¹ Rulon Wells, Immediate Constituents, Lg. 23.81–117 (1947), especially §53.

² K. L. Pike, Grammemic Theory, General linguistics 2.35–41 (1957). The reader should be alerted here not to confuse Pike's use of tagmeme with Bloomfield's earlier use of the same term. Pike's use has almost nothing in common with Bloomfield's—except that both men, at the time that they began to use the term, were searching for an operational concept that would prove basic to grammar just as the phoneme had proved basic to phonology and the morpheme to the lexicon.

³ Hypertagmeme as used here needs to be distinguished from Pike's earlier use of the same term (then called hypergrameme): ‘... the higher-layered unit beyond a grameme would be constituted of a hypergrameme, and a hypergrameme is comprised of the manifested forms of two (or more) gramemes ...‘ Language in relation to a unified theory of the structure of human behavior 1.120b (Glendale, 1954).

⁴ For two articles written under Pike's tutelage and using the tagmeme (formerly grammeme) on various levels, see Olive Shell, Cashibo II, IJAL 23.179–218 (1957), and H. Hart, Hierarchical Structuring of Amuzgo Grammar, IJAL 23.141–64 (1957).

⁵ Pike, on first devising the tagmeme as an operational concept, envisioned a simple straightforward division of the ‘utterance’ into tagmemes on analogy with the division of the ‘utterance’ into phonemes and morphemes, rather than a thoroughgoing hierarchical approach of the sort envisioned in this paper. See Language 1.136 ff., where Pike refers repeatedly to ‘utteremic-grammemic formulas’.

^5a Fred Householder, who recently read an earlier draft of this paper, makes the following comment: 'The cut between the and pioneer's family is a mistake; it must be between the pioneer's and family, then between the pioneer and 's, and only then between the and pioneer.' The illustrative value of the example is not materially affected by the choice of one analysis or the other.

⁶ Cf. fn. 1.

⁷ Clause needs to be defined with reference to the language that one is describing. A tentative definition of clause might be as a class of hypertagmemes of a hierarchical order ranking above such hypertagmemes as the word and/or phrase on the one hand and below such hypertagmemes as the sentence and paragraph on the other hand. It typically nucleates around a noncoordinate nonsubordinate construction and expresses some such relationship as predication or equation. For noncoordinate nonsubordinate constructions, see Nida, Morphology 95 (Ann Arbor, 1949).

⁸ In using English here for illustration, I am handicapped by the fact that English has never been studied from the point of view here proposed.

⁹ A tentative definition of phrase might be as a class of hypertagmemes of a hierarchical order ranking above such hypertagmemes as the word and/or stem on the one hand and below such hypertagmemes as the clause and sentence on the other hand. It may be of coordinate, subordinate, or noncoordinate nonsubordinate structure (cf. Nida, Morphology 94–5), and expresses such relationships as modification of a head, linkage of elements, or relation of an element to the clause by means of an overt axis.

¹⁰ I do not apologize here for suggesting no ‘discovery procedure’ for obtaining the tagmas of a language. On the other hand, it is hardly necessary to adopt the extreme position of Chomsky, who in rejecting ‘discovery procedure’ states that ‘it is unreasonable to demand of linguistic theory that it provide anything more than a practical evaluation procedure for grammars’ (Syntactic structures 51–2). Chomsky, in effect, tells us to write a grammar by hook or crook, then linguistic theory will help us evaluate it. The point of view here is that if we first obtain tagmas by hook or crook, linguistic theory will guide us in arriving at a tagmemic analysis and in writing a grammar.

¹¹ That such nouns preceded by ˀet- manifest a clause-level subject tagmeme with passive verbs is considerably obscured in Gesenius, Kautzsch, Cowley, Hebrew grammar (Oxford, 1910).

¹² This may be true even in some relatively opaque cases, such as the problem of ‘subject’ of active transitive verb versus ‘subject’ of passive verb in English. At first glance, there seems to be no good reason for considering that I in the first clause below is tagmemically different from I in the second: (1) I hit him; (2) I was hit by him. With a large enough corpus one would undoubtedly find that for all practical purposes the items manifesting one ‘subject’ manifest also the other ‘subject’. This results from two facts: the subject of an English passive verb is a transform of the object of an active transitive verb; and the subject and object of a transitive verb have no systematic difference in manifesting class in English. However, this very observation is fit cause for reflection: (1) if subject and object of the same clause type can have identical manifesting classes, it is evident that in principle two tagmas are not to be tagmemically identified simply because they have identical manifesting classes, provided that the function of the two tagmas can be shown to be different; (2) that the functions of ‘subject’ of active transitive verb and ‘subject’ of passive verb are different is demonstrable in that the latter is a transform of the object of the active transitive verb—and hence in some sense related to the latter—while at the same time occurring in the typical ‘subject’ position in the English clause. In brief, if we take grammatical transforms seriously, there is no reason to believe that a transform relation is a lesser factor in distinguishing the two ‘subjects’ than a difference in manifesting classes or in physical position of occurrence. I therefore posit an actor tagmeme as ‘subject’ of English transitive clauses, but a subject tagmeme in English passive clauses.

¹³ Ixil data are from an unpublished paper by Ray Elliott, Ixil Clause Structure.

¹⁴ The K'ekchi' data are from an unpublished paper by Francis Eachus, K'ekchi' Clause Structure.

¹⁵ The K'ekchi' tagmas here mentioned are posited on the basis of assumed meaning-function differences. Since in some languages such meaning-functions as cause, agent, instrument, and accompaniment are found to be structurally distinct, it was assumed that they might be distinct here as well. An alternative course of action would have been to have set up five tagmas each determined by the occurrence of a different particle. Either alternative leads to the same analytical result, provided that the data are duly inspected for tagmemic contrasts. Semantic distinctions that have no formal correlates will not, at any rate, survive in the finished analysis.

¹⁶ IJAL 23.129–49 (1957).

¹⁷ The Trique forms n·i ²³ ‘my mother’ and dre ²³ ‘my father’ are here considered to be phrases in that they consist of the stem forms n·i³ ‘mother’ and dre ³h ‘father’ plus a fused 1 p.s. ending which is mutually exclusive with ‘free’ forms. n·i ³ ‘mother’ plus this fused ending gives n·i ²³ ‘my mother’, in a construction parallel to n·i ³ ‘mother’ plus zi ²lihi ⁵⁴ ‘boy’, giving n·i ³ zi ²lihi ⁵⁴ ‘the boy's mother.‘ For a description of this and other fused elements in Trique see my Trique Tone Morphemics, Anthropological linguistics 1: 4.5–42 (1959).

¹⁸ This and similar elements are described under Macromorpheme 2 in Trique Tone Morphemics.

¹⁹ This morpheme could well be added as Macromorpheme 11, along with the ten macromorphemes already described in Trique Tone Morphemics.

²⁰ Cf. §5 and §5.6.

²¹ A tentative definition of sentence might be as a class of hypertagmemes of a hierarchical order ranking above such hypertagmemes as the clause on the one hand and below such hypertagmemes as the paragraph on the other. It may consist of a single clause, of a patterned combination of clauses, or of a clause fragment (usually of phrasal structure and often dependent in sense on other sentences in the linguistic or situational context). It tends to be characterized by more closure and grammatical independence than the clause, as evidenced by introducing and closing particles as well as the incidence of pause. Relations such as affirmation, question, and conditional propositions are typically expressed in sentence structures. For a discussion of problems raised by the occurrence of ‘dependent sentences’ see a forthcoming paper of Viola Waterhouse, Independent and Dependent Sentences.

²² In some respects both phonology and grammar seem to exhibit parallel sorts of hierarchical structuring. Among other considerations it may be pointed out that (1) while some levels in either the phonology or the grammar of a given language may be characterized by both minimal and complex strings, nevertheless, (2) one or more levels in either hierarchy may have the peculiarity of being characterized by only complex strings. Thus, while in Mixtec phonology a (noninitial) syllable may consist of but one segmental phoneme (plus a tone), the stress group—which is a higher level of phonological structure—must always consist of at least two syllables. In Trique phonology, on the contrary, an utterance such as o ²³ ‘I'm shelling it’ is (a) a syllable composed of but one segmental phoneme—although presence of phonemic pitch in this language again makes our illustration imperfect; (b) a stress group composed of but one syllable; (c) a breath group composed of but one stress group, etc. In brief, aside from the complication presented by the presence of pitch phonemes, the same phonological stretch in Trique may simultaneously manifest successive levels of the phonological hierarchy from the phoneme up. Hierarchical grammatical structuring in Trique is somewhat parallel to hierarchical phonological structuring in Mixtec in that there is one grammatical level in Trique (the lower sentence level, the level of clause-clause combination) which is necessarily complex. But aside from this one level, hierarchical grammatical structuring in Trique is rather parallel to hierarchical phonological structuring in the same language, in that such a Trique item as a ³mq³⁵ˀ ‘it's raining’ is simultaneously (a) a minimal word consisting of but a stem which manifests one word-level tagmeme; (b) a minimal phrase consisting of but one word which manifests one phrase-level tagmeme; (c) a minimal observation clause, which alone among Trique clause types may consist of a single verb manifesting one clause-level tagmeme; and (d) a sentence consisting of but one clause which manifests one sentence-level tagmeme—although as a sentence a ³mą ³⁵ˀ ‘raining’ would probably be followed by one or more sentence-final particles as in the following example: a ³mą ³⁵ˀ ną ⁴h a ³ ‘(it's) raining, sure enough!‘

²³ a ³mą ³⁵ˀ ‘raining’, ži ³gi̯ ³⁵ ‘shining’, ni ³⁵ ‘to become night, nightfall’, and a few other words referring to meteorological phenomena occur without subjects or actors in strings which seem in every other respect to be clausal in structure. Cf. a ³mą ³⁵ˀ ne ³⁴ˀ žeˀe ⁵ a ³kwα ⁴ˀ nï ⁴³ ‘(it's) raining outside right now’:

gą ³ˀą ³⁴h zi ³ ne ³⁴ˀ žeˀe ⁵ a ³kwą ⁴ˀ nï ⁴³ ‘he went outside right now’.

Our definition of clause in fn. 7 is expressly made broad enough to include such structures as the Trique observation clause: a clause is not necessarily a bipartite structure; it does not even necessarily contain a bipartite structure, but only ‘typically nucleates around a noncoordinate nonsubordinate construction and expresses some such relationship as predication or equation’.

²⁴ Sequences of this sort are considered to be one sentence in spite of their length and in spite of the possible rhythm break or pause which may occur between the clauses; for (a) the whole sequence seems to be but one structured unit—if structure is at least partially equatable with limitation in respect to selection of constituent parts—in that the verb is repeated in both clauses, with identity of referent of actor in both clauses as well; and (b) consistent absence of a conjunction between the two clauses is presumably significant, since most sentences in Trique narrative text begin with a conjunction or conjunction complex.

²⁵ Shell 182–84, 186–202.

²⁶ William Wonderly, Zoque Phonemics and Morphology, reprints from IJAL 17.1–9, 105–23, 137–61, 235–51; 18.35–48, 189–202 (1951, 1952). The pertinent sections here are §4.1 and §4.2, pp. 137–38, and also Table 2, pp. 139–40.

²⁷ An unpublished paper of Clarence Church, The Jacaltec Noun Phrase.

²⁸ Without discussing in detail the junctural phenomena involved here it may be noted that admission of such particles as the last three cited above to the category of bound forms would greatly augment the number of consonant clusters occurring in phonological words in Jacaltec.

²⁹ I have analyzed Trique verb phrase structure and noun phrase structure by the immediate constituent approach in an unpublished student paper (University of Pennsylvania, 1954). This older analysis seems to be not as elegant as the one here posited: (a) it occasionally separates phrase types which are hypertagmemically identical, while (b) it occasionally joins phrase types that are hypertagmemically distinct. As far as I am concerned, it obscures Trique phrase structure at several points.

³⁰ Shell 193a.

³¹ Zellig Harris, Yokuts Structure and Newman's Grammar, IJAL 10.203b (1944).

³² 202b.