Abstract not found

Chomsky and Halle (1968) and many formal linguists rely on the notion of a universally available phonetic space defined in discrete time. This assumption plays a central role in phonological theory. Discreteness at the phonetic level guarantees the discreteness of all other levels of language. But decades of phonetics research demonstrate that there exists no universal inventory of phonetic objects. We discuss three kinds of evidence: first, phonologies differ incommensurably. Second, some phonetic characteristics of languages depend on intrinsically temporal patterns, and, third, some linguistic sound categories within a language are different from each other despite a high degree of overlap that precludes distinctness. Linguistics has mistakenly presumed that speech can always be spelled with letter-like tokens. A variety of implications of these conclusions for research in phonology are discussed.* The generative paradigm of language description (Chomsky 1964, 1965, Chomsky & Halle 1968) has dominated linguistic thinking in the United States for many years. Its specific claims about the phonetic basis of linguistic analysis still provide the cornerstone of most linguistic research. Many criticisms have been raised against the phonetic claims of the Sound pattern of English (Chomsky & Halle 1968), some from early on

Temporal properties associated with the speech signal are potentially important for understanding spoken language. Five hours of spontaneous American English dialogue material (from the SWITCHBOARD corpus) were hand-labeled and segmented at the phonetic-segment level; a fortyfive -minute subset was also manually annotated (at the syllabic level) with respect to stress accent. Statistical analysis of the corpus indicates that much of the temporal variation observed at the syllabic and phonetic-segment levels can be accounted for in terms of two basic parameters: (1) stress-accent pattern and (2) position of the segment within the syllable. Segments are generally longest in heavily accented syllables and shortest in syllables without accent. However, the magnitude of accent's impact on duration varies as a function of syllable position. Duration of the nucleus is heavily affected by accent level (heavily accented nuclei are, on average, twice as long as their unaccented counterparts), while the duration of the onset is also significantly affected but to a lesser degree. In contrast, accent has relatively little impact on the duration of the coda. This pattern of durational variation is incommensurate with segmental models, but rather implies the importance of syllable structure (and stress accent) for understanding spoken language.

This paper explores onset-sensitive stress from a typological, phonetic and phonological perspective. A phonetic study of three onset-sensitive stress systems suggests a close match between onset weight distinctions and a phonetic measure of perceptual energy, such that phonological weight criteria are the phonetically most effective ones. Perceptual considerations also offer an explanation for other typological observations, including the relative rarity of onset-sensitive stress, the greater weight of low sonority onsets, and the subordination of onset-sensitive weight distinctions to rimal based ones in languages with both types of weight distinctions. Onset-based weight criteria are effectively modelled using a skeletal slot model of the syllable referenced by a family of prominence constraints requiring that heavy syllables be stressed and that light syllables be unstressed. 1.

Languages are felt to be spoken with different kinds of rhythm. Traditional accounts have proposed that some languages are "stress-timed", while others are "syllable-timed". Despite the intuitive appeal of this typology, however, there is little phonetic evidence for the distinction. Meanwhile, generative phonology fails to provide clear explications on how formal representations of linguistic rhythm are to be phonetically interpreted. This thesis compared the rhythmic organization of English and Japanese using a novel experimental method, called "speech cycling", in which subjects produce a given phrase repeatedly in time with a controlled metronome. The task induces overtly rhythmic forms of speaking, and overcomes difficulties encountered by earlier studies in finding reliable physical correlates of speech rhythm and its variation across languages. In Experiment 1, subjects were asked to repeat phrases at a continuum of speaking rates, in order to examine whether they would fall in ...

I argue in this paper that vowel length plays no role in the synchronic phonology of Tigrinya: processes affecting vowels should be treated in featural terms only. The evidence in favor of synchronic vowel length is weak, and stronger evidence favors an analysis in which vowel length is phonologically irrelevant. While some researchers have made use of synchronic vowel length to account for ostensible closed-syllable shortening, I show that the relevant alternations are very limited in scope and represent at best the residue of historical vowel length. The evidence presented includes word minimality, vowel coalescence, word-final fronting, guttural lowering, and low dissimilation, with analyses of these phenomena in purely featural terms. Against vowel length in Tigrinya The analysis of the Tigrinya vowel system raises basic questions of abstractness and productivity. While the inventory, on the surface, consists of seven vowels well distinguished by quality (i.e. features), much recent generative work on Tigrinya phonology and morphology makes use of vowel length in analyzing certain patterns, for example vowel alternations construed as closed-syllable shortening. However, these patterns are quite limited in scope, and the assumption of vowel length in order to account for them

In this study we compare human ability to identify vowels with a machine learning approach. A perception experiment for 14 Hungarian vowels in isolation and embedded in a carrier word was accomplished, and a C4.5 decision tree was trained on the same material. A comparison between the identification results of the subjects and the classifier showed that in three of four conditions (isolated vowel quantity and identity, embedded vowel identity) the performance of the classifier was superior and in one condition (embedded vowel quantity) equal to the subjects’ performance. This outcome can be explained by perceptual limits of the subjects and by stimulus properties. The classifier’s performance was significantly weakened by replacing the continuous spectral information by binary 3-Bark thresholds as proposed in phonetic literature [8]. Parts of the resulting decision trees can be interpreted phonetically, which could qualify this classifier as a tool for phonetic research.

The standard way to represent language within linguistics is using letter-like units, i.e., with consonant and vowel phones and phonemes. But this representation, despite its powerful intuitive appeal is not supported by any experimental data. There are 4 straightforward predictions implied by this model. All are shown to fail. One of the main problems is the inability of segments to permit representing actual values of time. One realtime property of speech is the periodically produced speech found, e.g., in song and chant. Several audio clips of spontaneously produced rhythmic speech will be analyzed and speaker performance on the related laboratory task of speech cycling will be reviewed. In this phrase-repetition task speakers quickly adopt nested periodic timing patterns that are surprisingly rigid in time. Letter-based descriptions of speech make such rhythm invisible and irrelevant to anything `linguistic’. This is further evidence that phones and phonemes are only categories of patterns in the speech of a community, and not the psychological symbol tokens employed in the realtime production and perception of speech. 1. Phones and Phonemes are Letters in Disguise The conventional idea about the representation of language in the mind by professional linguists and psychologists as well as by laymen is based on letters – either orthographic letters or letters in a technical alphabet such as the IPA alphabet. For over a century scientists have assumed that speakers employ phones

Infants have a remarkable ability to perceive all manner of phonetic contrasts. The phonological categories of a language, however, have to be learned from experience. Two learning paradigms are contrasted – supervised learning (where learners receive feedback on their categorization attempts) and unsupervised learning (where learners rely only on properties of the input). It is argued that unsupervised learning may be the appropriate paradigm, at least for the initial stages of acquisition. Thereafter, the emergence of phoneme categories draws on various kinds of knowledge available to the learner, including knowledge of articulation, and of literacy conventions. A concluding section emphasizes the taxonomic nature of the phoneme, and suggests that the special salience of a phonemic representation reflects the status of the phoneme as a basic level category.

Abstract not found