We present laryngographic data from four speakers of English English (EE), four speakers of Western Libyan Arabic (WLA), and four speakers of South-West Saudi-Arabian Arabic (SWSA) producing words ending in VC. They were recorded to capture synchronised acoustic and laryngographic signals whilst producing single words in which the final prepausal C varies systematically across five manner classes: voiceless stops, voiceless fricatives, voiced stops, voiced fricatives, and voiced sonorant consonants. Each speaker produced 3 tokens of each word. Using Speech Studio software, the closed quotient (Qx) value of the final regular cycle on the larynx waveform (Lx waveform) of the vowel before the start of the consonantal constriction was computed, taking the upper 70% of the wave’s height as the closed phase of the cycle. \ud Results show very clear but different patterns for EE and SWSA and a different but less clear pattern for WLA. In EE, Qx is significantly higher immediately before voiceless stops (mean of 51%) compared to all other classes (means between 33–37%); this is consistent with anticipating the glottalisation/glottal reinforcement typically found in these phonemes in British English (Roach 1979). In SWSA, by contrast, Qx values are significantly higher in vowels before all the voiced classes (means of 49–50%) compared to those before the voiceless classes (mean of 33%); this is consistent with anticipating the prepausal glottalisation characteristic of SWSA dialects (Watson & Asiri 2008). Those classes associated with higher Qx values also have high incidences of Lx waveform irregularities in the vicinity of the final consonant which often, but not always, sound creaky. When there are no irregularities and no auditory evidence of creaky phonation, the Qx values are still high. Tokens exhibiting overt irregularities and creakiness in the Lx waveform are therefore a subset of tokens exhibiting high Qx values, raising the question whether irregularity/creakiness is the ‘target’ which other tokens undershoot or whether the target is merely high glottal tension which often manifests as irregularity and creakiness. \ud In WLA, vowels before sonorant consonants have significantly higher Qx values (mean of 41%) than the other classes, and voiceless fricatives have significantly lower values (mean of 33%). Although sonorant consonants are associated with the highest vocalic Qx values of all the classes in WLA (mean of 41%), these values are significantly lower than for the SWSA speakers (mean of 49%), and significantly higher than for the EE speakers (mean of 33%). Vowels before voiceless stops are not significantly different in their Qx values from those before voiced stops and voiced fricatives. In WLA the relationship between Qx values, Lx irregularity and creakiness is highly variable, whereas in EE and SWSA they are highly consistent.\ud Phonemes classed as ‘voiced’ in many languages are often realised without voicing, and it is notable that in both varieties of Arabic represented in our data, two stops usually classed as voiceless (Watson 2002), /T/ (emphatic coronal) and /q/, have the same Qx values as the voiced sounds, not as the other voiceless sounds. Interestingly, the 8th Century grammarian Sibawayh did indeed group them with the other voiced sounds (Al-Nassir 1993). Our results indicate that there is a similar degree of glottal tension in vowels immediately prior to realisations of /T, q/ as there is prior to realisations of /d, z, m, n/ and other voiced phonemes. It may make more sense in Arabic to avoid ‘voiced’ (and its binary partner ‘voiceless’) in favour of terms which sit more comfortably with the phonetic facts, e.g. ‘tense’ and ‘lax’ as proposed by Trubetzkoy (1969), or ‘unbreathed’ and ‘breathed’ (Garbell 1958).\ud The main point arising from this research is that glottal tension associated with different manner classes, as quantified by the Qx measure, varies systematically within and across languages and language varieties not only in absolute terms but also in relative terms. The implication of this variation is that learners acquire these systematic differences, meaning that they are encoded in the grammar and are part of the learned phonetic realisation of the manner categories. That is to say, in EE speakers learn to realise final voiceless stops with, among many other things, of course, a high level of glottal tension, whereas in SWSA they learn to realise them with a low level of tension, and vice versa for voiced phonemes. Subtle but systematic differences such as these, which show that purportedly the ‘same’ phonological category may be realised by distinct mixtures of phonetic properties, undermine notions of the universalism of phonetic and phonological categories in favour of the view that sounds and phonological categories cannot be assumed to be the same from language to language and from variety to variety, at least in non-auditory terms.
No related research data.
No similar publications.