Scott Myers

Surface underspecification of tone in Chichewa

In Chichewa, a Bantu language spoken mainly in Malawi, there is a contrast between high and low tone, as illustrated by such minimal pairs as mtengo ‘price’ vs. mtengo ‘tree’. But there is a strong asymmetry between the two tones in their phonological behaviour: high tone is phonologically active, while low tone is phonologically inert. Tone changes occur in Chichewa only if there is a high tone present in the phrase; a phrase composed only of low-toned morphemes is always realised unchanged with all low tones. The tonal phonology of the language can be described completely without reference to low tone (Kanerva 1989), as is typical for the Bantu languages (Stevick 1969). I argue in this paper that this asymmetry is due to underspecification. The contrast in Chichewa is a privative one between high tone and no tone. Low tone is phonologically inert because it is simply the absence of tone. In particular, low tone is absent from surface representation. Syllables that are not specified as high-toned are assigned F0 by a non-linear transition function, as proposed for English intonation by Pierre-humbert (1980).

F₀ Timing in Kinyarwanda

In Kinyarwanda (Bantu, Rwanda), a high tone can occur on either the first or the second mora of a syllable. A study of f₀ timing in the language was undertaken to determine how such a contrast is realized phonetically, and how this realization is affected by phrase position, word position, and distance from a following high tone. The study also aimed to explore a reported effect of ‘tone anticipation’ (raising of pitch in a syllable before a high tone) in the language. The study found significant effects of the position of a tone within the syllable, its position within the phrase, and its position with respect to following high tones. Models of f₀ timing in Kinyarwanda are proposed, demonstrating the role of moras in phonetic implementation of tone. These models clarify issues regarding the phonological representation of tones in Kinyarwanda, and the division between phonetic patterns and phonological ones.

Tonal transfer in Chichewa

What happens to tone when a form is reduplicated in a tone language? In Marantzs (1982) model of reduplication, it is only segmental melody that is copied from the base. This approach predicts that no tones of the base will appear on the reduplicant. In other models, the whole base is copied, including prosody (Steriade 1988; McCarthy & Prince 1988, forthcoming). This approach predicts that the tone of the base will always appear on the reduplicant, i.e. there will be ‘transfer’ of the tone (Clements 1985).

Vowel duration and neutralization of vowel length contrasts in Kinyarwanda

Abstract In Kinyarwanda, a Bantu language spoken in Rwanda, there is a contrast in vowel length, as in [ɡutaka] “to scream” and [ɡutaːka] “to decorate”. This contrast is neutralized in various positions: there are only short vowels word-initially or -finally, and there are only long vowels before a nasal-obstruent (NC) sequence or after a consonant-glide (CG) sequence. An acoustic study was performed to determine how vowel duration varied according to (a) vowel type (short, long, pre-NC, post-CG), (b) vowel height (low, high), (c) position in the word (initial, penultimate, final), and (d) position in the phrase (initial word, final word). It was found that pre-NC vowels were intermediate in duration between short vowels on the one hand and long or post-CG vowels on the other. Low vowels were longer than high vowels. Word-penultimate vowels were longer than word-initial or -final vowels, and phrase-penultimate vowels were longer than other word-penultimate vowels. The duration of phrase-final vowels depended on the extent of final devoicing. Implications for the representation and implementation of the vowel length contrast are discussed.

Domain generalisation in artificial language learning

Many languages have restrictions on word-final segments, such as a requirement that any word- final obstruent be voiceless. There is a phonetic basis for such restrictions at the ends of utterances, but not the ends of words. Historical linguists have long noted this mismatch, and have attributed it to an analogical generalization of such restrictions from utterance-final to word-final position. To test whether language learners actually generalize in this way, two artificial language learning experiments were conducted. Participants heard nonsense sentences in which there was a restriction on utterance-final obstruents, but in which no information was available about word-final, utterance-medial obstruents. They were then tested on utterances that included obstruents in both positions. They learned the pattern and generalized it to word-final utterance-medial position, confirming that learners are biased toward word-based distributional patterns.

Regressive voicing assimilation: Production and perception studies

Many languages have a phonological pattern of regressive voicing assimilation, according to which an obstruent is required to match a following obstruent in voicing (e.g. Russian, Sanskrit). This restriction on the distribution of categories has parallels in the phonetic fact that an obstruent has a longer interval of glottal pulsing when it occurs before a voiced sound than when it occurs before a voiceless sound. It is proposed that the phonological pattern arises diachronically through a reanalysis of the phonetic pattern, beginning with a tendency for listeners to identify an obstruent before another obstruent as matching the latter in voicing. This paper reports on two experiments designed to test premises of this account. A production study explores how obstruent voicing in English is affected by voicing in a following segment. A perception study explores how the identification of voicing categories is impacted by the acoustic effects of following segment context. It is found that listeners tend to identify a fricative as voiceless if it is drawn from the position before a voiceless obstruent, but that a following voiced segment has no significant effect on voicing class identification. Implications for the diachronic account of regressive voicing assimilation are discussed.

F0 Timing and Tone Contrasts in Luganda

Scholars of Luganda (Bantu, Uganda) have described a contrast between falling and high tone, which is limited to syllables with a long vowel or a coda. The contrast has been represented with H on the first mora of a falling-tone syllable and on both moras in a high-tone syllable. This article explores this contrast through an instrumental study of the timing of f0 events in Luganda. It was found that the 2-tone classes differed in the timing of both the f0 rise and the subsequent f0 fall, supporting a reanalysis of the contrast as one between an early high tone and a late high tone. It was also found that for the speakers in this study, the contrast was limited to syllables with long vowels. The timing of f0 events was sensitive to the duration of segments in the CVC interval centered on the high-toned syllable, including non-moraic elements and consonants outside the syllable. The association of a tone to a syllable identifies the segments relevant for the timing of the f0 contour, but that timing is not directly sensitive to the interval of the syllable.

The consonant length contrast in Persian: Production and perception

In Persian, length contrasts occur in all classes of consonants, including obstruents, glides and laryngeals. The results of an acoustic study of Persian consonant length are reported, showing that geminates differ from singletons across speaking rates and manner classes in constriction duration, preceding vowel duration, formant transition duration, and the intensity drop from the preceding vowel. Two perception studies tested the claim that consonant length contrasts are more difficult for Persian speakers to perceive when the consonant is more similar to neighboring vowels, as is the case with glides and laryngeals. In one of these experiments, Persian speakers identified the length class of consonants differing in constriction duration and consonant type. Identification was most consistent and rapid when the test consonant was an obstruent, and least so when it was laryngeal. In the other perception experiment, subjects identified the length class in consonants differing in both constriction duration and formant transition duration. Consonants with longer formant transitions were more likely to be identified as long than consonants with shorter transitions, but also were identified less consistently and more slowly. These results suggest a phonetic explanation for the avoidance of length contrasts in more vowel-like consonants in many languages.

An acoustic study of Luganda liquid allophones

Abstract Transcription-based descriptions of Luganda (Bantu, Uganda) have stated that there is a rhotic (r) in the language that occurs only after a front vowel, in complementary distribution with a lateral (l) that occurs only elsewhere. However, these descriptions of the pattern have disagreed about the nature and distribution of the sounds involved. Such a pattern has been reported in a number of Bantu and non-Bantu languages, and it is intriguing because its phonetic basis is not obvious, since there is nothing rhotic about a front vowel or lateral about a non-front vowel. To clarify the facts and provide insight into this kind of pattern, a production study was conducted in which Luganda speakers produced the liquid in varying vocalic contexts. F1, F2, duration, mean harmonics-to-noise ratio (HNR), and intensity were measured. The results support the characterization of the liquid after a front vowel as a nonlateral tap, and elsewhere as a lateral sonorant. But the intensity data also show a continuum in the degree of tongue tip constriction, ranging from approximant to tap, depending on the frontness of both the preceding and following vowels. The pattern of distribution is thus more complex and gradient than has been evident from transcriptions alone.

Perceptual studies of two phonological voicing patterns.

Two of the most common phonological patterns in the worlds languages are final devoicing and regressive voicing assimilation. In final devoicing, there is a contrast in obstruent voicing generally, but in word‐, phrase‐, or syllable‐final position, only voiceless obstruents are allowed (e.g., Dutch, Russian, Walloon). In regressive voicing assimilation, the first of a sequence of two obstruents is required to agree with the second in voicing (e.g., Russian, Catalan, Hebrew). These phonological patterns can be related to patterns of speech production: anticipation of the spread glottis posture in pause and laryngeal coarticulation within consonant clusters. The studies reported here investigate the perceptual consequences of these coarticulatory effects. Native speakers of English listened to words contrasting in the voicing of the final consonant (cease/seize) and identified which word they heard. The words were excised from utterance‐final position or from phrase‐medial position preceding a vowel, nasal...