may-/elsewhere
2. {nay}, *p>st' = /na-/before/n/, and /nay-/elsewhere
3. {-sia}, 'hi,' == /-sia/ if no other subject, and 1-0/
elsewhere.
4. {-R-}, 'plural' = a reduplication of the second -CV-
syllable.

SUPPLEMENTARY READINGS S
Bloomfield, Leonard, Language, New York, Holt, Rinehart and Winston,
Inc., 193S. Definition of lexicon, 162, and word, 178. Chap. 13, "Mor-
phology," 207-226; Chap. 14, "Morphologic Types," 227-246.
Elson, Benjamin, and Pickett, Velma, An Introduction to Morphology
and Syntax, Santa Ana, Summer Institute of Linguistics, 1962.
Chaps. 9 and 1Q, "Tagmemes and Constructions at Word (and Stem)
Level," 75-81; Chaps. 14 and 15, "Survey of Stem Formation, and
Word Types," 05-102. Word level includes inflection, stem level,
derivation, and so on.
Fries, Charles C, The Structure of English, New York, Harcourt, Brace
and World, Inci, 1952. Chap. 7, "Parts of Speech, Formal Charac-
teristics,'' 110-141, including lists of derivational affixes.
Hockett, Charles F., A Course in Modern Linguistics, New York, Mac-
millari Company, 1958. Chap. 24, "Inflection," 209-213, and Chap.
28, "Derivations," 240-245, restrictive and governing types.
Lees, Robert B., The Gramrnar of English Nominalizations, The Hague,
Mouton & Co., reprinted 1960. For the deep structure under-
lying nominal constructions in English.
 140 INTRODUCTION TO TAGMEMIC ANALYSIS

Marchand, Hans, The Categories and Types of Present Day English

Word Formation, Wiesbaden, Otto Harrasowitz, 1960. The most
extensive work on English word formation to date.
Nida, Eugene A., Morphology: A Descriptive Analysis of Words, Ann
Arbor, Mich., University of Michigan Press, 1949. Chap. 1, "Prin-
ciples of Morphemic Analysis." In Chap. 4, the processes of inflec-
tion and derivation are contrasted, 99.
Sapir, Edward, Language, New York, Harcourt, Brace and World, Inc.,
1921. Chap. 2, "The Elements of Speech," 24-41, word denned, 34;
Chap. 4, "Form in Language, Grammatical Processes," 57-81; and
Chap. 5, "Form in Language, Grammatical Concepts," 82-119.
Thorndike* Edward L., and Lorge, Irving, The Teacher's Word Book of
30,000 Words, New York, Columbia University Press, 1944; fifth
printing, 1968. Frequency lists of English word forms; for use in
establishing a basic vocabulary list on the basis of actual occur-
rence.
 TABLE 6: STRUCTURAL DESCRIPTIONS

1. Sentence Level

2. Clause Level

3. Phrase Level

4. Word Level

5. Morpheme Level

Preterminal String (np + + np) + (tvs + tm) + (det + aj + + (ajs + avzr))

Terminal String Tom Gerry learn-ed the new'lesson very 'quick

6 MORPHEME LEVEL

The morpheme level is the ultimate level of the grammar, but it is

not a level of analysis. Beginning with the symbol for sentence
formulas are established at the levels of sentence, clause, phrase
and word. The morphemes are entered in the lexicon and pro-
grammed into the given formulas at the four levels of the gram-
mar. The resulting i morpheme sequences, with their underlying
structural relationships, constitute what we call the morpheme
level of the grammar.
Grammar has been defined by Koutsoudas as "a finite set of
rules, which generates an infinite number of grammatical sen-
tences and no ungrammatical ones, and assigns to each sentence
generated its proper structural description" (1966:4). At the mor-
pheme level, it is useful to understand the relationships gen-
erated between the morphemes in the strings that are generated
by the grammar.
1. Grammatical patterns are generated, and the morphemes
are arranged in strings according to these grammatical
patterns. These patterns may be expressed as genera-
tive rewrite rules. In tagmemic type grammars, these
relationships are expressed in terms of the following:
a. The functional slots of the tagmemes in construction
b. The filler classes that are found in these functional
slots
c. The occurrence of morphemes as members of these
filler classes
2. Structural descriptions are generated and each mor-
pheme sequence generated has a well-defined underly-

143
144 INTRODUCTION TO TAGMEMIC ANALYSIS

ing structure applied to the string. Structures implicit

in tagmemic formulas can be made explicit:
a. Underlying phrase markers can be constructed from
the formulas.
b. These phrase markers can be reduced to preterminal
strings.
c. Morphemes are programmed into strings to produce
sentences.
In this final step, concrete morphemes can be pro-
grammed into arrays to produce a generative machine
with an explicit generative capacity.
3. Generative possibilities of the model. Not only does the
grammar generate patterns, complete with underlying
phrase markers and preterminal strings, but the genera-
tive potential of the grammar can be calculated exactly
for every finite language problem. The generative power
of the model is demonstrated by calculating the number
of:
a. Combinations: the number of elements in fixed
order. The law of combinations enables us to reduce
every conflated formula to an exact number of ele-
ments in a fixed formula. By an extension of this
law not only the number of formulas but the num-
ber of elements in each of the formulas generated
can be calculated.
b. Permutations: the number of elements in freely
movable order. After the number of formulas and
the number of elements in each formula are known
the number of permutations—that is, the number
of formulas that can be formed if the elements are
freely movable—can be calculated. This number will
be a maximum and represent the greatest number
of formulas possible with a given number of ele-
ments.
c. Manifestations: after the number of formulas and
the number of morpheme class symbols in the for-
mulas are known concrete morphemes can be intro-
duced from the lexicon, to produce sentences in the
language. Manifestation is, in the first instance,
the number of different filler classes in a given slot
but ultimately is the number of morphemes that
can occur within a given slot. The generative power
of the resulting formulas is calculated as a simple
product of the numbers involved.
The generative possibilities of the tagmemic model were
 MORPHEME LEVEL 148

firsit suggested by Longacre, in Grammar Discovery Procedures

(1964:24), and were called: (R) readings; (P) permutations; and
(E) exponents. We have chosen the terms combination and
permutation, as these are the formulas used in the process taken
from any college algebra--text. Of the two terms manifestations
and exponents, bojth suggested by Longacre, we preferred the
tagmemic term rfyanifestatian rather than the mathematical
term exponent, as the process does not deal with exponents in
the sense in whteli this term is used in mathematics. The mani-
festation process <:an be further subdivided into manifestation
of classes^-which generates preterminal abstract strings of mor-
pheme classes, and substitution, whereby concrete morphemes
are introduced in place of class symbols to form the terminal
strings wMchlare the actual sentences of the language. Further
references to the generative power of the model may be found
in Longacre "Some Fundamental Insights of Tagmemics,
Language, 41:65-70, 1965; in On Tagmemes and Transforms,
Georgetown University Press, 1964; "The Generative Power of a
Tagmemic Grammar," Georgetown University Press, 1967 18th
Annual Round T?at)le, No. 20, 27-41; and Longacre, "Reply to
Postal's Review," NAL, 3Sr323-328 1967.

GRAMMATICAL PATTERNS
Tagmemic analysis) begins with the sentence and analyzes all of
the utterances of the data as far as the ultimate morpheme
level, The formal statement of the resultant grammar estab-
lishes a set of well-defined relationships between grammatical
functions, between filler classes, and between the morphemes
themselves.
Functional Slots Related
g grammars consist of a set of formulas at the five
levels of the grammar: sentence, clause, phrase, word, and mor-
pheme. The units of (this system are tagmeme units, correlations
of functional slot and filler class. These units are strung into
construction patterns, and the patterns occur at definite levels
in a grammatical system.
Functional Slot. Traditional grammar has always dealt with
grammatical functions. From the earliest grammars, we find
terms I such &a subject, predicate, modifier, and object. Chomsky
(Aspects, 1965:60) recognizes these functional terms and calls
them grammatical relationships. In Chomsky's view, these
"grammatical relationships" need not be explicitly stated in the
 146 INTRODUCTION TO TAGMEMIC ANALYSIS

formulas of the grammar, as long as they are well defined in

the context of the theory. For example, a formula such as
S -* NP -ft VP, purely in terms of form, implies that the NP is
subject of the VP and likewise implies that the VP is predicate
of the NP. Therefore a notation which specifies S:NP and P:VP
is redundant.
Chomsky's argument is probably valid for immediate constit-
uent type phrase structure components. If there are two and
only two elements in the phrase, then there exists a binary rela-
tionship that can be predefined in the theory. Note, however,
that these relationships are important to grammar, and prior
to 1965 were not explicitly part of the theory. Chomsky's further
observation (1965:69) that only "higher-level" relationships re-
ceived explicit names is also a valid observation.
In a tagmemic grammar, however, the analysis is fundamen-
tally different. The branching is not binary only, but follows a
multiple type branching, string analysis technique. In this
string, there may be many noun phrases, which have to be dis-
tinguished according to function. The labeling of function in
such a grammar is no longer redundant; it is necessary. Such a
labeling does not lead to a confusion of function and form. Noth-
ing could be less confusing than explicit function-form symbols.
Functions in Construction. The functional slots of a tag-
memic grammar are related to each other in the construction
string. These relationships are indicated by:
1. Position in the String. Tagmemes mark definite posi-
tions within the construction frame —points in the pat-
tern that can be recognized. For example, the subject in
English normally precedes the predicate.
2. Concatenation Symbols. Tagmemes, even the first in the
string, are marked with concatenation symbols, which
indicate whether they are (+) obligatory or (±) optional
to the construction in which they occur.
3. Function Label. The tagmeme is written in the form
slot: filler. The symbol to the left of the ratio sign (:)
symbolizes the function. This function is opposed to all
other functional labels within the same string.
Consequently, any particular grammatical unit can be de-
scribed, from the formal statement, in terms of position (where it
occurs), relative importance (obligatory or optional, peripheral or
nuclear), and also in terms of its grammatical meaning (subject
as opposed to predicate). For example, the subject in English
may be described as that unit which, in a simple complete state-
ment occurs before the predicate, is obligatory and nuclear to
 MORPHEME LEVEL 147

the construction, is opposed to the other noun phrases of the con-

struction, and carries the meaning "topic-of-sentence."
Functions in Grammar. The functional slots, related in con-
structions and specified at the various levels of the grammar, set
up a grid of relationships into which the morphemes of language
and sequences of morphemes are programmed. At the lowest
levels, morphemes enter into word constructions, then words into
phrases, phrases into clauses, and clauses into sentences. At
each stage of the process, every form class entering into con-
struction is named, and every grammatical function which unites
these forms into constructions is also explicitly named. These
names, or labels, are partially redundant at the lower levels of
the grammar in such notations as tense: tense marker (t:tm),
but are necessary at the phrase, clause and sentence levels to
adequately distinguish forms which enter into different kinds
of constructions in a multiple branching system.
In summary, grammatical patterns are functional patterns;
they are the blueprints of language. Into these patterns, the
morphemes, or meaningful units of language, are programmed.
As they enter construction, these forms take on the added gram-
matical meanings attached to the functional slots so that the
resulting meaningful utterance can be understood only if the
structural meanings of the patterns are understood.
Filler Classes Related
Tagmemic grammars are functional grammars, with the
underlying pattern manifest in the relationships established
between the functional slots. These slots are filled by concrete
forms. Because of the slot these forms, or filler classes are
related to each other throughout the construction patterns in
which they occur.
Filler Classes. Traditional grammars dealt not only with
functional names, such as subject, predicate, and so on. They
also named the various words and word groups, with such terms
as noun verb, noun phrase, and verb phrase. These form sym-
bols, Chomsky identifies as "grammatical categories" (1965:64);
and the transformational grammar he postulates is formulated
in terms-bf form Claris symbols only, with function implicit in the
theory. Rather thanispecify S:NP and P:VP, Chomsky leaves the
functional definitions to the theory and states his generative
grammar in terms df form class only, in such formulas as S -*•
NP+VP.
However, as a generative grammar develops, more and more
 148 INTRODUCTION TO TAGMEMIC ANALYSIS

functional terms are introduced into the form class symboliza-

tion, so that the distinction between function and form be-
comes obscure. Thus in giving an illustration of a fragment of
a base component, Chomsky (1965:106), uses formulas no longer
binary and no longer formal. For example:
1. Sentence -» NP + Predicate Phrase
2. Predicate Phrase -* Aux + VP (Place) (Time)
3. VP -» V (NP) (Prep-Phrase) (Manner)
4. Prep Phrase -* Direction, Place, and so on
In a tagmemic grammar, terms such as predicate(phrase), place,
time manner, and direction are considered functional; terms
such as NP, VP, Aux, Prep-Phrase (relater-axis-phrase) are con-
sidered formal. These are carefully distinguished in a notation
that specifies both form and function for every unit in the con-
struction. The functional slots form a grid of grammatical rela-
tionships, and this grid is filled by words or word-groups from the
various formal grammatical categories. Thus:
Function: tCl = +S +P ± 0 ±L ±T ±M
The transitive clause type of construction has a subject and
predicate, with optional object, place, time and manner slots in
the string.
Form: tCl=+NP +VP ±NP ±Av/RA-phrase
This same clause is manifested in a set of forms consisting of a
noun phrase, followed by a verb phrase, followed by an optional
noun phrase, and a series of optional adverbial or prepositional
phrases.
Filler Classes in Constructions. Filler classes are not tag-
memes but they manifest the tagmemes which are specified as
to both function and form. The functions set up the grammatical
relationships, which are then realized, in each single construc-
tion, by the occurrence of form-filling-the-function.
1. Case Marking. Single fillers are often marked for case
a case mark which coincides with the functional slot.
Subjects are generally nominative; objects are either
accusative, if there is such a case, or in a form of the
direct case, as opposed to oblique genitive and adverbial
cases.
2. Relater Words. Multiple, or word-group, fillers are often
marked for the constructional slot in which they occur.
Thus prepositions determine and mark the place, time
and manner slots in many different languages.
3. Subcategorization Rules. The fillers of slots may often be
 MORPHEME LEVEL 149

categorized according to the slots in which they occur.

Thus one set of pronouns (pnj may occur as subjects,
while another set (pn2) occurs as objects. The functional
slot, and the grammatical relationships it implies, help
us to subdivide classes into grammatically significant
groups.
Filler: Classes and Agreement. Agreement, or concord, is be-
tween forms in different slots. Using concord as a general term,
the following types of concord may occur:
1. Agreement occurs in endocentric phrases or between
the subject and predicate adjective in an equational
clause. Concord is marked by a tie bar between the slots
concerned and must be specified in the grammar, as
gender-number-case concord or any combination of
these.
2. Cross-reference occurs between subjects and predicates
in topic-comment constructions (Hockett, 1958:217).
This is marked by a tie bar between the subject and
predicate slots and must be specified. English has this
concord only in the third person singular present.
3. Governmeiit occurs between verb and object, or prepo-
sition and: object, and need not be marked. This type
of syntactic linkage is already adequately marked by
the case markings (for example: accusative) of the filler,
in accordance with the paragraph on case marking
above. For example, in Latin, one set of prepositions
takes the ablative, one the accusative.
In summary, the filler classes which manifest functional
slots are tied together by the functional grid. In addition, they
often reflect these functional ties in a formal signaling device
consisting of ease markings, relater words, and subclasses within
a class; or they are tied together by various types of concord or
syntactic linkage, such as agreement, cross-reference, or govern-
ment with appropriate tie-markings.

Morphemes Related in the String

The tagmemic grammar is stated in terms of function and
form; each tagmeme is a functional slot, a point in the grammati-
cal pattern filled by a form class symbol. The correlation of func-
tion and form class constitutes the tagmeme, which enters into
constructions «t all levels of the grammar.
But the actual occurrences of tagmemes are the occur-
rences of fdrms-fllliflg-a-function. In speech, it is the morphemes
 160 INTRODUCTION TO TAGMEMIC ANALYSIS

of the language which occur in morpheme sequences and mani-

fest tagmemic function. These individual morphemes are pro-
grammed into the grammar from the lexicon, into those slots
which admit morphemes of a given class.
Morpheme Fillers. The complete tagmemic solution includes
grammar and lexicon. The lexicon lists all the morphs of the lan-
guage or sublanguage, with their classification and gloss. The
grammar consists of tagmemic formulas in terms of function
and form class. The morphemes can then be programmed into
the grammar in that position in which the form class is specified,
and in no other position. The form classes of grammar and lexicon
must be accurately specified, so that automatic programming of
morphemes into constructions is possible.
A generative grammar is a grammar that generates all the
grammatical sentences of a language and none of the ungram-
matical ones. A tagmemic grammar is generative in this sense.
By programming concrete morphemes into the formulas of the
grammar, the grammar will generate all and only the grammati-
cal sequences of the language. In a limited corpus, all of the
morphemes of this sublanguage can be programmed into all the
appropriate slots, with the result that the maximum generation
potential (MGP) of the grammar is realized. The MGP is the max-
imum number of utterances which can be generated by a finite
grammar with a finite morpheme inventory. It includes all com-
binations manifestations, and substitutions possible, but ex-
cludes permutation possibilities. It is calculated by program-
ming the number of morphemes into the appropriate slots and
adding +1 for each optional slot to account for zero occurrence.
The MGP of each construction is carried to the next higher level
when the construction sequence occurs as a filler. The resulting
total, under sentence, is the MGP of the grammar. The MGP an-
swers the question: How many sentences can be generated by
the grammar? It defines "all" numerically. Each sentence gen-
erated is grammatical; it has the right parts in the right order.

Morpheme Fillers: Restricted Generation Potential. The

maximum generation potential of the grammar may generate
nonsensible sentences as well as sensible ones. To restrict this
output and to come closer to the "all and only" requirement,
whatever formal restrictions necessary are imposed upon the
grammar. Restrictions are of the form:
1. Morph of class A excludes morph of class B or
2. Morph of class A requires morph of class B.
With these restrictions stated and built into the grammar, the
MGP of the grammar is reduced. Certain sentences are not al-
 MORPHEME LEVEL 151

lowed, and the RGP (restricted generation potential) = MGP —

restricted sentences. Restrictions are built in by concord ties
agreement markers, either/or notation, or by cover statements.
The grammar is now a generative machine with a reduced poten-
tial and represents what the analyst considers to be the real
grammatical output of the grammar. As long as the restrictions
are .formal* the sentences are grammatical, but still are not nec-
essarily sensible sentences in the language (see also p. 83).
Morphemes Lexically Restricted. The tagmemic grammar as
a generative machine is a formalized model which produces
grammatical sequences automatically. Each of the sentences of
the language now becomes grammatical because it is a product
of the grammar. JNonsense sentences may occur; the only way to
obviate this difficulty is to use a highly sophisticated lexicon or
semantic component, which will establish co-occurrence re-
strictions between various types of lexical items.
Current transformational theory approaches the lexical
problem through the use of semantic features in much the same
way as it approaches the phonological problem through fea-
tures. But features occur in bundles; they can be programmed
into the grammar as unit morphemes in classes cross-classified
according to such features as:
animate human personal
inanimate nonhuman impersonal
Such features may be part of universal thought, but they are
not manifest in every language. Whether the feature approach
or the unit approach is used, attention must be paid to which
features are grammatically pertinent. These lexical features
with overt syntactic manifestation can then be used as a basis
for subclassification of morpheme classes.
In an integrated theory of language, the fillers of various
slots can be reclassified into subclasses for pertinent lexical
features, whenever it is found that only items of one class such
as animate nouns, occur in this particular grammatical slot.
STRUCTURAL DESCRIPTIONS

The tagmemic grammar is also a generative grammar. By the

use of programming techniques, the morphemes of the lexicon
can be introdueed into the formulas of the grammar to produce
all and only the grammatical sentences of the language. But a
generative grammar should also assign a structural description
to each sentence generated by the grammar (Koutsoudas, 1966:4).
How does the tagmemic grammar assign a structural description
to each of the sentences generated?
 152 INTRODUCTION TO TAGMEMIC ANALYSIS

Underlying Phrase Markers

A tagmemic grammar consists of a series of syntactic state-
ments at the sentence, clause, phrase, and word levels. Each
formula consists of a construction name, a rewrite or equals
sign, and a string of tagmeme units marked as obligatory or
optional to the construction. Such a grammar is not only a gen-
erative machine; it assigns a structural description, an under-
lying phrase-marker or tree diagram, to each sentence generated.
Patterns Generated. The tagmemic grammar with a finite
morpheme inventory generates a well-defined set of grammatical
sentences. We now ask the further question: How many patterns
are generated by the grammar, and how many of the sentences
generated fall into each of the generated patterns?
The tagmemic model is a closed system beginning with the
symbol #Sent# and proceeding as far as the ultimate constit-
uents or morphemes of the language. This closed system is an
ordered set of rules or formulas, stated from higher to lower
at various levels of the grammar.
A simple formula is a formula that contains only obligatory
units. A conflated formula is a formula that contains at least
one optional unit. By definition, a conflated formula represents
a set of simple formulas. When a series of formulas from sentence
to morpheme is assembled in an ordered way, this series con-
tains an underlying structural description or tree diagram. If
any of the formulas are conflated formulas, then the resulting
tree is a conflated tree, that is, it contains optional elements.
Conflated trees represent a set of simple trees —a set which is
determined by the exercise of each yes/no option in the tree. In
the conflated tree, optional elements are enclosed in parentheses.
Tagmemic models are normally of the conflated type, since they
are designed to describe sets of sentences. A tagmemic grammar
may be reduced to its underlying structural description by a
rigid set of procedures.
Procedure for Generating Trees. Within the tagmemic model,
we prefer to ignore the functional slots in the grammar and
use only the form class symbols of the grammar in the repre-
sentation of structural descriptions, or tree diagrams:1
1. Each equals, sign in the grammar represents a branching
node in the tree. Construction names are node l b l s f
1
If function is included in the tree, the functions are inserted as labels on the branches.
"A labeled branch (function) which terminates in a node (constituent) is not an inappro-
priate representation of the function-set notation" (Longacre, "Reply," 1967:327).
 MORPHEME LEVEL 163

and the fillers in the construction string are the branches

which spread out from the construction node.
2. In conflated trees optional elements are enclosed in
parentheses; the relationships fixed by the functional
grammar are then evident in the structural tree, rep-
resented in terms of form only.

Sent = +Base:tCl -Into:ICF

+S:N+P:tV±O:N
+H:n±Mod:aj
±Mbd:av+H:tv

Simple Trees from Conflated Trees. Conflated trees represent

a set of simple itrees. To determine how many simple trees are
represented by j a conflated tree diagram, simply program the
numeral "one" into the formulas and calculate the MGP of the
solution. The resulting "unitary manifestation," which supposes
one and only one morpheme for each morpheme class, is equal
to the number of1 underlying tree structures. To determine these
structures in th;e concrete, redraw the tree as many times as
required, exercising each yes/no option represented by the op-
tional elements in the tree. The present tree diagram is reduced
to 12 different Simple tree diagrams, that is the MGP of the
grammar is realized in 12 grammatical patterns.
Once the trees are constructed the anatomy of the tagmemic
model becomes apparent; it is different from other models.
Branching is not: binary, but multiple; and this branching occurs
at a series of natural levels: the sentence, clause, phrase, word
and morpheme. The tagmemic tree is, as it were overlaid on a
structural grid, $ functional grid. Any single horizontal line in
the tree may be fread as units at the same level in the grammar.
It is the structure of this tagmemic tree, more than any other
fact, that justifies Longaere's contention (1964:7) that tagmemic
grammars represent a radical departure from former American
structuralism.
In summary, the tagmemic grammar is an ordered set of rules
which generates all and only the grammatical sentences of the
language and assigns structural descriptions to the sentences
generated. We must now inquire which patterns belong to which
sentences.
 154 INTRODUCTION TO TAGMEMIC ANALYSIS

Preterminal Strings of Grammar

Once the underlying phrase markers of the grammar have

been established first as a conflated tree which is then re-
duced to a set of simple trees this structure is mirrored in a set
of preterminal strings. These preterminal strings are repre-
sented first by a single conflated string, for the conflated tree;
then, each simple tree is reduced to a simple preterminal string.
The strings and trees give the same information; they show
relationships between morpheme classes.

Preterminal String Denned. A preterminal string is a string

of morpheme classes which in a tagmemic grammar represents
the correct morpheme classes in the correct order. The term is
taken from Chomsky (Aspects, 1965:84), who defines preterminal
string as a string that consists of "grammatical formatives and
complex symbols," the result of applying the system of rewrite
rules of the phrase structure grammar. This string, as we under-
stand it is an abstract string, derived from branching rules
before the application of any of the lexical rules of the phrase
structure grammar; it becomes concrete only when concrete
morphemes are programmed into the string in place of the mor-
pheme class symbols.
Just as there is one conflated tree representing many (here,
12) simple trees, so there is one preterminal conflated string,
representing 12 simple preterminal strings. The conflated pre-
terminal string is:
Sentence = +(n ±aj) +(±av +tv) ±(+n ±aj)
The minimum preterminal string is the string formed when all
of the optional elements are deleted from the formula, in this
case:
Minimum String = n + tv
The maximum preterminal string is the string formed when all
possible elements are included. In this case, the maximum string
consists of:
Maximum String = n + aj + av + tv + n + aj
Between these two extremes are sentences of three four and
five words. Simple strings are derived from the conflated string
by patiently exercising each yes/no option until all simple strings
are formed; or, the simple strings may be formed directly from
simple tree diagrams. Once the set of simple trees is determined
the patterns of the language are clear in the abstract representa-
tion of preterminal strings.
 MORPHEME LEVEL 155

To make these strings terminal, that is, to generate sentences

with these patterns, simply program the morphemes of the lexi-
con into each of the preterminal strings, to obtain the terminal
strings, which are the sentences of the language, each with its
own pattern.
Simple Preterminal Strings. The patterns of the language
are assigned to each sentence that is generated by the grammar.
These patterns are called structural descriptions and are rep-
resented both by tree diagrams and formulas. The formulas or
trees if conflated, represent a set of simple formulas. Thus m
the present case, the set of underlying strings is as follows:
1. n + tv
2. n + tvHm
3. n + tvHkn + aj
4. n + aj^-tv
5. n + ajH- tv -l-n
6. n + ajH- tv -t-n + aj
7. n 4- av -f t v
8. n + av -f tv-+ n
9. n + av -ftv- •+• n + a j
10. n + aj-\- a v ••l-tv
11. n + aj-l- av -•f-tv + n
12. n + ajH- a v •4- tv + n + aj
The grammar of the language, as formulated in tagmemics,
generates these 12 patterns, and only these 12 patterns, for the
language. The number of sentences formed in the language by
each pattern depends upon the number of morphemes of each
class in the lexicon of the language.
Terminal Strings of Morphemes. A terminal string is a string
of morphemes in the correct order. The preterminal strings of
language are reduced to terminal strings by substituting con-
crete morphemes for each of the morpheme class symbols in
the preterminal string. The result is a terminal string, which is
also a sentence in the language. This sentence, through mor-
phophonemic and phonological rules, is reduced to a pronounce-
able sentence in the language.
The number of terminal strings produced by each single
pattern can be calculated by programming the number of mor-
phemes of each class into each pattern, to obtain the number of
sentences. If these totals are added together, the sum of the
sentences produced by each pattern will add up to the MGP of
the solution, already calculated. Suppose, for example, that there
were exactly two morphemes of each class in the above set of
strings. Then the number of sentences produced is:
 156 INTRODUCTION TO TAGMEMIC ANALYSIS

Type 1: 2 X 2 = 4
Type 2: 2 X 2 X 2 = 8
Type 3: 2 X 2 X 2 x 2 = 16
Type 4: 2 X 2 X 2 = 8
Type 5: 2 X 2 X 2 x 2 = 16
Type 6: 2 X 2 X 2 x 2 x 2 = 32
Type 7: 2 X 2 X 2 = 8
Type 8: 2 X 2 X 2 x 2 = 16
Type 9: 2 X 2 X 2 x 2 x 2 = 32
Type 10: 2 X 2 X 2 x 2 = 16
Type 11: 2 X 2 X 2 x 2 x 2 = 32
Type 12: 2 X 2 X 2 x 2 x 2 x 2 = 64
Total number of sentences generated by all patterns together
would be equal to 252, which coincides with the MGP of the tag-
memic grammar.

Sentence Generators
After the underlying tree markers have been established
and the structure reduced to a set of preterminal strings, show-
ing the same structure in formula form the concrete sentences
of the language can be generated by arranging the elements of
sentences in a matrix form. These matrices, arrays of rows and
columns are called sentence generators.

Sentence Generator Denned. Preterminal strings represent

the output of the grammar in an abstract way. When the con-
crete morphemes of the language are then programmed into
the preterminal strings, the strings become terminal. In a tag-
memic grammar, as opposed to a transformational grammar
(Chomsky, 1965:84), the terminal strings are sentences of the
language given in phonemic script. However, the terminal string
here is analogous to that of Chomsky; the terminal string is
formed from the preterminal string by insertion of lexical forma-
tives. To generate the terminal strings of the grammar, we first
set up a sentence generator.
A sentence generator is a matrix, or array of rows and col-
umns. The columns represent the morpheme classes in the
correct order and are labeled with a set of dummy indices such
as I, J, K, L, M, N, and so on. The elements in each row are num-
bered and represent the stock of morpheme fillers belonging to
each morpheme class. When the same class occurs more than
once in the string, it is entered completely into the generator
as often as it occurs.
The generation of sentences from such a device is purely
mechanical and is easily performed by a computer or other
 MORPHEME LEVEL 157

mechanical device. The morphs could, for example, be printed

on the wheels of slot machines, so that each pull of the lever
would generate, at random, a sentence of the language which is
programmed into the machine. With digital computers, we enter
the data, identify each morph by number and class, then call
the pattern we desire, and give the command to execute the
process. For example, the command DO I = 1,2, DO L = 1 2 would
produce four different sentences of the I-L pattern. It is signifi-
cant that, in thie sample generation of the maximum pattern in
which form classes are repeated, the computer refused to repeat
operations on the noun and adjective classes unless these
classes were given a new dummy index. Telling the machine to
generate the pattern I-J-K-L-I-J, produced the answer, "But I
have already done I and J." However, after rewriting the pro-
gram with the last two sets marked as M-N, the full output of
the matrix was generated without further complicating the
machine programming.
Method of Generation. Language problems solved by tag-
memic methods have been run on an IBM 1620 computer, using
Fortran II programming. The results have been the generation
of the complete IjlGP of the problem as calculated by the methods
already suggested. The method of generation of sentences from
a sentence generator matrix is described below.
The computer works with a series of DO-loops, one for each
morpheme class. To generate the complete output, each class is
represented by a DO-loop, and all the morphemes of the sub-
language are read into the computer and identified as members
of these classes. The first item in each category is chosen by the
computer, beginning with the first morpheme in the sentence
and proceeding from left to right. The computer picks the first
morpheme of each group and prints the first sentence, using
the top line of the matrix generator.
The second sentence is the same as the first, except that the
morpheme at the end of the sentence is changed. For succeeding
sentences all of the possibilities in the far right column are first
exhausted, then the computer works on the possibilities in the
next to last column, and so on, until every possible combination
of these morphs in this order are completely exhausted. The
printout then contains, in a definite fixed order, all of the sen-
tences that can be generated by the solution.
Making Grammars Generative. Using all of the above pro-
cedures, it is possible to take a tagmemic grammar, determine
the underlying structural descriptions, write out the preterminal
strings, and generate the MGP of the grammar mechanically.
This generative ppwer results from the model as stated and is
 158 INTRODUCTION TO TAGMEMIC ANALYSIS

already implicit in that model. It is not dependent on the pro-

cedures by which the model was established as a tentative guess
at the underlying grammar.
Longacre once suggested that future tagmemic grammars
might well employ rewrite operations by "(a) stating them in the
front of the grammar, (b) carrying them out in illustrative
fashion here and there throughout the grammar, and (c) incor-
porating sections of rewrite exercises for the reader" (1964:32).
We suggest, further, that future tagmemic manuals might carry
mathematical answers to problems, the MGP and RGP of the
solutions with structural descriptions, preterminal strings, and
sentence generators illustrated here and there throughout the
series of exercises.
After calculating the MGP for many language problems,
determining the underlying phrase markers and the preterminal
strings, and actually generating these solutions on computers
with favorable results, it is difficult not to admit the generative
power inherent in the tagmemic model.

GENERATIVE POSSIBILITIES OF THE MODEL

The generative possibilities of the tagmemic model are reduced
to a series of operations by Longacre (Grammar Discovery Pro-
cedures "Symbols and Rewrite Operations," 1964, 24-32). These
operations, in order, are: (1) readings (or combinations), the
number of combinations of elements in a fixed order; (2) permu-
tations, the number of combinations possible where the order of
elements is changeable; and (3) exponents (or manifestations),
the number of fillers that may occur in each particular tagmemic
slot.

Combinations: Elements in Fixed Order

The tagmemic formula is a formula made up of symbols, and
these symbols are inherently mathematical. This formula, or
any such formula which contains optional elements, may be read
in a variety of ways. It represents, in summary form, many
different formulas. Each concrete reading either includes or
excludes each optional element.
Law of Combinations: C = 2" — 1. In mathematics, the law of
combinations states that the number of possible combinations
of n elements in a fixed order is represented by the formula:
C = 2" — 1. In this formula the numeral 2 is a binary base which
represents the choice of occurrence or nonoccurrence for each
 MORPHEME LEVEL 159

element; the exponent n represents the number of times this

choice is to be made; and the —1 excludes the possibility of
choosing nonoccurrence for all elements, and thus producing a
null string.
Problem: Given a string of 4 optional elements, ± A ± B ± C ± D ,
how many combinations are possible, with the elements in fixed
order?
Solution: Let n =>= the number of optional elements in the string,
then:
C = 2"4 - 1
C* 2 - 1=
C = 16 — 1 = 15 Answer
Check: Write the number of combinations of these elements.
ABCD
ABC ABD, ACD BCD
AB, AC, AD, BC, BD, CD
A, B, C D Total:15
This law of combinations and its applications may be found in
any standard textbook for college mathematics. For example, in
Schaum's Outline Series (College Algebra, Murray R. Spiegel,
1956, p. 231), it is suggested that a quarter, a dime, a nickel and
a penny may occur m 15 combinations.

Number of Combinations in a Formula. In tagmemics, each

formula has one or more obligatory elements. Group all obliga-
tory elements under one symbol, K = constant element. Let n =
number of optional tagmemes in the string. Then the number of
combinations possible, the number of ways the formula may be
read is C =2". With a constant element (K) in the formula, there
is no need to include the (—1) designation, as there is no possi-
bility of a null string.
K C = 2° = 1 combination
K ±A C = 2*2 = 2 combinations
K ±A ±B C = 2 = 4 combinations
K±A±B±C C = 243 = 8 combinations
K ±A ±B ±C ±D C = 2 = 16 combinations
The constant element, K, occurs in every combination, and n,
m C =* 2n, represents the number of yes/no choices the number
of optional tagmemes.

Number of Elements in Each Combination. Not only the

number of combinations, but the precise nature of each combi-
 160 INTRODUCTION TO TAGMEMIC ANALYSIS

nation may be determined. If we consider our combination for-

mula as a binomial expansion, then the kinds of combinations
may be determined by the laws for finding the coefficients of
binomial expansions. Thus C = 2n is considered as C = (a + b)n,
where a = yes, b = no. One method for determining the coeffi-
cients is known as Paschal's triangle.
Paschal's Triangle: Coefficients of Binomial Expansion
n=0 1 C = 2° = 1

n=l 1 1 C = 21 = 2 = 1+1
n=2 1 2 1 C = 22 = 4 = 1+2+1

n=3 1 3 3 1 C = 23 = 8 = 1+3+3+1

n=4 1 4 6 4 1 C = 24 = 16 = 1+4+6+4+1
If for example, the number of combinations is 16 the kinds of
combinations are broken down into 1+4+6+4+1, that is, 1 C with
five elements KABCD, 4 C with four elements, 6 C with three
elements, 4 C with two elements, and 1 combination with one
element K. These 16 combinations are: KABCD; KABC, KABD,
KACD, KBCD; KAB, KAC, KAD, KBC, KBD, KCD; KA, KB, KC
KD and the minimum string K.

Permutations: Elements Freely Movable

The Law of Combinations deals with elements of a formula
independent of the order in which they occur. This may be under-
stood simply by considering the elements in one fixed order. The
Law of Permutations is then used for each single combination,
if the order of elements in the formula is free. Given these freely
movable elements, how many different permutations of each
single formula are possible?

Law of Permutations: P = m! In mathematics the law of

permutations states that the number of possible permutations
of a string of m elements, when the order of those elements is
free, is represented by the formula:
P = m!
Read: The number of permutations equals the factorial of
the number of elements. The symbol m! (m-factorial) means
that the number m is the product of all of the integers from
one to the number m; thus 3! = 3 X 2 x l = 6.
Problem: Given a simple string of elements, such as KABC
with the elements freely movable, how many permutations will
be possible?
 MORPHEME LEVEL 181

Solution:
Let m = the total number of elements in the string
Then
P = m!
P==4 x 3 X2 x 1
P = 24, Answer
Check: Write the 24 permutations of the given formula.
KABC KACB, KBAC, KBCA, KCAB KCBA
AKBC AKCB, ABKC, ABCK, ACKB ACBK
BKAC, BKCA, BAKC, BACK, BCKA, BCAK
CKAB CKBA, CAKB, CABK, CBKA, CBAK
In tagmemics, the Law of Combinations is first applied to all
formulas which have optional elements (conflated or condensed
formulas) to reduce these to many simple formulas with only
obligatory elements. Second, the law of permutations is applied
to any simple formula which has elements that are freely mova-
ble. Note that in applying the Law of Combinations n = the
number of optional elements whereas in the Law of Permuta-
tions m = the total number of elements present. This total
includes the constant element K as well as A, B, C and so on.
This Law of Permutation may be found in the already cited
College Algebra (Schaum's Outline Series, Murray R. Spiegel
1956, p. 229), where the formula represents a special case of
permutations—of m elements, taken m at a time, for example:
5 elements, taken 5 at a time.

Permutations of Each Combination. If the order is fixed the

number of combinations alone is to be calculated. If, however,
the order of elements is freely movable then the number of
permutations must be calculated for each combination. 4Suppose
for example, there are 4 optional elements, then C = 2 = 16. If
the order of elements is freely movable each of these 16 combi-
nations has a definite number of permutations, calculated as
follows:
Number of Number of Number of
Combinations Elements Permutations s
1 1 1 x 1! = 1 X 1 == 1
4 2 4 x 2 ! = 4 X 2 == 8
6 3 6 x 3! = 6 X 6 == 36
4 4 4 X 4! = 4 X 24 == 96
i-i

5 1 X5! = 1 X 120 == 120

C = 16 P == 261
The formula, +K ±A ±B ±C ±D, will represent 16 combinations
of elements if the elements are in a fixed order, but 261 permu-
tations if the order of elements is freely movable. That is the
one tagmemic formula represents a maximum of 261 formulas.
 162 INTRODUCTION TO TAGMEMIC ANALYSIS

Maximum Number of Permutations. The answers resulting

from applying the Law of Permutations involve the maximum
number of permutations if the order of optional elements is
freely movable with the constant K. If some obligatory elements
are movable with respect to each other, these elements must be
counted separately in the calculation of permutations. If the
order is not entirely, but only partially, free, the permutation
formula will give only the maximum number of permutations.
If some of these permutations do not occur then restrictions
must be imposed upon the permutations possible. Note that the
maximum number of permutations is constant and may be cal-
culated and tabulated as follows:
With no optional elements, C= l P=1 1
With 1 optional element, C=2 P=3 1 - t-1
With 2 optional elements, C=4 P =9 1- f 2-1-1
With 3 optional elements C=8 P = 49 1 -1-3-(-3 + 1
With 4 optional elements, C = 16 P = 261 1 - l - 4 - i 1-6 + 4 + 1
With 5 optional elements C = 32 P=1631 1 - h 5 - i h l O + 10 + 5 + 1

Manifestations: Strings Made Terminal

The law of manifestation deals with the fundamental prin-
ciple underlying the mathematical theory of combinations and
permutations. In tagmemic formulas various manifestations of
the formula are realized by programming the filler classes into
the functional slots, whose combinations and permutations have
already been calculated.

Law of Manifestation: M = m x n. The fundamental principle

of combination and permutation theory is: given an element
that occurs m times, followed by an element that occurs n times
then the two elements in succession can occur m x n different
ways. The total number of manifestations possible is equal to
the product of each set in the string.
Problem: Given a string of Noun + Adjective + Adverb + Verb,
with 2 nouns 2 adjectives, 3 verbs, and 3 adverbs what is the
total number of manifestations possible of such a string?
Solution: Sentence = Noun + Adjective + Adverb + Verb
=2x2x3x3
= 36 possible manifestations, Answer
Where the string contains optional elements, the number of
manifestations and the number of combinations (excluding
permutation possibilities) can be calculated at the same time.
Each optional element has +1 added to account for the possibility
of zero occurrence at that position.
 MORPHEME LEVEL 163

Problem: Given the above string, in which the adjectives and

adverbs are optional elements, how many manifestations of the
string are possible?
Solution: Sentence = +Noun rtAdjective ±Adverb +Verb
= 2 x (2 + 1) x (3 + 1) x 3
= 2 x 3 X4.X3
= 72 possible manifestations Answer
Check: ;The number of manifestations when all elements were
present has. already been calculated: 36. With the adjective non-
oc4urrerit, the noym + adverb + verb manifestations = 18; with the
adverb nonoecurrent, the noun + adjective + verb manifesta-
tions = 12; with bjoth adjective and adverb nonoccurrent, noun +
verb manifestations = 6. The total of all manifestations possible
= 3S + 18 + 12 + 6 = 72, QED.
For manifestation of the preterminal strings of the grammar,
program the numeral 1 into the formula to obtain the number
of preterminal strings (— number of structural descriptions).
For the teririinal strings, program the number of morphs of each
class from the grammar into the formula to obtain the number
of sentences,
Slots with Multiple Fillers. For the calculation of the preter-
minal strings or the terminal strings, of the grammar, manifesta-
tions that occur within the same slot must be added together;
the resulting sum is the number of manifestations at that point
of the pattern. This sum then enters into the product that gives
the itotal number of manifestations in the string. For example:
given a clause formula, with S:N/pn/np, read as a subject slot
filled by either a noun phrase, or a pronoun, or a proper noun
the number entered as the "number of subjects" is 3 = 1 + 1-1-1
for preterminal strings. If these elements are manifested by
240 noun plwases, 6 pronouns, and 2 proper nouns then the
sum 240 + 6 + 2 = 248 is entered in this slot, in order to calculate
the number of sentences that are generated.
Similarly, when two slots are tied by an either/or notation
these two slots are considered to be effectively the same slot
for either one or the other must occur, but not both. In the for-
mula ±t-:tro'+-t:tmi tense must be expressed, but only once. In
this case, the sum of the tense markers, tm, and tm is entered
as the manifestation of that slot. If for example, tm, = 3 and
tm2 = 1, the number of manifestations is 4.
Calculations with Restrictions. When the formulas of the
grammar express cp-oceurrence restrictions, the mathematical
methods thus far described generate the maximum number of
sentences, without taking restrictions into account. The only
 184 INTRODUCTION TO TAGMEMIC ANALYSIS

rule of thumb for calculating restrictions is to break down the

manifesting number into subgroups and calculate the number
of strings that are generated with each subgroup.
For example, given a restriction of gender-number concord
on the formula for the noun phrase N = +Det:det +H:n ±Mod:aj
in Spanish, with 4 articles, 4 adjectives, and 4 nouns (which are
all different numbers and genders), the manifestation is not
4 X 4 x 4 = 64 because not all of these elements combine in the
phrase. Rather, calculations must be performed separately for
masculine/feminine and for singular/plural groups. Each group
has 1 x 1 x 1 = 1 phrase, with a total of 4 phrases in all; the other
60 phrases are not generated, as they violate the concord re-
striction concerning gender and number.
In summary, the tagmemic model has definite generation
possibilities, so that the exact number of patterns produced,
the number of combinations, permutations, and manifestations
that singly or together give us the full generation power of the
grammar, can be accurately calculated. The solutions of con-
crete problems are reduced to mathematical answers.

PRACTICE 6: ASSIGNING STRUCTURAL DESCRIPTIONS

For the following sentences adapted from Problem #99, Viet-

namese (Laboratory Manual, Merrifield, 1967):
1. Draw the conflated tree (structural description).
2. Generate the preterminal strings (sentence patterns).
3. Set up a sentence generator for computer generation.

VIETNAMESE
1. cho sem cum to
'The dog sees the big bird.'
2. cho to xawng sem cho nyo
'The big dog does not see the little dog.'
3. cho nyo thay chim nyo
'The little dog perceives the little bird.'
4. chim ku?ng sem cho
'The bird also sees the dog.
5. chim ku?ng thay
'The bird also perceives.
6. chim xawng thay
'The bird does not perceive.
7. cho thay chim
'The dog perceives the bird.'
8. cho sem
'The dog sees.
 MORPHEME LEVEL 165

9. chim to thay cho to

'The big bird perceives the big dog.'
10. chim nyo kul?ng thay chim to
'The little bird also perceives the big bird.'

GRAMMAR LEXICON
Sent = +Base:tCl -Into:ICF chim n. 'bird'
252 252 x 1 cho n. 'dog'
tCl=+S:N+P:tV±O:N nyo aj. 'little'
252 6 x 6 x (6 4-1) to aj. 'big'
N = +H:n ±Mod:»j ku?ng av. 'also'
6 2 x(2 + l) xawng av. 'not'
tV = ±Mod:av +H:tv sem tv. 'see
6 (2+1) x 2 thay tv. 'perce

STRUCTURAL DESCRIPTION
Sent

PRETERMINAL STRINGS

1. n -f- tv = 4 Sent I L types

2. n -1- tv 4- n = 8 Sent, I L M
3. n -( • • t v + n -1- aj = 16 Sent I L M N
4. n -H aj + tv = 8 Sent, I J L types
5. n ih aj + tv -h n = 16 Sent I J L M
6. n -iKaj + tv -h n + aj = 32 Sent I J L M N
7. • n Hh a v + tv = Sent,
8 I K L types
8. n Hf- a v+ tv Hh n = 16
Sent I K LM
9. n Hf- a v+ tv Hh n + aj = 32
Sent I K L MN
10. n HK aj + avHh tv = 16
Sent, I J K L types
11. n -1H aj + av H1- tv + n = 32
Sent, I J KLM
12. n -h a j + av -h tv + n + aj = 64 Sent, I J K L M N

MGP = 252 Sent, 12 patterns

 166 INTRODUCTION TO TAGMEMIC ANALYSIS

SENTENCE GENERATOR
I J K L M N
1. chim nyo ku?ng sem chim nyo
2. cho to xawng thay cho to

SUPPLEMENTARY READINGS 6
Chomsky, Noam, Syntactic Structures. The Hague, Mouton & Co.,
1957. Grammar denned as generative, 13; all and only requirement,
18; phrase structure rules, derivations, and structural trees, 26-
27.
, Aspects of the Theory of Syntax. The M.I.T. Press, 1965. Struc-
tural descriptions assigned, 9; branching vs. lexical rule, 68; ter-
minal vs. preterminal string, 84; grammatical relations vs. gram-
matical category, 68. Extra-node tree, 69, not acceptable.
Cook, Walter A., S. J., On Tagmemes and Transforms. Washington, D.C.,
Georgetown University Press, 1964. Rules for generation, 53-56
developed from prepublication edition of Longacre, 1964 (infra).
, "The Generative Power of a Tagmemic Grammar," Monograph
Series on Languages and Linguistics, No. 20, 27-41. Washington
D.C., Georgetown University Press, 1967. Methods for generating
sentences and assigning structural descriptions.
Katz, Jerrold J., and Postal, Paul, An Integrated Theory of Linguistic
Description. The M.I.T. Press, 1964. For semantic component re-
strictions on concurrence based on semantic features.
Koutsoudas, Andre, Writing Transformational Grammars. New York,
McGraw-Hill, Inc., 1966. For notion of the conflated rule 9 flf. and
the construction of P-markers and PS rules, 14 ff.
Longacre, Robert E., Grammar Discovery Procedures. The Hague,
Mouton & Co., 1964. "Introduction: Symbols and Rewrite Opera-
tions," 24-34, in terms of readings, permutations, and manifesta-
tions.
"Some Fundamental Insights of Tagmemics." Language, 41:
65-76 (1965). Generative power of tagmemic grammar, 71 ff.
, "Reply to Postal's Review of Grammar Procedures," UAL,
33:323-328 (1967), with sample tagmemic trees, 325.
Merrifield, William R., "On the Form of Rules in a Generative Gram-
mar," Monograph Series on Languages and Linguistics, No. 20,
43-55. Washington, D.C., Georgetown University Press, 1967.
Postal Paul M., "Constituent Structure: A Study of Contemporary
Models of Syntactic Description," UAL, vol. 39. no. 1, Part III,
33-51 (1964).
, "Review of Robert E. Longacre's Grammar Discovery Pro-
cedures," UAL, 32: 93-98 (1966). For objections to the tagmemic
system.
 TABLE 7: LANGUAGE DESCRIPTION

LANGUAGE UNITS VARIATION CONTRAST SYSTEM

Trimodally Manifestation Mode Feature Mode Distribution Mode
Structured Wave-Dynamic Particle-Static Field-Systematic

I 1 2 3
PHONOLOGICAL ALLOPHONES PHONEME PHONEMIC SYSTEM
COMPONENT Set of etic sounds Emic units of sound Paradigmatic
(Phonology) Phoneme variants Differential function Matrix of Phonemes

II 4 5 6
LEXICAL ALLOMORPHS MORPHEME MORPHEMIC SYSTEM
COMPONENT Set of etic forms Emic units of form Paradigmatic
(Lexicon) Morpheme variants Referential function Matrix of Morphemes

III 7 8 9
SYNTACTIC ALLOTAGMAS TAGMEME TAGMEMIC SYSTEM
COMPONENT Set of etic patterns Emic pattern units Paradigmatic
(Grammar) Tagmeme variants Syntactic function Matrix of Tagmemes
7 LANGUAGE DESCRIPTION

Language description has three basic requirements. It must

deal with sounds with forms, and with the arrangements of
forms in sentences. "If language is a structure," says Sapir
(1921:24), "and if the significant elements of language are the
bricks of that structure then the sounds of speech can only be
compared to the unformed and unburnt clay of which the bricks
are fashioned." The problem of language description is to isolate
the units of sound, that are used to form the units of referential
meaning, which then fit into the blueprints of language.
Phonology, Lexicon, and Grammar. The tagmemic model
deals with sounds, forms, and arrangements in a triple gram-
matical hierarchy of phonology, lexicon, and grammar. These
three hierarchies are systems that are "semi-autonomous but
interlocking" (Longacre, 1964:7). The triple hierarchy, as ex-
plained by Pike (1958:275), is as follows:
1. The phonological hierarchy, with the phoneme as the
minimum unit, and syllables, stress groups, and so on
as higher units in the hierarchy.
2. The lexical hierarchy, with the morpheme as the mini-
mum unit and with morpheme sequences or special
collocations as higher units.
3. The grammatical hierarchy, with the tagmeme as the
minimum unit, and various tagmemic constructions as
higher units in the hierarchy.
Language description according to the tagmemic model has
three components. The syntactic component, or tagmemic gram-
mar lists the arrangements of language in terms of basic tag-
 170 INTRODUCTION TO TAGMEMIC ANALYSIS

meme units; the lexical component, or tagmemic lexicon lists

the forms of language and their meanings; the phonological com-
ponent lists the sound units of language and their allophonic
realization in language utterances. The language description is
in terms of "a phonological statement, a grammatical statement,
and a highly sophisticated dictionary" (Longacre, 1964:8).

Particle, Wave, and Field. The units of language description

may be considered as simultaneously particle, wave, and field
as explained by Pike (1959:37-54; 1967:510):
1. Particle refers to units in their feature mode. It is a
static view of language, with units described as clear-cut
particles, well denned.
2. Wave refers to units in their manifestation mode. It is
a dynamic view of language, with variants subject to
blurring and overlapping.
3. Field refers to units in their distribution mode. It is a
systematic or functional view of language, with units
as part of a set of oppositions.
The three different ways of viewing the unit are not contra-
dictory but complementary. The unit is "structured three ways
at once" (Pike, 1967:93). The unit is not divided into three parts,
but considered from three different viewpoints: as a discrete
unit as manifested in variants, and as part of a set of opposi-
tions within a matrix or field. Within this trimodal structuring,
particle and wave are more closely related. In the formula LS =
(M/F)D, the linguistic sign, or unit, is composed of a manifesta-
tion-feature complex which is distributed in a field. The F mode
corresponds to comprehension or definition and the M mode to
the extension or set of things defined. It is the comprehensive-
extensive unit which is distributed as a single cell in a matrix of
oppositions.

Contrast, Variation and Distribution. The three modes of the

unit may also be considered under the process titles of contrast
variation and distribution. In an equivalence set up by Pike
(1967:85, fn. 3):
1. For feature mode, one can read contrast (contrastive-
identificational components). The feature mode is de-
fined as composed of "simultaneously occurring identi-
ficational-contrastive components" (Pike, 1967:85).
2. For manifestation mode, one can read variation (range
of free, conditioned, or complex variants with physical
component obligatory). The manifestation mode is com-
posed of "nonsimultaneously occurring physical vari-
ants" (Pike, 1967:85).
 LANGUAGE DESCRIPTION 171

3. For distribution mode, one can read distribution (oc-

currence in a class as member of that class, in a se-
quence of segments organized hierarchically, and in
.cells of a matrix made up of intersecting dimensions).
The distributional mode is denned as composed of "re-
lational components" (Pike, 1967:85), and includes distri-
bution in class, sequence, and matrix.
No justification is given for the three modes except the very
practical justification of their usefulness in description. The
basic modes seem to be linked to the discovery principles of (1)
phonetic and semantic similarity; (2) contrast and complementa-
tion; and (3) pattern congruity. The contrast, variation and
distribution of units used in the discovery process appear as
verified modes in the final language description.
THE PHONOLOGICAL COMPONENT
The first component of a language description deals with
sounds —sounds as manifested in the forms of language, sounds
grouped as units according to identifying-contrastive features
and sounds as part of a system. In the phonological component
of a language description, we consider (1) the manifestation
mode, (2) the feature mode, and (3) the distribution mode of the
sound units of the language, the phonemes. The three modes of
the phoneme ar# developed in Pike (1967, Chap. 8).

The Manifestation Mode: Phonetics

In the discovery process, we begin the study of the sound
system with the study of phonetics — the isolation and classifica-
tion of all of the! sounds that occur in the given language. In the
presentation of language descriptions, these sounds are listed
as etic variants of the sound unit; they are grouped as allophones
of the phonemes of the language.
Manifesting Set of Allophones. In a language description we
establish a phonemic inventory—a list of the essential sound
units or phonemes, of the language. Each phoneme has a well-
defined set of manifesting forms, or allophones. Even if the
phoneme has only one allophone this one sound is considered as
a set of manifesting forms. For example, in a partial description
of the phoneme 7p/ in English, defined as a "voiceless bilabial
stop," the allophones may be listed as:
[p] unaspirated, in initial clusters, CCV-
For example, spot, spill, span
[ph] aspirated, in initial position, CV-
For example, pot, pill, pan
 172 INTRODUCTION TO TAGMEMIC ANALYSIS

The phoneme is manifested at the concrete level of phonetic

transcription. The allophones represented in that transcription
are nonsimultaneously occurring physical variants of the
phoneme. They are nonsimultaneous because in each "occur-
rence" of the phoneme, only one of these etic units occurs at
a time. They are physical variants in the sense that they are
an approximation of the physical event. Even in phonetic tran-
scription, some of the physical sound is lost; but the features
worth noting are based on wide experience with many languages
and roughly correspond to the limits of transcription of the
International Phonetic Alphabet. They are the variants of the
same sound because they occur either in complementary distri-
bution or in partial complementation with some free variation
or in free variation; but they never occur in contrast with each
other in the language. They are noncontrastive variants.

Allophones as Waves. The manifestation mode of an emic

unit has wavelike features. It is a dynamic view of language,
closer to the physical reality, and accounts for overlapping and
blurring characteristics in language. In phonology, sounds are
produced by a continuous movement with the borders of indi-
vidual sounds largely indeterminate. A word with a CVC pattern
is not pronounced as a sequence of discrete sounds, but as a
continuous wavelike movement in which the sounds overlap.
The units are perceived as the peaks of sound but the sound
borders are indeterminate. For example, a word, such as Ipinl,
/pin/, might be represented by waves:
[Ph] [i] [n]

The initial /p/ sound is aspirated and is in a forward position

because of the following HI sound; the /i/, in turn, is partially
prenasalized in anticipation of the following /n/. The overlapping
of the sounds can be demonstrated in a phonetics laboratory by
comparing the qualities of the vowels and consonants in dif-
ferent sequences. In recognizing the sounds as units, we recog-
nize the peaks of the waves and ignore the overlapping indeter-
minate borders of the sounds.

Allophonic Variation. The manifestation mode of an emic

unit is also characterized by the noncontrastive distribution of
its variants. Allophones of the same phoneme may be in comple-
mentary distribution or in partial complementation with some
free variation, but these allophones are never found in contrast
with each other. The principles of phonemic analysis deal with
possible variants:
 LANGUAGE DESCRIPTION 173

1. The Principle of Phonetic Similarity. Similar sounds

may belong to the same phoneme. Dissimilar sounds
must belong to different phonemes. This principle deals
with the form of the sound and declares that there
must be some degree of similarity between two sounds
before th^y can be considered as possible variants of
the same phoneme.
2. The Principle of Complementation. Sounds not in con-
trast, which are either in complementary distribution
in mutually exclusive environments or in partial com-
plementation with some free variation, may belong to
the same phoneme. Sounds in contrast must belong to
different phonemes. This principle deals with the distri-
bution of the sounds and declares that two sounds may
be grouped as allophones of the same phoneme if they
are neve* found in contrastive distribution.
According to these principles, allophonic variants are var-
iants that are phonetically similar, suspicious pairs that are
always.found in ttoncontrastive distribution in the forms of the
language.

The Feature Mode: Phonemics

The study of the sound system proceeds in two steps. First
the etic sounds which occur in the language are isolated and
classified in the study of phonetics. Second these etic sounds
are grouped into emic units in the study of phonemics. In the
presentation of the description, the essential sound units are
listed as the phonemes of the language.
The phoneme is defined as a minimum unit of sound which
has the capacity for changing the meaning of a linguistic form.
Sounds in contrast belong to different phonemes; sounds not in
contrast belong to the same phoneme. The test for contrast is
whether the sound change is capable of producing a change in
the meanings of linguistic forms. For example, in testing for
English consonants, use a test frame. Each consonant that
forms a new word belongs to a new phoneme. Thus:
-ILL: pill, bill; till, dill; chill, Jill; kill, gill
Phonemes: /p, b/ /t, d/ /c, j / /k, g/
In testing for English vowel nuclei, use a test frame. Thus:
B__LT: bit, beat; bet, bait; bat, but; boot, bought
Phonemes: HI, /iy/, lei, /ey/, /se/,/A/, /uw/,/o/
The use of many such frames would isolate all of the consonants
and vowels of the English language as minimum essential units.
 174 INTRODUCTION TO TAGMEMIC ANALYSIS

Distinctive Features of Phonemes. In a language description

all of the phonemes are listed with their identifying-contrastive
features. If the description is based on articulatory phonetics,
these features are articulatory. For consonants the features
are generally voicing, place of articulation, and manner of articu-
lation; for vowels, the features are tongue height, tongue posi-
tion and lip rounding. In acoustic phonetics, the features are
limited to 12 to 15 acoustic features universal in language.
The identifying-contrastive features, whether articulatory
or acoustic constitute the feature mode of the phoneme; these
features are simultaneously occurring components of the unit.
They are identificational because they positively identify the
physical characteristics of the phoneme. They are at least
implicitly contrastive, in that each feature used to identify
a phoneme also sets it apart from other phonemes of the lan-
guage.
Essential sound units in this view are not constituted
merely by a set of oppositions. Each sound unit has positive
identifying characteristics and because of these identificational
characteristics, it is opposed to other units of the system which
have positive characteristics. The /p/ phoneme, a voiceless
bilabial stop, is opposed to nonvoiceless, nonbilabial and non-
stops. The features are first identificational; their contrastive
nature only becomes apparent when the phoneme is viewed
within the phonemic system.
Phonemes as Particles. The feature mode of an emic unit
has particle-like features. The particle view is a static view of
language, in which the essential units appear as clear-cut build-
ing blocks, out of which the meaningful forms of language are
constructed. The sounds which actually occur may overlap, but
these etic sounds are grouped under units called phonemes. A
phoneme is said to "occur" when one of its manifesting forms
occurs.
The phoneme is an abstraction. It is abstracted from the set
of allophones in which it is realized. Because it is abstract it is
a clear-cut particle, defined by the features common to its al-
lophones. It is part of organized language, belonging to language
rather than speech, to competence rather than performance. In
abstraction essential features are retained and nonesssntial
features are ignored. Thus, the word /pin/, 'pin,' might be repre-
sented phonemically as made of particles:

/ i / J + I /n/
In this representation, the overlapping characteristics of the
sounds are ignored, /p/ is a voiceless bilabial stop. The feature
 LANGUAGE DESCRIPTION 175

of aspiration is ignored, /i/ is a high front unrounded vowel. Its

prenasalization before /n/ is ignored, /n/ is a voiced alveolar
nasal.

Phonemic Contrast. The feature mode of an emic unit is also

characterized by the fact that in this mode the unit appears in
contrast. Phonemes are in contrastive distribution. They are
essential sound units which have the capacity for changing
meaning. Because they can change meaning, they must be rec-
ognized, or the message can be distorted. The principles of
phonemic analysis isolate the essential phonemes:
1. The Principle of Phonetic Similarity (Negative Norm).
Sounds which are dissimilar are not grouped as allo-
phones of the same phoneme. This principle prevents
grouping of any sounds except those listed as suspicious
pairs. Thus, the sound /h/ and the sound /q/, despite
their perfect complementary distribution, do not belong
to the same phoneme.
2. The Principle of Contrast. Sounds in contrast in identi-
cal environments where the sound difference is paral-
leled by a meaning difference, must belong to different
phonemes. If two sounds occur in the same slot in a
frame in parallel distribution, and if, while occurring
in this slot, they change the meaning of the form, they
are different phonemes. Once the sound unit has demon-
strated its capacity to change meaning, by actually
changing meaning in at least one case, it is a phoneme.
Once a phoneme, always a phoneme. Pairs of sounds
occurring in identical frames, with parallel change in
meaning, are called minimal pairs.

The Distribution Mode: Phonemic Systems

After the study of phonetics and phonemics the results of
the phonemic analysis are presented as a list of phonemes or-
ganized into a phonemic system. The consonants of the language
are arranged in a contrastive chart of consonant phonemes; the
vowels of the language are arranged in a contrastive chart of
the vowel phonemes. In this mode, each phoneme is seen as part
of the sound system of the language.
The phonemic system of a language is a list of all the pho-
nemes of the language arranged in some systematic pattern.
These sound systems generally range from 30 to 60 phonemes,
segmental and suprasegmental. One standard analysis of
English lists 24 consonants, 9 vowels, and 12 suprasegmentals,
or 45 phonemes total (Dinneen, 1967:44). Hawaiian has only 13
 176 INTRODUCTION TO TAGMEMIC ANALYSIS

segmental phonemes, 5 vowels, and 8 consonants the least num-

ber of phonemes recorded for a language. The highest number
estimated for any language is about 75.
The distribution mode of the phoneme may be interpreted
as a set of syntagmatic relations, a set of paradigmatic relations,
or both. The syntagmatic relations of the phoneme are its pos-
sibilities of occurrence in linear sequences and are sometimes
referred to as phonotactics. For a clear understanding of pho-
nemes as a system, it seems preferable to leave phonotactics
aside and concentrate on paradigmatic relations.
Phonemic Systems. In a language description, phonemes
are charted within a systematic framework, generally with
separate charts for the consonants, the vowels, and possibly the
suprasegmental phonemes. The distribution mode of the emic
unit is the system of oppositions of which the emic unit is a part.
It is only within this relational system that the unit, with its
identifying features, is seen as contrasting with the other units
of the system. The features, within the system, become explicitly
contrastive. To completely understand the emic unit we must
know:
1. What is it? What are its identifying contrastive features?
2. How is it manifested? What comprises its existential set?
3. How does it fit into the system? What is it opposed to?
As in Gestalt psychology, the figure emerges more clearly
when seen against the background of its immediate environ-
ment within the system or matrix of which it is a part. Within
the phonemic system, phonemes may differ by a minimal con-
trast of one feature, or they may differ by several features. The
/p/ phoneme, in English, is seen as opposed to nonvoiceless, non-
bilabial, and nonstops in a system where these sounds occur. The
phoneme is partially determined by these contrasts. Likewise
the range of allophones of the phoneme is limited by the sounds
that surround it in its own phonemic matrix.
Phonemic Fields. The distribution mode of an emic unit has
the characteristics of a matrix or field. The field view is a func-
tional view of language, in which the emic units, already recog-
nized as clear-cut units, are seen as functioning particles within
a system. When the phoneme occurs, it occurs as part of a pho-
nemic system. It brings with it a set of identifying features,
seen within the system as explicitly contrastive.
Once the phonemic chart is established the occurrences of
the phonemes are easier to chart in linear sequence. First, the
choice of essential units is limited to a concrete inventory.
Second, the phonemes of the inventory are so closely interrelated
 LANGUAGE DESCRIPTION 177

that the choice of phoneme occurrence is limited by, and con-

trasts with the nearest phonemes. Thus, in English, the pho-
neme 1$ is contrasted with other units in the following chart:
c k Voiceless stops
j g Voiced stops
Like a navigator locating the position of his ship on the sea the
analyst uses two intersecting lines to get a fix on the position —
on the one hand, the position of the sound in the linear string, a
horizontal line, and on the other hand, the place of the phoneme
in its paradigm, a vertical line. In this way, the sequence /pin/
is distinguished from /bin/ and /tin/.
Phonemic Symmetry. The distribution mode of the emic unit
seen as paradigmatic, places the phoneme in a symmetrical ma-
trix, or at least in aimatrix which tends towards symmetry. These
essential units make up the total set of units of the language,
arranged in a symmetrical system. They constitute a set of op-
positions which is the sound system of the particular language.
The principles of symmetry and economy are used to help isolate
phonemic units:
1. The Principle of Symmetry. Phonemic systems tend to-
wards symmetry or neatness of pattern. Asymmetrical
solutions proposed for phonemic inventories are con-
sidered less likely. According to this principle, if a
language shows a set of contrasts, such as /p,b/, /t,d/,
and has the phoneme /k/, one would expect to find a cor-
responding phoneme /g/ to produce a balanced /p,t,k/
versus /b,dj.g/ system of stops in the language.
2. The Principle of Economy. Interpretation of phonemic
systems teinds towards the least number of phonemes.
However as Hockett has pointed out (1958:110), the
least number of phonemes per utterance is also part
of economy. The principle of economy does not prevent
the analysis of phoneme sequences as units, such as
IM for [t§]. Such an analysis may add the new phoneme
f&f to the inventory, but it simplifies each of the pho-
nemic utterances in which [ts] occurs in the language.

THE LEXICAL COMPONENT

The second compohent of a language description deals with
forms—forms as they are manifested in language, forms in their
identifying-contrastive features, forms as part of the lexical
system. The minimum meaningful form unit is the morpheme. In
 178 INTRODUCTION TO TAGMEMIC ANALYSIS

the lexical component of a language description, we consider:

(1) the manifestation mode, (2) the feature mode, and (3) the dis-
tribution mode of the form units of language, the morphemes.
The three modes of the morpheme are developed by Pike (1967,
Chap. 6).

The Manifestation Mode: Morphetics

In the discovery process, we begin the study of the lexical sys-
tem of language with the study of morphetics, which is the study
of forms that occur in language. In the presentation of language
description, the results of the study of morphetics become the
manifestation mode of the morpheme unit. Each etic morph
discovered is listed as an allomorph of one of the emic units the
morphemes of the language.
Manifesting Set of Allomorphs. In a language description we
establish a lexicon for the language, which is an inventory of the
"total stock of morphemes in a language" (Bloomfield, 1933:162).
This listing may be extended to include other lexemes such as
polymorphic words and idiomatic expressions. Within the lexi-
con, each single morpheme is realized in a well-defined set of
manifesting forms, which are called allomorphs of that mor-
pheme. Even when the morpheme is invariant and is always
manifested by the same form,the morpheme is realized in a set of
allomorphs, which, in this case, is limited to a single member.
These morphemes are realized on the concrete level of phonemic
transcription as morphs, which are minimum meaningful forms.
For example, in a description of English, the "plural" morpheme
{Z,} is manifested by a concrete set of actually occurring plural
forms such as the forms:
/-iz/ after sibilants and fricatives, s, z, s, z, c, j
as in horses, roses, churches, judges
/-s/ after other voiceless sounds, (except s, s, c)
as in, cats, tacks, cliffs, myths
l-zl after other voiced sounds, (except z, z, j)
as in dogs, pads, boys, girls
The manifesting set of allomorphs corresponds to the manifesta-
tion mode of the emic unit. The morphs, in phonemic transcrip-
tion are the "nonsimultaneously occurring physical variants"
(Pike, 1967:85) in the lexical component. They are nonsimultane-
ous because one and only one morph occurs at a given time. The
selection of the proper morph depends upon the phonological,
lexical, or grammatical environment of the form.
 LANGUAGE DESCRIPTION 179

Allomorphg as Waves. The manifestation mode of an emic

unit has wavelikei features. In the dynamic view of language, in
which forms are considered at the concrete level of occurrence
overlapping and blurring characteristics may be manifested.
Forms in actual speech run together so that their occurrence
appears like the overlapping of wave structures rather than as
a continuous chain of discrete meaningful particles. For ex-
ample:
Did you enjoy it sounds like /djuwnjoyit/
How did you like it sounds like /hawj A/+/laykit/
To understand these phrases in which the borders of morphs are
not well defined we must revert to that level of analysis where
the meaningful forms of language are abstracted and systemati-
cally grouped as well-defined lexical particles within the lexical
system.
Allomorphic Variation. The manifestation mode of the emic
unit is also characterized by the noncontrastive distribution of
its variants. Morphophonemics is the study of the representation
of alternate forms of a morpheme within different environments.
Hockett (1958:273) distinguishes sporadic and systematic alterna-
tion:
1. Sporadic alternation is the nonpredictable variation of
the forms of language, due to history, style, dialect, slips
of the tongue, and so forth. Such alternates do not form
part of systematic description.
2. Systematic alternation is predictable in terms of the
environment in which the form occurs, either phonologi-
eal, or morphological, or grammatical. This alternation
is an essential part of description.
Allomorphs of the same morpheme may be in complementary
distribution, or in partial complementation with some free varia-
tion but not in full free variation (which would make them spo-
radic), and never in contrast in the language. The following
principles are useful:
1. The Principle of Phonetic-Semantic Similarity. Invari-
ant morphs with one form and one meaning in all occur-
rences are morphemes. Variant morphs with the same
meaning must belong to the same morpheme. Semantic
similarity is necessary and sufficient to establish morphs
as allomorphs of a morpheme; phonetic similarity is
generally found, but is not absolutely required. Where
phonetic similarity is missing, as in the grouping of
 180 INTRODUCTION TO TAGMEMIC ANALYSIS

go and went or / and me, strong arguments from pattern

congruity are required to justify grouping these as
allomorphs.
2. The Principle of Complementation (Negative Norm).
Forms not in contrast which are either in complemen-
tary distribution in mutually exclusive environments
or m partial complementation with some free variation
may belong to the same morpheme. This principle states
that morphs may be grouped as allomorphs, provided
they are never found in positive contrast in the lan-
guage. It will in fact be the case that, if the morphs are
to be grouped as allomorphs of a single morpheme,
they will appear as noncontrastive variants, with well-
defined environments for the occurrence of each variant
of the morpheme.

The Feature Mode: Morphemics

Discovery of morphemes proceeds in two steps. First, in the
study of morphetics, the minimum meaningful forms of the
language are isolated with their proper meaning. Second, the etic
morphs are grouped, in the study of morphemics into minimum
meaningful form units, which may be represented by one or more
allomorphs. In the presentation of the lexical component, the
morphemes are listed in the lexicon, together with their form
class, and gloss. Where these morphemes have more than one
allomorph, the allomorphs are listed in a set of morphophonemic
rules which give the proper environment for each allomorph.
These rules are appended to the lexicon.
The morpheme is denned as a minimum meaningful unit in
the language. Forms with the same meaning belong to the same
morpheme; forms with different meanings belong to different
morphemes. The test for contrast is whether, given that the
forms occur in identical environments, they have different mean-
ings within the same environment. Thus, even homophonous
forms, such as pair/pare/pear, are listed as different morphemes.
If meanings are similar, an attempt should be made to set up a
statement of complementary distribution which describes the
environment in which each form occurs. According to this test,
so-called synonyms, such as big I large, fail to complement each
other and are listed as different morphemes.

Distinctive Features of Morphemes. In the language descrip-

tion the morphemes are listed in the lexicon together with their
form class, and gloss. For the form-meaning unit, the list of vari-
ant forms and their respective environments, the form class to
which the morpheme belongs and the meaning carried by the
 LANGUAGE DESCRIPTION 181

morpheme, constitute its distinctive features. Meanings are

often difficult to describe; morphemes are generally considered
to have an "area of meaning" that is in contrast with the "areas
of meanings" of other morphemes in the language.
The identifying-contrastive features of the morpheme cor-
respond to the feature mode of the lexical unit. These features
are the "simultaneously occurring identificational-contrastive
features" (Pike, 1967:85). In the morpheme unit the referential
inward meaning occurs simultaneously with the outward form
or set of forms of the morpheme. These forms, and the meaning
they carry, contrast with other morphemes of the language. They
constitute the morpheme as a form-meaning composite.
The features of the morpheme are first identificational and
then contrastive. The morpheme of the language must be estab-
lished as a form-meaning composite, with one meaning and the
forms which carry this meaning; this unit is then seen as con-
trastive with thei other form-meaning composites within the
structure of the same language.

Morphemes as Particles. In the feature mode, the morphemes

of the language are viewed as particles. This is the static view of
language, in which units are seen as distinct and free of the
overlapping characteristics that occur at the etic level. The
form-meaning composites are the discrete building blocks of
which utterances are made, independently of the concrete form
which the morpheme assumes in a particular utterance.
The morpheme is an abstraction. It is abstracted from the
set of allomorphs by which it is realized. Being abstract, it is a
discrete type of particle, denned by the features of form and
meaning common to its allomorphs. This abstract unit is part of
organized language, belonging to competence rather than to per-
formance. The essential features of the form-meaning composite
are retained and the nonessential features ignored. Words may
be blurred in actual speech, but the morphemes are distinct.
Thus the sentence* which phonologically is /djuwnjoyit/, is repre-
sented:
do + |past + you| + en
Each of the units is a morpheme, manifested in a particular
allomorph. The morpheme sequence is particle-like, independent
of the allomorphic representation chosen, and the phonological
realization of the allomorphs.

Morphemic Contrast. In the feature mode of the emic unit

the unit is contrastive. Each form-meaning composite is opposed
to all the other form-meaning composites of the language. Mor-
phemes are essential form-meaning units. The message carried
 182 INTRODUCTION TO TAGMEMIC ANALYSIS

by each form must be recognized and the meanings behind

forms are in fact, recognized by the speech community, which
enables them to communicate via language. The principles of
morphemic analysis help isolate the morphemes:
1. The Principle of Phonetic Similarity (Negative Norm).
Forms that are dissimilar phonetically normally are
not grouped as allomorphs of the same morpheme. The
presumption is that if two different forms occur with
closely related meanings, they belong to different mor-
phemes. We assume the so-called synonyms are different
in the language. However, allomorphs differ from almost
perfect identity of form to suppletion, in which the whole
form is replaced. Without similarity, strong arguments
are required from pattern congruity to establish mor-
phemic identity.
2. The Principle of Contrast. Forms in contrast in identical
environments, where the form difference is paralleled
by a meaning difference, must belong to different mor-
phemes. This is the positive principle of contrast. If two
morphs occur in the same grammatical slot, in parallel
distribution and if, while occurring in this slot, they
signal a change in meaning, then they belong to differ-
ent morphemes, that is, the morphemes are in contrast.

The Distribution Mode: Morphemic Systems

In the discovery process, after the study of morphetics and
the grouping of allomorphs in morphemics the results of mor-
phemic analysis are presented in a systematic arrangement in
the lexicon. In this lexicon, morphemes are seen as part of a
contrastive system, in which morphemes are grouped into mor-
pheme class, system. The lexical units are separated into content
classes are opposed to each other.
The lexical system of a language is a part-of-speech, or mor-
pheme class, system. The lexical units are separated into content
classes such as noun, verb, adjective, and adverb; and function
word classes, such as determiner, connector, relater, intensifier,
negative, and auxiliary. In the form-class-gloss listing of the
lexicon the form class label sorts the morphemes according to
classes; and within each class, the morpheme units differ accord-
ing to their form and meaning.
The distribution mode of the morpheme considers the dis-
tribution of the morpheme units in sequence, class, and matrix.
The distribution of morphemes in sequences is handled in the
tagmemic grammar. The distribution of morpheme units into
classes and in matrices within the classes, belongs to the para-
digmatic relations of the lexicon.
 LANGUAGE DESCRIPTION 183

Morphemic Systems. Morphemes are grouped into classes

and the classes are opposed to each other in the morphology of
the language. A morpheme marked as "noun" contrasts with all
nonnouns in the language; it also contrasts with all other nouns
within the noun class. This mode is "composed of relational com-
ponents, including class membership" (see Pike, 1967:85).
The tagmemic lexicon is the semantic system of the language
description; there is no semantic component divorced from the
lexical (and grammatical) systems. Although this system has
received little attention to date, Longacre suggests the need for
a ''highly-sophisticated lexicon" (1964:8) in order to generate
both grammatical and sensible sentences. The detailing of
semantic substrata need not be done by semantic features in
isolation. Just as phonemes are units composed of a bundle of
features, so the semantic features can be treated as bundles of
features, and the morpheme classes broken down into sub-
classes.; The semantic feature approach parallels the use of se-
mantic subclasses:

± Count ± Human ± Common

count noun human noun common noun
mass noun nonhuman noun proper noun

The features chosen as norms for subdivision into classes would

be only those which represent grammatical features of the lan-
guage, and only those grammatical features which are not
overtly marked by other formal means. The subclass categories
must be selective, not inflective, categories.
Lexical Fields. The distribution mode of the emic unit of the
lexicon may be viewed as the distribution of lexical units in
classes within a lexical field. The total lexicon may be viewed as
the network of relationships between the forms of language. The
language system is like a net that casts its shadow upon a blank
screen called the purport (Hjelmslev, 1963:57), which represents
all of reality. It is the lexical system that divides up reality ac-
cording to the mind of the native speakers. The lexical units are
the functioning parts of a language system. In the field view the
morphemes of language are seen as explicitly contrastive.
Not only the whole lexicon, but individual parts of the lexicon,
such as a single class, can be set up as a separate matrix. Some
work has been done by lexicographers in establishing lexical
fields: for example, within pronominal systems, kinship terms,
tense-aspect contrasts, and so on. Languages with first person
inclusive and exclusive, and with dual and plural numbers, have
a different pronominal system from that used in English. For
example, the Mundari language has a system of one inanimate
 184 INTRODUCTION TO TAGMEMIC ANALYSIS

and 11 animate pronouns. The pronoun we has four equivalents,

using dual/plural, inclusive/exclusive features.
aling, 'we two, you (the hearer) and I'
alang, 'we two, he (not the hearer) and I'
abu 'we plural, you (the hearers) and I'
ale 'we plural, they (not the hearers) and I'
Once lexical systems are established for a language, the rela-
tions of class to class, and the associative relations within the
class are evident. The choice, for example of one of the above
pronouns in a grammatical slot, is the implicit exclusion of all
the other pronouns of the language system.
Lexical Symmetry. The distribution mode of the lexical emic
unit places each morpheme within a paradigmatic system which
is generally symmetrical for the language. The lexical units con-
stitute a system of oppositions, of class to class, and of units
within each class. Forms are arranged in contrastive sets or
within specific word class paradigms. The principle of symmetry
is used to isolate morpheme units:
1. The Principle of Symmetry. Morphemic systems tend to-
wards symmetry, or pattern congruity. The range of
each morpheme should be nonunique (Hockett, 1958:
275); it should fit the emerging grammatical patterns of
the language. According to this principle, we group
/gow/ and /wen-/ as allomorphs of a single morpheme for
together they make a verb paradigm. Without this
grouping, /gow/ would have no past, and /wen-/ would
have no present.
2. The Principle of Economy. Morphemic systems tend to
be as economical as possible. In applying this principle,
simplicity of syntax takes precedence over morphological
simplicity. For example, it is syntactically convenient to
speak of one plural in English, even though this means
grouping widely disparate forms. Likewise, the grouping
of / and me as allomorphs fits the pronouns into the
noun paradigm and eliminates the objective case from
English.

THE SYNTACTIC COMPONENT

The third component of a language description deals with the
patterns of arrangements —with arrangements as manifested in
the forms of the language, arrangements in their identifying-
contrastive features, and arrangements as part of a syntactic
system. The minimum unit of this component is the tagmeme,
 LANGUAGE DESCRIPTION 185

considered in its manifestation mode, feature mode and distri-

bution mode. The tagmeme is treated under these three modes
by Pike (1967, Chap. 7).

The Manifestation Mode: Tagmatics

In the discovery process, we begin the study of grammar with
the study of tagmatics, which is the isolation of the etic slot:class
correlatives, the tagmas. These tagmas are then grouped as allo-
tagmas of the emic units, the tagmemes. In the presentation of
the results of analysis, the etic units constitute the manifesting
set of function-form units and make up the manifestation mode
of the tagmeme.
Manifesting Set of Allotagmas. In language description we
set up a series of tagmemic formulas at the natural levels of the
sentence the clause, the phrase and the word. The formulas are
composed of tagmeme units. Each tagmeme is manifested by a
set, containing one or more slot:class correlatives, called tagmas.
They may vary as follows (see also pp. 20-21):
1. Form Variants. Allotagmas of a tagmeme may differ
from each other in form alone. Since the tagmeme is a
slot:class correlation, the slot may remain constant but
manifest different filler classes. Each filler class repre-
sents a different allotagma of the same tagmeme. Thus
the subject tagmeme S:N/pn has two variants S:N and
S:pn, with different fillers.
2. Positional Variants. Allotagmas of a tagmeme may
differ from each other in position alone. Since the slot:
class unit has a distribution in the language, any change
of position, without corresponding change of form and
without change in grammatical meaning, is a separate
allotagma. In a language with SPO and POS order the
subject has two positions in the string, and these con-
stitute two positional variants of the tagmeme.
3. Meaning Variants. Functional meaning of the tagmemic
slot is the main identifying characteristic of the tag-
meme and should be kept constant. In a formal analysis,
where the generic meaning of the tagmeme is the same
and the manifesting form and position of the tagmeme
are constant then there may be different shades of
meaning within the same tagmeme. In order to avoid
being misled by a priori categories, the analyst must
consider the outward form, position, and type of con-
struction as the main clues to the identity of the tag-
meme.
 186 INTRODUCTION TO TAGMEMIC ANALYSIS

Allotagmas as Waves. The manifestation mode of the tag-

meme should show wavelike characteristics. However in the
syntactic part of the grammar, we are at a higher level of ab-
straction. The tagmeme is manifested by morpheme sequences;
and the morphemes are, in turn, reduced via morphophonemics
to phoneme sequences. Whenever the morpheme boundaries
overlap, the tagmemes also overlap. But if we restrict tagmemic
overlap to clear-cut sequences of morphemes, the only overlap
is one of function not of form. In those cases where the same
morpheme sequence has more than one function, or where there
is a portmanteau representation of two functions in one form
the wavelike characteristics of the grammatical unit are evident.
In the sentence John asked her to go to the dance tonight the
form her simultaneously manifests the object tagmeme, with
respect to the verb ask, and a subject tagmeme, with respect to
the verb go. In a portmanteau representation, such as the verb
form won't = will not the auxiliary and the negative tagmeme
are manifested in a single form. Grammatical constructions are
represented as waves (Pike, 1967:1-14), with attention to an un-
changed wave peak or nucleus, and with deletion or severe re-
strictions on form and meaning at the edges of the wavelike
structure, occurring at all levels of the grammar.

Allotagmatic Variation. The manifestation mode of the tag-

meme is characterized by the noncontrastive distribution of the
allotagmas which are its manifesting set. A judgment must be
made as to when two tagmas are the same (belong to one tag-
meme) and when they are different (belong to different tag-
memes). The main differences to be considered are: (1) differences
in slot meaning, (2) differences in filler class, and (3) differences
in position within a construction, or occurrence in different con-
structions.
1. Grammatical Similarity. Two tagmas with distinct slot
meaning and distinct filler classes belong to different
tagmemes, for example, S:N and P:tV. Tagmas with
similar functional meaning and similar filler classes
may belong to the same tagmeme, for example S:N
and S:pn.
2. Complementation. Two tagmas which never contrast
that is which never occur within the same construction
may be in complementary distribution if one occurs in
one construction and another in a different construction
with similarity of slot meanings and some similarity of
filler class.
Following this principle of complementation it would seem
that all subjects in English could be reduced to a single subject
 LANGUAGE DESCRIPTION 187

tagmeme, with the meanings subject-as-actor, subject-as-goal,

and so on, depending upon the construction in which it stands.
Pike, however, seems to prefer to call each different subject type
a separate tagmeme in English (1967:196). In this view, the many
subject tagmemes would constitute a class of subject tagmemes,
which are analogously the same (see also, "Tagma," 1967:219;
"Homophonous T&gmas," 1967:231).
The Feature Mode: Tagmemics
In the discovery process, we first isolate the etic slot:class
units or tagmas; the next step is to group these into emic units
the tagmemes. T&gmatics is a cutting process, the isolation of
the first approximations to the grammatical units. Tagmemics
is a grouping process, which involves human judgment, an at-
tempt to group tagmas into units essential to the language, as
the language appears to the native speaker.
The1 tagmeme is defined as the correlation of a grammatical
function, or slot, with the list of mutually substitutable items
that fill that slot (Elson and Pickett, 1962:57). It is a slot:class
correlation. The grammatical function is represented by the
slot symbol and irefers to the grammatical meaning attached
to a functional position in a frame. The filler class refers to all
the morphemes and morpheme sequences that may be used to
fill that slot in the construction frame. Both form and function
are explicit in the concept of the tagmeme.
Slot, or function, refers to such notions as subject, predicate,
object, and agent,iwhich, in transformational grammar, are called
grammatical relationships. Filler class refers to such notions as
noun, noun phrase, verb, transitive clause, and so on, which in
transformational r grammar, are called grammatical categories
(see Chomsky, 1965:64). In the complex notation of tagmemics,
for example, S:N^ function and form are never confused. Func-
tional (relational) symbols are written to the left of the ratio
mark (:); form (category) symbols are written to the right.
Distinctive Features of Tagmemes. The distinctive features
of the tagmeme are the identifying-contrastive features of the
unit. These are the features of meaning, form, and distribution
that is, (1) functional meaning of the slot, (2) the list of forms
filling the slot, and (3) the position of the tagmeme within a
given construction. We ask:
What is it? a noun phrase (filler class)
What does it do? it acts as subject (functional slot)
Where does it occur? at clause level (distribution)
All three of these features are used to identify the tagmeme, al-
 188 INTRODUCTION TO TAGMEMIC ANALYSIS

though within any one feature there may be variation. Each tag-
meme will contrast with others in the system, according to these
contrastive features.
The functional meaning of the tagmeme is vaguely denned
and this is an advantage in initial analysis. Subject is an intui-
tive notion; but it is opposed to predicate, object, location and
so on, which occur in the same construction. Subjects which occur
in different constructions are only the same by analogy: active
subjects are to active predicates as passive subjects are to pas-
sive predicates. The relation is the same; the meanings of the
various subjects need not be univocal.
Tagmemes as Particles. The feature mode of the slot:class
correlative establishes the tagmeme as a particle. Tagmemes
are "structurally significant points within a given pattern"
(Longacre, 1964:17). In this static view, there is no overlapping,
but a clear-cut abstracted view of definite points in the construc-
tion pattern filled by definite filler classes. Further, the patterns
occur at levels, so that we may speak of clause level tagmemes
or phrase level tagmemes.
The points within the construction pattern are grammati-
cally significant, in that they assign grammatical meaning to
fillers of the slot. The grammatical meaning is determined by
subtracting the lexical meaning of occurring items from the total
linguistic meaning. This meaning is associated with the slot
not with the particular lexical item. By this method we isolate
such meanings as subject, predicate, object, and so on at the
clause level, or head and modifier at the phrase level.
There is correlativity between the slot meaning isolated and
the meanings of the other slots in the same string. Point and
pattern are correlatives; there is no construction without con-
stituent tagmemes, and no tagmemes independent of the con-
struction in which they occur. The "subject" is subject because it
is opposed to the predicates, objects, and adverbial adjuncts
within the same string.

Tagmemic Contrast. The feature mode of the tagmeme unit

is also characterized by the fact that the tagmemes of the gram-
mar are in contrast with each other. These are units of arrange-
ment which carry their own contrastive grammatical meanings
in the language. These grammatical meanings must be under-
stood by anyone using the particular language.
1. Grammatical Similarity (Negative Norm). Dissimilar
tagmas are not grouped together into the same tag-
meme. There must be some foundation of similarity, in
form or function before consideration is given to re-
ducing two tagmas to the same tagmemic unit.
 LANGUAGE DESCRIPTION 189

2. Principle of Contrast. Tagmemes contrast within the

grammar. First, tagmemes at different levels are dis-
tinct, and not comparable. Second, within a given level,
tagmemes are in contrast with each other within the
same construction string. The fact of co-occurrence of
two similar tagmas in the same string is prima facie
evidence that there are two tagmemes present, al-
though it is possible, in some languages, to have rep-
etitions of a tagmeme within the string. Third more
subtle differences of meaning in the same tagmeme may
require separation of one generic functional notion into
two or more tagmemes, if there are other differences.
If formal differences are minimal and the functional
differences are analogous, the variants may remain as
allotagmas of the one tagmeme.

The Distribution Mode: Tagmemic Systems

In the discovery process, the tagmas, or initial slot:class
units, are first isolated. Then those tagmas which are func-
tionally the same are grouped into tagmeme units. Finally, the
tagmeme units are viewed as arranged in a systematic way
within the grammar. This system involves: (1) the unit, the tag-
meme, as a slot:class correlative; (2) the construction, or syntag-
meme, which is a string of tagmemes; and (3) the levels of the
grammar at which constructions occur.
The systematic distribution of the tagmeme occurs within
the larger framework of grammatical space called the grammati-
cal hierarchy. Each tagmeme is seen against the background of
the total grammar. The tagmemes are arranged like the beads on
an abacus, or counting board. The beads are the units. These
units are arranged on wires to form strings. The strings are
arranged from higher to lower within a frame. To locate a par-
ticular bead, one must specify which level, which string, which
unit. Likewise, the tagmemic unit is part of a grammatical
system in which the level, construction, and place within the
construction determine its place in the total system.
Tagmemic Systems. The distribution mode of the tagmeme is
its position within the tagmemic system. The system in which
the tagmeme is distributed is the total grammar of the language.
The tagmeme unit forms part of a set of oppositions which com-
prise the whole system. This set of oppositions is similar to
phonemic and lexical systems, but there are differences:
1. The tagmeme unit is a correlation of function and set
and of slot and filler class. Just as phonemes differ by
one or many distinctive features, so tagmemes may dif-
 190 INTRODUCTION TO TAGMEMIC ANALYSIS

fer by function or form or both. Function and form are

correlative and inseparable. Forms do not occur in
grammar without having a specific grammatical func-
tion; and functions do not occur in grammar without
forms to manifest that function.
2. The tagmeme construction is a string of tagmemes. The
construction is a pattern; the tagmemes are the con-
trastive points in the pattern. This is a second correla-
tivity. The tagmeme exists only in the string, and the
string does not exist without its component tagmemes.
The set of oppositions is between tagmemes within the
same string.
3. Constructions occur at well-defined levels in the gram-
mar; the level specifies the tagmeme as a clause level
tagmeme, occurring in a clause level construction. Sub-
ject at word level and subject at clause level are distinct
tagmemes, not by grammatical function, but according
to the level of grammar at which they occur.
Tagmemic Fields. The distribution mode of an emic unit has
the characteristics of a matrix, or field. The field view is func-
tional and represents the units as functioning elements within
the total system. When the tagmeme occurs, it occurs as part of
a grammatical system, in explicit contrast with all other tag-
memes of the system.
Once the tagmeme is clearly defined as part of a field the
actual occurrences of the unit are easier to recognize. The tag-
meme will be clearly contrastive as to (1) the range of functional
meaning allowable within the tagmeme unit; (2) the type of
fillers allowable within the unit including special case markings
or recognizable sets of function words; and (3) the type of con-
struction in which the unit occurs and the level at which the
construction occurs.
To locate a specific tagmeme within the system, the tag-
meme should be at a definite rank and depth within the system.
Proceed to the proper level, go to the construction at the proper
layer within that level, and follow the tagmemic string to the
given tagmeme. It is fixed in position at a definite level in the
hierarchy and at a special position within the level.
Tagmemic Symmetry. Language systems tend toward sym-
metry at the phonological and lexical levels. They also tend to-
ward symmetry in grammar. Tagmemic constructions may be
set up as matrix displays, showing elements of constructions as
explicitly contrastive. Construction types may be charted in vari-
ous dimensions, as was done at sentence and clause level to
determine whether there are gaps within the overall system.
 LANGUAGE DESCRIPTION 191

The object of the language description is to reveal the facts

of the grammatical structure of unknown languages in the
simplest and most economical way. We have assumed five levels
to be "natural" to language, but a particular language might
require more or fewer levels. Within these levels we expect rela-
tional, coordinate and subordinate structures, yet all strata
might not be represented. The language, for example, might
lack an inflectional system. Finally, in the individual construc-
tions the order of elements is as we find it. The individual tag-
memes are specified functionally according to the range of gram-
matical meaning allowed for the language under consideration;
the filler classes that manifest these functions are those forms
that actually occur fulfilling this function. In the overall lan-
guage analysis, the symbols for function and form the means
for combining these in constructions, and the methods for order-
ing constructions at levels from higher to lower are present in
the theory. It is up to the individual analyst to so use the tech-
niques provided in tagmemics as to describe accurately a given
language, and to so apply tagmemic theory as to reveal in the
simplest possible way, the structure of that language.

PRACTICE 7: WORKING FROM A GIVEN CORPUS

For the following sentences adapted from Problem #165, Sierra
Popoluca, (Laboratory Manual, Merrifield, 1967):
1. Set up the grammar for the language, including formulas at
the four levels of sentence clause, phrase, and word. Include
any transformation rules, restrictions, or assumptions that
are necessary.
2. Set up the lexicon of the language, including all the mor-
phemes which occur in the data. For morphemes which have
variants list the morpheme variants and summary rules for
the morphophonemics of the language.

Procedures for the Grammar

1. Sentence Level. Final intonation assumed for each utter-
ance #1-75.
2. Clause Level. The following clause level types occur:
a. Statements—transitive
— intransitive
— equational
b. Commands—transitive
— intransitive
 192 INTRODUCTION TO TAGMEMIC ANALYSIS

3. Phrase Level. The following phrase types occur:

a. Head noun and modifiers
{he?m} 'that,' {turn} 'a,' {ha?yang} 'much.'
b. Head adjective and modifier
{ca-m} 'very.
4. Word Level. The following composite word types occur:
a. Noun stem with possessive.
b. Verb stem with subject, tense, and aspect.
c. Imperative verb stem with mood marker.
d. Noun stem with adverbializers expressing loca-
tion or manner.

Restrictions and Assumptions

1. Subject Permutability. The subject is movable in tCl and

iCl. It may occur before or after the predicate-object
complex.
2. Adjunct Permutability. The adjuncts, LMT, are freely
movable. They are optional and occur in any order after
the SPO complex.
3. Negative Transformation. Affirmativey statements be-
come negative by the addition of {d a} 'not'; negative
commands take {odoy} 'don't.'
4. Cross-Functional Usage. Some words have two uses:
a. Indeclinable nouns are used as locationals
{playa} 'beach' and {atebet} 'Soteapan.
b. The locational adverb {huuma} 'far away is
used in the predicate attribute position with
the adjective class.

Lexicon and Morphophonemics

1. The Lexicon may be set up as a classified lexicon.

a. The noun system, including nouns, pronouns,
determiners, adjectives, intensifiers and noun
suffixes.
b. The verb system, including verb stems, tense
aspect, and mood; and subject pronouns, nega-
tives introducers, locatives, and temporals.
2. Morphophonemics. Some of the morphemes have variant
forms.
a. Morphemes with variants are listed in the lexi-
con in braces{ }.
b. Allomorphs are listed for each of these in the
lexicon, with a statement of the conditioning
environment.
 LANGUAGE DESCRIPTION 193

3. Summary Rules. The1 principal changes can be summa-

rized in the following rules:
a. Assimilation of prefix-final nasals, except palatal
nasals.
b. Loss of stem final vowel in nouns, loss of prefix
initial /h-/.
c. Palatalization of initial consonants, /t,n/ after
prefixes with high front vowels.
d. Vowel changes in the imperative form of the
verb stem.
Maximum Generative Potential
The maximum generative potential (MGP) can be reduced by
these restrictions:
1. Subject and predicate have person and number concord.
2. Equational clauses require a noun phrase which in-
cludes {he?m},'that.'
3. Manner adverbs in {-yukma}, and {-ma}, are in com-
plementary distribution, with one form for transitives,
one for intransitives.
Even with these restrictions the restricted generation potential
(RGP) is over 280 billion sentences, excluding all permutations.
The combinations of the adjuncts, LMT alone account for
40,000 possibilities with each transitive and intransitive verb
form, whether statement or command. Since the subject and
object slots are filled by almost 250 forms, the output is neces-
sarily large.
Conclusion. The modes of procedure are given for one par-
ticular problem to indicate the information which is sought in a
particular language corpus. First, sort the corpus according to
sentence and clause types. This reduces the present problem of
75 sentences to a set of five different problems dealing with tCl
i d , eqCl, C*tCl, aind C-iCl. Second, each of these 5 isolated sets of
data is solved for grammar and lexicon, then the grammars and
lexicons are grouped in one solution. Restrictions across clause
types, for example {he?m} above, should be noted. Third mor-
phemes with variants are checked, and summary rules are con-
structed to account for all the morphophonemic changes.

SUPPLEMENTARY READINGS 7
Cook, Walter A., S;J., On Tagmemea and Transforms. Washington, D.C.,
Georgetown University Press, 1964. For grammatical hierarchy,
10-12; particle, wave, and field, 12-14; and discussion, 27-39.
 194 INTRODUCTION TO TAGMEMIC ANALYSIS
Conklin, Harold C, "Lexicographical Treatment of Folk Taxonomies."
UAL, 28, Part IV (1962). Problems in Lexicography, 119-141.
Methods for establishing: distribution mode for lexical units.
Longacre, Robert E., Grammar Discovery Procedures. The Hague,
Mouton & Co., 1964. "Introduction: The Notion of Grammar," for
the triple hierarchy of "semiautonomous and interlocking" sys-
tems.
, "Prolegomena to Lexical Structure." Linguistics, An Interna-
tional Review, 5:5-24. The Hague, Mouton & Co., 1964. For a differ-
ent viewpoint on field structure, with a matrix showing phonology,
lexicon, and grammar, opposed to particle, string, field.
Pike, Kenneth L., Language in Relation to a Unified Theory of the Struc-
ture of Human Behaviour: The Hague, Mouton & Co., 1967. Lan-
guage unit, with feature, manifestation, distribution modes.
Chap. 6, Morpheme 6.4 (154) 6.5 (163) 6.6 (169)
Chap. 7, Tagmeme 7.3 (196) 7.4 (228) 7.5 (236)
Chap. 8, Phoneme 8.3(293) 8.4(306) 8.5(318)
."Language as Particle, Wave, and Field," The Texas Quarterly,
2:37-54 (1959). Original presentation of particle, wave, field theory
showing that the three views may be reconciled in single units.
, "Grammar as Wave." Monograph Series on Languages and
Linguistics, No. 20, 1-14. Washington, D.C., Georgetown Uni-
versity Press, 1967. Applied to Kasem language of Ghana.
, "Non-linear Order and Anti-Redundancy in German Morpho-
logical Matrices." Journal of Dialectology, 32 (1965), Wiesbaden.
For application of matrix theory in the lexical component.
, "Tagmemic and Matrix Linguistics Applied to Selected African
Languages." HEW Contract No. OE-5-U-065 Report (1966).
Sapir, Edward, Language: An Introduction to the Study of Speech. New
York, Harcourt, Brace and World, Inc., 1921. Chap. 2, "The Ele-
ments of Speech (Lexicon)," and Chap. 3, "The Sounds of Language
(Phonology)."
SUPPLEMENTARY EXERCISES

The problems suggested as supplementary exercises are from

the Laboratory Manual for Morphology and Syntax by William
R. Merrifield, Constance M. Naish, Calvin R. Rensch and Gillian
Story, published by the Summer Institute of Linguistics, Santa
Ana, California, 1967 Revision.

chapter 2: Sentence Level

This is beat illustrated by problems involving all levels of structure,
with some mixed clause types, commands, and nested construc-
tions (to follow Chapters 3, 4, and 5).
#96. Mezquital Otomi
#97. Michoacan Aztec
#106. Coatlan Mixe
#107. Turkish

chapter 3: Clause Level

This is best illustrated by problems that require sentence and
clause level formulas, but require no formulas at the level of the
phrase or wordi
#92. Mixtec Of San Miguel
#93. Awa of New Guinea
#94. Lotuko of Africa
#95. Palantla Chinantec

chapter 4: Phrase Level

This is best illustrated by problems that require sentence, clause,
and phrase level formulas, but require no formulas at the word
level.
#98. Palantla Chinantec
#99. Vietnamese

195
 196 INTRODUCTION TO TAGMEMIC ANALYSIS

#100. Apinaye of Brazil

#102. N. Puebla Totonac

chapter 5: Word Level

This is best illustrated by problems that require sentence, clause,
phrase, and word level formulas. Introduction of concord ties in
formulas.
#101. Chontal of Oaxaca
#103. Otomi of Mexico
#104. Western Warburton
#105. Hausa of Africa

chapter 6: Morpheme Level

This is best illustrated by problems that require sentence, clause,
phrase, and word level formulas. Introduction of relater-axis
phrases.
#108. Mixtec of San Miguel
#109. Pocomchi
#110. Guajarara
#111. Mezquital Otomi

chapter 7: Language Description

This is best illustrated by problems with 75 or more sentences,
involving many clause types; this is suggested as a review of all
the tagmemic procedures.
#164. Cashinahua
#165. Sierra Popoluca
TAGMEMIO SYMBOLS

aj adjective word class 3

Aj adjective phrase group with adjective head
ajzr adjectivizer derivational affix
App: appositive slot in item-appositive phrase
asp: aspect slot at word level, verbs
aspm aspect marker word level affix
aux auxiliary verb subclass of verbs
Aux: auxiliary slot at phrase level, verb
phrase
av adverb word class 4
Av adverb phrase group with adverb head
avzr adverbializer derivational affix
Ax: axis slot in relater-axis phrase
Base: base slot at sentence level
connector function word
C: connector slot at levels above the word
core: core slot at word level
det determiner function word
Det: determiner slot at phrase level, noun
phrase
eqCl equational clause clause type, equational
verb
eqy equational verb subclass of verbs
H; head slot in head-modifier phrase
i introducer function word
I: introducer slot at clause level
IA item-appositive phrase multiple head, one
IC1CI GUI'
IGF final intonation contour suprasegmental
iCI intransitive clause clause type, intransitive
verb
Id: identifier slot at phrase level for clauses

197
188 INTRODUCTION TO TAGMEMIC ANALYSIS

int lntensifier function word

Int: intensifier slot at phrase level
Into: intonation slot at sentence level
It: item slot in item-appositive phrase
IV intransitive verb subclass of verbs
iV intransitive verb phrase group with intransitive
verb head
1VS intransitive verb stem stem of intransitive verb
L: location slot at clause level, adjunct
loc locational adverb subclass of adverbs, place
M: manner slot at clause level, adjunct
Marg: margin slot at sentence level
md: mood slot at word level, verbs
mdm mood marker word level affix
Mod: modifier slot in head-modifier phrase
n noun word class 1
N noun phrase group with noun head
Nco coordinate noun phrase multiple head, many
referents
ns noun stem stem of nouns
neg negative function word
Neg: negative slot at phrase level, verbs
nnuc: noun nuclear slot at word level, nouns
nom nominalizer derivational affix
np proper noun subclass of nouns
num numeral subclass of adjectives
num: number slot at word level, sg/pl
numm number marker word level affix
o: object slot at word level
0: object slot at clause level
om object marker word level affix
P: predicate slot at clause level
PA: predicate attribute slot at clause level, equational
clause
pn pronoun substitute, bound or free
poa: possessive slot at word level
Pos: possessive slot at phrase level
posm possessive marker word level affix
Qn: quantifier slot at phrase level, nouns
R: relater slot in relater-axis phrase
RA relater-axis phrase exocentric phrase type
rel relater function word
s: subject slot at word level
S: subject slot at clause level
Sent sentence utterance with final
intonation
sm subject marker word level affix
t: tense slot at word level, verbs
T: temporal slot at clause level, adjunct
tci transitive clause clause type, transitive
verb
tern temporal adverb subclass of adverbs
 TAGMEMIC SYMBOLS 199

tm tense marker word level affix

tv transitive verb subclass of verbs
tV transitive verb phrase group with transitive verb
head
tvs transitive verb stem stem of transitive verb
vbzr verbalizer derivational affix
vnuc: verb nuclear slot at word level verbs
SELECTED BIBLIOGRAPHY

Allen, Robert Livingston, The Verb System of Present-Day American

English. The Hague, Mouton & Co., 1966.
, "Sector Analysis: From Sentence to Morpheme in English."
Monograph Series on Languages and Linguistics No. 20, 159-174.
Washington, D.C., Georgetown University Press, 1967.
Becker, Alton L., "Conjoining in a Tagmemic Grammar of English."
Monograph Series on Languages and Linguistics No. 20, 109-121.
Washington, D.C., Georgetown University Press, 1967.
Belasco, Simon, "Tagmemics and Transformational Grammar in
Linguistic Analysis." Linguistics, 10:5-15 (1964).
, "The Role of Transformational Grammar and Tagmemes in
the Analysis of an Old French Text." Lingua, 10:375-390 (1961).
Blansitt, Edward Lee, The Verb Phrase in Spanish: Classes and Rela-
tions. Ph.D. Dissertation, University of Texas, 1963.
Bloomfield, Leonard, Language. New York, Holt, Rinehart and Winston,
Inc., 1933.
Boas, Franz, Introduction to the Handbook of American Indian Lan-
guages. Washington, D.C., Georgetown University Press 1963.
Bochenski J. M., The Methods of Contemporary Thought. Dordrecht,
Holland, D. Reidel Publishing Co., 1965.
Brend, Ruth M., A Tagmemic Analysis of Mexican Spanish Clauses. The
Hague, Mouton & Co., 1968.
Chafe, Wallace L., "Review of Longacre's Grammar Discovery Proce-
dures 1964." Language, 41:640-647 (1965).
Chomsky, Noam, Syntactic Structures. The Hague, Mouton & Co., 1957.
, Aspects of the Theory of Syntax. Cambridge, Mass., The M.I.T.
Press, 1965.
Conklin, Harold C, "Lexicographical Treatment of Folk Taxonomies."
UAL, 28:2, IV: Problems in Lexicography, 119-141 (1962).

200
 SELECTED BIBLIOGRAPHY 201
Cook, Walter A,,: On Tagmemes and Transforms. Washington, D.C.,
Georgetown; University Press; 1964.
, A Deseiiptive Analysis of Mundari. Ph.D. Dissertation,George-
town University, 1965.
, ""Hie Generative Power of a Tagmemic Grammar." Monograph
Series on Languages and Linguistics, No. 20, 27-41. Washington,
D.C., Georgetown University Press, 1967.
DeSaussure, Ferdinand, Course in General Linguistics, trans, by Wade
Baskin. Geneva, 1916. New York, Philosophical Library, Inc., 1959.
Dinneen, Francis; P., S.J., An Introduction to General Linguistics. New
York, Holt, Rinehart and Winston, Inc., 1967.
Elson, Benjamin, and Pickett, Velma, An Introduction to Morphology
and Syntax, Santa Ana, Summer Institute of Linguistics, 1964.
Frllmore, Charles J., "The Case for Case." Universal in Linguistic
Theory, eds.| Emmon Bach and Robert Harms. New York, Holt,
Rinehart and Winston, Inc., 1968.
Fodor, Jerry A., and Katz, Jerrold J., The Structure of Language: Read-
ings in the Philosophy of Language. Englewood Cliffs, N.J., Pren-
tice-Hall, Inc., 1964.
Fries, Charles C.,The Structure of English. New York, Harcourt, Brace
and World, Inc., 1952. \
, Linguistics and Reading. New York, Holt, Rinehart and Win-
ston, Inc., 1962.
Gleason, Henry A., Jr., An Introduction to Descriptive Linguistics.
New York, Holt, Rinehart and Winston, Inc., 1955; revised, 1961.
, Linguistics and English Grammar. New York, Holt, Rinehart
and Winston, Inc., 1962,
, "Contrastive Analysis in Discourse Structure." Monograph
Series on Languages and Linguistics, No. 21, 39-63. Washington,
B.C., Georgetown University Press, 1968.
Greenfeerg, Joseph H., Universals of Language. Cambridge, Mass., The
M.I.T^Press> 1963.
Halliday, M. A. K|., "Categories of the Theory of Grammar." Word, vol.
17(1961).
, Mclntosh, Angus, and Strevens, Peter, The Linguistic Sciences
and Language Teaching. London, Longmans Green, Ltd., 1964.
, "Syntax iand the Consumer." Monograph Series on Languages
and Linguistics, No. 17,11-14. Washington, D.C., Georgetown Uni-
versity Presfe, 1964.
Harris, Zellig, String Analysis of Sentence Structure. The Hague,
Mouton & Co., 1962.
Hjelmslev, Louis, Prolegomena to a Theory of Language, trans, by
Francis J. Whitfield. Madison, Wise, The University of Wisconsin
Press, 1963.
Hockett, Charles F., "Problems of Morphemic Analysis." Language,
23:321-343 (1947).
 202 INTRODUCTION TO TAGMEMIC ANALYSIS

"Two Models of Grammatical Description." Word, 10:210-231

(1954).
A Course in Modern Linguistics. New York The Macmillan
Company, 1958.
, "Linguistic Elements and Their Relations." Language, 37:
29-53 (1961).
Huddleston, R. D., "Rank and Depth (in Scale-and-Category Grammar)."
Language, 41:574-586 (1965).
Joos, Martin, The English Verb: Form and Meanings. Madison, Wise,
University of Wisconsin Press, 1964.
Katz, Jerrold J., and Postal, Paul M., An Integrated Theory of Linguistic
Description. Cambridge, Mass., The M.I.T. Press, 1964.
Kemeny, John G., A Philosopher Looks at Science. Princeton, N.J.,
D. Van Nostrand Company, Inc., 1959.
Koutsoudas, Andre, Writing Transformational Grammars. New York
McGraw-Hill, Inc. 1966.
Lado, Robert, Linguistics Across Cultures. Ann Arbor, Mich., University
of Michigan Press 1957.
, Language Teaching: A Scientific Approach. New York, McGraw-
Hill, Inc., 1963.
Law, Howard W., "The Use of Function-Set in English Adverbial Classi-
fication." Monograph Series on Languages and Linguistics, No. 20,
93-102. Washington, D.C., Georgetown University Press, 1967.
Lees, Robert B., "The Grammar of English Nominalizations." UAL,
26:3,11:3,1-205 (1960); fourth printing, The Hague, Mouton & Co.,
1966.
Liem, Nguyen Dang, English Grammar: A Combined Tagmemic and
Transformational Approach, A Contrastive Analysis of Vietnamese
and English. Vol. I, Linguistic Circle of Canberra, 1966.
Lonergan, Bernard J., S.J., Insight: A Study of Human Understanding.
New York, The Philosophical Library, Inc., 1958, reprint, 1961.
Longacre, Robert E., "String Constituent Analysis." Language, 36:
63-88, (1960).
, Grammar Discovery Procedures. The Hague, Mouton & Co.,
1964.
"Prolegomena to Lexical Structure." Linguistics, 5:5-24
(1964 b).
, "Some Fundamental Insights of Tagmemics." Language, 41:
65-76 (1965).
, "The Notion of Sentence." Monograph Series on Languages
and Linguistics, No. 20,15-25. Washington, D.C., Georgetown Uni-
versity Press, 1967.
, "Reply to Postal's Review of Grammar Procedures." UAL,
33:323-328 (1967).
and Williams, Ann F., "Popoloca Clause Types," Ada Linguis-
tica Hafniensia, 10:161-186 (Copenhagen, 1967).
Marchand, Hans, The Categories and Types of Present-Day English
Word Formation. Wiesbaden, Otto Harrassowitz, 1960.
 SELKCTED BIBLIOGRAPHY 203

Matson, Dan M., "Tagmemic Description of Agreement." Monograph

Series on Languages and Linguistics, No. 20,103-108. Washington,
D.C., Georgetown University Press, 1967.
Merrifield, William R., Laboratory Manual for Morphology and Syntax,
with Constance M. Naish, Calvin R. Rensch, and Gillian Story.
Santa Ana, Summer Institute of Linguistics, 1962; revised, 1967.
, "On the Form of Rules in a Generative Grammar." Monograph
Series on Languages and Linguistics, No. 20, 43-55. Washington
D.C., Georgetown University Press, 1967.
Morgan, James 0.* "English Structure above the Sentence Level."
Monograph Series on Languages and Linguistics, No. 20, 123-132.
Washington, D.C., Georgetown University Press, 1967.
Nida, Eugene A., "The Identification of Morphemes." Language, 24:
414-441 (1948).
, Morphology: A Descriptive Analysis of Words. Ann Arbor,
Mich., University of Michigan Press, 1949.
, Learning a Foreign Language. New York, Friendship Press,
National Council of Churches, 1957.
A Synopsis of English Syntax. Norman, Okla., Summer Insti-
tute of Linguistics, 1959; 5th edition, 1964.
O'Brien, Richard J.r S.J., A Descriptive Grammar of Ecclesiastical Latin.
Chicago, Loyola Press, 1965.
Pickett, Velma, The Grammatical Hierarchy of Isthmus Zapotec. Lan-
guage Dissertation 56. Baltimore, Waverly Press, 1960.
Pike, Kenneth L., Language in Relation to a Unified Theory of the Struc-
ture of Human Behaviour. Glendale, Calif., Summer Institute of
Linguistics, Part 1(1954), Part 11(1955), Part 111(1960); reprinted
in one volume, The Hague, Mouton & Co., 1967.
, "On Tagmemes nee Gramemes." UAL, 24:273-278 (1958).
, "Language as Particle, Wave and Field." The Texas Quarterly,
2:37-54,(1959).
, "Dimensions of Grammatical Structure." Language, 38:221-
244, (1962).
"A Syntactic Paradigm." Language, 39:216-230 (1963).
, "Discourse Analysis and Tagmemic Matrices." Oceanic Lin-
guistics, 3:5-25 (1964).
, "Non-linear Order and Anti-Redundancy in German Morpho-
logical Matrices." Journal of Dialectology, 32: 193-221 (1965).
, Tagmemic and Matrix Linguistics Applied to Selected African
Languages. HEW Report, No. OE 5-14-065 (November, 1966).
, "A Guide to Publications Related to Tagmemic Theory." Cur-
rent Trends in Linguistics, Vol. Ill, ed. Thomas A. Sebeok 365-
394. The Hague, Mouton & Co., 1966.
, "Grammar as Wave." Monograph Series on Languages and
Linguistics, No. 20, 1-14. Washington, D.C., Georgetown Univer-
sity Press, 1967.
Postal, Paul M., "Constituent Structure: A Study of Contemporary
Models of Syntactic Description." UAL, 30:2, III (1964).
 204 INTRODUCTION TO TAGMEMIC ANALYSIS

, "Review of Robert E. Longacre's Grammar Discovery Pro-

cedures." UAL, 32:93-98 (1966).
Sapir, Edward, Language: An Introduction to the Study of Speech.
New York, Harcourt, Brace and World, Inc., 1921.
Stageberg, Norman C, An Introductory English Grammar. New York
Holt, Rinehart and Winston, Inc., 1965.
Thorndike, Edward, and Lorge, Irving, The Teacher's Word Book of
10,000 Words. New York, Columbia University Press, 1944.
Trager, George L., and Smith, Henry Lee, Jr., An Outline of English
Structure. Studies in Linguistics, Occasional Papers No. 3. Wash-
ington, D.C., American Council of Learned Societies, 1957.
Waterhouse, Viola, "Independent and Dependent Sentences." UAL,
29:45-54(1963).
Warriner John E., English Grammar and Composition. New York,
Harcourt Brace and World, Inc. 1951; revised, 1965.
Weinreich, Uriel, Languages %n Contact: Findings and Problems. Lin-
guistic Circle of New York, 1953; reprinted The Hague, Mouton &
Co., 1963.
, "Lexicographic Definition in Descriptive Semantics." UAL,
28:2 IV: Problems in Lexicography, 25-43 (1962).
INDEX
abbreviated language, 59 Bochenski, J. M., 3, 200
active voice, 41-42, 47, 49, 51, 58
66, 72, 80 calls, 38 58
addition sentence, 56-57 case, 16, 94, 99, 103, 106, 109-110
adjectival, 6-7, 69, 76, 96 121-122, 125 148-149
adjective, 122-124,197 case grammar, 69
adjective compound, 135 Chomsky, N., 1, 5,24,41,43,60, 62,
adjective phrase, 107-108, 197 100, 112, 145, 146, 147, 148, 154
adjectivizer, 128-131,197 166, 187, 200
adverb 124,197 clause 65-66
adverb compound, 135 clause analysis, 79-86
adverb phrase, 107-108, 197 clause level, 65-86
adverbial, 6-7, 69, 76, 95 clitic, 30 98
adverbializer 128-431,197 closure, 4 23-24
affirmative, 41-42,47, 49,51,58,66, combinations, 5,144, 158-160
72 80 command, 42, 49, 50, 51, 52, 71-72
affix 118 competence, 1-3
agent, 27, 72 complex sentence, 39-46
agreement, 68,70,105,109-110,149 complex stem 130
-AL designation, 6, 69 component, 169-170
allomorph, 19, 85,168,178-180 compound sentence, 39-46
allophone, 19,168,171-172 compound word 131-138
allotagma, 19-20,168,185-187 computation, 82 158-164
appositive, 92, lOOi-101, 103, 105- computer generation, 156-158
106 concord, 68,83,99,103,105,109,149
aspect, 112, 125, 197 conflated tree, 152-153
assumption, 82-83 conjoining, 44 99-100
auxiliary, 107, 111-112,123, 197 connector, 45, 100 197
axis, 74-78, 93-99,197 construction, 21-26
contrast 168,175,181, 188
base, 12,38,40,44-47,52-54,56-57, contrast, variation, and distribu-
61, 86,113, 138,165, 197 tion, 170-171
Becker, A. L., 45,100, 114, 200 coordinate phrase 99-106
Bloomfield, L., 39, 4D, 48,57,62,83, coordinate strata 31-32
117, 118,139, 178, 200 core slot, 129, 135 197

207
 208 INDEX

cross-reference, 67, 105, 109, 149 greetings, 38, 58

guess-and-check method 4
deep structure, 11 68
deletion 55,105 Halliday, M. A. K., 30, 91, 201
dependent clause, 73-79 head-modifier phrase, 106-112
derivation 125-131 head slot, 103, 108-110
derived clause, 71-72 hierarchy, 27, 31 169
derived sentence 41-42, 47-50 Hjelmalev, L., 183, 201
DeSausaure, F., 1, 21, 201 Hockett, C. F., 39, 40, 92, 108, 114
determiner, 107, 197 122, 126, 139, 149, 177, 179, 184
dimension, matrix, 49-50, 69 201
Dinneen, F. P., 175, 201 Huddleston, R. D., 33,100,114 202
discourse analysis, 40 hypothesis, 3-4
discovery process, 4
distribution, 171, 175, 182,189 IC (immediate constituent) analy-
drama analogy, 68-69,106 sis, 7, 14 22
identifier slot 44, 62, 76, 98 105
ellipsis, 55, 78 imperative mood, 48, 50, 52 71
Elson, B., and Pickett, V., 14,15,19, independent clause 67-73
21, 27, 31, 36, 40, 54, 62, 65, 71,87, inductive cycle, 3-4
91, 97,104,107,110,112,114,117, infinitive, 78 96
124, 139, 187 201 inflection, 119-125
embedding, 44, 73 initial string, 46
emphatic clause, 72 inscriptions, 38 58-59
endocentric, 23, 92 134 insight, 3-4
equational clause, 50 70 intensifier, 108 198
etic-emic construction, 25-26 interjection, 58
etic-emic level 30-31 intonation, 12, 38-39,41,43,45-47,
etic-emic unit, 19-20 52-58 197
exclamatory sentence 57 intransitive clause, 50-54 64 66
69-70
feature mode, 170, 173, 180 187 introducer slot, 68 197
field structure, 31 intuition 3-4
filler class 16, 147-149 item slot, 103, 105, 111 198
Fillmore, C, J., 69, 201 item-appositive, 90, 103, 105-106
fixed order tagmemes, 18 197
focus, 72 item-possessive, 110-111
form, class, and gloss, 84 182-183
Fries, C. C., 6, 10, 32, 54, 62, 69, 87, Kemeny, J. G., 2-3, 202
95 101, 114, 139, 201 kernel sentence, 41-42 47-48
function words, 6 Koutsoudaa, A., 78,87,109,143,151,
functional model 5-9 166 202
functional slot, 15, 145-147
Lado R., 95, 96, 202
generative grammar, 157 language, 1
generation potential, 144, 159-164 language description, 169-191
Gleason, H. A., 62, 69, 87,114, 201 layering, 31, 67 98-99
governing derivation, 126-127, 130 level, 8, 27-34
government, 68, 94, 99, 149 level-skipping, 31, 97 99
grammar, 1-5 lexical component, 168 177-184
grammatical pattern, 143,159-164 lexical fields, 183
Greenberg, J. H., 87, 94, 201 lexicon, 83-86,177-184
 INDEX 209
Idem, N. D., 62, 71, 87, 202 notation, 9,197-199
linguistics, 1 noun, 6,121-122 124-125
location slot, 68, 76, 95,198 noun compound, 137-138
Longacre, R. E., 4, 5| 7,8,14,15,18, noun phrase, 106-107
22, 26, 27,28,29, 31,32, 34,36,39, noun subclasses, 108 180
40,44, 50,63,65,68,69, 73,74,79, nuclear slot 119-125
88 104, 145, 153, 158, 166, 169, nuclear tagmeme, 19, 26-27 67
170,183,188,194, 202
loopback, 31, 44, 67,119 object slot, 67, 96
Obligatory tagmeme, 17 25
major sentence, 38, 40, 47-54 O'Brien, R. J., 120, 122, 203
manifestation mode, 170-178, 185 open construction, 23 104
manifestations, 5, 144, 162-164 operator, matrix 50-51
manner slot, 68, 95,198 optional tagmeme, 17 25
margin slot, 45-47, 55, 61,198 order linear, 15,18, 20
marginal sentence, 55
matrix, clause, 69 paradigm, 122
matrix, sentence, 49 parallellism, 19 169-170
Matson, D. M., 110,114, 203 participle, 78-79
meaning, use of, 10 particle, 168,174,181 188
metalanguage, 59 particle, wave, and field, 168 170
Merrifield, W. R., 9, 14, 52, 73, 81, passive voice, 41-42, 47, 49, 51, 58
164 166,191,195-196,203 66,72 80
MGP (maximum generation po- pattern, 21-26 152
tential), 83,150,153,155,157-158 patterned dependency, 39
middle voice, 41-42,; 49, 66, 72,80 performance, 1-3
minor sentence, 38, 40, 54-60 peripheral tagmeme, 18 68
models of grammar, 5-8 permutations, 5,144,160-162
modification structure, 24,106-112 phone,19 171-173
modifier slot, 106-111, 198 phoneme, 168 173-175
mood slot, 72,121-122,198 phonemics 173-175
morph, 19,178-180 phonemic system, 175-177
morpheme, 149-151, 168, 180-182 phonetics 171-173
morpheme level, 143-164 phonological component, 168 171-
morphemics, 180-182 177
morphemic system, 182-184 phonology, lexicon, and grammar,
morphetics, 178-180 170
morphology, 177-184 phrase clitic, 30 98
morphophonemics, 85,192-193 phrase level, 91-112
mottoes 38, 58 phrase marker 152-153
movable order, 18,192 Pike, K. L., 4, 7,8,13,14,15,22,27,
multiple clause, 40,45-47 28, 30, 32, 36, 39,40, 50, 51, 69,92,
multiple head phrase, 99-106 110 169, 170, 171, 178, 181, 183,
185 186, 187,194 203
negative, 41-42, 49, 51, 66, 72, 80, portmanteau, 74, 77 186
108 111-112 possessive, 107, 110 198
Nida E. A., 127,140, 203 potential string, 22
nominal, 6-7, 69, 76, 97 practice and theory, 9-11
nominalizer, 128-131 predicate attribute slot, 66, 68-71
nonclause structure, 58 198
nonfinite verb, 55, 78-79 predicate slot, 65-77, 80-82, 198
nontypical language use, 59 preterminal string, 154-156
 210 INDEX

process rule, 85 suprasegmental, 45 137

pronoun, 105, 107, 123 symbols, 10, 197-199
purpose clause, 61 symmetry, 177, 184 190
syntactic component, 168,184-191
quantifier, 107 syntagmeme, 21-26
question, 42, 49-52, 53-54, 71-72 systems of words, 121
question marker 53
question word, 71 tagma, 19-20
tagmatics, 19, 185-187
readings, 5,145 tagmeme, 15, 168 187-189
readout, 82 tagmemics, 187-189
reciprocal voice, 49, 66, 72 tagmemic analysis, 13-34
recursive layering, 24 tagmemic model 8-9
reduction 43-45, 60-62 tagmemic system, 189-190
relater-axis clause 32, 77-78 temporal slot, 68, 76, 95 198
relater-axis phrase, 32, 93-99 tense, 123-124
relater class, 94-95 terminal string, 155
relational strata 32, 77, 93, 120 theory, 4 9-11
relative clause, 44, 75, 76, 77-78 Thorndike, E., 95, 140 204
repetition of tagmemes, 103-104 titles, 58
restrictions, 83,150, 163 Trager, G., and Smith, H. L., 6, 69,
restrictive derivation 126 129-131 204
RGP (restricted generation poten- transformations 42
tial), 83, 150 transformed sentence 48
roles in grammar, 68-69 transitive clause, 50 70
rule of two, 26-27, 69-70 tree diagrams, 142, 152-153 165
trimodal units 168, 169-170
Sapir, E., 1, 6, 118, 169, 194, 204 triple hierarchy, 170
sentence, 39-40 typical levels, 29
sentence generator, 156-158 typical mapping, 30 99
sentence level 39-60
sequence sentence, 54 units, 168 169-190
similarity, phonetic, 174 unity of construction 22
similarity, semantic, 179 universals 87 94
slot-class correlative, 13-20, 171-
176 variation, 168, 170, 172, 179 186
sorting clauses, 79 verb,122-125 199
sorting sentences, 41-46, 60-62 verb compound, 134
specialized language, 59 verb phrase, 111-112 199
statement, 48, 50-51 verbal 6-7, 69, 76 95
stem, 118 verbalizer, 128-131
stress 137 verification process, 4 11
string constituents, 7, 21 voice, 41-42, 47, 49, 51, 58,66, 72, 80
structural descriptions, 5, 11, 143
151-159 wave, 168, 170, 172, 175, 186
subject, 16, 19, 22, 67, 79-80, 167, word, 117-119
185-188 198 word level, 117-138
subordinate clause, 78 Warriner, J., 66, 114 204
subordinated clause 77
subordinate strata, 32, 127 zero morphs, 70 83-84