The morphome in constructive and abstractive models of morphology

Although the definition and usage of the term ‘morphome’ differs in the academic literature, the original definition of a morphome by Aronoff (1994) is that it is a function which determines the distribution of form within the inflectional paradigm and beyond. More importantly, however, morphomes suppose the existence of what Aronoff terms ‘a morphomic level’ which embodies an empirical claim about the structure of language: ‘the mapping from morphosyntax to phonological realization is not direct but passes through an intermediate level’ (Aronoff in Morphology by itself, MIT Press, Cambridge, 1994:25). This is a strong claim concerning all types of morphological exponence. In this article, after an analysis of Aronoff’s model of morphology, I draw upon the distinction which Blevins (J. Linguist. 42:531–573, 2006) makes between constructive and abstractive models of morphology. I note that the introduction of a morphomic level merely constitutes an incorporation of a partial paradigmatic dimension into a constructive model of morphology. With reference to a semantically and phonologically unmotivated distribution of allomorphy in Romance I argue that although the introduction of the morphomic level is beneficial, since it can formalise systematic and ‘psychologically real’ homonymies in synchronic grammars in a way which is not merely coincidental, a constructive morphomic theory of morphology does not present significant advantages over abstractive theories as regards the formalisation of the synchronic facts and presents disadvantages as regards the motivation of diachronic tendencies which this particular distribution of allomorphy shows cross-linguistically. I argue that these historical tendencies support a more paradigmatic theory of morphology inherent in abstractive models, which consider the word as the basic unit of analysis and word formation a matter of analogical deduction based on the predictable capacity of stored exemplar paradigms and principal parts.