Edward Flemming

Auditory representations in phonology

Preface Acknowledgments 1. Introduction 2. The Dispersion Theory of Contrast 3. Ways of Maximizing Distinctiveness 4. Consonant-Vowel Assimilation 5. Neutralization 6. Minimization of Allomorphy 7. Conclusions References Index

Rosa's roses : reduced vowels in American English

Beginning phonetics students are taught that some varieties of American English have two contrasting reduced vowels, transcribed as [e] and [i], illustrated by the unstressed vowels in the minimal pair Rosas vs. roses (e.g. Ladefoged 2001, 2005). However, little seems to be known about the precise nature or distribution of these vowels. This study explores these questions through acoustic analysis of reduced vowels in the speech of nine American English speakers. The results show that there is a fundamental distinction between the mid central [e] vowel that can occur in unstressed word-final position (e.g. in Rosa ), and high reduced vowels that occur in most other unstressed positions, and might be transcribed as [i]. The contrast between pairs like Rosas and roses derives from this difference because the word-final [e] is preserved when an inflectional suffix is added, so the schwa of Rosas is similar to the final vowel of Rosa , whereas the unstressed vowel of roses is the high [i] reduced vowel quality found elsewhere. So the standard transcription of the reduced vowel contrast is justified, but the widespread use of [e] to transcribe word-internal reduced vowels is misleading – mid reduced vowels are generally only found in stem-final position.

The relationship between coronal place and vowel backness

It has been proposed that coronal consonants and front vowels are both specified [coronal] (e.g. Hume 1992). This leads to the prediction that all coronals can condition fronting of adjacent vowels. While fronting of vowels by coronals is attested (e.g. in Cantonese), not all kinds of coronals can condition fronting retroflexes condition retraction of vowels. We propose an account of the effects of coronals on vowels according to which they involve simple assimilation to the tongue body position of the coronal ‐ i.e. coronals that condition fronting are produced with a fronted tongue body, while coronals which condition backing of vowels are produced with a more retracted tongue body. Tongue body position is affected by the position of the tongue tip and blade because these articulators are physically connected, so for each type of coronal there are preferred tongue body positions that minimize the effort of producing the coronal constriction. This analysis also accounts for the fact that vowel backness can affect coronal place.

Phonetic analyses of word and segment variation using the TIMIT corpus of American English

This paper reports a set of studies of some phonetic characteristics of the American English represented in the TIMIT speech database. First we describe some relevant characteristics of TIMIT, and how we use the non-speech files on the TIMIT CD with a commercial database program. Two studies are then described: one using only the non-audio parts of TIMIT (segmental transcriptions and durations, and speaker information), and one using the audio signal for acoustic analysis. Results of such studies should be useful not only to linguistic phoneticians but also for speech recognition lexicons and text-to-speech systems.

Phonetic structures of Montana Salish

Abstract Montana Salish is an Interior Salishan language spoken on the Flathead Reservation in Northwest Montana by an estimated population of about 40 speakers. This paper describes the basic phonetic characteristics of the language based on data from five speakers. Montana Salish contains a number of typologically unusual consonant types. Including glottalized sonorants, pre-stopped laterals, and a series of pharyngeals distinguished by secondary articulations of glottalization and/or labialization. The language also allows long sequences of obstruent consonants. These and more familiar phonetic characteristics are described through analysis of acoustic, electroglottographic, and aerodynamic data, and compared with related characteristics in other languages of the world.

Rosa’s roses: Reduced vowels in American English

Beginning phonetics students are taught that American English has two contrasting reduced vowels, transcribed as [ə] and [ī], illustrated by the unstressed vowels in the minimal pair Rosa’s versus roses. However, little seems to be known about the precise nature or distribution of these vowels. This study explores these questions through acoustic analysis of reduced vowels in the speech of 12 American English speakers. The results show that there is a fundamental distinction between the mid central [ə] vowel that can occur in unstressed word‐final position (e.g., in Rosa), and high reduced vowels that occur in most other unstressed positions, and might be transcribed as [ī]. The contrast between pairs like Rosa’s and roses derives from this difference because the word‐final [ə] is preserved when an inflectional suffix is added, so the schwa of Rosa’s is similar to the final vowel of Rosa, whereas the unstressed vowel of roses is the high [ī] reduced vowel quality found elsewhere. So the standard transcrip...

The phonetic specification of contour tones: evidence from the Mandarin rising tone

This paper investigates the phonetic specification of contour tones through a case study of the Mandarin rising tone. The patterns of variation in the realisation of the rising tone as a function of speech rate indicate that its specifications include targets pertaining to both the pitch movement and its end points: the slope of the F0 rise, the magnitude of the rise, and the alignment of the onset and offset of the rise. This analysis implies that the rising tone is overspecified, in that any one of the target properties can be derived from the other three (e.g. slope is predictable from the magnitude and timing of the rise). As a result, the targets conflict, and cannot all be realised. The conflict between tone targets is resolved by a compromise between them, a pattern that is analysed quantitatively by formulating the targets as weighted, violable constraints.

Cues to stop place in stop‐liquid clusters

The acoustic correlates of stop place contrasts in prevocalic position have been studied intensively, but much less is known about stop place contrasts in consonant clusters. An understanding of the nature and distribution of cues to contrasts across a variety of contexts is important to the phonological analysis of the distribution of contrasts. This study extends the characterization of stop place contrasts to stop‐liquid clusters /br, dr, gr, bl, gl/, based on acoustic analysis of American English. We examine how well established cues relating to burst duration, burst spectrum, and formant transitions generalize to the stop‐liquid context. The relative perceptual weight of burst and formant transitions will be assessed through identification of cross‐spliced stimuli.

Targetless schwa revisited

It has been suggested that schwa vowels in English lack vowel quality targets, that is, they are realized as interpolations between the preceding and following segmental contexts or are the result of separating two consonantal constriction gestures, without an inherent vocalic gesture. Tests of this hypothesis have produced mixed results. The present study revisits the question of schwa targets based on acoustic analysis of schwa vowels in a wider range of segmental and morphological contexts. The main conclusions are: (i) It is important to distinguish stem‐final reduced vowels (as in ‘‘pandA’’) from other schwa vowels (e.g., ‘‘About,’’ ‘‘bEgin’’). The former clearly have vowel quality targets they are generally mid central vowels whereas the latter are much more contextually variable. (ii) The variable schwa vowels have a target, but this target is not a particular vowel quality, it is to indicate the presence of a vowel. This is achieved through a variety of means, including duration and realization of...

Statistical Learning Theory and Linguistic Typology: a Learnability Perspective on OT’s Strict Domination

This paper develops a learnability argument for strict domination by looking at the generalization error of learners trained on OT and HG target grammars. The argument is based on both a review of error bounds in the recent statistical learning literature and simulation results on realistic phonological test cases.