WORD

ISSN: 0043-7956 (Print) 2373-5112 (Online) Journal homepage: www.tandfonline.com/journals/rwrd20

On Defining the Morpheme

Dwight L. Bolinger

To cite this article: Dwight L. Bolinger (1948) On Defining the Morpheme, WORD, 4:1, 18-23,
DOI: 10.1080/00437956.1948.11659323
To link to this article: https://doi.org/10.1080/00437956.1948.11659323

Published online: 04 Dec 2015.

Submit your article to this journal

Article views: 6631

View related articles

Citing articles: 7 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=rwrd20
 ON DEFINING THE MORPHEME
DWIGHT L. BOLINGER

Implicit in Bloomfield's definition of the morpheme as 'a linguistic form which

bears no partial phonetic-semantic resemblance to any other form', and of the
morpheme's semantic content (sememe) as 'a constant and definite unit of
meaning, different from all other meanings, including all other sememes, in the
language', 1 is the assumption that a given phonetic concourse is either entirely
with or entirely without meaning. The passage from the 'ultimate constituent'
to the 'meaningless sub-unit' is abrupt: there is no meaning at all below the
morpheme (if there is, we shall be suspected of not having analyzed far enough);
and there is as much meaning, qualitatively, in morpheme plus zero as in mor-
pheme plus morpheme plus morpheme ....
I believe that this concept of the morpheme needs examination because (1)
it is important to constituent analysis; (2) it is a crossroads between diachronic
and synchronic morphology; and (3) it shows the necessity for a more rigorous
treatment of meaning as it applies to synchronic analysis. I shall maintain:
(1) That the transition from sub-unit to morpheme is, as regards meaning, not
abrupt, altho there is a point below which we contemplate a world that is dead,
or nearly so, but above which the degree of fluidity, the degree of animation,
jumps upward at a rate far exceeding its increase elsewhere; this degree of
animation, by which I mean the statistically determinable readiness with which
an element enters into new combinations, is the only sure linguistic evidence that
the element has a meaning of its own. (2) That this increase continues as
utterances are augmented in extent and still further as they are combined with
other utterances, tho at an infinitely slower rate. In other words, instead of
describing the access of meaning as a rectangle open at the growing end and
closed squarely at the end that represents morpheme plus zero, I should diagram
it as a parabola with the narrowest part of the curve standing for a least element
capable of entering into new combinations (having meaning), and with the
morpheme either at the same point or slightly to the right of it. The slow
widening of the open end of the parabola is only a way of depicting the fact that
a + b will join more readily into new combinations than a or b alone, that
(a+ b) + (c + d) will combine more freely in new arrangements than a+ b
or c + d alone, etc.-in other words, that the linguistic environment of a smaller
unit tends to be more predictable than that of a larger unit. (3) That the mor-
pheme, as defined, is a variable, and scarcely easier to pin down than a word.
And (4) that the morpheme needs redefinition, as it represents at present a
curious survival of the confusion of contemporary and historical analysis.
The application of the definition that I have quoted is illustrated by the
statement that 'unhesitating is not a morpheme' .2 Since un-, -hesitat-, and -ing
1 L. Bloomfield, Language 161-162; New York, 1933.
2 R. S. Wells, Immediate Constituents, LANG. 23.81 (1947).
18
 ON DEFINING THE MORPHEME 19

are all encountered elsewhere with meanings similar to the ones that they reveal
here, unhesitating is not an ultimate constituent, or morpheme.
The difficulty arises when we attempt to deal with words like away or disease.
We know, of course, etymologically, that there are two components in disease;
but this knowledge is diachronic, and cannot be invoked in a synchronic analysis.
As far as the contemporary meanings of dis- and -ease are concerned, they are
irrelevant to the contemporary meaning of disease-it would be impossible for a
modern speaker of English to create disease out of dis- and -ease as we now use
them, as he might, for example, create de-hair or de-sugar. Stimulated by our
etymological information we may imagine to ourselves how the meaning of
disease developed from the combined meanings of its etymological components;
but this in no way represents any picture that the vast majority of the users of
the language carry about with them. If we were limited to usage we could no
more divide disease (as spoken, diziiz) into dis- plus -ease than we can divide
curfew into cur- and -few, or copper into cop- and -er. Etymology has undoubt-
edly motivated attempts at synchronic constituent analysis of many words, but
needs to be carefully separated from it.
Now when we pass from words like disease in which the combination is dif-
ferent semantically from its elements (and the difference is not attributable to
any tagmeme, such as order or modulation, but is a psychological transfor-
mation related to the frequency of the combination), to words in which the
combination is clearly the sum of the parts, such as unhesitating, we traverse a
zone in which there is every imaginable degree of relationship between the part
and the whole. In some, the relationship is dim-one scarcely knows whether to
affirm it, or to call the totality a morpheme-word; as indicated by the stressed
-sai- versus the unstressed -si- of motorcycle and bicycle, we seem to have two
cognate forms one of which is clearly separable and the other may or may not
be. This wavering continues all the way up into fairly complex combinations,
with of course fewer and fewer examples the farther we go. To most unso-
phisticated users of the language a short circuit has nothing to do with either
short or circuit (except in so far as the phrase itself has been clipped to a short) ;
and to not a few of these it has come to signify merely some kind of electrical
mishap, completely removed from even that technical meaning which might, on
reflection, be traced to short plus circuit. Ask one of these persons to account for
the contrast short circuit versus long circuit and he will only look astonished.
If we abandon the etymological standard of analysis we resign ourselves to the
fact that the cept of receptive, concept, and except is no more 'a morpheme',
synchronically speaking, than is the taf of taffeta, taffy, and distaff, for neither
meets the test of meaning. May we go a step farther? Suppose that a form
which under many conditions does meet the test of meaning, such as the re- of
recall, reclaim, rebate, return, remand, and a host of neologisms, under other
conditions has its primary meaning swallowed up, as in repertory, religion, recipe,
or again has it contrasting with itself as in re-creation versus recreation, or, finally,
relates to a secondary meaning of the etymon as in research. Unless we resort
to etymology there is no way to identify all these instances as a single morpheme.
20 DWIGHT L. BOLINGER

The re- of research ('diligent, intensive') of recall ('back'), of rewrite ('again'),

and of religion (zero) are, synchronically, merely homonyms or near-homonyms.
Besides identity and mere homonymy there may be partial synonymy, as in the
un- of undetermined and the un- of unwind, which are related as negatives but
distinguished by their peculiar connotations of 'yet to be' and 'in reverse'.
This raises the all-important question of WHOSE meaning, since meaning is the
criterion. Clearly in the speech of the person who says a three-wheeled bicycle
we cannot analyze bicycle. There are speakers of English who could never see
a resemblance between the com- of compare and the com- of compourul except the
resemblance of sound, tho these same speakers would readily note the kinship of
the co- in co-worker and co-defendant; there are others who might be taught to see
the connexion in compare and compound, but would never think of it otherwise.
For these people, who probably make up the bulk of the speakers of the language,
can we rightly say that com- (con-) is a morpheme? It is doubtful whether for
them any collocation of phonemes can be called a morpheme (as defined) unless
it is still an active formative in the language, such as un-, re-, anti-, de-, in many
or most of their combinations.
Obviously we cannot use meaning to determine an element in speech until we
decide whose meaning, and what kind of meaning, we mean. As for whose, it
can scarcely be other than that of the majority of speakers. As to what kind, it
should be the kind that the majority would recognize as constituting a basis of
similarity among complex forms that are otherwise dissimilar in meaning.
For the latter, we might speak of 'proper meaning', referring to the meaning
that can be assigned to a segment taken separately. The in- of infer and intense
would not have proper meaning because the majority of speakers would never
take it separately; it would not, therefor, be a morpheme. It follows that
proper meaning, as the determinant of a morpheme, is intimately connected
with freedom. If it is a bound form, the element must -in order to be a mor-
pheme-be active; for the moment that it becomes inert the new generation of
speakers take it merely as a sound element, not a meaning element, of the larger
signal. How proper meaning begins to dim the moment a combination becomes
stereotyped is illustrated by a class in which the twelve students attempting to
use the he (him) + who (that) construction to translate Spanish al que were
divided equally between those who called it he who (as verb object!) and those
who chose him who; fifty per cent made the wrong choice because he who had
partially lost its active relation to the independent use of the pronoun.
Attempts to identify morphemes by formal means will probably bear little
fruit, for juncture and stress are too erratic and bear no simple relationship to
meaning. The open juncture in an aim and the close juncture in a name do not
distinguish an as a morpheme and a as a non-morpheme, but rather distinguish
an aim from a name as wholes. Tho holiday is not a phonemic phrase by the
Bloch-Trager definition, the -day is a morpheme in the sense that most speakers
would immediately use the word day in defining it, and would associate the
similar sounds with similar meanings. Freedom rather than form is what marks
the morpheme, tho the form is affected in loosely predictable ways. An
 ON DEFINING THE MORPHEME 21

utterance can perhaps be speeded up until all open junctures disappear, and yet it
is understood, because its morphemes are identified thru memory. Holiday
has often been encountered alone, and it is tied to a similar verbal habit or
memory in birthday and washday; freedom in the sense of not being phrasally
bound, and freedom in the sense of the manipulability of its parts, both RE-
MEMBERED, enable us to identify it as a word and its ending as a morpheme.
We thus arrive at a definition of the morpheme which parallels that of the
word. If a word is a least element that can be used by itself, a morpheme is a
least element that can enter into new combinations. Potentiality for new
combination has two distinct advantages, as criterion for the morpheme. In the
first place, it enables us to replace the ill-defined meaning with a measurable fact,
the recurring appearance in new environments. In the second place, we shall
discover that it is necessary in our definition of the word; for if a 'minimum free
form' is one which merely HAS appeared in varied contexts, it would actually be
BOUND to those (extensive but finite) contexts; only its potentiality for new
combination keeps it from being phrasally bound. The actual number of new
combinations made out of any given morpheme may be extremely small, but the
appearance of only one in the lifetime of a speaker is still sufficient proof that the
element has proper meaning, that its user views it as something existing at least
partly to itself. Admittedly such a definition will not be altogether easy to
apply; but it is an improvement on the definition that it replaces, which is just as
difficult in application and is impossible in theory.
The definition, however, rules out meaningless residual forms as morphemes-
such as the cran- of cranberry-as well as etymological components. There is
no way by which they can be included without opening the door to forms that
we should not wish to include. . This is an inconvenience, as it flies in the face of
usage of the term morpheme, and the change would involve correcting too many
things already written. Since very little constituent analysis has been done, it
will be easier to leave morpheme alone and to give a new name to the KIND of
morpheme that I have described as pertinent to constituent analysis. I therefor
propose formative in place of morpheme as I have defined it, and component for
an etymological entity (as used by Bloomfield, component and constituent are
precise synonyms, so that we can utilize the surplus term), whence a morpheme
is 'a formative, residue, or component'. But we must remember, if this is done,
that synchronic meaning is no longer the criterion for the morpheme; tho meaning
of some sort there would be, whether diachronic or synchronic. Formatives
would include morpheme words (whence formative might be defined also as 'a
minimum active form'). (Tho there is wide duplication between residue and
component and between formative and component, we still need to distinguish
them, for there are residues which are not components such as certain portions
of discombooberate, and there are even formatives which are not components, such
as the -aroo as encountered in the originating word buckaroo).
Constituent analysis is more and more hemmed in as it moves from the open
end of the parabola toward smaller and smaller units at the closed end. Th<:>
smaller the unit, the more likely it is to be partially or wholly bound. I do not
22 DWIGHT L. BOLINGER

refer now to the kind of bondage which mechanically limits certain forms to one
or a few environments, without altering materially the value of the parts, such
as brand, adj., limited to new (or span new), or hard of limited to hearing; I mean
the bondage which makes the whole radically different from the sum of its parts.
By all means in present-day speech belongs to the focus class of yes or certainly,
not to that of in every way possible; it is even less analyzable than certainly, where
-ly affects certain- just as it affects glad- in gladly, whereas by that method is not
semantically parallel to by all means. How do you do? belongs to the focus class
of Hello, not to that of How do you know? Like nobody's business belongs to the
focus class of like sixty, like fun, like hell, not to that of like my brother. Why
don't you be careful'! (admonition or reproach) versus Why not be careful? (sug-
gestion or hypothesis)-unlike the semantically related Why don't you try it'!
and Why not try it? or What do you say we-'/ paralleling Suppose we-?, etc., are
to be analyzed or not depending on how much the analyzer insists upon fidelity
between meaning of the whole and meanings of the parts. The analyst will
generally elect to analyze, and rightly so, for he cannot assume the impossible
burden of identifying all the stereotypes in a language. Bondage-in the sense
of uniqueness of meaning-is virtually complete by the time we reach down to
the word, and quite complete when we reach the formative. It is true that there
are hints of meaning with vague resemblances of form at inferior levels, such as
the n of un-, in-, non-, nude, numb, nix, no, or the vowel of goof, boob, google, etc.,
the occupants of the sharpest part of the parabola that describes the access of
meaning; but constituent analysis should stop before it reaches this stage. Its
problems are too specialized to be included.
CONCLUSIONS

1. Constituent analysis as undertaken up to the present works successfully-

reflecting not perfectly but with a high degree of accuracy actual practise in the
language-in larger groups, less successfully as groups are made smaller, until, at
the point of the 'morpheme,' it breaks down a large part of the time. This
necessitates a redefinition of the morpheme so as to separate morphemes that are
valid for constituent analysis (formatives and residues) from those which are
valid for diachronic morpholoogy (components). The pluralizing component of
Cincinnati and the genitive component of Evans are irrelevant to a constituent
analysis of contemporary English.
2. The redefinition of the morpheme suggests a clarification of freedom and
bondage. This is called for also because bondage is now used in two different
senses: a sub-word formative is 'bound' by the very fact that it is not a word,
that is by never appearing alone, tho the variety of its environments may be
almost infinite; a word, on the other hand, is 'bound' when it is restricted as to
its environments. Let us say, then, that: (1) Components are locked in 'inert
bondage.' (2) Formatives and residues may be locked in 'active bondage.'
(3) Words are locked in 'phrasal bondage' when the combination is mechanical
and the meaning of the parts answers to the meaning of the whole.
The phrasally bound gob in Shut your gob has the same meaning as mouth in
 ON DEFINING THE MORPHEME 23

this context, and admits of the same ornamentation as in Shut your silly gob
(or mouth). Phrasal bondage is in tum divided into (a) 'complete phrasal
bondage' when the word is used only in enumerable combinations such as full
adv. in full well and full many or tapis in on the tapis; (b) 'partial phrasal bondage'
when the word is used only in certain types of context such as budge (largely in
negative contexts); and (c) 'complex phrasal bondage' (itself either complete or
partial) when a phrase is phrasally bound, such as the no uncertain of in no
uncertain terms (words, phrases) or suffice it in suffice it to say (point out). (4)
Words are locked in 'semantic bondage' when a set phrase, made up of words
which may or may not be perfectly free under other circumstances, has a meaning
which does not answer to the sum of the meanings of the .parts. Semantic
bondage comprises most of the so-called 'idioms' in the language. Since no
expression is ever quite as free as the focus class to which it belongs, semantic
bondage affects in greater or lesser degree every utterance in the language--this
is to say that (even disregarding supra-segmental modifications) the whole is
never quite the same as the sum of the parts. In practice, the difference can
as a rule be safely ignored. Thru analogic creation, any form of bondage may be
released into its corresponding form of freedom: the phrasally bound hard of
hearing (not hard of seeing, hard of smelling) may become 'Is your car hard of
starting?' 3 The inertly pound suffix in delicious, luscious may become actively
bound in galuptious, curvaceous, crematious.
University of Southern California
8 PDQ commercial announcement on Abbott and Costello program, 19 Nov. 1947.