(This is a sample cover image for this issue. The actual cover is not yet available at this time.) This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright Author's personal copy Journal of Phonetics 41 (2013) 133–143 Contents lists available at SciVerse ScienceDirect Journal of Phonetics journal homepage: www.elsevier.com/locate/phonetics Acoustic differences between lexical and epenthetic vowels in Lebanese Arabic Nancy Hall n California State University Long Beach, 1250 Bellflower Blvd, Department of Linguistics (PSY 114), CSULB, Long Beach, CA 90840, United States ARTICLE INFO ABSTRACT Article history: Traditionally, epenthetic vowels in Lebanese Arabic are transcribed [i], and are assumed to be acoustically Received 23 November 2011 indistinguishable from lexical [i]. A production experiment finds variation among speakers: some do produce Received in revised form the vowels identically, others produce a schwa-like epenthetic vowel that is categorically distinct from lexical 6 September 2012 [i], and others produce clouds of epenthetic and lexical vowel tokens that partially overlap. Accepted 13 December 2012 & 2013 Elsevier Ltd. All rights reserved. 1. Introduction Levantine vowel epenthesis: In colloquial Levantine Arabic dialects (spoken in Lebanon, Syria, Israel, Palestine, and Jordan), underlying consonant clusters are frequently broken up by epenthetic vowels. Epenthesis is possible in CCC and word-final CC clusters, but only the latter are examined in this study. For many clusters, epenthesis is optional, and it is common for a speaker to produce the same word" sometimes with epenthesis and sometimes without. For example, ‘girl’, historically and underlyingly /bint/, can be pronounced either [bint] or ½ binit in Lebanese.1 This process receives much attention in the phonological literature because of its complex, often opaque interactions with other phonological processes. For example, Levantine dialects normally stress a CVC penult, as in (1a), but not if it contains an epenthetic vowel, as in (1b). The pattern can most easily be described through rule ordering, as shown here. a. ‘we wrote’ b. ‘our clothing’ /katab-na/" " /libs-na/ (1) Stress ka: tab:na "libs:na Epenthesis - - - li:bis:na " " [ka: tab:na] [ li:bis:na] Similar opaque interactions occur between epenthesis and other processes. For example, there is a vowel shortening pattern that affects CV:CC words; in the derivation =ja bÀP=-jabP-½jabÀiP ‘he didn’t bring’, epenthesis counter-bleeds shortening. Because opacity is challenging for standard Optimality Theory, currently the dominant phonological framework, Levantine epenthesis is often cited in arguments for alternative frameworks such as Stratal Optimality Theory (Kiparsky, 2003), Optimality Theory with candidate chains (McCarthy, 2007), and Harmonic Serialism. Also, in any phonological framework, it is a challenge to explain how opaque interactions are acquired: Alderete and Tesar (2002) point out that Arabic’s surface stress irregularities could mislead speakers into positing a lexical stress pattern. However, there has been little study of the phonetic characteristics of the epenthetic vowels. According to impressionistic descriptions, the quality of the epenthetic vowel varies within the Levant, from [ ] in Damascus (Cowell, 2005, p. 29) to [e] in Palestine (Elihay, 2004) to [i] in Lebanon (Haddad, 1984). It is usually described as being phonetically identical to one of the short lexical vowels in the same dialect. The inventory of lexical vowels also varies within the region: every Levantine dialect contrasts at least short [a i u], although there are few [i/u] minimal pairs; some dialects also have phonemic [e] and [o], and some additionally have [ ]. According to Cowell (2005, p. 13), the pairs [e/i] and [o/u] never contrast before a single word-final C (the environment that will be examined in this experiment), and [ ] is rarely contrastive in general. For all dialects, the short vowels are phonetically more centralized than the traditional transcriptions (used here) would imply. n Tel.: +1 562 985 2656. E-mail address: [email protected] 1 Throughout this paper, epenthetic vowels will be underlined for clarity. 0095-4470/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.wocn.2012.12.001 Author's personal copy 134 N. Hall / Journal of Phonetics 41 (2013) 133–143 Table 1 Underlying forms of paired target items. Environment Words with epenthesis Words with underlying /i/ b_l (1) /Pibl/ ‘lion cub’ /dibil/ ‘withered’ b_l (2) /<ibl/ ‘camel (collective)’ /<ibil/ ‘accepted’ b_s (1) /libs/ ‘clothes’ /libis/ ‘wore’ b_s (2) /dibs/ ‘syrup’ /yibis/ ‘dried up’ d_m /ridm/ ‘debris’ /nidim/ ‘regretted’ f_l /kifl/ ‘adoption’ /kifil/ ‘adopted’ k_b /rikb/ ‘riding (n.)’ /rikib/ ‘rode’ k_r /bikr/ ‘firstborn’ /sikir/ ‘got drunk’ l_d / ild/ ‘leather, skin’ /wilid/ ‘was born’ l_m /film/ ‘movie’ /silim/ ‘was safe’ l_s /kils/ ‘whitewash, calcium residue’ / ilis/ ‘was straightened’ m_P /rimP/ ‘eyelashes’ /kimiP/ ‘caught’ r_b /Pirb/ ‘drinking (n.)’ /Pirib/ ‘drank’ s_k /misk/ ‘musk’ /misik/ ‘held, grasped’ z_b /kizb/ ‘lying (n.)’ /kizib/ ‘lied’ In the case of Lebanese, phonological descriptions by native speakers (Haddad, 1983, 1984; Abdul-Karim, 1980) use the symbol [i] for both epenthetic and" lexical high front vowels, so that underlyingly different words like /libs/ ‘clothes’ and /libis/ ‘wore’ are transcribed as if pronounced identically, as [ libis]. However, some descriptions of Levantine dialects suggest that epenthetic and lexical vowels differ in duration or in other ways. According to Elihay (2004, 12), the epenthetic [e] of Palestinian is ‘‘the same as [lexical e], or even shorter.’’ Similarly, Haddad (1984, 37) reports of Beiruti speech: ‘‘I have often represented the epenthetic vowel as /i/. This is done only for typographical purposes. In some cases, however, this representation is rather inadequate since the epenthetic vowel is often very short as compared to an underlying /i/. Moreover, an epenthetic vowel is more easily prone to suprasegmental features such as [guttural]...and emphasis...which considerably alter its quality.’’ In a small-scale instrumental study, Gouskova and Hall (2009) found that some speakers of Lebanese differentiate epenthetic and lexical vowels in production. The epenthetic vowels in words like [libis] ‘clothes’ were significantly shorter in duration and lower in F2 than the corresponding lexical vowels in words like [libis] ‘wore’. There was also a non-significant trend towards higher F1 in epenthetic vowels. However, not all speakers produced a difference; for some, the vowels were identical. The present study builds on Gouskova and Hall (2009), by comparing epenthetic and lexical vowels across a larger number of speakers and items, and with a refined elicitation methodology. It largely confirms the previous results, and adds new details as to how speakers vary from one another. 2. Method Materials: Fifteen pairs of words were elicited, as shown in Table 1. Each pair consisted of a /CiCiC/ verb and a /CiCC/ noun, which according to standard descriptions of Lebanese should be realized as [CiCiC] with epenthesis. The experiment was designed to compare the second syllable vowels, although first syllable vowels were subsequently analyzed as well. Both verb and noun are unsuffixed, to avoid any possible effect of morpheme boundaries. The use of different parts of speech is unavoidable; Lebanese has a templatic morphology in which each part of speech corresponds to a limited number of shapes. " Seven of the" pairs are true minimal pairs. For five of these, both members of the pair are based on the same triconsonantal" root (e.g., [ libis] ‘clothes’" vs. [ libis] ‘wore’); the other two minimal pairs are based on homophonous but semantically unrelated roots (e.g., [ misik] ‘musk’ vs. [ misik] ‘held’). The remaining eight pairs are near-minimal pairs, in which only the word-initial consonant is different. This difference is not expected to affect the vowel in the second syllable. In all words, stress is on the first syllable. None of the words contain pharyngeal or ‘emphatic’ consonants, to avoid the effects these consonants may have on vowel quality.2,3 It was expected that each speaker would produce only a subset of the pairs in usable form. There are two reasons for this. First, vowel epenthesis is optional. If a speaker did not epenthesize in a particular /CiCC/ word, that pair was discarded. Second, many Lebanese words have variant lexical forms, even for speakers from similar demographic backgrounds. In many /CVCC/ nouns, the lexical vowel varies between /a/, /i/, and /u/, and many /CVCVC/ verbs vary between /CaCaC/ and /CiCiC/. So, for example, ‘riding (n.)’ can be [rikib] or [rakib], and ‘adopted’ can be [kifil] or [kafal]. If either member of a pair was produced with a lexical vowel other than /i/, the pair