1 An acoustic study on non-local anticipatory effects of Italian 2 length contrast 3 Giuseppina Turco a) 4 Laboratoire de Phon etique et Phonologie (UMR 7018) CNRS/Universit  e Sorbonne Nouvelle, 5 19 rue des Bernardins, 75005, Paris, France 6 Bettina Braun 7 Universit € at Konstanz, Department of Linguistics, Universit € atsstraße 10, Konstanz, 78457, Germany 8 (Received 14 November 2015; revised 27 July 2016; accepted 26 August 2016; published online xx 9 xx xxxx) 10 The present study investigates non-local temporal adjustments before an upcoming length contrast 11 in Italian minimal pairs that differ only in the length of the medial consonant (e.g., geminate word 12 palla “ball” vs singleton word pala “shovel”). This contrast is reportedly signaled by the duration 13 of the singleton/geminate consonant and of the preceding vowel. Here, it is shown that the duration 14 adjustment extends further to the word-initial consonant, e.g., the [p] in palla is significantly longer 15 than that in pala (experiment 1). In experiment 2, an effect of syllable structure is ruled out, an 16 unavoidable confound when comparing singleton and geminate words. The comparison of geminate 17 words with cluster words (e.g., as palco “stage”), both of which have a closed first syllable, shows a 18 similar lengthening. Implications for models of speech production are discussed. V C 2016 Acoustical Society of America.[http://dx.doi.org/10.1121/1.4962982] [MAH] Pages: 1–10 19 I. INTRODUCTION 20 It is well established that word-initial consonants are 21 lengthened in certain prosodic positions. For instance, in a 22 number of languages, stops are produced with longer voice 23 onset time (VOT) or closure duration in domain-initial posi- 24 tion compared to domain-medial position (e.g., Pierrehumbert 25 and Talkin, 1992; Jun, 1993; Hsu and Jun, 1996; Cho and Jun, 26 2000; Keating et al., 2003; Onaka, 2003). 27 In the literature, we find accidental reports of a further 28 kind of word-initial lengthening (accidental in the sense that 29 the findings appear to be by-products of a different research 30 question). A word that contains a geminate in medial posi- 31 tion (henceforth “geminate word” such as the Japanese word 32 kitte “postal stamps,” cf. Han, 1994) is produced with a lon- 33 ger word-initial consonant than an otherwise identical string 34 of sounds that contains a singleton in that position 35 (“singleton word” such as kite “listen”). This kind of length- 36 ening has been reported in Han (1994) and Idemaru and 37 Guion (2008) for Japanese voiceless stops and fricatives, and 38 in Ohala (2007) for Hindi stops. The duration differences are 39 typically small (between 8 and 11 ms), at least compared to 40 the domain-initial strengthening effects quoted above (with 41 duration differences between word-initial position and into- 42 national phrase initial position of on average 60 ms as, for 43 instance, in Japanese; Onaka, 2003). The small effect size 44 may be one of the reasons why this lengthening has not been 45 studied further or discussed systematically yet. However, the 46 initial lengthening differences in those studies appear to be 47 consistent (i.e., they are statistically significant) and are 48 hence not coincidental. 49 The first goal of this study is to replicate the word-initial 50 lengthening for Italian, a language in which, like Japanese 51 and Hindi, duration is the primary acoustic correlate of the 52 geminate-singleton contrast (Esposito and Di Benedetto, 53 1999; Pickett et al., 1999; Payne, 2005). To this end, in 54 experiment 1, we will first investigate the duration of word- 55 initial consonants in disyllabic geminate vs singleton words 56 (e.g., palla vs pala). The word pairs were matched for lexical 57 frequency, a factor that has been shown to strongly influence 58 segment durations (Pluymaekers et al., 2005). To create a 59 strong test case, we included variation by using a varied set 60 of word-initial and word-medial consonants, as well as 61 speakers from different Italian regions (some of which are 62 said to encode the length contrast more consistently than 63 others, cf. Bertinetto and Loporcaro, 2005). 64 The second goal is to investigate the mechanism behind 65 the effect. We see two plausible explanations that could 66 account for longer word-initial consonants in geminate 67 compared to singleton words. The lengthening may either 68 be caused by the upcoming length contrast (“gemination 69 account”) or by differences in syllable type (“syllable 70 account”), since the first syllable is open in singleton words 71 and closed in geminate words (cf. Gili Fivela and Zmarich, 72 2005). In other words, it is typically assumed that the [l] in 73 palla closes the first syllable and starts the second syllable 74 (Loporcaro, 1990), while the [l] in pala constitutes only the 75 onset of the second syllable. 76 What speaks in favour of the gemination account is that 77 there are other kinds of coarticulatory effects that span lon- 78 ger domains and manifest themselves on non-adjacent 79 sounds. For instance, there are well-known phenomena such 80 as vowel-to-vowel coarticulation (e.g., € Ohman, 1966), labi- 81 alization (e.g., Benguerel and Cowan, 1974), and nasaliza- 82 tion (Moll and Daniloff, 1971), but also less well understood PROOF COPY [JASA-00200] 058609JAS a) Electronic mail: giuseppina.turco@univ-paris3.fr J_ID: JASMAN DOI: 10.1121/1.4962982 Date: 28-September-16 Stage: Page: 1 Total Pages: 10 ID: aipepub3b2server Time: 08:15 I Path: D:/AIP/Support/XML_Signal_Tmp/AI-JAS#160549 J. Acoust. Soc. Am. 140 (3), September 2016 V C 2016 Acoustical Society of America 1 0001-4966/2016/140(3)/1/10/$30.00