A Quantitative Study of Spanish Paradigm Gaps

Paradigm gaps pose an interesting paradox in the generative capacity of native speakers. On the one hand, inflectional morphology tends to be automatic and prolifically productive, even for rare or made up words (Berko 1958; Bybee and Moder 1983). Once in a while, however, inflection fails for particular existing words. Pinker (1999) points out that even an inflectionally simple language like English has verbs for which many speakers cannot confidently produce a past tense form, such as forgo or bespeak. Similar examples can be found in many languages, including Spanish. The 1sg present indicative is virtually always marked in Spanish by the suffix -o, accompanied in some verbs by an additional change in the root vowel, or by insertion of a velar stop before the suffix: cant-ar/cant-o ‘sing-infin/1sg’, viv-ir/viv-o ‘live-infin/1sg’, cont-ar/cuent-o ‘count-infin/1sg’, sal-ir/sal-g-o ‘leave-infin/1sg’. In general, Spanish speakers have no trouble producing 1sg forms for rare or even made-up verbs (Albright, Andrade and Hayes 2001). For a handful of existing verbs, however, there is no 1sg present form, and all possible outcomes are deemed unacceptable. For example, for the verb abolir ‘abolish’, speakers are typically unsatisfied with any possible 1sg form (*abol-o, *abuel-o), and likewise for as-ir ‘grasp’ (*as-o, *as-g-o). In both Spanish and English, there is no apparent semantic reason why these particular forms should not exist, and for this reason, I will refer to the phenomenon as arbitrary lexical paradigm gaps (though I will ultimately argue that they are neither arbitrary nor lexical).1 As Hetzron (1975) observes, ∗ This work has benefitted greatly from the helpful comments and suggestions of many people, including especially Bruce Hayes, Junko Itô, Armin Mester, Jaye Padgett, Carson Schütze, Donca Steriade, Michael Wagner, Kie Zuraw, and audiences at UCLA, UCSC, MIT, and WCCFL 22. I am also indebted to Argelia Andrade for her help in collecting the experimental data that is reported here, and to the participants who took part in the study. All remaining errors and oversights are, of course, my own. 1. This phenomenon has gone under many names in the literature. A traditional term for such words is defective, but this fails to distinguish between forms that are missing for purely semantic/syntactic reasons (such as of impersonal verbs), and those with morphophonological difficulties. Fanselow and Féry (2002) and others adopt the more general term ineffability to refer to all cases in which the grammar fails to produce a usable output in syntax, morphology, or phonology. I use the term arbitrary