Phonological judgements are often gradient: blick>?bwick>*bnick>**bzick. The mechanisms behind gradient generalisation remain controversial, however. This paper tests the role of phonological features in helping speakers evaluate which novel combinations receive greater lexical support. A model is proposed in which the acceptability of a string is based on the most probable combination of natural classes that it instantiates. The model is tested on its ability to predict acceptability ratings of nonce words, and its predictions are compared against those of models that lack features or economise on feature specifications. The proposed model achieves the best balance of performance on attested and unattested sequences, and is a significant predictor of acceptability even after the other models are factored out. The feature-based model's predictions do not completely subsume those of simpler models, however. This may indicate multiple levels of evaluation, involving segment-based phonotactic probability and feature-based gradient phonological grammaticality.

The assumption that there is a common set of linguistic principles underlying both spoken language and sign language phonology, which forms part of the human language capacity, is shared by most phonologists working on sign language. See Sandler (1993a) for an extensive discussion of these issues. But even though this assumption is reasonable, since both spoken and signed languages are products of the same human brain and fulfil the same function, it is not clear that theories of representation which have been proposed for spoken languages can be directly applied to the structure of sign languages. Such representations have been developed on the basis of the spoken language modality only. They are often so close to the phonetics of spoken languages that we cannot rule out the possibility that non-trivial aspects of them are modality-specific. Therefore, rather than, for example, attempting to test various competing (spoken language-based) theories of syllable structure, we must first investigate the structure of sign language in its own right. This strategy need not be pushed too far, however. In developing a model of signs we can benefit from general principles which have proved successful in the study of spoken languages, especially if these principles do not seem to be directly based on ‘spoken phonetics’.

It is generally believed that Turkish stress is always word-final. Closer examination, however, reveals several types of exceptions to this pattern involving both roots and affixes. This paper proposes a unified analysis of regular and irregular stress in Turkish that crucially depends on our definition of the Phonological Word. In addition, we discuss stress in constituents beyond the word, and provide evidence for the Clitic Group as well as the Phonological Phrase. Finally, we also briefly discuss vowel harmony and a set of syllabification phenomena, and show how the latter, in particular, provide independent support for the proposal we advance here.

It is well known that initially and when preceded by a word that ends with a voiceless sound, German so-called ‘voiced’ stops are usually voiceless, that intervocalically both voiced and voiceless stops occur and that syllable-final (obstruent) stops are voiceless. Such a distribution is consistent with an analysis in which the contrast is one of [voice] and syllable-final stops are devoiced. It is also consistent with the view that in German the contrast is between stops that are [spread glottis] and those that are not. On such a view, the intervocalic voiced stops arise because of passive voicing of the non-[spread glottis] stops. The purpose of this paper is to present experimental results that support the view that German has underlying [spread glottis] stops, not [voice] stops.

Our colleague Charles Fillmore invented the following parable (personal communication) to characterise the two dominant ways linguists attempt to solve problems of language structure and behaviour.

One of the most widespread – and widely studied – properties of speech fundamental frequency (Fo) is a tendency to decline gradually during the course of utterances. This tendency has been given a variety of names, of which the best known is probably DECLINATION. This is the term I shall use in this paper. The purpose of the paper is to review past work on declination as it affects the phonological and phonetic modelling of Fo contours, and to outline some ideas for the empirical resolution of the issues that emerge from the discussion.

Vowel sounds may be inserted into a word by two mechanisms: insertion of a vocalic articulatory gesture (epenthesis), or retiming of existing gestures to produce a vowel-like transition between consonants (intrusion). I argue that epenthetic vowels are phonological units but intrusive vowels are not. A representation using abstract gestures as well as segments can capture facts about the typology of vowel intrusion.

Coronals have been claimed to behave as unmarked consonants in epenthesis. However, it is well known that the glottal consonants ([h ʔ]) are frequently epenthetic, and the empirical basis for the claim about coronal epenthesis has been weak, with only a single example commonly cited. I will show that coronals can in fact occur as epenthetic segments, but only in specific situations showing the classic signs of constraint conflict. I will argue that these patterns can be accounted for in Optimality Theory using fully specified Place in the representations and extending Smolensky's (1993) universally ranked *PLACE hierarchy so that the glottals have the least marked Place. The result will be that when all other things are equal, glottals will be the ideal epenthetic consonant, but when some higher-ranked requirement makes epenthesising a glottal impossible, the still relatively low-ranked position of *COR means that coronals will be the next best way to satisfy Place markedness.

This paper presents a theory of morphophonology based on a development in the theory of faithfulness in Optimality Theory. A new constraint type, anti-faithfulness, is proposed that evaluates a pair of related words and requires an alternation in the shared stem. This constraint type is motivated initially by a set of problems, e.g. morphological deletions, segmental exchanges and non-structure preserving processes, which show that morphophonology must encompass more than markedness–faithfulness interactions. The anti-faithfulness thesis is then applied to accentual processes in which affixes idiosyncratically cause deletion of accent in a neighbouring morpheme. It is argued that anti-faithfulness both motivates the observed deletion and accounts for its properties with principles that are generally available in phonological theory. Anti-faithfulness is then shown to extend naturally to the analysis of other affix-induced alternations, including accent insertions, shifts, and retractions of stress and tone, a result which distinguishes this theory from plausible alternatives.

Many phonological phenomena require reference to information beyond the immediate context of segments and boundaries. Most commonly, the syllable has been employed as an additional phonological context, but determining the exact nature of syllables and their boundaries has proved notoriously difficult. One approach is to add resyllabification (Selkirk 1982, Borowsky 1986) or ambisyllabification (Kahn 1976, Gussenhoven 1986, Rubach 1996) to basic syllabification in an attempt to account for phenomena where syllabification alone is inadequate. In this paper we argue against this approach and argue for a second approach, the PROSODIC approach. The essence of the prosodic approach is that there is a HIERARCHY of prosodic categories, of which the syllable is one, any one of which can serve as the context for phonological rules. Essentially, the argument is that prosodic categories, independently required for a wide variety of phonological processes, are sufficient for accounting for all the phenomena that resyllabification or ambisyllabification have been used for, whereas the reverse is not true. Resyllabification and ambisyllabification are invoked ad hoc for certain phenomena and cannot account adequately for the full range of processes that can be understood in terms of prosodic categories. Resyllabification and ambisyllabification are therefore superfluous and can be dispensed with, indeed must be dispensed with, in an explanatory theory of phonology.

A great many languages of the world exhibit phenomena of FO DOWNTREND – phenomena whereby, other things being equal, the fundamental frequency (Fo) of the speaking voice declines over the course of an utterance. That much is uncontroversial; further details are either simply unknown or the subject of considerable debate. The purpose of the study reported here was to shed light on some of these unknown or uncertain matters by the controlled investigation of pitch realisation in Yoruba.

In some languages, syllable weight depends exclusively on vowel length, while in others, coda consonants add weight to syllables. In this paper we assume that syllable weight is reflected in moraic structure, and that weight-bearing coda consonants are the exclusive dependents of a mora, while weightless consonants share a mora with the preceding vowel. We consider whether the durations of vowels and coda consonants reflect the distinction between a segment which occupies its own mora and a segment that shares a mora. We examine three patterns of coda weight, reflected in stress assignment: in Hindi, codas always contribute to syllable weight; in Malayalam, coda consonants are always weightless; and in Levantine Arabic, coda weight is contextually determined, with word-internal codas contributing to syllable weight following a short vowel, but weightless following a long vowel. These phonological patterns translate into different moraic representations of CVC and CVVC syllables across the different languages. We examine the durations of vowels and coda consonants in CV, CVC, CVV and CVVC syllables in Hindi, Malayalam and Levantine Arabic, and find that in all three languages, segments that we represent as mora-sharing are significantly shorter than segments that we represent as occupying an independent mora. The striking differences in durational patterns across the three languages correlate with the different moraic representations proposed on the basis of phonological patterning.

Japanese shows an asymmetry in the treatment of word-final [n] in loanwords from English and French: while it is adapted as a moraic nasal consonant in loanwords from English, it is adapted with a following epenthetic vowel in loanwords from French. We provide experimental evidence that this asymmetry is due to phonetic differences in the realisation of word-final [n] in English and French, and, consequently, to the way in which English and French word-final [n] are perceived by native speakers of Japanese. Specifically, French but not English word-final [n] has a strong vocalic release that Japanese listeners perceive as their native vowel [ɯ]. We propose a psycholinguistic model in which most loanword adaptations originate in perceptual assimilation, a process which takes place during perception and which maps non-native sounds and sound structures onto the phonetically closest native ones. We compare our model to alternatives couched within phonological theory.

The study of phonological structure and patterns across languages is seen by contemporary phonologists as a way of gaining insight into language as a cognitive system. Traditionally, phonologists have focused on spoken languages. More recently, we have observed a growing interest in the grammatical system underlying signed languages of the deaf. This development in the field of phonology provides a natural laboratory for investigating language universals. As grammatical systems, in part, reflect the modality in which they are expressed, the comparison of spoken and signed languages permits us to separate those aspects of grammar which are modality-dependent from those which are shared by all human languages. On the other hand, modality-dependent characteristics must also be accounted for by a comprehensive theory of language. Comparing languages in two modalities is therefore of theoretical importance for both reasons: establishing modality-independent linguistic universals, and accounting for modality-dependent structure and organisation.

This paper examines the relationship between tone and prosodic positions. I show that prosodic heads prefer higher tone over lower tone, while non-heads exhibit the opposite preference. These generalisations are expressed within Optimality Theory as a family of constraints in a fixed ranking. One set regulates the relation of tone to heads: *HD/L [Gt ] *HD/M [Gt ] *HD/H; the other deals with tone on non-heads: *NON-HD/H H [Gt ] *NON-HD/M [Gt ] *NON-HD/L. These constraints are used to account for the stress system of Ayutla Mixtec: in this language, stress is attracted to a syllable based on its tonal content, but is also influenced by the post-tonic syllable's tone. The implications of the theory for other tone–stress interactions – metrically influenced tone placement and neutralisation – are also examined.

Optimality Theory explains typological markedness implications by proposing that all speakers possess universal constraints penalising marked structures, irrespective of the evidence provided by their language (Prince & Smolensky 2004). The account of phonological perception sketched here entails that markedness constraints reveal their presence by inducing perceptual ‘repairs’ to structures ungrammatical in the hearer's language. As onset clusters of falling sonority are typologically marked relative to those of rising sonority (Greenberg 1978), we examine English speakers' perception of nasal-initial clusters, which are lacking in English. We find greater accuracy for rising-sonority clusters, evidencing knowledge of markedness constraints favouring such onset clusters. The misperception of sonority falls cannot be accounted for by stimulus artefacts (the materials are perceived accurately by speakers of Russian, a language allowing nasal-initial clusters) nor by phonetic failure (English speakers misperceive falls even with printed materials) nor by putative relations of such onsets to the statistics of the English lexicon.

The present paper explores whether the metrical foot is necessary for the description of prosodic systems. To this end, we present empirical findings on the perception of German word stress using event-related brain potentials as the dependent measure. A manipulation of the main stress position within three-syllable words revealed differential brain responses, which (a) correlated with the reorganisation of syllables into feet in stress violations, and (b) differed in strength depending on syllable weight. The experiments therefore provide evidence that the processing of word stress not only involves lexical information about stress positions, but also (quantity-sensitive) information about metrical structures, in particular feet and syllables.

A growing body of work on exemplar-based theories of learning suggests the possibility of formal models of phonological representation which will offer deeper explanations of basic phonological properties than current models allow. The main purpose of this paper is to shed light on near merger, a recalcitrant problem in sound change and in phonological theory, with this newer perspective, through a case study of tonal near merger in Cantonese.

A number of recent papers have demonstrated the advantages of using a phonological model incorporating the timing and magnitude of articulatory gestures to account for alternations involving segments such as the English nasals, liquids and glides (e.g. Krakow 1989, Browman & Goldstein 1992, 1995, Sproat & Fujimura 1993, Gick, in press). Some of these works (McMahon et al. 1994, McMahon & Foulkes 1995) have made specific reference to the well-known phenomenon of English intrusiver, shown in (1).

formula here

However, previous analyses have not linked the intrusive r explicitly to other similar processes, nor viewed all of these processes as the natural results of more general principles of phonological organisation. Thus, the intrusive r has remained, in the eyes of most linguists, an isolated quirk of English history, or, as one phonologist (McCarthy 1993: 191) has called it, ‘the phonologically unnatural phenomenon of r-epenthesis’.

The present paper introduces into the discussion of intrusive r a recently documented related phenomenon known as intrusivel (Gick 1991, 1997, in preparation, Miller 1993). It is argued that these new facts, in conjunction with current advances in the understanding of articulatory factors in syllable structure, support a view in which the intrusive r and l are synchronically underlyingly present.

The theory of Lexical Phonology proposed in Kiparsky (1982a) is a major step forward in generative phonology with respect to the problem of the interaction of phonology and morphology. Its basic claim is that morphological rules and word level phonological rules are interspersed. A rule of word phonology (i.e. a lexical phonological rule, which exclusively applies within words) may apply as soon as the required environment for its application has been created by some morphological rule. That is: ‘morphology and phonology go hand in hand’.