No CrossRef data available.
Article contents
The Identification of Morphemes
Published online by Cambridge University Press: 22 April 2026
Extract
Recent developments and publications concerning the analysis and classification of morphemes have been of great importance: they have provided a variety of new approaches to old problems, refined the statements of relationships, demonstrated the applicability of the method to a number of languages, emphasized the importance of distributional criteria, and pointed out a number of unsolved or perhaps only partially solved problems. All this has been a genuine gain. Nevertheless, a close examination of some of these developments is rather disturbing, and certain rather basic objections can be raised. For one thing, the term morpheme has come to be used by Hockett, and to a somewhat lesser extent by others, as a designation for forms which are related almost wholly in terms of distribution. I do not wish to deny the significance of tactics in linguistic structure; but when the primary criterion for the choosing of alternative possibilities is stated by Hockett as being ‘tactical simplicity’, I am inclined to doubt the validity of quite so much stress being laid upon distributional characteristics, especially when such a choice leads to the conclusions which Hockett has described. In both Bloch's and Hockett's recent papers on morphemic problems there is a conspicuous tendency to make covert distinctions more important than overt ones, though the treatment is quite different as regards detail.
Information
- Type
- Research Article
- Information
- Copyright
- Copyright © 1948 Linguistic Society of America
References
1 Zellig S. Harris, Morpheme alternants in linguistic analysis, Lang. 18.169–80 (1942); id., Discontinuous morphemes, Lang. 21.121–7 (1945); id., From morpheme to utterance, Lang. 22.161–83 (1946); Rulon S. Wells, Immediate constituents, Lang. 23.81–117 (1947); C. F. Voegelin, A problem in morpheme alternants and their distribution, Lang. 23.245–54 (1947); Charles F. Hockett, Problems in morphemic analysis, Lang. 23.321–43 (1947); Bernard Bloch, English verb inflection, Lang. 23.399–418 (1947).
2 Op.cit. 342.
3 [The use of the word ‘replacement’ evades the objection raised in Lang. 23.400.—BB]
4 The phonemic writing of English is here adapted to conform to my own speech, in order that all citations of English in this article may be consistent.
5 Leonard Bloomfield, Language 161 (New York, 1933).
6 Op.cit. 331.
7 Op.cit. 162.
8 I am indebted to my colleagues Kenneth L. Pike and William L. Wonderly for many helpful suggestions made during the development of the system described here.
9 Op.cit. 161 and 138.
10 For a definition of ‘common semantic distinctiveness’ see Principle 3, fn. 15.
11 In this paper the symbol ~ is used only to designate phonologically defined alternation. The symbol ∞ is used to relate alternants whose distribution is defined morphologically (see Principle 3).
12 The space left in writing between proclitics and nouns is not phonemic.
13 Morphemic alternants can conveniently be called allomorphs. Accordingly, allomorphs are related to morphemes as allophones are related to phonemes. In the process of analyzing a language there might be occasion to use the term morph to designate a structural unit which had not as yet been assigned to any morpheme; but in the description of a language (as distinct from the procedure of analyzing it) every structural element except features of arrangement is either a morpheme or part of a morpheme. Hence every element is also an allomorph or part of an allomorph.
14 In a descriptive grammar of Classical Greek such a statement concerning the potential sequences of aspirates would be included in the morphophonemic section, and would apply to all situations described in later sections where such sequences might arise.
15 To say that forms possess ‘a common semantic distinctiveness’ does not mean that all the occurrences of these forms must have an identical meaning, but rather that they possess some semantic feature in common which remains constant in all their meanings and which sets them apart from all other forms in the language. The phrase ‘a common semantic distinctiveness’ is essentially a negative formulation: it implies merely that the forms which possess that distinctiveness are in semantic contrast with other forms. It is this definition by exclusion that makes it possible to deal with certain sub-morphemic semantic differences, which will be treated under Principles 5–7.
16 It is necessary to distinguish between ‘immediate’ and ‘non-immediate’ tactical environments, and between ‘identical’ and ‘different’ tactical environments. The distinctions overlap. In the expressions this man and that boy, the forms this and that occur in identical immediate tactical environments. In the sentence This man hit that boy, the immediate tactical environments of this and that are identical, but the non-immediate tactical environments are different. In the expressions for this man and by that boy, both the immediate and the non-immediate tactical environments of this and that are identical; but in the sentence That boy worked for this man, the forms this and that again occur in different non-immediate tactical environments. In the sentences They left me and They left late, the non-immediate tactical environments of me and late are identical (both occur in subject-predicate constructions), but their immediate tactical environments are different: me occurs in a verb-goal construction, late occurs in a construction of verb plus temporal attributive. For any one occurrence of a morpheme there is usually only one immediate tactical environment; but the number of its non-immediate tactical environments is limited only by the number of sets within sets of immediate constituents in which that morpheme is included. Hence immediateness of tactical environment is determined by the analysis into immediate constituents.
17 Restriction 5 is covered by a separate principle (see §3.07).
18 This complementation holds good except for such phrases as It's me: It's I, for you and me: for you and I, which are treated as special problems under Principle 7.
19 The relation between I: me, we: us, she: her is not the same as the relation between any set of six completely unrelated morphemes in the language. Rather, me, us, and her constitute a type of replacive morpheme (see §4.32).
20 These hypothetical illustrations are intentionally extreme, since they are designed to point up the theoretical limits of the restriction.
21 I am indebted to John McIntosh of the Summer Institute of Linguistics for this illustrative material.
22 Note, however, that Bloch (Lang. 23.401) does not use this type of analysis, but combines all formally identical items.
23 There is admittedly a certain difference in meaning between ex- and the past-tense morpheme of verbs. But since our science does not provide us with an adequate measuring-rod for handling such differences, and since our basic definition of the morpheme is primarily negative and admits of sub-morphemic distinctions, my main reason for refusing to combine ex- in one morpheme with /-əd ~ -d ~ -t/ is distributional.
24 For the Makushi data I am indebted to Neill Hawkins, a research fellow of the University of Oklahoma and the Summer Institute of Linguistics.
25 This type of pattern furnishes a reason for choosing /-əd/ as the basic alternant in the series /-əd ~ -d ~ -t/, the past-tense morpheme in English verbs. Compare also the parallel atonic alternants of would and had.
26 Bloomfield, op.cit. 217.
27 Any science of classification becomes involved in the same difficulty. Thus, taxonomists in the fields of botany and zoology are faced with the same problem of determining the degree and the nature of structural likeness or difference which are to be regarded as significant in establishing classes and subclasses.
28 The change of vowel here need not be taken as a morpheme, since it is not independently meaningful (does not occur as the only difference between related forms, or as the only difference aside from a zero difference). The element /le . . . k-/ could be taken as a single discontinuous morpheme (cf. Harris, Lang. 21.121–7), except for the fact that one of the two parts occurs without the other. See §4.2.
29 The so-called direction of derivation (strive derived from strife or strife derived from strive) depends on factors which are not discussed here. See §4.32.
30 In this use of the term syllabic I follow Bloch (Lang. 23.414), who says, ‘we shall regard every syllabic as a unit, whether it consists of a vowel alone or of a vowel and a semivowel.‘ In my own English, only the postvocalic semivowels /y/ and /w/ combine with vowels to constitute single syllabics.
31 On the basis of Bloomfield's principle of over-differentiation (see §3.03), we might consider the second try a different morpheme, since in the verb to be the regular third-person form is is but after a principal verb such as move the form instead is be.
32 Any body of material sufficiently extensive to serve adequately as a basis for descriptive analysis must include fluctuations of forms, which are symptomatic of different degrees of acceptability. Synchronic contrasts of this kind are closely allied to contrasts of productivity vs. non-productivity.
33 I admit readily that the viewpoint here expressed is different from the traditional descriptive approach; but we are forced to treat fluctuations of forms in some way within the frame work of our descriptive statements. It does not seem to me wise to go to the extremes proposed by Bloch within the rigid confines of his method.
34 When two fluctuating forms are shifting their status—when the originally more frequent form is growing less frequent and the other is climbing up—there is undoubtedly a point at which the two are equally in use. At such a point one might insist that free variation actually exists between the forms. But the semantic difference between them derives from the whole pattern of the language; and even though the forms may have the same frequency statistically, they can and will nevertheless be different in meaning, if this is measured by their distribution in terms of the language as a whole.
35 One might also consider that the terms in this series exhibit sufficient semantic-phonetic resemblance to each other to justify a different treatment. One could take /hæv/ as the basic form, and regard the other forms as consisting of the morpheme /hæv/ plus various morphemes of replacement (/ə ← æ/) or loss (/— h/ or /— hæ/). Then the form /əv/ would consist of three morphemes: /hæv/ plus /ə ← æ/ plus /— h/. Such an analysis, however, not only would be awkward but would seem to be untenable in the absence of parallel formations.
36 The occurrence of particular allomorphs here is not determined solely by phonological criteria, nor does the characteristic distribution depend on purely phonetic features. The distribution involves both phonological and morphological criteria, and the alternation between allomorphs is accordingly symbolized by ∞.
37 The fluctuation between It's I and It's me is covered by this principle, but the form me is related to the overt series him, them, whom. See §3.03.
38 The expression ‘distributional classes’ is used to avoid the possible ambiguity of the term ‘form-classes’. The latter is now applied to two fundamentally different sets of forms: not only to distributional classes in the present sense, but also sometimes to classes of forms with a particular phonological structure (e.g. CVCV as against CV, or high-low tone as against low-high). Classes of the second kind may or may not be distributionally pertinent. There is often a correlation between the two kinds of classification, but it is nevertheless important to distinguish them.
39 Criteria for distinguishing ‘related’ and ‘distinct’ are given in the next section.
40 I admit that my classification is to some extent open to criticism, since I here classify in two or more morphemes certain forms which are identical in shape and related in meaning, but which do not occur in complementary distribution. It is true that such forms exhibit partial phonetic-semantic resemblances to each other. If I adopt the formal criterion of distribution as basic, it is because we have no technique as yet which will enable us to define the various degrees of semantic difference.
41 Cf. Hockett's discussion of this problem, Lang. 23.333.
42 Hockett treats a similar problem in his discussion of tense-mode and person-number elements in Spanish verbs, Lang. 23.338.
43 This paper makes no attempt to treat the significance of various classes of structure, such as derivational vs. inflectional formations, criteria for class determination, the relative order in descriptive statements, and the relation of constituent elements to the sets of immediate constituents. For these subjects see my book Morphology (Ann Arbor, 1946).
44 Juncture is intentionally excluded from this classification. Junctures are to be treated on the same level as order, which is the other formal feature of arrangement. [On juncture cf. Lang. 24.41 and fn. 34; on the dubious value of the distinction between segmental and suprasegmental see esp. Lang. 24.46 fn. 43.—BB]
45 See Kenneth L. Pike, The intonation of American English (Ann Arbor, 1945).
46 In some positions within the intonational pattern the phoneme of full stress may be reduced. [For a different view of stress in English see Rulon S. Wells, Lang. 23.108–14.]
47 The term ‘canonical form’ is Hockett's; see Lang. 23.333–4.
48 Zellig S. Harris, Linguistic structure of Hebrew, JAOS 61.143–67 (1941).
49 The situation with respect to French ne ... pas is not analogous. The element ne occurs without pas (in ne ... rien, ne ... point, ne ... que, etc.), and pas occurs without ne (in pas du tout, pourquoi pas, etc.).
50 The direction of derivation depends primarily on semantic criteria; but there are parallel formal patterns to guide the choice.
51 Bloomfield, op.cit. 217. William L. Wonderly has proposed in discussion that these French forms might be most economically handled as morphophonemic. Cf. George L. Trager, The verb morphology of spoken French, Lang. 20.131–41 (1940).
- 58
- Cited by