As discussed in Chapter 1.2, the primacy of grammatical words in WP models is not due to any a priori preference for words over smaller (or larger) units of form but reflects the view that there is a sharply diminishing return from pushing morphological analysis below the word level. This utilitarian perspective is expressed succinctly by Robins (1959:128) when he acknowledges that “grammatical statements are abstractions” before continuing “but they are more profitably abstracted from words as wholes than from individual morphemes”. At a practical level, the descriptive utility of words is mostly taken for granted in teaching grammars, dictionaries and reference grammars. Even Bloomfield (1933) freely concedes the advantages that words enjoy over sub-word units for the preparation of these types of practical materials:
For the purposes of ordinary life, the word is the smallest unit of speech. Our dictionaries list the words of a language; for all purposes except the systematic study of language, this procedure is doubtless more useful than would be a list of morphemes. (Bloomfield 1933:178)
It is indisputable that written grammars and dictionaries impose constraints and introduce idealizations that have no counterparts in a speaker’s linguistic system. Nevertheless, it is not clear that the main distortion arises at the level of units, rather than in the correspondence between the organization of a description and the structure of a linguistic system. In particular, there appears to be no compelling reason to believe that the notions of compactness and complexity that apply to written descriptions have any direct relevance to the linguistic system described. The Bloomfieldian attempt to remove redundancy from the lexicon, like the classical goal of identifying unique principal parts for any word or inflection class, conforms to a grammar ‘design aesthetic’ that has advantages for written descriptions but no established linguistic or psycholinguistic motivation. Hence, as argued in more detail in Chapter 4.4, it is not the recognition of words as basic units that makes dictionaries unsuitable for “the systematic study of language”. Instead, the primary discrepancies lie in the idealizations that have been tacitly incorporated into the notions of the ‘lexicon’ or ‘mental lexicon’ since Bloomfield’s time. Foremost among these is the conception of a lexicon as a static, redundancy- free repository of discrete items, with no encoding of frequency or distribution.