Volume 23, Issue 3

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 exemplarbased 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.

In mainstream phonology, contrastive properties, like stem-final voicing, are simply listed in the lexicon. This article reviews experimental evidence that such contrastive properties may be predictable to some degree and that the relevant statistically gradient generalizations form an inherent part of the grammar. The evidence comes from the underlying voice specification of stem-final obstruents in Dutch. Contrary to received wisdom, this voice specification is partly predictable from the obstruent's manner and place of articulation and from the phonological properties of the preceding segments. The degree of predictability, which depends on the exact contents of the lexicon, directs speakers' guesses of underlying voice specifications. Moreover, existing words that disobey the generalizations are disadvantaged by being recognized and produced more slowly and less accurately, also under natural conditions. We discuss how these observations can be accounted for in two types of different approaches to grammar, Stochastic Optimality Theory and exemplar-based modeling.

Phonological variation in spoken words is a ubiquitous aspect of spontaneous speech and presents a challenge for recognition of spoken words. We discuss two classes of models, abstract and episodic, that have been proposed for spoken word recognition. Abstract theories rely on inference processes and/or underspecified representations to account for spoken word recognition. Episodic theories assume a lexical representation that encodes each spoken word event with exposure frequency linked to strength of a lexical entry. A model is proposed that posits a frequency-driven phonological variant lexical representation. The model assumes that a word may have more than one variant representation and that exposure to phonological variant form influences the strength of a given variant representation. Evidence for the proposed model is reviewed for a number of variants (nasal flaps, schwa deletion and medial flaps).

Evidence supporting a rich memory for associations suggests that people can store perceptual details in the form of exemplars. The resulting particulate model of category contents allows the application of evolution theory in modeling category change, because variation in categorized percepts is reflected in the distribution of exemplars in a category. Within a production-perception feedback loop, variation within an exemplar-based category provides a reserve of variants that can serve as the seeds for shifts in the system over time through random or selection-driven asymmetries in production and perception. Here, three potential pathways for evolutionary change are identified in linguistic categories: pruning of lines of inheritance, blending inheritance and natural selection. Simulations of each of these pathways are shown within a simple exemplar-based model of category production and perception, showing how consideration of evolutionary processes may contribute to our understanding of linguistic category change over time.

The early phases of syntactic acquisition are characterized by many input frequency and item effects, which argue against theories assuming innate access to classical syntactic categories. In formulating an alternative view, we consider both prototype and exemplar-learning models of categorization. We argue for a ‘hybrid’ usage-based view in which acquisition depends on exemplar learning and retention, out of which permanent abstract schemas gradually emerge and are immanent across the summed similarity of exemplar collections. These schemas are graded in strength depending on the number of exemplars and the degree to which semantic similarity is reinforced by phonological, lexical, and distributional similarity.

Exemplar-based models of language propose that human language production and understanding operate with a store of concrete linguistic experiences rather than with abstract linguistic rules. While exemplar-based models are well acknowledged in areas like phonology and morphology, common wisdom has it that they are intrinsically flawed for syntax where infinite generative capacity is needed. This article shows that this common wisdom is wrong. It starts out by reviewing an exemplar-based syntactic model, known as Data-Oriented Parsing, or DOP, which operates on a corpus of phrase-structure trees. While this model is productive, it is inadequate from the point of grammatical productivity. We therefore extend it to the more sophisticated linguistic representations proposed by Lexical-Functional Grammar theory, resulting in the model known as LFG-DOP, which does allow for meta-linguistic judgments of acceptability. We show how DOP deals with first language acquisition, suggesting a unified model for language learning and language use, and go into a number of syntactic phenomena that can be explained by DOP but that challenge rule-based models. We argue that if there is anything innate in language cognition it is not Universal Grammar but “Universal Representation”.

This article considers the exemplar theories which are independently developing in phonetics and in syntax, and argues that they jointly make some predictions that neither does alone. One of these predictions is explored in the context of two sound changes which occurred in the history of New Zealand English. We show that both of these phonetic changes were affected by phrase-level factors. The raising of /æ/ was more advanced in the word hand when it referred to a limb, than when used in phrases such as give a hand or lend a hand . And the centralization of the / i / vowel was more advanced in utterances of give involving abstract themes ( give a chance ), than when it had a meaning of transfer of possession ( give a pen ). We argue that existence of such effects lends support both to the idea (from syntactic exemplar theory) that phrases are stored, and the idea (from phonetic exemplar theory) that lexical representations are phonetically detailed.

Niedzielski (1999) reports on an experiment which demonstrates that individuals in Detroit ‘hear’ more Canadian Raising in the speech of a speaker when they think that speaker is Canadian. We describe an experiment designed to follow up on this result in a New Zealand context. Participants listened to a New Zealand English (NZE) speaker reading a list of sentences. Each sentence appeared on the answer-sheet, with a target word underlined. For each sentence, participants were asked to select from a synthesized vowel continuum the token that best matched the target vowel produced by the speaker. Half the participants had an answer-sheet with the word ‘Australian’ written on it, and half had an answer-sheet with ‘New Zealander’ written on it. Participants in the two conditions behaved significantly differently from one another. For example, they were more likely to hear a higher fronter / i / vowel when ‘Australian’ appeared on the answer sheet, and more likely to hear a centralized version when ‘New Zealander’ appeared – a trend which reflects production differences between the two dialects. This is despite the fact that nearly all participants reported that they knew they were listening to a New Zealander. We discuss the implication of these results, and argue that they support exemplar models of speech perception.