Bruce Hayes

Empirical Tests of the Gradual Learning Algorithm

The Gradual Learning Algorithm (Boersma 1997) is a constraint-ranking algorithm for learning optimality-theoretic grammars. The purpose of this article is to assess the capabilities of the Gradual Learning Algorithm, particularly in comparison with the Constraint Demotion algorithm of Tesar and Smolensky (1993, 1996, 1998, 2000), which initiated the learnability research program for Optimality Theory. We argue that the Gradual Learning Algorithm has a number of special advantages: it can learn free variation, deal effectively with noisy learning data, and account for gradient well-formedness judgments. The case studies we examine involve Ilokano reduplication and metathesis, Finnish genitive plurals, and the distribution of English light and dark /l/.

A Maximum Entropy Model of Phonotactics and Phonotactic Learning

The study of phonotactics is a central topic in phonology. We propose a theory of phonotactic grammars and a learning algorithm that constructs such grammars from positive evidence. Our grammars consist of constraints that are assigned numerical weights according to the principle of maximum entropy. The grammars assess possible words on the basis of the weighted sum of their constraint violations. The learning algorithm yields grammars that can capture both categorical and gradient phonotactic patterns. The algorithm is not provided with constraints in advance, but uses its own resources to form constraints and weight them. A baseline model, in which Universal Grammar is reduced to a feature set and an SPE-style constraint format, suffices to learn many phonotactic phenomena. In order for the model to learn nonlocal phenomena such as stress and vowel harmony, it must be augmented with autosegmental tiers and metrical grids. Our results thus offer novel, learning-theoretic support for such representations. We apply the model in a variety of learning simulations, showing that the learned grammars capture the distributional generalizations of these languages and accurately predict the findings of a phonotactic experiment.

THE PROSODIC HIERARCHY IN METER

Publisher Summary This chapter discusses the prosodic hierarchy in meter. In prosodic hierarchy, the essence of the theory is that utterances are phrased in the same sense that musical passages are phrased. As in music, phrasing in language is hierarchical: The lowest units are grouped into small phrases that in turn are grouped into larger phrases, and so on through several levels. This phrasing, or prosodic hierarchy, governs the way in which sandhi rules may be applied. The prosodic hierarchy of an utterance is determined by its syntactic structure but is not identical to it. The hierarchy is derived from syntactic structure by a set of rules that alter bracketing and provide labels for the various levels of phrasing.

Bengali intonational phonology

This paper proposes a phonological analysis of the Bengali intonational system, using a descriptive framework developed by Pierrehumbert (1980) and others. Our analysis bears on a number of theoretical points. We argue that the Bengali facts support a typology of intonational tones that includes only pitch accents and boundary tones, and that the docking sites for boundary tones are the phrase edges provided under the theory of the Prosodic Hierarchy (Selkirk 1980). We show that Bengali intonational contours are governed by the obligatory Contour Principle (OCP), which forbids adjacent identical tones. Underlying contours that violate the OCP are converted to permissible surface forms by a phonological rule. We also bring Bengali data to bear on a long-standing controversy concerning phrasal stress: Bengali can be shown to have a default, phonologically assigned phrasal stress pattern; thus phrasal stress assignment cannot be reduced exclusively to focus and other semantic factors.

Stochastic Phonological Knowledge: The Case of Hungarian Vowel Harmony

In Hungarian, stems ending in a back vowel plus one or more neutral vowels show unusual behavior: for such stems, the otherwise-general process of vowel harmony is lexically idiosyncratic. Particular stems can take front suffixes, take back suffixes, or vacillate. Yet at a statistical level, the patterning among these stems is lawful: in the aggregate, they obey principles that relate the propensity to take back or front harmony to the height of the rightmost vowel and to the number of neutral vowels. We argue that this patterned statistical variation in the Hungarian lexicon is internalized by native speakers. Our evidence is that they replicate the pattern when they are asked to apply harmony to novel stems in a “wug” test (Berko 1958). Our test results match quantitative data about the Hungarian lexicon, gathered with an automated Web search. We model the speakers’ knowledge and intuitions with a grammar constructed under the dual listing/generation model of Zuraw (2000), then show how the constraint rankings of this grammar can be learned by algorithm. *

Modeling English Past Tense Intuitions with Minimal Generalization

We describe here a supervised learning model that, given paradigms of related words, learns the morphological and phonological rules needed to derive the paradigm. The model can use its rules to make guesses about how novel forms would be inflected, and has been tested experimentally against the intuitions of human speakers.

Explaining sonority projection effects

The term sonority projection refers to behavioural distinctions speakers make between unattested phonological sequences on the basis of sonority. For example, among onset clusters, the well-formedness relation [bn]>[lb] is observed in speech perception, speech production and non-word acceptability (Davidson 2006 , 2007 , Berent et al. 2007 , Albright, ms). We begin by replicating the sonority projection effects in a non-word acceptability study. Then we evaluate the extent to which sonority projection is predicted by existing computational models of phonotactics (Coleman & Pierrehumbert 1997 , Hayes & Wilson 2008 , inter alia ). We show that a model based only on lexical statistics can explain sonority projection in English without a pre-existing sonority sequencing principle. To do this, a model must possess (i) a featural system supporting sonority-based generalisations, and (ii) a context representation including syllabification or equivalent information.

REDUPLICATION AND SYLLABIFICATION IN ILOKANO

Abstract This article presents an analysis of the productive morphophonemic phenomena in a dialect of Ilokano (N. Philippines). The analysis bears on three theoretical points. First, we argue that the rules creating well-formed syllable structure, hence syllabification itself, must apply cyclically. Second, a reduplication pattern of this dialect provides support for the theory of Prosodic Morphology (McCarthy and Prince (forthcoming)). The template for reduplication can be straightforwardly characterized in prozodic terms, but not segmentally. Third, we suggest that Ilokano exhibits ‘antitransfer’, i.e. the copying of a glide as a vowel in reduplication, and discuss the implications of Ilokano antitransfer for the theory of reduplication proposed by Steriade (1988).

The role of phonological phrasing in sung and chanted verse

Abstract This article is a study of the metrics of sung and chanted verse, based on a data corpus of 670 English folksong lines, as well as chanted renditions of the corpus by ten native speaker consultants. Our theoretical focus is on the role of phonological phrasing in metrics. We find that in sung and chanted verse, an especially tight correspondence to the metrical pattern is imposed on linguistic material that is either bounded within a tight phrasal domain or located at the right edge of a high-level domain. These patterns have been observed earlier for spoken verse. But for the phenomenon of metrical inversion, the behavior of sung and chanted verse is quite different from spoken verse. We develop an explanation of the difference, based on the idea that inversion in sung and chanted verse occurs only in those cases where it is the best available metrical option. A further finding is that sung and chanted verse tends to match the number of beats alloted to a syllable to that syllables natural linguistic duration. We suggest that the relevant measure of duration is phonetic, not phonological; and that the tendency to phonetic duration matching in sung and chanted verse has a categorial, phonologized analogue in the phenomenon of spoken-verse Resolution.

Phonological Naturalness and Phonotactic Learning

We investigate whether the patterns of phonotactic well-formedness internalized by language learners are direct reflections of the phonological patterns they encounter, or reflect in addition principles of phonological naturalness. We employed the phonotactic learning system of Hayes and Wilson (2008) to search the English lexicon for phonotactic generalizations and found that it learned many constraints that are evidently unnatural, having no typological or phonetic basis. We tested 10 such constraints by obtaining native-speaker ratings of 40 nonce words: 10 violated our unnatural constraints, 10 violated natural constraints assigned comparable weights by the learner, and 20 were control forms. Violations of the natural constraints had a powerful effect on ratings, violations of the unnatural constraints at best a weak one. We assess various hypotheses intended to explain this disparity, and conclude in favor of a learning bias account.