The meaning of -ize derivatives Conclusion

In the foregoing discussion I have put forward a unified analysis of the meaning of the suffix -ize and its derivatives in English. The investigation of a large corpus of neologisms has provided strong evidence for the semantic coherence of -ize formations, a coherence hitherto not recognized in the linguistic literature. On the basis of the LCS suggested in (28), the polysemy of -ize verbs can be described, predicted and formalized in a straightforward manner. The semantic interpretation of a given derivative can be construed by mapping the different participants and the base onto the semantic representation as expressed in the LCS.

This analysis has certain advantages over previous approaches. First, the proposed model can not only account for the semantic diversity between different derivatives, but can also explain the attested or possible divergent meanings of individual formations, which is a problem largely neglected by previous authors. Second, it was shown explicitly how the different meanings of -ize formations are related, with the consequence that the apparent semantic diversity of productively formed -ize derivatives turned out to be a case of polysemy, not homonymy. Third, the specification of the syntactic category of the base word was shown to be superfluous, which leads to a welcome simplification of the morphological machinery. These findings have a number of implications for some theoretical issues in morphology such as the status of affixes and word formation rules in the lexicon.

The analysis of -ize derivatives put forward here provides strong arguments against purely syntactic models of derivational morphology (such as E. Williams 1981a, 1981b, Lieber 1992, Hale and Keyser 1993). These arguments are essentially the same as those provided and discussed in Lieber's (1996) analysis of -ize, the main point being that the intricacies of the semantics cannot be captured by purely syntactic formalisms. Since this point is laid out convincingly and in considerable detail in Lieber (1996), it need not be elaborated here, and I will focus my attention on another theoretical problem, the format of word formation rules.

In standard generative approaches word formation rules contain, among other things, information on the semantics of the suffix and the syntactic category of the base. Furthermore, word formation rules are considered as operations on a possible input with a predictable output. Both assumptions are called into question by my findings.

Consider again Gussmann's (1987) approach, which was already briefly mentioned. He claims that there is a so-called rule of derivation, i.e. a "uniform semanto-syntactic formula whereby adjectives become verbalized" (1987:82) with the meaning 'make (more) X'. This rule of derivation is spelled out by the rival affixes -ize, -ate, -ify, etc. While the postulated meaning 'make (more) X' is intuitively plausible, these rules can neither account for the other widely-attested meanings nor can they explain why many of the verbalizing affixes (including -ize) also take nouns and bound roots as their bases. In Gussmann's framework one would have to posit yet another - denominal - rule (plus a number of additional derivational rules that express the other possible meanings). Streamlining the denominal and deadjectival rules by making reference to a syntactic feature like [+ N]¹common to both adjectives and nouns would achieve greater descriptive adequacy, but would not eliminate the stipulation as such. Under the approach presented above nothing needs to be stipulated, but it follows from the LCS itself that the syntactic category may be a noun or an adjective, and not, say, a verb.

Thus, the case of -ize strongly challenges the traditional practice (generative and non-generative alike) of treating the information about the syntactic category of the base word as crucial for derivational processes. Generalizing from -ize, one could even come up with the strong hypothesis that with any given productive affix, the syntactic category of potential base words is only a by-product of the semantics of the process. Such a hypothesis would find some support in a recent study of the person noun forming suffix -ee, in which it is shown that some existing denominal derivatives can naturally be accounted for if the semantics of -ee is properly defined (Barker 1996²). Future research will show whether such a strong hypothesis is indeed tenable.

The problem of the syntactic specification of the base word is symptomatic of a more general problem, namely the architecture of word formation rules. Such rules are usually seen as operations on a more or less well-defined class of possible input structures which derive a class of possible output structures. The account of -ize I propose here shifts emphasis from the possible input to the possible output by rephrasing the traditional input-output relation between base, affix and derivative as a general condition on the output, with desirable empirical and theoretical consequences. Such an analysis of derivational morphological processes has been advocated earlier for example by Plank (1981), and, in regard to inflection, by Bybee and Slobin (1982), Bybee and Moder (1983), and links up with similar recent trends in other areas of linguistics. Thus, in phonology and syntax frame-works like Optimality Theory or the Minimalist Program (e.g. Prince and Smolensky 1993b, Chomsky 1993, respectively) incorporate a major shift from derivational to representational mechanisms. That output-oriented approaches lead to new insights into the nature of morpho-phonological alternations has already been shown in a growing number of studies³, when the phonology of -ize derivatives will be dealt with. The findings presented above call for an extension of representational mechanisms to the syntactic and semantic aspects of morphological processes.

¹ Following Aronoff (1976:48), and Scalise (1984:139).

² Note that Barker himself does not advocate this in his paper. Rather, he assumes that the denominal forms are based on an underlying zero-derived verb even if such a verb is not attested (e.g. festschriftee < *festschrift v + -ee). This analysis seems rather ad hoc, especially in view of the fact that the pertinent data can be accounted for in the purely semantic terms developed by Barker.

³ See e.g. McCarthy and Prince (1993b), Golston and Wiese (1996), Neef (1996), Raffelsiefen (1996), Booij (1998), and many of the papers listed in the Rutgers Optimality Archive.