Title: When allomorphy meets infixation: Cyclicity and separation
How are abstract (morpho)syntactic structures realized as linear phonological sequences? Is there a serial derivation separating (at least) morphology and phonology, or are morphological and phonological calculations made simultaneously? To shed light on these questions, I investigate crosslinguistic interactions between infixation on the one hand and three types of allomorphy—suppletive allomorphy (go/went), morphophonological allomorphy (leaf/leaves), and surface allomorphy (dog[z]/cat[s])—on the other hand. While much literature has established the behavior of each of these phenomena independently of the others (see, e.g., Moravcsik 1977, Carstairs 1987, 1990, Paster 2006, Yu 2007, Bobaljik 2012), I aim to show that a consistent and illuminating picture emerges, too, from examining cases where they overlap.
From a crosslinguistic survey of 28 languages where infixation interacts with one or more types of allomorphy, I propose five (to my knowledge) novel empirical generalizations:
(i) When a morpheme has multiple allomorphs, at least one of which is infixal, all the allomorphs orient with respect to the same edge of the stem.
(ii) Infixes never supplete based on their surface (infixed) environment.
(iii) Infixes undergo morphophonological and surface allomorphy exclusively in their surface (infixed) environment.
(iv) Suppletive and morphophonological allomorphy in the stem of infixation (the stem the infix combines with) are unaffected by the infix.
(v) Surface phonology within the stem of infixation (the stem the infix combines with) is affected by the infix.
Some morpho(phono)logical theories take the relative ordering of infixation and allomorphy to be variable (e.g., Wolf 2008), while others take them to occur simultaneously (e.g., McCarthy and Prince 1993a,b) or in a fixed universal order (e.g., Embick 2010, Bye and Svenonius 2012). The empirical findings summarized above point to the latter type of approach: exponents are chosen cyclically starting from the most deeply embedded node, exponent choice precedes infixation, and infixation precedes (morpho)phonology.