We must furthermore be guided by the frequent warning against an absolute imposition of the formal structure of a language on a description of semantic structure. For example, when the focus is on componentially isolating such features as ‘ plural ’ and ‘ male ’ in the meaning of men, it is irrelevant that ‘ plural ’ is given a separate morphemic status for men (but not for people) in the description of formal structure to account for certain syntactic patterns.

The morphemic question of homonymy and polysemy also assumes a different complexion. Take raise ‘rear, bring up’, raise ‘cause to rise’, and lift. A generative semantic theory would include metalanguage symbols for meanings or components. A certain combination of such semantic symbols would ultimately be rewritten as the object-language form raise. A second combination would also produce a form raise. A third combination, more similar to the second than is the first, would yield lift. That the first and second sets of rules both generate raise is a fact of the formal side of the language, to be noted, for example, in accounting for lexical ambiguity (e.g., He raised the child). By historical criteria, raise is treated as one polysemous form and ear (for hearing) and ear (of corn) as homonyms. Our purely synchronic approach makes no such distinction, treating all as homonymy. Eliciting techniques that tap speakers’ intuitions about polysemy and homonymy would result in preserving the distinction, though not necessarily where historical criteria would: some speakers might consider ear one form, application to a spike of grain being a metaphorical extension (‘ because it sticks out of the stalk like an ear ’). But polysemy from extreme historical divergence which speakers reinterpret as homonymy, or folk etymology which results in the reverse, supports the treatment of speaker intuition about the distinction under a different aspect of linguistic description.^a

In the dictionary, then, each definition would be a separate entry represented by its own form. Two different definitions that share some semantic components may or may not be represented by identical forms, but their relation to one another in the semantic system remains the same. Since the meaning of a polymorphemic lexeme is by definition unpredictable from the meanings of its constituent morphemes, the unit form to be defined is the lexeme, whether mono- or polymorphemic, again keeping semantic and formal structure separate.

¹ This thought is a departure from the original monograph, p. 12, last paragraph, and is further clarified by the discussion of cognition. Further against polysemy see Weinreich 1966 and McCawley 1968: 125-7.