Parsing and Productivity1

It has often been argued that the (type or token) frequency of an affix in the lexicon cannot be used to predict the degree to which that affix is productive. Affix type frequency refers to the number of different words which contain an affix, token frequency refers to the summed lexical frequency of those words. The observation that neither of these counts relates straightforwardly to productivity, raises difficult questions about the source of different degrees of productivity, making the nature of morphological productivity one of the “central mysteries of word-formation” (Aronoff 1976:35). If productivity does not arise as a function of frequency, then where does it come from? This paper argues that frequency and productivity are, in fact, intimately linked. Type and token frequency in the lexicon are not good predictors of productivity. But frequency counts of decomposed forms in the lexicon can predict the degree to which an affix is likely to be productive. The problem with doing a straightforward frequency count of forms containing an affix, is that not all affixed forms contain the affix to the same degree. Some affixed words are highly affixed, and are highly decomposable (e.g. tasteless). Other affixed words appear more opaque, and tend to be characterised by whole word access, rather than parsing (e.g. listless). We argue that the former set facilitate productivity much more strongly than the latter set. Decomposed forms in the lexicon arise from parsing in perception. By coming to a clear understanding of the types of factors which tend to lead to parsing in perception, then, we can predict the degree to which an affix is represented by decomposed forms in the lexicon, and so (we argue), the degree to which it is likely to exhibit productivity. Thus, we argue that there is a strong relationship between parsing in perception, 1We are indebted to Andrew Carstairs-McCarthy, Wolfgang Dressler, Anke Luedeling, Janet Pierrehumbert, Ingo Plag and Robin Schafer, whose comments have greatly improved the quality and coherence of this paper. Remaining errors or incoherencies are the sole responsibilities of the authors.