Special thanks to John McCarthy for detailed discussion of virtually every issue raised here and for a fine-grained skepsis of the entire first draft of the ms., which resulted in innumerable improvements and would have resulted in innumerably more, were this a better world. We are particularly grateful for his comments and suggestions in r' Chs. 7 and 9. We also wish to thank

No abstract

A set of hypotheses is formulated for a connectionist approach to cognitive modeling. These hypotheses are shown to be incompatible with the hypotheses underlying traditional cognitive models. The connectionist models considered are massively parallel numerical computational systems that are a kind of continuous dynamical system. The numerical variables in the system correspond semantically to fine-grained features below the level of the concepts consciously used to describe the task domain. The level of analysis is intermediate between those of symbolic cognitive models and neural models. The explanations of behavior provided are like those traditional in the physical sciences, unlike the explanations provided by symbolic models.Higher-level analyses of these connectionist models reveal subtle relations to symbolic models. Parallel connectionist memory and linguistic processes are hypothesized to give rise to processes that are describable at a higher level as sequential rule application. At the lower level, computation has the character of massively parallel satisfaction of soft numerical constraints; at the higher level, this can lead to competence characterizable by hard rules. Performance will typically deviate from this competence since behavior is achieved not by interpreting hard rules but by satisfying soft constraints. The result is a picture in which traditional and connectionist theoretical constructs collaborate intimately to provide an understanding of cognition.

No abstract

No abstract

In this article we show how Optimality Theory yields a highly general Constraint Demotion principle for grammar learning. The resulting learning procedure specifically exploits the grammatical structure of Optimality Theory, independent of the content of substantive constraints defining any given grammatical module. We decompose the learning problem and present formal results for a central subproblem, deducing the constraint ranking particular to a target language, given structural descriptions of positive examples. The structure imposed on the space of possible grammars by Optimality Theory allows efficient convergence to a correct grammar. We discuss implications for learning from overt data only, as well as other learning issues. We argue that Optimality Theory promotes confluence of the demands of more effective learnability and deeper linguistic explanation. How exactly does a theory of grammar bear on questions of learnability? Restrictions on what counts as a possible human language can restrict the learner's search space. But this is a coarse observation: alone it says nothing about how data may be brought to bear on the problem, and further, the number of possible languages predicted by most linguistic theories is extremely large. 1 It would clearly be a desirable result if the nature of the restrictions imposed by a theory of grammar could contribute further to language learnability. The central claim of this article is that the character of the restrictions imposed by Optimality Theory (Prince and Smolensky 1991, 1993) have demonstrable and significant consequences for central questions of learnability. Optimality Theory explains linguistic phenomena through the complex interaction of violable constraints. The main results of this article demonstrate that those constraint interactions are nevertheless restricted in a way that permits the correct grammar to be inferred from grammatical structural descriptions. These results are theorems, based on a formal We are greatly indebted to Alan Prince, whose challenges, insights, and suggestions have improved nearly every

In this article we show how Optimality Theory yields a highly general Constraint Demotion principle for grammar learning. The resulting learning procedure specifically exploits the grammatical structure of Optimality Theory, independent of the content of substantive constraints defining any given grammatical module. We decompose the learning problem and present formal results for a central subproblem, deducing the constraint ranking particular to a target language, given structural descriptions of positive examples. The structure imposed on the space of possible grammars by Optimality Theory allows efficient convergence to a correct grammar. We discuss implications for learning from overt data only, as well as other learning issues. We argue that Optimality Theory promotes confluence of the demands of more effective learnability and deeper linguistic explanation.Keywords: Optimality Theory, learning, acquisition, computational linguistics How exactly does a theory of grammar bear on questions of learnability? Restrictions on what counts as a possible human language can restrict the learner's search space. But this is a coarse observation: alone it says nothing about how data may be brought to bear on the problem, and further, the number of possible languages predicted by most linguistic theories is extremely large. 1It would clearly be a desirable result if the nature of the restrictions imposed by a theory of grammar could contribute further to language learnability.The central claim of this article is that the character of the restrictions imposed by Optimality Theory Smolensky 1991, 1993) have demonstrable and significant consequences for central questions of learnability. Optimality Theory explains linguistic phenomena through the complex interaction of violable constraints. The main results of this article demonstrate that those constraint interactions are nevertheless restricted in a way that permits the correct grammar to be inferred from grammatical structural descriptions. These results are theorems, based on a formalWe are greatly indebted to Alan Prince, whose challenges, insights, and suggestions have improved nearly every

Do speakers know universal restrictions on linguistic elements that are absent from their language? We report an experimental test of this question. Our case study concerns the universal restrictions on initial consonant sequences, onset clusters (e.g., bl in block). Across languages, certain onset clusters (e.g., lb) are dispreferred (e.g., systematically under-represented) relative to others (e.g., bl). We demonstrate such preferences among Korean speakers, whose language lacks initial C1C2 clusters altogether. Our demonstration exploits speakers' well known tendency to misperceive ill-formed clusters. We show that universally dispreferred onset clusters are more frequently misperceived than universally preferred ones, indicating that Korean speakers consider the former cluster-type more ill-formed. The misperception of universally ill-formed clusters is unlikely to be due to a simple auditory failure. Likewise, the aversion of universally dispreferred onsets by Korean speakers is not explained by English proficiency or by several phonetic and phonological properties of Korean. We conclude that language universals are neither relics of language change nor are they artifacts of generic limitations on auditory perception and motor control-they reflect universal linguistic knowledge, active in speakers' brains.optimality theory ͉ phonology ͉ sonority ͉ syllable T he ''nature vs. nurture'' debate concerns the origin of speakers' knowledge of their language. Both sides of this controversy presuppose that people have some knowledge of abstract linguistic regularities. They disagree on whether such regularities reflect the properties of linguistic experience, auditory perception, and motor control (1, 2) or universal, possibly innate, and domain-specific restrictions on language structure (3-5, **). Empirical support for such restrictions comes from linguistic universals: regularities exhibited across the world's languages. These universals, for example, assert that the sound sequence lbif makes a poor word, whereas the sequence blif is better: Languages always make use of words like blif before (as in Russian) resorting to words like lbif. But the significance of such observations is unclear. One view holds that language universals form part of the language faculty of all speakers (5-7). The alternative denies that speakers have knowledge of language universals. Rather, speakers simply know regularities (either structural or statistical) concerning words in their own language. Language universals are not mentally represented-they are only statistical tendencies, shaped by generic (auditory and motor) constraints on language evolution (8). For example, words beginning with lb have a tendency to decline relative to those beginning with bl because the former are more frequently mispronounced or misperceived. The question at hand, then, is whether language universals are active in the brains of all speakers, or mere relics of systematic language change and its distal generic causes?The matter is difficult to resolve bec...