The Decade of the Morpheme
Ultimately, the problems that occupied the Post-Bloomfieldians during what Hock- ett (1987: 81) later termed “the Decade of the Morpheme” derived from their attempt to associate grammatical meaning with sub-word units rather than from the particular morphotactic assumptions they adopted or the specific techniques they employed. This orientation was part of Bloomfield’s morphological legacy, reflecting the influence of Panini and the Sanskrit grammarians on Bloomfield’s thought (Emeneau 1988). Bloomfield had famously regarded Panini’s concise description of Sanskrit as “an indispensable model for the description of languages” (Bloomfield 1929) and even appears to have attached theoretical importance to the economy of presentation achieved by Panini.
However, by the end of the decade Hockett had begun to distinguish the properties of languages from the properties of language descriptions. This led him to question whether techniques for achieving economy of exposition in a written morphemic analysis had any genuine status within the grammar of a language, or any relevance to the use or acquisition of language by speakers:
Morphophonemes, morphs, phones, and acoustic phones are artifacts of analysis or conveniences for description, not elements in a language. (Hockett 1961:42)
By 1967, Hockett had ceased to see morphemic analysis as anything other than a linguist’s concise shorthand for more psychologically plausible descriptions. For a more ‘realistic’ alternative, Hockett offers a description that conforms to a classical WP model, consisting of sets of word forms organized into paradigms and extended by processes of analogy:
To cover the complex alternations of Yawelmani by principal-parts-and-paradigms would take much more space than is occupied in the first sections of this paper by the morphophoneme-and-rewrite-rule presentation. But there would be a net gain in realism, for the student of the language would now be required to produce new forms in exactly the way the native user of the language produces or recognizes them—by analogy... A correct principal-parts-and-paradigms statement and a correct morphophoneme-and-rule statement subsume the same actual facts of an alternation, the former more directly, the latter more succinctly. We ought therefore to be free to use the latter, provided we specify that it is to be understood only as convenient shorthand for the former. (Hockett 1967:22if.)
Hockett’s endorsement of the WP model in part reflected his view that the problems that had arisen in elaborating morphemic analysis were artifacts of the Post- Bloomfieldians’ narrow frame of reference (Hockett 1987:82ff.).19
Within this frame of reference, the status of the morpheme was reinforced by parallels with the phoneme and by the fact that morphemes could be identified on the basis of procedures of distributional analysis. Indeed, the morpheme can be seen more as a product of a motivated method than as a motivated unit in its own right. Hence when examined outside this methodological context, the morpheme may strike the modern reader as a somewhat peculiar morphological unit. A description of a grammatical system must identify the properties that are distinctive within that system. But there is no clear reason for treating each individual property as a kind of ‘unit’. The strategy of parcelling individual properties into morphemes just appears to reflect the fact that the Post-Bloomfieldians had only rudimentary techniques at their disposal for representing feature information. A model that incorporates even simple feature ‘bundles’ (as nearly all contemporary models do) can represent individual distinctive properties within a larger set or structure. In such a model, there is no need to assume that every distinctive property has some discrete realization. Instead, it can be left as an empirical question which feature combinations are realized in a given language, and How they are realized: singly, cumulatively, multiply, etc. The observation that individual person and number properties have discrete realizations in a language such as
19 This position is echoed in the perspective that Matthews (1991) attributes to a proponent of a classical WP model who encounters a modern IA analysis:
An apologist for ancient grammar would answer that these elements [morphemes] are fictions. They are created by the modern method; and, if we foist them on a flectional system, we are bound to describe it as an agglutinating system that has somehow gone wrong. (Matthews 1991:204)
Turkish does not entail that the same features must have discrete realizations in, say, Russian any more than the fusional realization of person and number features in Russian determine an ‘underlyingly fusional’ structure for Turkish.
In a contemporary setting, there is neither formal nor empirical motivation for the assumption that individual properties should be parcelled into ‘units’. The idea that properties should be realized exactly once in a form seems even more out of place in a model of linguistic systems that serve a communicative function. It is well known that distinctive phonological contrasts tend to be multiply cued: an obstruent voicing contrast, for example, may be cued by voice onset time differences at consonant release, by durational differences in consonant closure, and by durational differences in preceding vowels. The same advantages accrue in the morphological domain, where multiple exponents of a grammatical property enhance communication over a noisy channel and contribute to robustness in general. It is unclear why one would make the apriori assumption that each distinctive property should be uniquely realized, let alone build such an assumption into the architecture of a morphological model. It is thus unsurprising that this assumption is falsified by patterns of extended exponence in various languages, and enigmatic that it was ever seriously entertained.