¹ ‘The non-uniqueness of phonemic solutions of phonetic systems’, Bulletin of the Institute of History and Philology, Academia Sinica 4:4.363–97 (1934); Readings in linguistics 38–54.

² Louis Hjelmslev, Prolegomena to a theory of language, translated by Francis J. Whitfield, revised English edition (Madison, 1961), 46.

³ Leonard Bloomfield, ‘A set of postulates for the science of language’, Lg. 2.153–64 (1926); Readings in linguistics 26–31.

⁴ See, for example, Morris Halle, The sound pattern of Russian (The Hague, 1959); Noam Chomsky, Current issues in linguistic theory (The Hague, 1964) ; Morris Halle, ‘Phonology in generative grammar’, Word 18.54–72 (1962); Noam Chomsky and Morris Halle, ‘Some controversial questions in phonological theory’, Journal of linguistics 1.97–138 (1965).

⁵ There may be some doubt as to whether this point is universally accepted, since one school of phonological thought, that of Chomsky and Halle (cf. footnote 4), has provided no direct means of accounting for such patterns of arrangement. An indirect account of some phonotactic features is to be provided by ‘redundancy rules’, which add features to individual segments, but it has not been shown how or whether the facts of the phonotactics of a language in general (e.g. the structure of the English syllable) could be described by such means. Nevertheless it is clear from various remarks in their discussions of redundancy rules that they would like to be able to account for such facts.

⁶ Cf. Roman Jakobson, Gunnar Fant, and Morris Halle, Preliminaries to speech analysis, M. I. T. Acoustics Laboratory Technical Report No. 13 (Cambridge, Mass., 1952).

⁷ Chomsky and Halle, Journal of linguistics 1.119 (1965): ‘We conclude ... that only feature notation has linguistic significance, and that segments are simply to be regarded as conventional abbreviations, utilized to cope with the exigencies of printing but having no linguistic significance in themselves.‘

⁸ Noam Chomsky, Aspects of the theory of syntax (Cambridge, Mass., 1965), 84: ‘ ... the lexicon is a set of lexical entries, each lexical entry being a pair (D, C), where D is a phonological distinctive feature matrix “spelling” a certain lexical formative and C is a collection of specified syntactic features ... ‘

⁹ I.e. the items which are treated by the tactics as atomic, items whose combinations are accounted for by it but whose composition, if any, is not.

¹⁰ Cf. Morris Swadesh, ‘The phonemic principle’, Lg. 10.117–29 (1934), Readings in linguistics 32–7; and numerous other articles. This principle, which follows from the distinctiveness principle discussed below, has been criticized by Chomsky and Halle (e.g. Journal of linguistics 1.128–30) on the grounds that if it is applied as the sole criterion in phonemic analysis, absurd solutions will result. It should be obvious that the principle was not intended to be applied in isolation from all others.

¹¹ Cf. footnote 2.

¹² Zellig S. Harris, ‘Simultaneous components in phonology’, Lg. 20.181–205 (1944); Readings in linguistics 124–38.

¹³ Bernard Bloch, ‘A set of postulates for phonemic analysis’, Lg. 24.3–46 (1948); ‘Studies in colloquial Japanese IV: Phonemics’, Lg. 26.86–125 (1950); Readings in linguistics 329–48.

¹⁴ Samuel E. Martin, ‘Korean phonemics’, Lg. 27.519–33 (1951); Readings in linguistics 364–71.

¹⁵ Charles F. Hockett, ‘Peiping phonology’, JAOS 67.253–67 (1947); Readings in linguistics 217–228; ‘Componential analysis of Sierra Popoluca’, IJAL 13.258–67 (1947).

¹⁶ Cf., for example, the treatments of English by Leonard Bloomfield, Language (New York, 1933), 130–5, and Benjamin Lee Whorf, Language, thought, and reality (New York' 1956), 223–30.

¹⁷ Charles F. Hockett, A manual of phonology, IJAL Memoir 11 (1955).

¹⁸ Cf. Current issues in linguistic theory (see footnote 4), 75ff.

¹⁹ Chomsky makes the surprising assertion, for which he offers no support, that ‘the invariance condition has no clear meaning unless the linearity condition is also met’, Current issues in linguistic theory, 79.

²⁰ Current issues in linguistic theory, 78. Evidently it is also to be understood that, according to this condition, phonemes and phones may occur only in linear sequences. Chomsky incorrectly asserts that the linearity condition follows from definitions of the phoneme as a bundle of distinctive features. It would follow only if there were requirements that (1) features occurring simultaneously may not belong to different bundles, and (2) no bundle can be discontinuous.

²¹ See the following sections (cf. footnote 13): 48.2, 52.4, 53.2, 53.6, 56.1, 56.2, 57.2. Note also the following statements from ‘Studies in colloquial Japanese IV: Phonemies’: 'The order of phones in a phrase is successive, simultaneous, or overlapping: segments occur only in succession; but spans, which are coextensive with qualities common to a train of several segments, occur simultaneously with the segments that compose the train, and may overlap with other spans.' 'Each phoneme is defined by the quality or combination of qualities present in all its members and absent from all other phones of the dialect; and every phone that contains such a quality or combination belongs to the phoneme which is defined by it. A phone that contains two or more such qualities or combinations accordingly belongs to two or more phonemes at once. '

²² Current issues in linguistic theory, 80.

²³ Lg. 27.340. The example chosen by Hockett in his discussion is a hypothetical one which is perhaps unrealistic in that it assumes no phonetic difference between prejunctural and postjunctural [t].

²⁴ This statement in no way denies the value of working out and teaching practical procedures of analysis. The point is that such procedures should be distinguished from expositions of linguistic theory and from criteria for evaluating proposed descriptions. Linguistic theory and practical linguistic analysis both benefit from such separation.

²⁵ Morris Halle, The sound pattern of Russian (The Hague, 1959), 22–3.

²⁶ For a more transparent example of this argument than Halle's Russian one, see Chomsky's discussion of the analysis of the centering offglide of English, Current issues in linguistic theory, 104–5.

²⁷ In other words, if a provisional solution had an element of voicing or unvoicing with each segment, we would find it possible to eliminate it on all but one of them.

²⁸ Also, it would be more difficult to handle the fact that v may follow voiceless obstruents if the voiceless condition were taken as neutral.

²⁹ Previously presented in my paper ‘On redefining the phoneme’ delivered at the 1963 annual meeting of the Linguistic Society of America.

³⁰ Current trends in linguistics III (The Hague, 1966), 119–20.

³¹ The graphic notation and associated terminology used in my article ‘The sememic approach to structural semantics’, AA 66:3:2.57–78 (1964) are related to the present version as follows: ‘vertical’ there corresponds to ‘or’ here, ‘horizontal’ to ‘and’; since these two types of relationship were distinguished by the two dimensions in the earlier notation, diagonally upward was used for what is here simply upward. Figure 1 illustrates composite realization; Figure 2 shows diversification, neutralization, zero realization, and empty realization.

³² For a fuller discussion of the differences between process and realization, see my ‘On alternation, transformation, realization, and stratification’, Georgetown University monograph series on languages and linguistics 17.105–22 (1964).

³³ Cf. footnote 8.

³⁴ In my Outline of stratificational grammar (Berkeley, 1962) ; ‘Stratificational linguistics as a basis for machine translation’, in Bulcsu Laszlo, ed., Approaches to language data processing (The Hague, in press); and ‘Kinship terminology and linguistic structure’, AA 67:5:2.37–64 (1965). See also the discussion of simplicity and related concepts in my ‘Epilegomena to a theory of language’, Romance philology 19:4 (1966).

³⁵ Graphic and algebraic notations may be designed to be mechanically interconvertible (cf. graphs of conic sections and the associated quadratic equations), so that they differ only as alternative notational implementations of the same underlying set of descriptive concepts. The implementation of the simplicity principle, which must yield comparable results for the alternative notations, is easier for the graphic notation than for the algebraic, since with algebraic notation it is necessary to provide for ignoring nonsignificant differences that result from mechanical limitations of algebraic notation, such as its essentially linear character. In fact the easiest way to implement the simplicity principle for the algebraic notation of stratificational theory is to specify the implementation for the graphic notation plus the rules for converting from the algebraic to the graphic.

³⁶ It might be supposed that nodes should be counted rather than (or in addition to) lines, but nodes are not mutually comparable in a simple way since they can differ from one another in complexity: a node with six lines leading out of one side is more complex than one with only two.

³⁷ Cf. my ‘On alternation, transformation, realization, and stratification’ (see footnote 32). Ordering of subrules of a realization rule often makes it possible to state conditioning environments more simply than if they were unordered, but not always. Nevertheless, as a matter of purely practical convenience, I have used the convention of always writing the subrules of a rule in some order (i.e. even when simplification does not result). In the graphic notation, however, ordering in an ‘or’ is specified only where relevant.

Realization rules as described in that paper have been simplified considerably as a result of work done since it was written, particularly that reported here.

³⁸ Cf. footnote 17.

³⁹ The listing omits various additional alternations related to juncture.

⁴⁰ For example, in my dissertation on North Fork Monachi, the anataxis of (10) and (11) was described as having effect between morpheme combinations and morphophoneme combinations. Thus the term morphophonemic was applied to the sequences V?BV (10), V?RV (11). Then the morphophonemic realization rule corresponding to (2) was applicable to (10). That treatment thus anticipates in part the formulation arrived at here.

⁴¹ On the suffix -on, see my ‘The sememic approach to structural semantics’, AA 66:3:2.57–78 (1964). Downward components must be distinguished from upward components. For example, the morpheme ^M/worse/ has morphons (^M/w/ etc.) as downward components, but has the lexons ^L/bad/ and ^L/er/ as upward components.

⁴² The morphophoneme ^M/</ is realized as a specification that the realization of the following morphon occur in the immediately preceding syllable (in the position specified for it by the syllable structure tactics). Examples are: (10) ^M/yaka <?'ki/ ^bp/ya?ka'ki/ ^cp/ya'ka'ki/ 'to ciy intermittently', (11) ^M/miya <?'ki/ ^bp/mi?ya'ki/ ^CP/mi?ya'ki/ 'to go little by little', ^M/ki'ma <?'ki/ ^BP/ki?ma'ki/ ^cp/ki?ma'ki/ 'to come little by little'; (12) ^m/yaka ^bp/yaki/ ^cp/yaki/ ‘to cry at’.

⁴³ A realizate is that which is realized—that which a realization is a realization of. Cf. my ‘On alternation, transformation, realization, and stratification’ (see footnote 32).

⁴⁴ Thus the statement of conditioning environment in a realization rule is actually only a repetition of part of the information already present in the associated tactic pattern; it is, however, a nonsignificant repetition of information stated only once in the graphic description.

⁴⁵ Chomsky, Current issues in linguistic theory 68–9: ‘So far as I can see, there is no other significant level that can be isolated in the phonological component. ... the existence of an additional level is highly dubious ...‘

There is no simple definition possible of just where ‘their type of morphophonemic level’ is, since it is characterized by them in several different and mutually contradictory ways. In one sense it is a conflation of everything from basic phononic to lexemic.