The Concatenative (IA) model

By reducing Bloomfield’s intricate taxonomy of forms and arrangements to a single hierarchy of part-whole relations, his successors arrived at a conception that was strikingly simple and transparent. At each linguistic level, elements could be organized into sequences that formed the primes for the next higher level.[1] The lowest, phonemic, level assigned classes of phones to phonemes. The morphemic level organized sequences of phonemes into morphemes. The syntactic level organized sequences of morphemes into larger constituents (Wells 1947). In this way, a uniform part-whole analysis could be extended from phone to utterance. Uniformity was not regarded solely as an end in itself, but reflected the Descriptivists’ practical and methodological interest in general procedures of analysis. A model in which levels differ solely in the nature of the elements they contain can obtain a complete analysis by applying general procedures of segmentation and classification. The morpheme lies squarely at the heart of this model. By combining sequences of sub-meaningful elements into meaningful units, morphemes provide the point of entry for meaning.

The resulting ‘Russian doll’ model is appealing in its sheer simplicity, so much so that it is still widely assumed in informal presentations of grammar and morphology, particularly in introductory textbooks.[2] However, in its most basic form, this model was immediately shown to be inadequate. A central problem arose in connection with even simple cases of stem allomorphy. An example considered by both Harris (1942) and Hockett (1961) involves English singular-plural pairs such as knife^knives and calf ^calves. The relevant feature of these pairs is just that the singular form ends in a voiceless fricative, /f/, whereas the stem of the plural ends in the voiced counterpart, /v/. The alternation cannot be treated as purely phonological, given pairs such as fife^fifes and gaff^gaffs, which represent the productive pattern. This pattern does not raise difficulties for Bloomfield, because he did not assume that morphemes were composed directly of phonemes. But in any model that does make this assumption, the fact that knife (/naif/) and kniv- (/naiv/) end in different phonemes means that they cannot be treated as the same morpheme, and hence that knife and knives share no morphemes in common. Bloomfield’s successors understandably regarded this outcome as unsatisfactory and realized that it called for a refinement of their initial assumptions.

Fortunately, a solution was near at hand. Hockett (1942) had earlier formulated principles of phonemic analysis that treated phonemes as abstract units, representing classes of phones with a non-contrastive distribution. Morphemic analysis could be established on exactly the same basis. A morpheme could be treated as an abstract unit, which represented classes of morphs with a non-contrastive distribution. Defining morphs as sequences of phonemes forged a more indirect link between morphemes and phonemes in a way that avoided the problems posed by morphologically conditioned allomorphy. To return to the earlier example, the morpheme {knife} would represent the two allomorphs /naif/ and /naiv/, whereas {fife} would represent just /faif/. Hence the morpheme {knife} could be a common element in the analyses of knife and knives, realized as /naif/ in the singular and as /naiv/ in the plural.6

The introduction of an abstract level mediating between morphological and phonological levels removed the main obstacle to the development of a general model of analysis based on simple classes and sequences. The removal of meaning from the grammar proper eliminated ‘sememes’ and ‘episememes’ from grammatical analyses. This led in turn to the collapse of Bloomfield’s distinction between meaning-bearing tagmemes and sub-meaningful taxemes. What remained then were just two fundamental elements: morphemes, representing classes of morphs, and phonemes, representing classes of phones:

Most linguists agree on the existence, or at least on the inescapable utility, of two kinds of basic elements in a language: morphemes and phonemes. (Hockett 1961:29)

6 This was not the only solution proposed at the time. An alternative discussed by Harris (Г942) and Hockett (2962) represented the final consonant in knife^knives by a ‘morphophoneme’ /F/ (though see Hockett (2987: §7.4) on ‘morphophonemes’):

We therefore create a morphophonemic symbol, say /F/, which represents /v/ before /-z/ ‘plural’ and /f/ elsewhere, and say that there is but one English morpheme /najF/. (Harris 1942:170)

Other alternatives include an item-specific “morphophonemic formula” that would ensure that “/f/ is replaced by /v/ before /-z/ ‘plural’ in the following morphemes—knife, wife, ...” (Harris 2942: 270). Although the specific mechanisms vary, all of these proposals introduce an abstract level mediating between morphemes and phonemes.

Given the reduction of syntax to selectional and ordering relations (Matthews 1993:148), words and larger syntactic constructions could be composed of sequences of morphemes. The model of grammar that emerges from these revisions is a recognizably modern, constituency-based model:

We summarize this by asserting that every language has its own grammar. The grammar, or grammatical system, of a language is (1) the morphemes used in the language, and (2) the arrangements in which these morphemes occur relative to each other in utterances. (Hockett 1958:129)

Within this model, morphological analysis reduces to the syntagmatic arrangement of morphemes. The general method of morphemic analysis is outlined in Harris (1942) and refined in Hockett (1947). This method proceeds in three steps. The first step examines the utterances of a language to identify “recurrent partials with constant meaning” (Hockett 1947:322). Those recurrent partials that are not composed of smaller forms are classified as morphs, or “morpheme alternants” in the terminology proposed by Harris (1942):

We divide each expression in the given language into the smallest sequences of phonemes which have what we consider the same meaning when they occur in other expressions, or which are left over when all other parts of the expression have been divided off... The resultant minimum parts we call not morphemes, but morpheme alternants. (Harris 1942:170)

The second step assigns morphs to a common morpheme if they satisfy the semantic and distributional criteria outlined in Section 2.2.3. The final step sets up the morphophonemic conventions that regulate the selection and shape of the allomorphs that realize a given morpheme in a particular context. In effect, these steps trace a loop that extracts morphemes from surface forms and then invokes rules to reassociate these abstract representations with their surface realization. At each step in this procedure the IA model is confronted with problems created by the disassembly and reassembly of forms.

  • [1] This conception is carried over into transformational models (Chomsky 1975).
  • [2] A good recent example is Bloomer et al. (2004:180), who explain that “each sentence consists ofclauses, each clause consists of phrases, each phrase consists of words, each word consists of morphemesand each morpheme consists of phonemes”.
< Prev   CONTENTS   Source   Next >