Modeling ungrammaticality in optimality theory (review)

The articles in this collection address an old problem that has received renewed attention in the recent linguistic literature (e.g. Baerman et al. 2010), namely, how to account for the fact that we find gaps in otherwise productive morphosyntactic constructions. For example, in English morphology, even though all verbs are expected to have a past-tense form, native speakers avoid using forego in the past tense, as no grammatical form seems to be available. Similarly, in English syntax, multiple wh-questions are grammatical in some cases (Who bought what?), but not in others (*What disappeared why?). It is a problem for all current linguistic models to account for such gaps, since one must explain not only why the usual rules or constraints are blocked in these contexts, but also why the grammar provides no alternative way of filling the gap: the construction remains ineffable.

For the optimality theory (OT) framework, the problem of gaps appears to represent a fundamental challenge, as the editors' introduction, 'Modeling ungrammaticality', makes clear. Since OT constraints are violable, some optimal output candidate will always emerge for any input. A special formal strategy must be found for optimizing the null output (i.e. a gap) over any possible repair. The eight papers in this volume take up the challenge of how to optimize gaps in OT, providing examples from a variety of languages, including Dutch, German, Hebrew, Hungarian, Norwegian, Russian, Spanish, and Turkish, as well as English. The range of constructions covered includes not only inflectional paradigms but also productive derivational morphology, such as diminutive formation (Dutch and Hebrew) and 'stress shifting' nominalizations (English), as well as the syntax of wh-questions.

All but three of the papers (those by Albright, Legendre, and Vogel) analyze phonologically motivated gaps, and show that the motivations fall into three categories: phonotactic violations, the obligatory contour principle (OCP), and minimality or maximality constraints. The Norwegian imperative paradigm discussed in Rice 2005, 2007 and reanalyzed in this volume by Matthew Wolf and John J. McCarthy, in 'Less than zero: Correspondence and the null output', illustrates a gap motivated by phonotactic violations. Normally, imperatives in Norwegian are identical to their roots. When the bare root ends in an unsyllabifiable consonant sequence (e.g. *sykl 'bike!'), it has no imperative form (if it occurs sentence-finally), even though, in nouns, identical consonant sequences are repaired through epenthesis (e.g. /sykl/ → [sykkel] 'bike'). Similarly, in 'Covert and overt defectiveness in paradigms', PÉter RÉbrus and MiklÓs Törkenczy show that defective paradigms arise in Hungarian with morpheme combinations that contain unsyllabifiable consonant sequences. A gap occurs when no syllabification repair (like epenthesis) is employed, although repair strategies are employed when similar consonant sequences occur in other morphological contexts.

Phonological motivation for gaps can also come from the OCP: if the base and affix are phonologically (nearly) identical, affixation is blocked. As Outi Bat-el notes in 'A gap in the feminine paradigm of Hebrew: A consequence of identity avoidance in the suffix domain', these gaps are closely related to those created by haplology, except that complete identity between the affix and the adjacent string is not required. Instead, identity between a single segment in the affix and an adjacent morpheme suffices to trigger OCP-related effects. Bat-el shows that in Hebrew several feminine suffixes have the form /-Vt/. When these suffixes combine, often an OCP-motivated repair can apply to derive an output: /rakav-et-ot/ → [rakavot] 'trains'. In some cases, however, no OCP-avoiding repair is possible, and a paradigm gap is obtained. Marc van Oostendorp, in 'Dutch diminutives and the question mark', provides another example of OCP-driven gaps from Dutch diminutives. The diminutive suffix is /-tjə/ and productively attaches, in principle, to any noun or proper name. One systematic class of exceptions is potential bases ending in a coronal plosive (/t/ or /d/), that is, the same consonant type found in the diminutive...