Aspiration length and intensity as potential acoustic cues to [voice] in final position: evidence from Shughni Yury Makarov HSE University / Institute of Linguistics of the Russian Academy of Sciences yurmak@iling-ran.ru ABSTRACT Final Shughni plosives tend to be released and aspirated, at least in citation forms. Data from 14 Shughni speakers showed that underlyingly voiced plosives have a shorter period of aspiration, and its intensity is weaker compared to that of voiceless ones. Although sometimes neutralized with regard to closure voicing, the contrast may be maintained by means of final aspiration length and intensity. Nevertheless, typological accounts do not usually treat these two characteristics as potential acoustic cues to underlying voicing in final position, and this study gives evidence that final aspiration length and intensity should be added to the list (cf. such cues as the duration of closure and preceding vowel, closure voicing, and F1 frequency). Moreover, possible implications for languages with final aspiration (like Kashmiri) are discussed. Keywords: neutralization, final devoicing, aspiration, acoustic cues, Iranian languages. 1. INTRODUCTION 1.1. Shughni language and speakers Shughni is an Iranian language spoken by ca. 100,000 in the Pamir Mountains in Tajikistan and Afghanistan. It has several varieties, constituting the so-called Shughni-Rushani group. From now on, the variety spoken in Khorugh, the capital of the Gorno- Badakhshan Autonomous Region in Tajikistan, is described. Shughni is used primarily as a spoken language, though there are a number of poetry books and even novels. It is usually the case that Shughni speakers are bi- or even multilingual, having some knowledge of Tajik (official language) and Russian (lingua franca) [1]. 1.2. Final laryngeal neutralization A typical example of final laryngeal neutralization is final devoicing. It occurs when consonants of the voiced and voiceless series are realized without vocal fold vibration in final position, e.g., Russian /kod/ ‘code’ and /kot/ ‘cat’ pronounced as [kot]. Although neutralization may be complete in respect of closure voicing, acoustic cues to underlying voicing are likely to be present (thus making the neutralization process incomplete). For example, utterance-final postvocalic stops in English may have closure voicing, but there are other means of signalling the [voice] feature [2]. In fact, for American English, vowel duration is claimed to be a perceptually sufficient cue to underlying voicing in word-final stops, fricatives and clusters [3], [4]. On the other hand, in German, it was shown to be significant only in production [5]. Well-established acoustic cues to voicing in English are summarized in Table 1. Cue Voiced Voiceless preceding vowel longer shorter closure voiced voiceless shorter longer F1 at the edges of adjacent vowels lower higher Table 1: Acoustic cues to underlying voicing in utterance-final postvocalic stops in English (after [2]). To be considered an acoustic cue, an acoustic characteristic must aid in the recognition of some feature of the segment [6], [7]. Until this perceptual significance is shown, this acoustic characteristic may be regarded as a potential cue, provided that it correlates with the feature in question. As for [voice], potential cues are the presence/absence of release (voiced stops are less often released) [8], [9], the intensity and spectral properties of release [10], and burst (expected differences are not specified) and aspiration (longer in underlyingly voiceless) duration [5], [7], [11]. There is also a claim that ‘any noise attendant upon consonant production will have greater intensity in the voiceless sound’ [12], though few works are focused on considering the linguistic implications of this acoustic fact, regarding final stops in particular. Final devoicing is not the only possible scenario of neutralization [13]. For instance, in Kashmiri (Indo-Aryan) the unaspirated and aspirated series are neutralized in favour of the latter, i.e., /wat/ ‘way’ and /katʰ/ ‘story’ are realized as [watʰ] and [katʰ] [14]. An interesting question is then what kind of cues (or at least acoustic correlates) are employed to maintain the underlying contrast when final aspiration occurs?