The productivity of ( non - ) productive morphology 1

Those of us who work with morphological productivity have fundamental principle which we apply in such discussions: some morphological processes are available, others are not. Those that are available may be used in the creation of neologisms, they are found (relatively) frequently in nonce-formations, they are in some sense psychologically salient, since the processes can be done and un-done (in the sense of comprehended) on-line in real time, and this leads to a relatively high number of types being found in corpora, with a relatively low number of tokens for each type. Those that are not available cannot be used in the creation of neologisms which will gain widespread acceptance in the language community. The result is that all words created with that process are already known (in some societal sense – they may not all be familiar to a given individual). All this being the case, we expect neologisms to be analysable in terms of existing processes, and we do not expect large numbers of neologisms or nonce words to arise which are formed in contravention of the known morphological rules. But what actually happens? In this paper I consider data from dictionaries of neologisms and my own files of observed words, to see how far the expectations are met. I focus on the creation of adjectives in English. The results are surprisingly uniform.