Introduction to the special issue on exemplar-based models in linguistics

Recent years have witnessed the emergence of a cluster of approaches in linguistic theory known as exemplar-based models. Such exemplar-based models are being developed in research domains as diverse as phonetics, phonology, morphology, historical linguistics, semantics, syntax, and language acquisition. This special issue brings together, for the first time, articles applying exemplar-based models to a wide range of linguistic subdisciplines. We hope that this collection will give readers a sense for the challenges and new research tools that exemplar-based models offer. The central idea behind exemplar-based models is that mental representations consist of memory traces of specific tokens. This idea runs counter to the goal of developing maximally simple, redundancy-free representations, a goal that has been central to many proposals within linguistic theory. For example, much phonological work in the 1980s went into the study of underspecification theory, which banned non-distinctive or predictable feature values from underlying representations. Similarly, highly influential models of syntax have taken economy to be a guiding consideration in their conceptions of the lexicon and some researchers have started exploring new models of the lexicon and its interface with other levels of linguistic competence which call into question the goal of redundancy-free representations. A category in exemplar-based models is defined by a collection of memorized tokens, or " exemplars ". Each exemplar may belong to many categories simultaneously. An exemplar-based speech processing system recognizes inputs and generates outputs by analogical evaluation across a lexicon of distinct memory traces of remembered tokens of speech. While specific exemplar-based models differ on how new experiences are assigned to relevant categories and integrated with the stored exemplars, all exemplar-based mod-214 Susanne Gahl and Alan C. L. Yu els assume that each experience alters the entire category system slightly. As the perceptual memories associated with a category accumulate and are incre-mentally updated, the distribution of these forms may shift. As a consequence, effects of frequency and recency of use provide crucial testing grounds for such models. Exemplar-based models have received empirical support from many sources in recent years. Studies in perception have shown that listeners may rely on fine-grained phonetic details in word recognition (Goldinger 1996; Johnson 1997). Pierrehumbert (2002) reviews a large body of literature showing that lexical representations must include rich, subphonemic detail. Several contributions in this volume offer further support for this information-rich conception of linguistic representations. Ernestus shows that subphonemic information affects the production and recognition of …