Rival morphological processes 4:

Where have all the rivals gone?

The central question I would like to answer is which mechanisms regulate the selection of the appropriate affix with a given base. The solution of this problem has important consequences for certain theoretical issues such as the role of paradigmatic mechanisms in morphology or the modeling of morphological processes. It will turn out that the distribution of affixes is largely governed by the individual properties of the processes in question, complemented only by token-blocking and local analogy. This analysis supports a sign-based view of morphology and challenges separationist theories like Beard's (e.g. 1995).

In order to investigate their rivalry, the processes will be compared with each other, starting with -ize and -ify. Before we turn to this task, I will briefly define what is meant by the term 'rival'. In general, morphological processes are regarded as rival if they are phonologically distinct but semantically identical. This phenomenon is also often discussed under the label of affix synonymy, and in separationist approaches to morphology synonymy is one of the central arguments for these theories. On closer inspection, however, it turns out that many of the putatively rival processes are not really rivals in this sense.

For example, Baayen and Lieber (1991) claim that the English adjectival suffixes -ous and -ish are rivals, one attaching to Latinate bases, the other to Germanic bases. Malkiel (1977) demonstrates however that the two are not synonymous, -ish deriving a qualitative adjective indicating similarity, -ous being purely relational. The notorious couple -ity and -ness is a parallel case, since, as convincingly argued by Riddle (1985), these two suffixes exhibit subtle semantic differences. A similar point is made by Doyle (1992) with respect to Irish nominalizations. From these studies one can draw the conclusion that the domains in which two processes are actual rivals are often much smaller than standardly assumed. In some cases no overlap exists at all, making the assumption of rivalry an artefact of an insufficient analysis. The remaining overlap between domains may be further curtailed by additional restrictions, e.g. phonological ones, with the consequence that the number of cases where there is indeed a choice between affixes is further reduced.

What does this mean for the verb-deriving processes in English? As already mentioned, the different affixes are not completely synonymous but merely overlap in meaning to varying degrees. This finding leads to a considerable reduction in the number of potentially rival forms. As we will shortly see, the phonological restrictions on the suffixes further diminish the potentially overlapping domains. The remaining truly rival domains are often very small, so that the number of actually competing derivatives is drastically diminished. It will become clear that in the remaining phonologically and semantically overlapping domains no additional restrictions can be discerned. In other words, the overall distribution of the affixes is only restricted by the particular combinatorial characteristics of each individual affix, i.e. its semantic and phonological properties.