The combinability of derivational suffixes Conclusion

It has been argued that Fabb's generalization concerning English suffixation, expressed in the title of his article, is substantially flawed, both empirically and theoretically.

The analysis of a large amount of data has demonstrated that there is an abundance of counterexamples to many of Fabb's claims. Although it is a well-known fact that lexical rules are often subject to exceptions, this point cannot be evoked to explain away the numerous and often systematically occurring patterns since in Fabb's framework we are not dealing with rules, but with selectional restrictions, i.e. idiosyncratic information of individual lexical items (in this case: suffixes). Thus, even if we accept Fabb's analysis, the counterexamples are not exceptions to proposed rules but violations of lexical requirements as stated in the individual suffix's lexical entry. And systematic violations of such requirements are a rather unexpected phenomenon, to say the least.

In my view, however, the most important arguments against Fabb's proposed selectional restrictions are of a conceptual nature. Thus, even if we did not count the numerous "exceptions", many, if not all failures of suffixes to attach to already suffixed forms are the natural consequence of phonological, morphological, semantic, and pragmatic constraints that have to be stated anyway in a reasonably adequate description of English suffixation. These constraints, in particular base-driven selectional restrictions and general morphological constraints like the Latinate Constraint and blocking, regulate the applicability of derivational processes to given domains, ruling out a great many logically possible combinations of stems and affixes. The kinds of selectional restrictions proposed by Fabb have been shown to be either empirically and theoretically inadequate or simply superfluous.

From the discussion of Fabb's last group of suffixes one can conclude that generalizations like "affixes that do not attach outside any affix/that attach outside one other affix" are of no theoretical significance. The fact that there are suffixes which share the property that they only attach to one or to two other suffixes does not make them a natural class, nor a morphologically relevant one. In fact, according to the restrictions that apply across the board, we would have to posit groups of suffixes that attach outside three, four, five, six, seven, and so forth, suffixes. In the present account the number of possible preceding suffixes is a mere artefact of the phonological, morphological, and semantic-pragmatic constraints at work, and thus theoretically insignificant.

The rejection of Fabb's proposal has wider theoretical implications. One of these implications concerns the problem whether word-internal structure is accessible to morphological processes. As was mentioned above, Fabb's model rests crucially on the assumption that suffixes are sensitive to the morphological make-up of potential bases. The rejection of Fabb's selectional restrictions does, however, not entail that we argue against the accessibility of word-internal structure to morphological processes. The approach taken here is rather neutral with regard to this controversy, although some of the processes discussed (e.g. -ment suffixation to prefixed bases) seem to involve access to the morphological make-up of the base word. At least some morphological processes are best explained by assuming access to word-internal structure (Carstairs-McCarthy 1993).

Future research will have to determine when word-internal structure is crucially visible and when it is not.

Finally, we have made some potentially controversial claims concerning the nature of selectional restrictions in morphology. For many of the suffixes it was shown that the incapability of a given suffix to attach to a certain class of stems was not due to some assumed property of the suffix, but rather due to a crucial property of the putative base. Taking this point seriously means to question any strictly affix-driven approach as put forward by standard lexical morphology or other stratum-oriented approaches. Most recently, Giegerich (1998) has proposed a base-driven stratificational model of English morphology, in which he argues against affix-driven morphology on different grounds. His idea of the bas-drivenness of morphological processes is independently supported by our findings. It seems, however, that we do not need a stratum model to account for the intriguing phenomena.

The foregoing presentation was designed as a tour d'horizon through English suffixation, giving an outline of what I consider the most important structural mechanisms that restrict the productivity of derivational processes. The following subjects will offer a more microscopic view of the matter through an in-depth analysis of verb-deriving processes in English.