Some Principles of Linguistic Methodology

Author(s): William Labov

Source: Language in Society , Apr., 1972, Vol. 1, No. 1 (Apr., 1972), pp. 97-120
Published by: Cambridge University Press

Stable URL: https://www.jstor.org/stable/4166672

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Cambridge University Press is collaborating with JSTOR to digitize, preserve and extend access
to Language in Society

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
Lang. Soc. I, 97-I20. Printed in Great Britain

Some principles of linguistic methodology

WILLIAM LABOV

Department of Linguistics, University of Pennsylvania

ABSTRACT

Current difficulties in achieving intersubjective agreement in linguistics

require attention to principles of methodology which consider sources of
error and ways to eliminate them. The methodological assumptions and
practices of various branches of linguistics are considered from the stand-
point of the types of data gathered: texts, elicitations, intuitions and observa-
tions. Observations of the vernacular provide the most systematic basis for
linguistic theory, but have been the most difficult kinds of data for linguists
to obtain; techniques for solving the problems encountered are outlined.
Intersubjective agreement is best reached by convergence of several kinds
of data with complementary sources of error.

Less than twenty years divide us from the time when the study of methods was
the reigning passion of American linguistics; yet the status of methodology has
fallen so fast and so far that it now lies in that outer, extra-linguistic darkness
where we have cast speculation on the origin of language and articles about slang.
It cannot be denied that the pursuit of methods was carried on with a certain
willful blindness to both theory and practice; there was a time when methods
cast in the canonical form of tough-mindedness and symmetry were elevated to
the status of ritual texts, though they were seen even then to be hopelessly
impractical. But even if methodology is no longer an O.K. word in linguistics, we
have no choice but to use it again. It is an open secret that the rough and ready
exploitation of grammatical intuitions has run its course from exhilaration to
despair. A more reflective approach to the problem of intersubjective knowledge
is needed if our arguments are to mirror anything but our own polemical
intent.
In our exploration of the use of language in the secular world, we find that
methods have steadily grown in theoretical importance: first as a necessity, then
as an art, and now as a route towards the development of a theory of speaking. A
course in secular linguistics must focus first on the act of speaking and methods
for observing it. But it should also include training in the several methods of
historical and synchronic linguistics, making use of the texts, elicitations, and
intuitions which are the mainstay of the scholarly world. This paper will present
some of the principles that govern the gathering of empirical data, in both the
secular and scholastic places of linguistics. The principles will be stated with a

G 97

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 LANGUAGE IN SOCIETY

minimum of supporting argument, referring the reader to th

of evidence in the sociolinguistic investigations cited.

SOME GENERAL NOTIONS ABOUT METHODS

Methodology as conceived here is not a complete program for converting ignor-

ance into knowledge, but rather a set of strategies for handling the rich data
from well-known languages. Linguists who are approaching a language for the
first time will have to make their way as best they can; in any case, their findings
will most likely be rewritten many times by those who come after them. With the
pleasure of being the first goes the certainty of being wrong, which is the con-
verse of the CUMULATIVE PRINCIPLE: the more that is known about a language, the
more we can find out about it.
The luxury of methodology first becomes a necessity when continued in-
vestigations produce several competing theories, and we have to find out which
one is right. This paper is in fact designed to hold up as the highest goal of
linguistics the possibility of being right. To be right on a matter of general prin-
ciple is certainly difficult in the extreme, a slim chance at best, but the reward is
that the work may form part of the continued construction of linguistic theory for
a long time to come. It is certainly worth trying, perhaps the only thing worth
trying for; but it is unfortunately true enough that this goal has been abandoned
by at least one major tradition in linguistics. Beginning with the elegant paper of
Chao on 'The non-uniqueness of phonemic solutions . . .' (I934) and culminating
in Harris's argument for the complementarity of string analysis, transformational
analysis, and constituent analysis gramrnmars (I965: 365) we observe a quietistic
tendency to claim that almost all our theories are notational variants of one
another, that each is true in its own way and has its own insight to contribute.'
But what is commendable in religion is self-defeating in science. My own view is
that such equivalent theories are trivial variants, and to confine ourselves to arguing
their merits is to engage in an aesthetic pursuit rather than a scientific one.
Among the other innovations which Chomsky brought with him was a note of
high seriousness in this respect. He is clearly interested in the structure of human
language and the capacities of the mind which learns it, not in different ways of
looking at the matter. Since Chomsky believes that linguistic theory is under-
determined by the data (I966), he proposes an internal evaluation measure,
hopefully isomorphic with the one that the language learner actually uses. But
the simplicity metric has had hard going; it is frequently misused by those
trying to prove that they are right and that someone else is wrong, and there is

[I] 'To interrelate these analyses, it is necessary to understand that these are not competing
theories, but rather complement each other in the description of sentences. It is not that
grammar is one or another of these analyses, but that sentences exhibit simultaneously
all of these properties' (Harris 1965: 365).

98

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 SOME PRINCIPLES OF LINGUISTIC METHODOLOGY

some question as to whether it has actually resolved any important issues (Lakoff
1970).
Here linguistics is in a position to benefit from the example of the developed
sciences. Scientific methodology can be thought of as the reverse procedure:
trying to prove to yourself that you are wrong. That is, methodology is careful
and conscientious search for error in one's own work, following Karl Popper's
principle that the best theories are the easiest to disconfirm (1959). To be right
means that you have finally abjectly, hopelessly failed to prove yourself wrong.
It is dangerous to assign this responsibility to anyone else, for no one will have
the same vested interest in this pursuit as you do.
This kind of methodological self-criticism leads to a continual refinement of
our methods, introducing safeguards, reliability tests, cross-checks, typical of the
scientific attitude which can profitably be practised even at the pre-scientific
stage where we now find ourselves. But such methodological rigor is often justi-
fiably identified with the dead end of a worn-out approach. By the time the
methods are perfected, the most important work has often been done, and the
methodologists continue oblivious to this fact, proceeding on the hopelessly
unrealistic program that everything which can be described should be. At this
point, we find cropping up a second kind of methodology, a revolutionary critic-
ism which identifies new problems and fundamental faults in the older methods
which cannot be repaired.
A new methodology with new kinds of data is then called for. But if new data
has to be introduced, we usually find that it has been barred for ideological
reasons, or not even been recognized as data at all, and the new methodology
must do more than develop techniques. It must demolish the beliefs and as-
sumptions which ruled its data out of the picture. Since many of these beliefs are
held as a matter of deep personal conviction, and spring from the well-established
habits of a lifetime, this kind of criticism is seldom accomplished without hard
feelings and polemics, until the old guard gradually dissolves into academic
security and scientific limbo.
We might approach the various methods available to linguistics by looking at
the activity of the linguists themselves, according to where they can be found. In
this search, we would find linguists working in the library, the bush, the closet, the
laboratory, and the street, and might so name each sub-division of the discipline.
But in this analysis we will take a different approach and examine the raw mater-
ials gathered by each variety of linguistics, distinguishing each linguist by his
product: texts, elicitations, intuitions, experiments, and observations. A full discus-
sion of the experimental method would carry us beyond our present scope;
some methodological problems in the application of controlled experiments to
verbal behavior are given in Labov I969. This paper will consider principles of
methodology centering about the use of tests, elicitations, intuitions, and observ-
ations of the vernacular.

99

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 LANGUAGE IN SOCIETY

TEXTS

The fundamental methodological fact that historical linguists have to face is that
they have no control over their data. Texts are produced by a series of historical
accidents; amateurs may complain about this predicament, but the sophisticated
historian is grateful that anything has survived at all. The great art of the
historical linguist is to make the best of this bad data - 'bad" in the sense that it
may be fragmentary, corrupted, or many times removed from the actual pro-
ductions of native speakers. He relies first of all on the canons of critical scholar-
ship - by-laws and safeguards against human fallibility and corruptibility. The
most important of these is reference - the act of making the original texts available
for the inspection of others who may have other biases and prejudices. In in-
sisting on the checkability of data, historical linguists are considerably ahead of
the average descriptive linguist. The historian tries to bring us as close to his
data as he can, while the descriptive linguist keeps us many times removed.
Between the reader and the native speaker there are interposed the training,
skills and theoretical orientation of the linguist; rarely is an effort made to bridge
this gap by publishing tapes or protocols.
The chief methodological principle of historical linguistics remains the
NEOGRAMMARIAN HYPOTHESIS:

every sound change, inasmuch as it occurs mechanically, takes place according

to laws that admit no exception. That is, the direction of the sound shift is
always the same for all the members of a linguistic community except where a
split into dialects occurs; and all words in which the sound subjected to the
change appears in the same relationship are affected by the change without
exception.
(Osthoff & Brugmann 1878)

It is no longer possible to defend this hypothesis as a substantive claim that word

classes actually do move intact and as a whole. Although the objections of nine-
teenth-century realists seemed to have been overwhelmed, decisive disproof has
now been provided by Wang and his associates who have demonstrated the
existence of lexical diffusion on a massive scale in the history of Chinese dialects
(Chen & Hsieh '97I; Cheng & Wang 1970). But as a methodological principle
the neogrammarian hypothesis has been more successful: it has provided the
basic incentive to search for regularity and underlying conditioning factors in
sound change, rather than accept surface variation on face value.
Unfortunately, most historical linguists felt it necessary to defend the neogram-
marian hypothesis as a substantive description of the process of sound change,
and this preoccupation brought them into conflict with the solid data of dialec-
tologists. When Gauchat (1905) demonstrated that sound changes in Charmey
proceeded across three generations by fluctuations and lexical oscillations, the

IOO

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 SOME PRINCIPLES OF LINGUISTIC METHODOLOGY

neogrammarians rejected this and other descriptions of sound change in progress

as mere 'dialect borrowing' (Goidanich I926; Bloomfield 1933: 36I). Yet given
the imperfect character of historical records, it seems inevitable that we must rely
on present data to interpret them. In fact, our current research on sound change
in progress (Labov 197oa) is based upon the UNIFORMITARIAN PRINCIPLE: the
linguistic processes taking place around us are the same as those that have operated to
produce the historical record.2
In weighing the limitations of the Uniformitarian Principle, we are forced to
ask whether the growth of literacy and mass media are new factors affecting the
course of linguistic change that did not operate in the past. But even if this should
be so, we can still isolate in the patterns of everyday speech the kind of factors
which have always operated on the spoken language and which determine the
main stream of linguistic evolution in the present (Labov I966a).
The mutual interpretation of past and present can be seen most clearly in the
classical problem of the Great Vowel Shift. Current controversy and the historical
evidence is summed up by Wolfe (I969). The traditional view of Jespersen and
Wyld accepted the evidence of Hart and other sixteenth-century orthoepists on
the route followed by the long vowels of English: in particular, that the diph-
thongized high vowel of die descended from [dii] to [dei]. But K6keritz, Dobson,
Stockwell and others found it difficult to accept this view: for at the same time,
the vowel of day rose from [dhei] to [dei]. Presumably the two would then have
merged, but in fact they did not. The counterclaim was put forward (Stockwell
I966) that die was first centralized and then fell from [di] to [dai], but there is
little hard evidence to support this view. Our current instrumental studies of
similar changes in progress show that in a wide range of dialects, the new high
diphthong /iy/ in see falls to /ey/ as a front vowel - but not in the extreme front
position typical of tense vowels. The nuclei of these falling diphthongs follow a
centralized track which is clearly in the region of front vowels but with more
moderate second formant positions. This view of current sound shifts cannot
give us certain knowledge of what happened in the sixteenth century; but it can
resolve the contradiction between theory and evidence summarized above. We
now know that there is no reason to expect a merger of die and day, even if
Middle English die followed the pattern of some current /iy/ vowels and fell to
[dci] and then to [dai].3 The two vowels can pass each other on the routes

[2] The term 'uniformitarian' is borrowed from geology, where it signifies the now generally
accepted principle of Hutton that processes now taking place around us - weathering,
sedimentation, volcanism, etc. - are the same as those that have operated in the past to
produce the geological record.
[3] Even if local observers had reported that die and day were 'the same' at that time, it
would not follow that they were in fact the same and would not prevent these word
classes from following their opposing paths without disruption. Recent research on
sound change in progress shows that native speakers do not perceive consistent second-
formant differences which effectively separate word classes in natural speech. See
Labov (I97oa) and below.

IOI

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 LANGUAGE IN SOCIETY

indicated in Figure i. In approaching this historical problem with very different

kinds of data than that originally used, we converge upon the problem with
different sources of error. Observations of current changes thus have a height-
ened value for the resolution of older problems, as indicated by the PRINCIPLE
OF CONVERGENCE: the value of new data for confirming and interpreting old data is
directly proportional to the differences in the methods used to gather it.
The problem of interpreting literary texts, letters, puns and rhymes has two
aspects: (I) determining the relation between the writing system and the spoken
language, and (2) determining the relation between normative responses and the
vernacular. The first problem has been explored and argued (Stockwell &
Barritt I96I; Kuhn & Quirk 1953) but the second is still largely neglected.

FIGURE I. Routes followed by falling (lap) nuclei and rising (tense) nuclei in sound
changes in progress.

Elegant investigations might be carried out by comparing current novels, letters,

puns, poems and dictionaries with the actual state of the language today, and so
reflect backwards on the interpretation of earlier documents. This would require
a joint effort of historical scholars and linguists working with natural speech -
a natural alliance, since these are the two branches of linguistics which are most
concerned with controlling their data and searching out error.

ELICITATIONS

For many anthropologists, linguistics as a whole is essentially a methodological

necessity. They learn language in order to enter the culture; if they report their
knowledge in grammatical sketch, it is to preserve this aspect of the culture
before it disappears, not for the sake of the general needs of linguistic theory.
The anthropological linguist looks at the matter differently, more with an eye to
the general problems of linguistics, but he also finds himself approaching the
language from outside the grammar and the culture. He makes no claim on

102

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 SOME PRINCIPLES OF LINGUISTIC METHODOLOGY

native intuition to defend his grammar, nor can he hope to seize the flow of
natural conversation until long past the usual stages of language learning and
reporting. Few anthropological linguists learn a language well enough to make
much use of the conversational data of the street and the marketplace. The normal
procedure is to transcribe texts - often traditional folk lore - to elicit translations
of sentences from bilingual informants, present minimal pairs and ask for
'sames' and 'differents'. It would be idle to criticize these methods because they
are limited; by such techniques, Boas and Sapir enriched tremendously the
range of data available to a linguistic theory which had been largely based on
texts and normative handbooks of a narrow group of European language families.
Yet a calm view of our current field techniques should make it evident that this
data is also normative, modified by conscious reflection and governed by cultural
norms of right and wrong, good and bad language. Many American linguists
assumed at first that such norms existed only in literary cultures, but in one of
the most candid and penetrating examples of self-criticism on record, Bloomfield
(I927) showed how mistaken he had been in this respect.
To the extent that there is a disjunction between norms and the pattern of
everyday speech, the traditional field approach will yield a rough and necessarily
imperfect first approach to the language. The minimal pair test or commutation
test, long considered the soundest of all behavioral tests, is a dramatic example of
this limitation. In our recent work on sound change in progress, we find that
minimal pairs can be doubly defective. It has been observed before that native
speakers can make distinctions in minimal pairs that they do not make in actual
speech; normative responses often preserve fanciful, archaic or mythical distinc-
tions. But it was not realized that native speakers can fail to recognize or register
distinctions which they regularly make in natural speech (Labov 197oa): their
self-reports often reflect the patterns of younger speakers rather than their own,
or blur regular phonetic distinctions that are too subtle to withstand the glare of
conscious examination.
The chief methodological principle here is that the linguist must be fully aware
of the nature of his data. A realistic methodology would not insist that he
abandon all description until he can trade insults with the man in the street or
dispute local theology in the full flood of an enlightened scepticism. But if the
linguist recognizes the existence of these higher levels of competence, he can use
his developing grasp of the language on second or third trips to the field to
locate the differences between norms and behavior; by doing so, he would
deepen the value of his original observations for an increasingly dynamic and
secular linguistic theory.
The courses infield methods taught in our linguistics departments are of course
quite domesticated; they are more garden variety than field. But their well-
informed informants provide students with their only serious practice in trans-
cription and analysis. The exercises of our traditional texts are even more

103

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 LANGUAGE IN SOCIETY

removed from the data of a secular linguistics, but the work they demand is
honest work. To the everlasting credit of structural linguistics, it took the
student seriously, and tried to give him all the help it could. Gleason (I96i),
Pike (1947) and Nida (949) assumed that the student they were addressing was
going to bring back some important data from the field, and they were anxious
to teach him the essential skills of phonetic transcription, segmentation, allo-
phonic grouping, recognition of conditioned variants, minimal pairs and com-
mutation tests. The success of this approach is evident in the best of the Cali-
fornia dissertations and SIL reports.
However, the formal methodology which grew out of the structuralist tradition
was alarmingly unrealistic. It fully deserved the criticism which Chomsky
turned on 'discovery procedures'. Although it is hard for us to believe it today, a
number of students took seriously the contention of Bloch and Harris that one
could analyze a language by beginning at the phonetic level without reference to
the meanings of the words.4 There was an unstated and informal methodological
principle that gained currency and influenced practice which we can call the
PRINCIPLE OF PREFERENTIAL IGNORANCE: the less the linguist knows about a langu-
age, the more accurate (objective? scientific?) his description will be. It is unlikely
that the theory of ignorance will ever be fully developed, since the questions we
ask today demand deep rather than shallow knowledge. But granted that the
linguist could apply his technical skills of segmentation and classification in a
reliable way, would the results be valid? What would a test of validity be in such
a procedure?
The kind of methodology reflected in the discovery procedures of the I940S
and 5os has little relation to the principles to be presented here. We do not see
reflected therein the careful concern with sources of error, the search for new
kinds of data, for convergent and confirming perspectives which is our concept of
methodology.
At a number of points throughout this discussion the phrase 'sources of error'
has been used, and it will be helpful to specify the kinds of error we are talking
about. There are of course errors of measurement, of memory, or of calculation,
all of which can be avoided by careful attention to procedures. Tests of reliability
help us to check such errors and eliminate them. But we are more concerned with
a different type of error which stems from a misapprehension of the nature of the
data. The data which may be cited as evidence of some underlying construct
such as a linguistic rule, when it may in fact be largely the product of many
factors and represent no single property at all. If the errors of misapprehension or
the contextual factors neglected are local (such as interaction with a particular
interviewer) we will obtain a loss of reliability when others repeat the work. But

[4] John Street informs me (personal communication) that under the influence of Bloch he
once spent many months trying to do a phonetic transcription of Mongolian before he
understood the meanings of any of the words. The attempt did not succeed.

104

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 SOME PRINCIPLES OF LINGUISTIC METHODOLOGY

if they are general (such as neglecting the tendency to report norms rather than
behavior) we may obtain reproducible results which are still erroneous in their
application to the theoretical problem. In order to detect and eliminate such
results, we need tests of validity.
Unfortunately, we do not find any concept of validity in the methodology
developed by descriptive linguistics to elicit linguistic data. The seriousness of
descriptive intent was offset by a fashionable and aimless relativism in theory. A
simple, commonsense adherence to the search for intersubjective agreement will
reject such defensive manoeuvers. It seems reasonable for anthropological
linguists to be guided by the growing concern within anthropology to reduce the
role of the observer and let their methods be as transparent as possible; and so
arrange matters that their own training, skills, and limitations be cancelled in the
final analysis. This is of course only a goal to be aimed at, not a practice to be
achieved, but it reflects a definition of validity which we will explore in a later
section: that our descriptions should apply to the language which was spoken
before we arrive and will still be spoken after we leave.

INTUITIONS

There is no doubt that Chomsky is responsible for the most important metho-
dological revolution in our field. He directed a withering criticism at the extreme
behavioral approach which denied the existence of mental constructs and
rejected intuitive evidence out of hand. Chomsky and his students have demon-
strated the existence of a vast, seemingly inexhaustible supply of data which the
linguist can draw from his own knowledge of language. The richness of the
results is beyond dispute. We know much more about English, and about
language in general than we knew before and this achievement will stand
whether or not the current form of generative theory survives in a recognizable
form.
The question remains as to whether generative grammar has any methodology
beyond the decision to exploit intuitions of grammaticality to the full. To begin
with, we should note that this strategy depends upon the successful exploitation
of the SAUSSURIAN PARADOX. Saussure argued (I962: 321) that the linguist must
concentrate upon the social aspect of language, langue, which is conceived as so
general that it is in the possession of every speaker. It follows that one can in-
vestigate langue by asking anyone about it, even oneself, which is what Chomsky
proceeded to do. On the other hand, the individual details of parole can be
ascertained only through a social survey in the midst of the population. The
SAUSSURIAN PARADOX, then, is that the social aspect of language can be studied
through the intuitions of any one individual, while the individual aspect can be
studied only by sampling the behavior of an entire population.
The development of generative grammar has brought about a steady enrich-

I05

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 LANGUAGE IN SOCIETY

ment of this intuitive data. Chomsky's early response to criticisms of the gram-
matical-ungrammatical opposition was to suggest an ordered scale of gram-
maticality (I96I), but in further developments each writer has followed his own
bent. As judgments became more refined, several intermediate designations
began to appear: in addition to ungrammatical '*', we observe questionable '?',
questionably ungrammatical '?*', and outstandingly ungrammatical '**'. In
addition, the kinds of intuitions to be cited as evidence were steadily enlarged.
(i) The original judgments of grammaticality (well-formedness) naturally
included, (2) judgments of ambiguity, and (3) judgments of correct paraphrase.
But even from the outset we also note claims for (4), judgments of sameness or
difference of sentence type, and (5) intuitions about immediate constituents
(Chomsky I96I). A new emphasis on the theory of markedness has brought
the citation of (6), native intuitions on marked and unmarked status. Finally,
there are the most powerful of all kinds of intuition, (7) feelings that a given
theory is the right one, or that another solution is 'counterintuitive'.
It is unfortunate that this proliferation of the intuitive data has not been
accompanied by a methodological concern for the reduction of error, or a search
for intersubjective agreement. The weaknesses of intuitive studies in this respect
are known to all of us, but not everyone knows what to do about it without
abandoning the advances we have made. Originally, Chomsky hoped that the
area of agreement on judgments of grammaticality would be so large that the
disputed areas could easily be resolved by following the general pattern. But this
has not worked out in practice. The search for critical arguments has driven
almost everyone to the use of examples which command no agreement at all. As
one of countless examples, I cite Jackendoff's article on 'Quantifiers in English'
(1968). Among the sentences given as grammatical without question we have
The three of the men that you met yesterday have not left yet and Of the men, the
three you met yesterday have not left yet. Chomsky himself has found it impossible
to avoid arguments based on admittedly personal judgments. In his paper
'Remarks on nominalizations' (1970) he cites pairs such as our election of 7ohn
(to the presidency) vs. *our election of John (to be) president and notes: 'Reactions
to these sentences vary slightly: [these] represent my judgments.' He then adds,
'Given such data, . . ' and proceeds with the argument. By 'data' he does not
mean the disagreements, but rather the evidence of his own decisions. As
valuable and insightful as such arguments may be, they cannot alone lead to the
sure sense of right and wrong that we have raised as our ultimate goal. To achieve
intersubjective knowledge, we will probably have to limit ourselves to intuitions
of types (I)-3) above, and refrain entirely from citing the intuitions of the
theorist himself as evidence. Any serious consideration of sources of error must
hold such data as the most suspect unless it coincides with other sources.
Nevertheless, linguists continue to use uncheckable examples and defend them
by asserting that they are only discussing their own dialect. If 'my dialect'

io6

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 SOME PRINCIPLES OF LINGUISTIC METHODOLOGY

means no more than 'people disagree with me', it is certainly an illegitimate and
unworthy escape from serious work. Perhaps the most alarming symptom of
this retreat into the theorist's introspections is that it is no longer considered
proper to doubt such intuitive data. At one meeting after another, such questions
have been shrugged off as speakers refer impatiently to their own dialects as the
only relevant source of evidence. As a result, one seldom hears questions about
data any more.
A number of generative grammarians are now actively investigating the nature
of syntactic dialects, most importantly Guy Carden. Carden has discovered
implicational relations among dialect differences in the interpretation of negatives
and quantifiers (I970). By demonstrating that a given interpretation of quantifiers
with negatives implies the forn of tag questions and other sentence types,
Carden has given meaning to the concept of syntactic dialect. He has also given
considerable attention to reliability (personal communication). He has examined
I25 cases where an informant was asked the same question in a second interview
and found

no change 99
change 20 (9 interview errors, 2 systematic pressure changes,
9 apparently random)
possible change 6 (where identical responses might have been coded
differently)

Given this degree of reliability, we must agree that the investigation of other
peoples' intuitions is on a sound footing; at the same time, Carden recog-
nizes that there is a problem of explaining and controlling the changes that
occur.
Other studies of informant judgments indicate that we are dealing with a
statistical phenomenon, at least as usually carried out. The studies of grammat-
ical acceptability by Quirk and his associates confirm our informal observations
that it is rare to find IOO per cent agreement or disagreement on any sentence.
Investigations of judgments on tag questions and other syntactic issues carried
out by Lehiste and by Wedge & Ingemann (1970) showed that such data is
variable and shifting in the extreme. We can find implicational relations within
the flux of responses (see Elliott, Legum & Thompson 1969), if we are prepared
for a certain number of irregularities. There may be emerging a whole new
calculus of variations in the study of intuitive judgments. However, there has
been no success so far in replicating regular patterns in such variation (Postal
I968; Labov 197ob, and Heringer's disagreement with Carden I968). In general
we must observe that it is the nature of language to produce categorical judgments,
and we should not forget that it is usually the difficult and disputed areas that
have been investigated. But when we enter variable areas, it appears that in-
tuitive judgments are less regular than behavior. We seem to move quickly

I07

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 LANGUAGE IN SOCIETY

from regular areas of social agreement (langue) to a region of intuitive parole. It

seems clear that a great many disputes in generative grammar do revolve about
an area of idiosyncratic behavior where the social compact has disappeared.
For rare sentence types, it is only natural that each individual should have to
solve the problem for himself; in so far as he can do this by extending his current
roles in a predictable manner, we are dealing with langue; in so far as individuals
diverge without any observable pattern, we are dealing with true idiolects. The
very concept of idiolect, of course, represents a defeat for the Saussurian notion of
langue as the general possession of the speech community. Our general aim is to
write the grammar of that speech community, with all of its internal variation,
style shifting, change in progress (Weinreich, Labov & Herzog I968). When the
data begins to fragment into unpatterned idiosyncrasies - for normative judg-
ments for actual behavior - then linguistics comes to a stop. This is not the kind
of data that we can rely on for a theory of language that would satisfy the least of
Saussure's ambitions.

OBSERVATIONS

There is obviously something odd in placing observations last in the types of data
used by linguists. But the observation of natural speech is in fact the most
difficult of all the methods discussed so far. Texts, elicitations and intuitions are
much more accessible, more easily segmented and classified; yet the wealth of
linguistic description and theory which has been built upon such data still
remains to be interpreted in its relation to language as a vehicle of communica-
tion in everyday life. In the gathering of elicitations and intuitions, there is no
obvious sense in which the work can be described as valid. If another linguist
obtains the same judgments from native speakers or from his own introspec-
tions, then we can say that the method is reliable. But reliability by itself does not
help us in developing a sound theory of language in the sense that we intend.
Very often the linguist is actually producing his own phenomena.5 He has
therefore created a further problem of relating these artifacts to natural
language.
What would it mean for elicitations or intuitions to be valid? One might

[5] As linguists increase their familiarity with the philosophy of ordinary language, there is
a growing tendency to use 'thought-experiments' as a means of manufacturing data.
Thus Jerry Morgan develops an idea about time machines (credited to McCawley)
as a means of producing data about the reflexive. What would an investigator, sitting in a
time machine, say if he saw an image of himself slapping himself? According to Morgan,
he could report 'I slapped myself'. 'If new me reaches out from the time machine and
slaps his younger counterpart, I can report it by "I slapped myself ". But if old me reaches
into the time machine and slaps new me, I cannot report it by "I slapped myself"'
(I969:55). If Morgan should succeed in getting reliable responses to such thought-
experiments, then he would have the problem of relating these responses to language as
we have conceived it here.

io8

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 SOME PRINCIPLES OF LINGUISTIC METHODOLOGY

reasonably demand that they match the language of everyday

linguist is not present. This demand follows from the fact tha
linguists and there are many speakers, an observation which m
as the PRINCIPLE OF THE VOCAL MAJORITY: many speak but f
our theories are merely the artifacts of our own analyzing a
little to tell us about the natural evolution of language. Eithe
about the language that ordinary people use on the street, a
or at home blaming their children, or they are about very litt
Those who gather literary texts are actually observing som
produced independently: historical linguists are certainly engaged in the
observation of language. Some believe that the full structure of language can
only be observed in its most literary developments, and that speech is relatively
impoverished. However, we retain the conviction of our predecessors in American
linguistics that texts can be understood only in their relation to the spoken
language - that the main stream of evolution of language is to be found in every-
day speech, even in highly literary cultures such as our own.
In order to introduce observations from everyday life, we have to carry out a
thorough-going criticism of beliefs and ideology more or less as Chomsky did
for the methodology of the Bloomfieldians. One widely propagated belief which
is used to discourage the study of ordinary language is that speech is incoherent.
Chomsky has often remarked that the child must discard the largest part of what
he hears as ungrammatical (I965: 58). This view is a myth based upon no
evidence at all, except perhaps a few transcripts of learned conferences. Anyone
who works with natural speech realizes this, and it has been shown systematically
that the majority of sentences spoken by ordinary people are well-formed with-
out any editing; all but a tiny percentage can be reduced to that form by the use
of simple and universal editing rules (Labov I966b).
Secondly, we find that most investigators describe their own community as
exceptional, rife with dialect mixture and chaotic variation as compared to the
homogeneous nature of traditional speech communities. But such homogeneous
communities are also myths. As Gauchat showed (1905), even the most remote
Swiss village shows systematic variation across sex and age group. More recent
investigations of speech communities in New England, New York City, Detroit,
Hillsboro, Salt Lake City, and Norwich show that this variation follows regular
patterns which tell us a great deal about the evolution of language as well as how
people use it. We find again and again that the grammar of a speech community
is more regular than the behavior of the individual (Labov I966a; Shuy, Wol-
fram & Riley I966; Levine & Crockett i966; Cook I969; Trudgill 1970).
We find a third ideological barrier in the claim that all such data belong to
some other far-away discipline called the study of performance, to be realized
when we have mastered the facts of competence. The distinction between
competence and performance may have its uses, but as it is now drawn it is

9og

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 LANGUAGE IN SOCIETY

almost incoherent. If performance factors are those which facilitate or impede

the production of sentences, then almost all of our transformational apparatus
would fall under that rubric: rules of extraposition, complementation, particle
movement, negative attraction, and so on. We begin with a multi-dimensional
deep structure, impossible to perform, and end with a linear organization that is
easy to say and to grasp. Instead of the left-embedded For anyone to do that is a
shame, we extrapose, and say It's a shame for anyone to do that; we then per-
form with greater ease a right-embedded complement as a result of this extra-
position.
There are also technical innovations which facilitate the study of everyday
speech. The magnetic tape recorder was introduced to this country just after
World War II. But most linguists have been slow to admit its importance, con-
tinuing to claim that data jotted down in person is more reliable than a tape
recording. Most linguistic students in graduate departments have access to an
aged Wollensak, if that, and have gotten no grasp of the difficult art of making
good recordings. It would be fair to say that a lack of professional orientation
towards equipment has been a serious impediment in the development of the
study of language in everyday life. The only serious relation to instrumentation
is found among phoneticians, and the general impression holds that good
recordings are important only in the laboratory. But in actual fact, much better
recording techniques are needed for the study of grammar than for phonology,
even better equipment is needed for the analysis of discourse in ordinary interac-
tion.6
The strongest constraints that prevent linguists from utilizing the wealth of
linguistic data with which they are surrounded are the barriers against inter-
action with strangers in one's own culture. The most common question addressed
to me after a lecture involving data drawn from speakers outside the university is:
'What do you say to these people?' This is a legitimate and important question.
But before we can answer it, we must recognize the nature of the problem: the
unnamed and unspecified fears that these strangers will somehow do us harm.
Each person is secretly convinced that he alone is fearful and isolated. To protect
themselves against accusations of incapacity, neglect or cowardice, many aca-
demic people manufacture a counter-ideology: maintaining that these other
people outside the university passionately want to be left alone; that it is immoral

[6] Some of the most recent developments in recording which are important for the study
of natural interaction are actually setbacks (possibly temporary) in fidelity. Cassette
tape-recorders running at I} inches per second cannot give us the frequency response to
capture all of the relevant acoustic properties of the speech signal. But their small size,
low cost, and inconspicuousness can hardly be discounted in the pursuit of naturalistic
observations. There have been advances in the insulation of built-in condenser micro-
phones from motor noise which are of some importance for fieldwork of this sort,
and it is likely that the combination of stereo separation with equipment of this type
will be a further step of consequence for the study of natural groups.

IIO

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 SOME PRINCIPLES OF LINGUISTIC METHODOLOGY

to invade their privacy by talking to them.7 It is necessary for

established (i) that what they are doing is what linguists should do, and (z) that it
is immoral, and not proper linguistics, for others to do otherwise. But no such
defense is necessary, since the fears that people feel about contact with strangers
are the product of regular rules of social interaction, common to everyone. To
overcome such barriers, one must have a general understanding of these rules,
how they can be used to advantage, and what their limitations are. With a few
exceptions, our society is composed of people who recognize the barriers be-
tween themselves and others, who would like to overcome them, and are de-
lighted when someone else takes the step which makes this possible.
There are personality differences among linguists which will inevitably lead to
specialization in the library, bush, street or closet. But it seems to me that one
must resist the tendency to redefine the limits of linguistics to suit one's own
personality. We see this tendency in the historian's rejection of spoken texts, the
structuralist's rejection of intuitions, the intuitionist's rejection of everyday
speech, and the anthropologist's rejection of his own society.
Our first steps in the study of everyday life allow us to say something about the
crucial question of validity for elicitations and intuitions. Under what conditions
do normative responses diverge from behavior? Under what conditions can we
ask direct questions about grammaticality and obtain responses related to the
language which is used in ordinary communication? On the whole, we cannot do
this with young children, although Lila Gleitman has demonstrated the ex-
traordinary capacity of certain children as grammarians (I970). The dialect used
by children is only one of many non-standard dialects which are in contact with a
dominant standard. Given such a condition, we can assert a general PRINCIPLE OF
SUBORDINATE SHIFT: When speakers of a subordinate dialect are asked direct
questions about their language, their answers will shift in an irregular manner
toward [or away from] the superordinate dialect. This principle operates whenever
we try to study the rules of working-class dialect, Black English, patois or creole
using formal elicitation or training native speakers to ask themselves questions.
That is not to say that such activities should not or cannot be carried on; but in
the absence of any other data, one must expect that the results will be invalid in a
number of unspecified and unforeseeable ways.8

[7] Over the past several years, I have collected a number of questions from academic
audiences which presuppose or imply this moral stance. E.g. 'Under what pretext do you
speak to these people?' One professor asked me, 'How do you insinuate your way into
these groups?' and a little later his wife asked, 'How do you accost them?' Pretext,
accost and insinuate appear to share a feature of [-propriety] which I have tried to
elucidate in the general formulation above.
[8] It is not only subordinate adult dialects which are governed by this principle. Language
forms used by children are non-standard dialects which are generally considered
incorrect, and one can expect that the direct elicitation of data from children is not
easily related to their actual production. It is a general finding that formal experiments
and elicitations from children show a linguistic competence which is below that dis-

III

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 LANGUAGE IN SOCIETY

Granted the ability to pass beyond ideological, technical, and

and the recognition of the disjunctions between norms and behavior, there
remains a crucial methodological paradox in the study of everyday language.
It follows from five principles which have been discussed elsewhere (Labov I 97ob),
and will be stated here quite briefly.
First is the PRINCIPLE OF STYLE SHIFTING: there are no single-style speakers.
Whenever we first encounter a speaker in a face-to-face situation, we must
assume that we are observing only a limited part of his entire linguistic reper-
toire. There may be some linguistic features that do not shift from one style to
another, but every speaker will have a configuration of linguistic variables that
shift from one context to another.
The PRINCIPLE OF ATTENTION asserts that Styles can be ordered along a single
dimension, measured by the amount of attention paid to speech. Despite the varied
nature of stylistic influences, and the multi-dimensional character of stylistic
rules, all of the patterns can be projected on a single ordered dimension which
has significance for our methodology. Casual and intimate styles can be stationed
at one end of this continuum, and frozen, ritualistic styles at the other. At present
we can control some of the factors which cause attention to be paid to speech
(see below), but we have not yet quantified the actual behavioral feature:
attention to or monitoring of speech.9
The third in this series is the VERNACULAR PRINCIPLE: that the style which is
most regular in its structure and in its relation to the evolution of the language is the
vernacular, in which the minimum attention is paid to speech. To justify this prin-
ciple fully would require a review of a large body of sociolinguistic data from a
great many sources (but see in particular Labov I966: xIv). This principle can
also be seen to follow quite naturally from the PRINCIPLE OF THE VOCAL MAJORITY
cited above. It is the high frequency and practiced automaticity of everyday
language which is responsible for its pervasive and well-formed character. The
word 'vernacular' has sometimes led to the misunderstanding that this principle
focuses only upon illiterate or lower-class speech. Most of the speakers of any
social group have a vernacular style, relative to their careful and literary forms of
speech. This most spontaneous, least studied style is the one that we as linguists
will find the most useful as we place the speaker in the overall pattern of the
speech community.
It can readily be seen that the fourth principle interferes with the third. The

played in their spontaneous performance (Brown 1971). Observations of language in

use were first developed in this field, and recent advances in the semantic interpretations
of children's language have utilized close attention to the behavioral context of speech
(Bloom 1970).
[9] Methods for quantifying attention have been developed by Broadbent (I962), but such
experimental techniques have not yet been applied to measure the amount of attention
paid to speech.

112

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 SOME PRINCIPLES OF LINGUISTIC METHODOLOGY

PRINCIPLE OF FORMALITY states that any systematic observatio

a formal context in which more than the minimum attention is
'systematic observation', we include more than the presence or absence of a
human observer. The tape recorder itself has a variable but persistent effect in
shifting speech towards the formal end of the spectrum.
We are then left with THE OBSERVER'S PARADOX: To obtain the data most
important for linguistic theory, we have to observe how people speak when they are
not being observed. The various solutions to this paradox define the methodology
for the study of language in context.
This methodology can be presented in the form of a brief history of socioling-
uistic methods. The point of departure is the traditional practice of dialectology,
in which the main concern is to elicit relatively small pieces of lexical or morph-
ological information. This implies a long question from the interviewer and a
short answer from the subject - just the opposite of our present practice which
limits any question to a maximum of five seconds. But the early interviews in
Martha's Vineyard (Labov I963) and Detroit (Shuy, Wolfram & Riley I966)
contained many such long, short-answer questions.
The sociolinguistic interview form was developed largely in work on the
Lower East Side of New York City and in Harlem (Labov I966a; Labov,
Cohen, Robins & Lewis I968). There we explored ways to break through the
constraints of the interview situation, as expressed in the FORMALITY PRINCIPLE.
These constraints can be sidestepped by setting up the assumption that they are
not relevant for a given time, or they may be overridden by more powerful
factors. In the Lower East Side study, five contextual situations were located in
advance where the vernacular would be most likely to emerge, and in such a
context casual speech was identified by the occurrence of one or more indepen-
dent channel cues: increase in volume pitch, tempo, breathing, or laughter. A
question that effectively triggers such responses may take six months to a year to
develop - for the theme is not the only important feature. Placement, wording,
timing and delivery all contribute to the likelihood of involving the speaker to
the extent that formal constraints are overridden. One of the most successful
questions of this type is on the Danger of Death: 'Have you ever been in a
situation where you were in serious danger of being killed, where you thought to
yourself, This is it? ... What happened?' A dramatic shift of style can be seen in
the following brief passage, from an interview with an eighteen-year-old Irish-
Italian boy in New York City. His careful style breaks up suddenly, with nervous
laughter, hard breathing, and rapid bursts of speech. The important linguistic
variables shift at the same time: (ing) shifts from [iP] to [in]; (th) and (dh)
shift towards [t] and [d]; consonant cluster simplification rises sharply, and
negative concord appears. In this transcript, suppressed laughter is indicated by
[hh] or [hhh], usually superimposed on the words, and the switch to casual
style and back again is indicated by italics.

H II3

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 LANGUAGE IN SOCIETY

(What happened to you?)

The school I go to is - uh - Food and Maritime, That's - uh - maritime
training. And I was up in the masthead, and the wind started blowing. I had
a rope secured around me to keep me from falling. But the rope parted [hh]
an' I was jus' hangin' there by my fingernails [hhh]. I never prayed to God so
fast [hh] and so [hh] hard in my life! But I came out all right.
(What happened?) Well the guys came up an' 'ey got me. (How long were you
up there?) About ten minutes [hhh]. (Jees I I can see you're still sweatin'
thinkin' about it.) Yeh [hhh]. I came down, I cou'n' hold a pencil in my han'
[hhh], I cou'n' touch nuttin'. I was shakin' like a [hhh] leaf. Sometimes I get
scared t'inkin' about it. But - uh - well it's training!

In constructing such questions, one must turn back to the basic sociolinguistic
conundrum: 'Why does anyone say anything?'1O There are three content themes
which have the greatest force for evoking speech from the broadest range of
speakers: (i) death and the danger of death, including any form of physical
violence (fights, accidents, sickness, operations); (2) sex and all of the machinery
for interaction between the sexes (proposals, dating, household negotiations);
(3) moral indignation (e.g. 'Did you ever get whupped for something you didn't
do?') Beyond such general considerations, there are a vast array of local issues,
humor and gossip which the field worker must seize as a by-product of partici-
pant-observation. Questions in specific local areas are constructed by a feed-
back technique which progressively assumes more as the field worker knows
more.1" An initial question. 'Do you play the numbers?' would thus give way to
'Did you ever hit big?' In talking to deer hunters, an initial question such as
'Where do you aim?' would give way to 'Is it worth trying a rump shot?' As the
outsider gradually becomes an insider, the quality of the speech obtained and the
speaker's involvement in it rises steadily. A field worker who stays outside his
subject, and deals with it as a mere excuse for eliciting language, will get very
little for his pains. Almost any question can be answered with no more informa-

[Io] That is, the study of methods involves us in basic questions of discourse analysis.
Naive approaches to eliciting speech rely heavily upon questions which superficially
force responses. Experimental methods used to assess the verbal competence of children
also utilize direct questions and obtain systematically misleading data. Further analysis
of the factors which lead children and adults to speak inevitably involve us with a con-
sideration of the actions being performed, and the underlying propositions about the
role of the speaker and addressee. Such considerations form the foundation of a study of
discourse which distinguishes sequencing rules from rules of interpretation which carry
us from 'what is said' to 'what is done'.
[iI] Such methods are similar to the techniques used by anthropologists in forming
questions which reflect the categories and vocabulary of the native culture (Black &
Metzger I965). But the extreme formality which has been used in the approaches
reported so far may give misleading results if there is a significant disjunction between
norms and behavior.

114

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 SOME PRINCIPLES OF LINGUISTIC METHODOLOGY

tion than was contained in it. When the speaker does give more, it is a gift,
drawn from some general fund of good will that is held in trust by himself and
the field worker. A deep knowledge implies a deep interest, and in payment for
that interest the speaker may give more than anyone has a right to expect. Thus
the field worker who can tap the full linguistic competence of his subjects must
acquire a detailed understanding of what he is asking about, as well as broad
knowledge of the general forms of human behavior.
Beyond the interview. The individual interview will remain as the foundation
of our investigations, since only there do we control the large bulk of speech and
the complex structures needed for the study of grammar. But the methods just
described for overriding the constraints of the formal interview are only sub-
stitutes for the real thing, and give us only fragments of the vernacular.
A more systematic approach to recording the vernacular of everyday life is to
allow the interaction of natural peer group itself to control the level of language
produced. The techniques used here are from the original work done by Gumperz
in Hemnes, Norway (I964) where he recorded the interaction of closed and
open networks of members of the community. The investigators provide the
initial setting, but gradually recede from the situation; the effect of recording is
never wholly absent, as our principles would predict, but it is largely overridden
by other factors - the same as those which operate in everyday life. Such tech-
niques were further developed in our work with adolescent peer groups in South
Central Harlem (Labov, Cohen, Robins & Lewis I968: I, 57). Though group
sessions give us the basic standard by which we can calibrate our other work,
they do not usually provide us with enough linguistic data on each individual. It
is possible to feed back some of the same interactive mechanism to interviews
with one or two speakers, if the interviewers are members of the same com-
munity. There are great differences in the techniques available to insiders, in the
quality of interaction, and in the type of information obtained, as compared to
the best interviews done by outsiders. Insiders cut deeper; but at the same time,
they are more limited in the range of speakers they can deal with. Since they have
a fixed location within the community, many members cannot speak to them as
freely as they would to a stranger. Narratives told to insiders tend to be more
fragmentary, less well-formed than those told to outsiders. The outsider can be
seen as a blunt instrument, a useful tool for all kinds of rough work, while the
insider can penetrate more deeply in a narrower range. No serious study of a
speech community should be planned without including both kinds of invest-
igators from the outset.
The kind of long-term observation just described would be most important in
studies of the acquisition of language. No such studies have been carried out yet
with a natural play group as the focus; the family has been the only target of
attention. Within that matrix, there has been considerable progress in the kinds
of observations that have been made, with increasing attention paid to the

II5

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 LANGUAGE IN SOCIETY

physical behavior of the child which offers the clues we ne

pretation (Brown 1970: ioo; Bloom 1970: 15-33).
The family as a whole is the focus of our attention in our recent studies of
sound change in progress (Labov 1970a). The main purpose is to trace age-
grading and the effects of change in progress. But we are also aware that our
individual interviews give us only an approximation to the vernacular, and
family sessions can give us observations of spontaneous interaction for calibrating
the main body of data. A single question, aimed at issues that are argued within
the family every day, can trigger an extended discussion in which the interviewer
gradually recedes from view. In the following example from a family in East
Atlanta, we can observe the beginning of such a lively argument, and the
emergence of family interaction that gives us the vernacular style we are looking
for. This was an after-dinner conversation at the house of Mr and Mrs Henry
G.; gathered around the dining room table were Henry Sr, 59, his wife Mrs
Henry, 55, their two daughters Gail and Barbara, and Barbara's husband Bill.
Bill works as a machinist in the railroad repair shop where Henry Sr is a fore-
man.

Interviewer: Do you think there's a natural life span for people, or is it possible
there's a way for you to live longer and longer?
Henry Sr: [head of household, 59] Yep. They're gonna do that. They're provin'
that every day.
Mrs Henry: [55] [overlapping] They are livin' longer and longer. People are
livin' longer. And that's health, you know. Back then, I don't see how they
lived as long as they did . . . back - years ago.

Both husband and wife respond together with the same point of view. The
interviewer finds that he has touched a live issue in the family as he is cut off
and the conversation proceeds on its own momentum. All five family members
join in as the theme moves to the familiar issue of the old ways versus the
new.

Interviewer: Yeh, well some -

Henry Sr: What is it? Used to be forty when I was young, and now it's sixty
something.
Mrs Henry: Oh - You take a woman back when - even when I was a chil',
a woman thirty-five was old!
Barbara: [daughter] Probably you took to a rocking chair.
Mrs Henry: Yeah, now that's really the truth, they really were, and they looked
old - thirty five.
Mrs Henry: And they were old.
Gail [daughter]: You had to b'eak forty, you were an old man [hh].
Henry Sr: Back then, look what a woman had to do.

II6

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 SOME PRINCIPLES OF LINGUISTIC METHODOLOGY

Mrs Henry: Tha's why I know; that's what I said.

Bill: [Barbara's husband] Nowaday all they got to do is th'ow it in here, 'n'
th'ow it in there, 'n' they got a machine to do it fo' em.

Until Bill entered the picture, sex was not the issue here. But Bill picks up
"woman" from Henry Sr and makes it the main point; his wife laughs at his
familiar stance. Mrs. Henry gives a straight answer, but when Bill keeps on
woman-baiting, she picks up a cue from her husband and carries the day with
a crushing reply.

Mrs Henry: Well that's everything - we get a -

Bill: [overlapping] They don't have to exert theyself. Y'got a vacuum cleaner 'n'
you push 'n' 'a's all you gotta do.
Barbara: [Laughs]
Mrs Henry: Well that's not healthy. I think it's good for you to do something -
work, far as that goes.
Bill: Back then - girls nowaday, back then, if they had to wash clothes -
Henry Sr: [overlapping] Well how about the men? How about the men?
Mrs G.: Men has never worked! [General laughter]

This conversation then continued for another five minutes without any inter-
vention of the interviewer. Bill and Henry Sr engaged in a long argument on
which was harder work-using a sledge hammer in the old days, or using a
pneumatic hammer today. The family members put themselves on display,
showing us their immediate concerns, their mode of argument, their system of
values, and the grammatical structure of their everyday language.
As we move away from the individual interview, our data is less complete, but
closer to the language of everyday life. Another step in that direction can be
taken through RAPID AND ANONYMOUS OBSERVATIONS. Here we may know very
little about the speakers, but we can observe a great many of them, and the effect
of the interview situation is nil. The survey of (r) in New York City department
stores (Labov I966a: III) provides one such model, and others have been devel-
oped in work with the telephone, asking directions, or street corner observation.
More recently we have used such techniques to observe the use of Spanish on
Harlem streets and check the birth place of those who use the language there.
Such rapid and anonymous studies have built-in sources of error, but the error is
complementary with that of interviews. When the two kinds of data converge,
we have in effect partialled out the effect of experimental and observational error.
There remains a residual and almost insoluble problem - the rarity of many
grammatical forms. It will no doubt always be necessary to extend our observa-
tions with intuitions. Yet we are only beginning to learn how to enrich the data of
natural conversation by minimal intervention. When we fully understand the use
of a given grammatical form, we are then able to elicit it in conversation without
using it ourselves, and without seeming to do anything odd or artificial. This has

117

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 LANGUAGE IN SOCIETY

been done only for a small number of items, such as the prete
future, and present perfect, and relative clauses. In general, we can say that the
future study of language in context will depend heavily upon the development of
means of enriching the data of natural conversation.

CONVERGENCE

This discussion has been limited to the methodology involved in gathering

linguistic data. A second paper of equal length is needed to deal with the methods
used in analysis. In such a discussion, we would be able to review the more
detailed evidence which supports the Vernacular Principle - that the most
systematic style is that used when the minimum attention is paid to speech. We
would also consider the ways in which different methods can be mutually con-
firming. In pages Io5-Io8, we noted that the study of intuitions has not achieved
this kind of convergence, and it seems unlikely that the current explorations of
variation and implications in intuitive judgments will yield that kind of result.
If we are right in thinking that most of this variation represents a kind of in-
tuitional parole, we can hardly expect uniform and convergent patterns to
emerge. The most encouraging aspect of work with observations within the speech
community is that we have achieved such convergence on principles of great
generality. A number of investigators have reproduced patterns of intricate and
regular interaction of stylistic and social stratification for stable linguistic
variables such as (ing) (Fischer 1958; Labov I966a; Shuy, Wolfram & Riley
I966; Labov, Cohen, Robins & Lewis I968; Trudgill I970). The cross-over
pattern associated with the hypercorrect behavior of the lower middle class has
been reliably reproduced in many independent studies (Labov I966a; Levine &
Crockett I966; Wolfram I969). We have independent confirmation of the fact
that women may lead men by almost a generation in on-going linguistic changes
but also shift more towards the older prestige standard in formal situations
(Gauchat I905; Labov I966a; Shuy, Wolfram & Riley I968; Wolfram 1969;
Trudgill I970). Variable constraints on the contraction and deletion of the
copula have been confirmed in several independent studies (Labov I969;
Wolfram I969; Mitchell-Kernan I970). A full presentation of the methodology
of analysis would show these convergent results in detail, but enough synthesis
has been done so far to justify the claim that with observations of natural speech
form an adequate base for intersubjective agreement in linguistics (Labov 197ob).
If this agreement has been achieved it is because attention was given to
possible sources of error in each study. The most effective way in which con-
vergence can be achieved is to approach a single problem with different methods,
with complementary sources of error. A number of the studies cited have used
observations, intuitions, elicitations, texts and experiments to achieve this result.
Here we must refer again to the PRINCIPLE OF CONVERGENCE: the value of new

i i8

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 SOME PRINCIPLES OF LINGUISTIC METHODOLOGY

data for confirming and interpreting old data is directly proportional to the
differences in the methods used to gather it.
Despite the fact that we have a variety of methodological approaches, the
unity of linguistics is not hard to conceive. It is not necessary for everyone to use
the same methods - indeed, it is far better if we do not. Otherwise we would not
benefit from the complementary principle. The unification of linguistics must
necessarily proceed from the understanding of linguists that the field need not be
defined to fit their personal style, but can expand to a broad attack on the com-
plexity of the problem. Data from a variety of distinct sources and methods,
properly interpreted, can be used to converge on right answers to hard questions.

REFERENCES

Black, M. & Metzger, D. (I965). Ethnographic description and the study of law. (AmA
67 (6) Pt 2, 141-65.) Reprinted in S. A. Tyler (ed.) (1969), Cognitive anthropology.
New York: Holt, Rinehart and Winston. 137-65.
Bloom, L. (I970). Language development. Cambridge, Mass.: MIT Press.
Bloomfield, L. (1933). Language. New York: Henry Holt.
(I927). Literate and illiterate speech. In American Speech 2. 432-9.
Broadbent, D. E. (1962). Attention and the perception of speech. Scientific American
206. 143-51.
Brown, R. (I970). Psycholinguistics. New York: The Free Press.
Brown, R. (I97I). A first language. Mimeographed.
Carden, G. (1970). A note on conflicting idiolects. Linguistic Inquiry I. 28I-90.
Chen, M. & Hsieh, H. (I97i). The time variable in phonological change. YL 7. i-I3.
Cheng, C. & Wang, W. (I970). Phonological change of middle Chinese initials. In Project
on linguistic analysis 2. io Berkeley: University of California.
Chao, Y. (1934). The non-uniqueness of phonemic solutions of phonetic systems. In
E. Hamp, F. Householder, and R. Austerlitz (eds), Readings in Linguistics II. Chicago:
University of Chicago Press.
Chomsky, N. (I96I). Some methodological remarks on generative grammar. Word 17.
219-39.
(i965). Aspects of the theory of syntax. Cambridge, Mass.: MIT Press.
(I966). Topics in the theory of generative grammar. In T. Sebeok (ed.), Current
trends in linguistics 3: linguistic theory. Bloomington, Ind.: Ind. University Press.
(I970). Remarks on nominalization. In P. Rosenbaum, and R. Jacobs (eds), Readings
in transformational grammar.
Cook, S. (I969). 'Language change and the emergence of an urban dialect in Utah'.
Unpublished University of Utah dissertation.
Elliott, D., Legum, S. & Thompson, S. (i969). Syntactic variation as linguistic data.
In R. Binnick et al. Papers from the fifth regional meeting, Chicago Linguistic Society.
Chicago: Department of Lingustics, University of Chicago.
Gauchat, L. (1905). L'unite phonetique dans le patois d'une commune. In Aus romanischen
sprachen und literaturen: festschrift heinrich morf. Halle: Max Niemeyer. 175-232.
Gleason, H. A. Jr. (I96I). Introduction to descriptive linguistics (2nd edition). New York:
Holt, Rinehart and Winston.
Gleitman, L. (I97i). The child as grammarian. Mimeographed.
Goidanich, P. G. (1926). Saggio critico sullo studio de L. Gauchat. In Archivio Glot-
tologico Italiano xx. 6o-7i.
Gumperz, J. (I964). Linguistic and social interaction in two communities. In J. Gumperz
and D. Hymes (eds), The ethnography of communication. (AmA 66 (6) Pt 2 ) Washing-
ton, D.C.: American Anthropological Association. 137-53.

II9

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms
 LANGUAGE IN SOCIETY

Harris, Z. (I965). Transformational theory. Lg 41. 363-401.

Heringer, J. T. (I970). Research on quantifier-negative idiolects. In Papers from sixth
regional meeting of the Chicago Linguistic Society. Chicago. 287-96.
Jackendoff, R. S. (I968). Quantifiers in English. FL 4. 422-42.
Kuhn, S. M. & Quirk, R. (1953). Some recent interpretations of Old English digraph
spellings. Lg 29. 143-56.
Labov, W. (i963). The social motivation of a sound change. Word I9. 273-309. Reprinted
in C. Scott, and J. Erickson (eds), (I968) Readings for the history of the English language.
Boston: Allyn and Bacon. 345-79.
Lakoff, G. (I970). Linguistics and natural logic. Ann Arbor: University of Michigan.
Levine, L. & Crockett Jr., H. J. (1966). Speech variation in a Piedmont community:
postvocalic r. In S. Lieberson (ed.), Explorations in sociolinguistics: sociological inquiry
36:2. Reprinted as Publication 44, I.J.A.L. I966.
Labov, W. ( I966a). The social stratification of English in New York City. Washington, D.C.:
Center for Applied Linguistics.
(ig66b). On the grammaticality of everyday speech. Paper given before the Lingu-
istic Society of America, New York City, 1966.
(I969). The logic of non-standard English. In J. Alatis (ed.), Linguistics and the
teaching of standard English to speakers of other languages or dialects. (Georgetown
University Monograph Series on Languages and Linguistics 22). Washington, D.C.:
Georgetown University Press, i-44. (Reprinted in F. Williams (ed.), Language and
poverty; perspectives on a theme. Chicago: Markham, 1970. 154-I91.)
(I970a). Proposal for continuation of research on sound changes in progress, sub-
mitted to National Science Foundation (NSF-GS-3287).
(197ob) The study of language in its social context. In Studium generale 23. 30-87.
Labov, W., Cohen, P. Robins, C. & Lewis, J. (I968). A study of the non-standard
English of Negro and Puerto Rican Speakers in New York City. Cooperative Research
Report 3288. Vols i and II. New York: Columbia University. (Reprinted by U.S.
Regional Survey, 3812 Walnut St, Eisenlohr Hall 202, Phila., PA 19104.)
Mitchell-Kernan, C. (1969). Language behavior in a black urban community. Working
paper No. 23. Berkeley, Cal.: Language-Behavior Laboratory, University of California.
Morgan, J. L. (I969). On arguing about semantics. Papers in Linguistics I. 49-70.
Nida, E. (1949). Morphology (2nd edition). Ann Arbor: University of Michigan Press.
Osthoff, H. & Brugman, K. (I878). Preface to Morphological investigations in the sphere of
the Indo-European languages i. Trans. W. P. Lehmann. In W. P. Lehmann (ed.) (I967),
A reader in nineteenth-century historical Indo-European linguistics. Bloomington, Ind.:
Indiana University Press. 197-209.
Pike, K. (I947). Phonemics. Ann Arbor: University of Michigan Press.
Popper, K. R. (1959). The logic of scientific discovery. New York: Basic Books.
Postal, P. (I968). 'Cross-over constraints'. Paper given at the Winter 1968 meeting of the
Linguistic Society of America, New York.
Saussure, F. (I962). Cours de linguistique ginerale. Paris: Payot.
Shuy, R., Wolfram, W. & Riley, W. K. (I967). A study of social dialects in Detroit.
Final Report, Project 6-1347. Washington, D.C.: Office of Education, 1967.
Stockwell, R. P. & Barritt, C. W. (i96i). Scribal practice: some assumptions. Lg 37.
75-82.
Trudgill, P. J. (197I). The social differentiation of English in Norwich. Unpublished Edin-
burgh University dissertation.
Wedge, G. & Ingemann, F. (I970). Tag questions, syntactic variables, and grammaticality.
In Papers from the Fifth Kansas Linguistics Conference. I66-203.
Weinreich, U., Labov, W. & Herzog, M. (I968). Empirical foundations for a theory of
language change. In W. Lehmann and Y. Malkiel (eds.), Directions for historical ling-
uistics. Austin, Texas: University of Texas. 97-195.
Wolfe, P. (I969). 'Linguistic change and the Great Vowel Shift in English'. Unpublished
University of California (Los Angeles) dissertation.

I20

This content downloaded from

202.43.95.117 on Mon, 12 Jun 2023 12:45:15 +00:00
All use subject to https://about.jstor.org/terms