The problem

It has often been noted that derivatives involving the suffix -ize in English are extremely heterogeneous in terms of their semantics, syntax and the types of bases the suffix attaches to. Thus adjectives and nouns may be verbalized by -ize, the resulting derivatives are transitive or intransitive, and they may have a whole range of different meanings, often paraphrased as 'render X, make X, convert into X, put into the form of X, give the character or shape of X, subject to the action, treatment or process of X, subject to a process connected with X, impregnate, treat, combine with X, act in a way characterized by X, imitate the manner of X' (cf., e.g., Jespersen (1942:319), Marchand (1969:320); X stands for the concept of the base word). Although this particular affix is generally regarded as the most productive overt verb-forming suffix in English, there is only one more detailed study of the semantic heterogeneity of -ize to date. In this study, Lieber (1996) proposes four different semantic structures for -ize, which are partly considered polysemous, and partly homophonous.

I will argue that all of the meanings suggested in previous studies are derived from one single semantic representation, which is claimed to be the underspecified Lexical Conceptual Structure (LCS) of possible -ize derivatives. It is shown that this LCS can not only account for almost every single -ize formation in the neologism corpus, but that it can also explain their relatedness in meaning in an explicit and straightforward manner. Furthermore, the polysemies of existing and possible derivatives can be predicted, which makes the claims easily testable against further data.

The case of -ize derivatives has implications for morphological theory, in particular the status of affixes and word formation rules in the lexicon, and the role of semantic and pragmatic information in word formation. The results of this study can be interpreted as evidence for an output-oriented model of the formation of -ize derivatives, in which the meaning of the derivative results from the interaction of the meaning of the stem with the semantic structure of possible -ize derivatives. An important consequence of the present analysis is that the syntactic category of the base is underspecified. This finding challenges the standard practice (generative and non-generative alike) of treating the information about the syntactic category of the base word as crucial for derivational processes.

Before turning to the analysis of the neologisms let us consider some of the earlier approaches. As mentioned above, standard sources like Jespersen (1942) and Marchand (1969) give a whole range of meanings for -ize derivatives, but do not try to explain the semantic relation between these meanings. The paraphrases they give often appear to be imprecise, or different paraphrases seem to encode rather identical concepts. Thus the two categories 'render, make X' (Marchand's example: legalize), and 'convert into, put into the form of, give the character or shape of X' (itemize) could easily be unified as change-of-states. Similarly, it is left unclear, what exactly the difference is between the paraphrase 'subject to the action, treatment or process of X' (propagandize), the paraphrase 'subject to a special (technical) process connected with X' (winterize), and the paraphrase 'impregnate, treat, combine with X' (alcoholize). A third group of paraphrases seem to be derivatives whose meaning is given as 'do as, act in a way characterized by X' (astronomize), or 'imitate the manner or style of X' (Petrarchize).

In a recent attempt to clarify this picture, Lieber (1996) proposes four different LCSs, which are given in (1):

(1) 

These structures show that "all -ize verbs are action verbs of some sort" (p. 8), i.e. they all share the first semantic function ACT, which Lieber borrows from Pinker (1989). In addition to this rather distant semantic relationship between all four types of derivatives, Lieber claims that at least the first three categories constitute a case of polysemy, because they have causative meanings. The exact nature of this polysemy is not further discussed. On the basis of the four LCSs, Lieber argues that the LCS of -ize derivatives is only partially determinate.

Lieber's account suffers from a number of weaknesses. First of all, it is unclear on which data she bases her claims. She mentions only 15 different derivatives altogether, of which very few are discussed in some detail. Furthermore, the status of these examples as productive derivatives is not discussed. Thus, it is difficult to assess the empirical adequacy of her claims. Second, the problem of polysemy is not solved by simply stating that different types of -ize derivatives share certain functions. How are these functions related? And why do -ize derivatives share just these functions and not others? Third, it is not clear why certain forms are cited as examples of certain LCSs. For example, why is unionize a case of (1a) and not of (1c), why is summarize a case of (1c) and not of (1a)? It seems that both analyses are possible, but if so, why should this be the case? And why is one structure, (1a), possible with verbal and nominal bases, whereas others are reserved for nominal bases? In addition, it is not obvious how some of her examples can be subsumed under the LCSs they are supposed to illustrate. For example, according to Lieber's 'Manner' structure (1d), economize would paraphrase roughly as something like 'act like (the) economy', which seems rather strange. Finally, it is unclear how a verb having a LCS as in (1d) can be used as a transitive verb, since there is no possible way to introduce a Theme argument into the LCS. Crucially, however, verbs of this kind are often, if not regularly used transitively, so that there should be an argument slot available for the object in the LCS.

In sum, the previous accounts of -ize are unsatisfactory in many respects. The following topic presents an attempt to develop an empirically and theoretically more adequate solution to the problems of -ize verbs.