On Ejective Fricatives in Omani Mehri

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159

brill.com/aall

On ejective fricatives in Omani Mehri

Rachid Ridouane* Laboratoire de Phonétique et Phonologie, cnrs/Sorbonne Nouvelle, Paris [email protected]

Cédric Gendrot Laboratoire de Phonétique et Phonologie, cnrs/Sorbonne Nouvelle, Paris [email protected]

Abstract

Ejective fricatives are extremely rare cross-linguistically. This infrequency is generally attributed to the incompatibility of two aerodynamic requirements: airflow to create noise frication and a high intraoral air pressure to implement ejectivity. Seeking to determine how this incompatibility is solved, this study presents an acoustic investigation of initial and intervocalic ejective fricatives in Mehri, a Modern South Arabian language spoken in Oman. Based on data from 5 Mehri speakers, the analysis of different temporal and non-temporal parameters shows a high degree of variability in the way ejectivity is implemented in fricatives. Much of this variability is shaped by the position of the segments within the word. In initial position, the ejectivity of fricatives translates into a frequent presence of post-frication glottal lags, higher intensity and higher center of gravity. These acoustic attributes are less frequently encountered in intervocalic position. In this position, it is argued, the systematic diphthongization of the following long vowel, induced by ejectivity combined with dorsopharyngealisation, is salient enough to allow the contrast of ejectivity to be recovered.

Keywords ejective fricatives – acoustics – Mehri – Modern South Arabian language

* This work was funded by a grant from the French Agence Nationale de la Recherche Scien- tifique (anr program OmanSam) and by the Labex efl (anr/CGI). We would like to warmly thank the Omani participants for their help and enthusiasm during the study visit of the first author in Salalah, and the reviewer for the careful and thorough review of this manuscript.

1 Introduction

Ejective fricatives are extremely rare in the world’s languages; only 3.7% of the languages were reported to have at least one such segment (Maddieson 2013). The rarity of ejective fricatives is attributed to aerodynamic constraints that make them acoustically instable: the continuing flow of air necessary to the production of frication is incompatible with the increasing of intraoral air pressure necessary to the production of ejectivity (Maddieson 1998, Shosted and Rose 2011). This incompatibility raises the question of whether these fricatives are real ejectives, or whether they should on closer examination prove to be dis- tinguished from their pulmonic counterparts by some other dimension. The present study examines the acoustic cues to ejectivity in fricatives of Mehri, a Modern South Arabian language spoken in Oman. The aim is to show how native speakers solve the problematic combination of glottal constriction and frication noise. In a short abstract entitled ‘Combining frication and glottal constriction: Two solutions to a dilemma’, Maddieson (1997) proposed two mechanisms to solve the incompatibility between frication noise and glottal constriction. These two solutions were described in some detail in two later studies (Maddieson 1998, Maddieson et al. 2001). The first solution was achieved through a spe- cific timing between frication noise and glottal constriction: a non-overlapping pulmonic frication followed by a glottal stop. Used in Yapese, an Austrone- sian language spoken in the Micronesian state of Yap, this sequencing strategy, while it enabled speakers to generate sufficient frication duration for place identification while maintaining distinctiveness from plain fricatives, raised the question of whether these fricatives should be termed ejectives in the canonical phonetic sense of this term or whether they should be considered as sequences of fricative + glottal stop (Maddieson 1998). The second solution, observed in Tlingit (Maddieson et al. 2001), a Na-Dene language spoken in the Pacific Northwest coast of North America, was to produce ejective fricatives with a much narrower constriction than was used in their pulmonic counterparts. This allowed for glottal closure to overlap the entire frication duration while producing high intra-oral pressure, suggesting that they were indeed ejectives. This ‘canonical’ ejection of a fricative translated acoustically into shorter duration and a tendency for “scrapiness” or pulsing during frication noise. Faced with the same dilemma, Shosted and Rose (2011), based on acoustic data from Tigrinya, an Ethio-Semitic language spoken in Ethiopia and Eritrea, posited a third solution. They showed that ejective /s’/ in Tigrinya was generally realized not as a fricative but as an affricate [ts’]. Affrication, manifested with

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access on ejective fricatives in omani mehri 141 pre-frication closure lag, was interpreted as a general strategy used to increase intraoral air pressure. As reported by Maddieson et al. (2001), Tlingit ejective fricatives also displayed (near-)silent periods preceding frication noise. The interpretation of this gap differed, however. While Maddieson et al. (2001) interpreted it as glottal closure, Shosted and Rose (2011) posited that the source of this closure is oral. The phonetic implementation of ejective fricatives has been instrumentally examined in some other languages, including Amharic (Demolin 2002), Kabar- dian (Gordon and Appelbaum 2006), and Upper Necaxa Totonac (Beck 2006). In an extensive survey of the characteristics of ejective fricatives in these languages, Shosted and Rose (2011) showed that they acoustically displayed shorter and ‘scrapy’ frication noise, generally preceded and/or followed by a silent interval. The shorter duration of the frication portion of fricatives produced with the ejective mechanism is expected given the small volume of air available and the limited mobility of the larynx. The short duration of ejective fricatives was not however systematically reported. In Tlingit, for example, Maddieson et al. (2001) reported averages well over 100 ms in duration. A range of duration similar to what is observed for pulmonic fricatives in other languages. The tendency for scrapiness during frication noise, which translates into a trill-like effect, is attributed to intermittent obstructions of the narrow construction, either due to articulators or saliva (Maddieson et al. 2001). The scrapy quality may also be caused by friction at the aryepiglottic folds (Esling 1996, Moisik 2008). In the present study we examine the acoustic characteristics of Mehri ejective initial and intervocalic fricatives and show how native speakers of this language solve the frication and ejectivity aerodynamic dilemma in light of what has been reported for other related and unrelated languages. We begin with some background information on Mehri and its consonantal system and review some of previous reports on ejective consonants in this language.

2 Ejectives in Mehri phonological system

Mehri belongs to the Modern South Arabian (msa) languages, a subgroup of the Semitic branch of the Afroasiatic family, spoken in Oman (and also in Yemen). In addition to Mehri, this family includes 5 other languages: Baṭħari, Harsūsi, Jibbali, Hobyot and Soqotri. Mehri, spoken by an estimated 50.800 speakers, is considered by the unesco as definitely endangered. The situation of the other msa languages is critical: Baṭħari (200 speakers), Harsūsi (600 speakers), Hobyōt (100 speakers), and Jibbali (25000 speakers). Mehri and the

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access 142 ridouane and gendrot

table 1 The consonantal system of Mehri (adapted from Bendjaballah and Ségéral 2014)

lab. interd. alv. lat. pal. vel. uvul. phar. lar. obstruents stops voiceless t k ʔ voiced b d g ejective t’ k’ fricatives voiceless f θ s ɬ ʃ χ ħ h voiced ð z ʁ (ʕ) ejective θ’ s’ ɬ’ ʃ’ sonorants nasals m n liquids (voiced) r l glides j w

other msa languages are particularly threatened by the prestigious status of Arabic, the only official and national language of Oman. Arabic and msa languages are both related since they are both Semitic languages. The exact nature of this relationship is still a matter of controversy (Hetzron 1974, Corriente 2003). The common classification, though not uncontroversial, views msa languages as part of a South branch of West Semitic and Arabic as part of a Central branch of West Semitic (Hetzron 1997). Arabic and msa languages are mutually incomprehensible. The variety of Mehri investigated in this study is spoken in Salalah, in the province of Dhofar. Its consonantal system is given in Table 1. In addition to voiceless and voiced stops and fricatives, the langue has ejective stops as well as a series of 4 ejective fricatives [θ’, s’, ɬ’, ʃ’]. Like other Semitic languages, it lacks voiceless labial stop /p/ and labiodental voiced fricative /v/. Each segment in Table 1 has a geminate counterpart which can be either lexically given or phonologically derived (Lonnet 1993, Rubin 2010, Watson 2012, Bendjaballah and Ridouane 2016). Ejectives were shown to pattern together with uvulars and pharyngeals as a natural class defined by the feature [+ low]. One very important characteristic of this class of segments is that it systematically triggers the diphthongization of following long high vowels /iː/ and /uː/ to [aj] and [aw], respec- tively, and the lowering of long /eː/ into /aː/ (Lonnet 1993, Rubin 2010, Bend- jaballah and Ségéral 2014). This is illustrated in Figure 1 for the following two forms:

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access on ejective fricatives in omani mehri 143

figure 1 Illustration of vowel formants for tokens [iɬuːbər] “to measure” and [iɬ’uːbəl] “to drink” for speaker s2, showing the diphthongization induced by ejectivity

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access 144 ridouane and gendrot

(1) Imperf, 3ms

√ ɬbr ‘measure’ /iɬuːbər/ [iɬuːbər] √ ɬ’bl ‘drink’ /iɬ’uːbəl / [iɬ’awbəl]

A large amount of variability was reported in the realization of Mehri ejective segments, depending on dialectal origin, place of articulation, manner of articulation, and prosodic position (Johnstone 1975, Lonnet 1993, Lonnet and Simeone-Senelle 1997, Watson and Bellem 2010, Watson 2012). The tran- scriptions used to represent these segments have been inconsistent, since the early twentieth-century Viennese expedition (Jahn 1902, 1905) to more recent researchers (Johnstone 1975, Sima 2009, Lonnet and Simeone-Senelle 1997, Sima 2009, Rubin 2010, Watson and Bellem 2010, Watson 2012). This variability inevitably raised some debate as to whether the description of these segments as ejectives was, in fact, correct. Lonnet (1993), although he uses the generic term “emphatics”, views these segments, both stops and fricatives, as being most commonly produced as glottalized or ejective segments as in Ethio- Semitic languages. Lonnet and Simeone-Senelle (1997) and Simeone-Senelle (1997), who also view these segments as ejectives, report some prosodically conditioned variation in their phonetic nature, leading to dorsopharyngealization, voicing, or creaky voice. Rejecting the term ‘ejective’,Watson (2012) argues that in addition to the velar stop /k’/ which is uniformly produced as ejective, only /s’/ showed ejective tokens utterance-initially and -finally. Otherwise, /s’/ was realized as an emphatic, i.e. dorsopharyngealized consonant. The other fricatives are reported to be dorsopharyngealized (as in Arabic dialects). The interdental fricative was, for example, reported to be invariably dorsopharyngealized in all environments. The great variability in the phonetic implementation of ejective consonants has led some researchers to argue for a gradual sound change which switches, given the overwhelming influence of Arabic, from a contrast of ejectivity to a contrast of dorsopharyngealization (Lonnet 1993, Watson and Bellem 2010, Watson 2012). Most of the reports on Mehri ejectives were limited to dialects spoken in Yemen. They also relied mostly on perception-based segmental transcription to capture the native speakers’ productions. Watson and Bellem (2010), who use the generic term ‘emphatics’ to refer to ejectives, provided a preliminary acoustic study of Yemeni Mehri ejective stops and fricatives, based on qualitative analysis of waveforms and spectrograms recorded from one native speaker. The only characteristic examined, however, was the presence/absence of a

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access on ejective fricatives in omani mehri 145 spike on the waveform, which was argued to be typical of ejectives. Their conclusion was that the acoustic manifestation of ejectives differed depending on place and manner of articulation of the consonants. The main correlates they observed were glottalic initiation in the case of /k’/, and tongue retrac- tion and pharyngeal contraction in the case of /t’/ and the ejective fricatives. According to the authors, the perception of these consonants as ejectives in pre-pausal position (due to the phenomenon of pre-pausal glottalization) combined with the fact that the velar stop /k’/ was systematically produced on a glottalic airstream presumably led researchers to the assumption that emphatics as a class were ejectives. This assumption is incorrect according to them since the only invariable ejective was the velar stop, all other ejectives were realized with at least partial pharyngeal contraction, and two of the sibilants /ʃ’/ and /ɬ’/ were affricated. For our study we recorded five Omani Mehri native speakers and examined several temporal and non-temporal characteristics that might acoustically differentiate [θ’,s’,ɬ’,ʃ’] from their pulmonic counterparts [θ, s, ɬ, ʃ]. More specif- ically, we seek to show how Mehri speakers solve the problematic combination of glottal constriction and frication noise, and to determine whether the cues used signal ejectivity or some other dimension such as dorsopharyngealization.

3 Method

3.1 Speakers Five male Mehri native speakers aged from 20 to 27 (mean 22.2, sd 2.86) were recorded during a fieldwork in Salalah, a city in southern Omani province of Dhofar. All of the participants reported being able to speak Omani Arabic and Standard Arabic, as is common for native Mehri speakers. None of them had any known history of hearing or speech disorders at the time of record- ings.

3.2 Procedure Table 2 shows the data used in this study. It consisted of 4 pairs of triliteral consonant roots contrasting each ejective fricative with its pulmonic counterpart (a larger data set was recorded including filler items and items contrasting ejective stops with their non-ejective counterparts). Each root was conjugated in three verbal paradigms (perf. 3ms, imperf. 3ms, subj. 3ms) and produced within a carrier sentence. Only the first two verbal forms were analyzed, yielding data with ejective and pulmonic fricatives in word-initial and

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access 146 ridouane and gendrot table 2 List of the Mehri items examined in this study

Root Perf. 3ms Imperf. 3ms

√sbħ səbuːħ isuːbəħ “to fix” √s’br s’əbuːr is’uːbər “to support” √θbr θəbuːr iθuːbər “to break” √θ’br θ’əbuːr iθ’uːbər “to blame” √ɬbr ɬəbuːr iɬuːbər “to measure” √ɬ’bl ɬ’əbuːl iɬ’uːbəl “to drink” √ʃbl ʃəbla iʃabla “to swallow” √ʃ’rm ʃ’əruːm iʃ’uːrəm “to slap” word-medial intervocalic positions. The third verbal forms contained word- final ejectives not examined here. Target fricatives in initial position were followed by a schwa vowel. In intervocalic position, they were preceded by /i/ and followed by the long vowel /uː/ (Recall that /uː/ following an ejective consonant is systematically realized as a diphthong [aw]). One item had a vowel /a/ as a following vowel due to a lack of actual Mehri verb with the relevant structure. Each verbal root was written down in a slightly modified Arabic alphabet commonly used to transcribe Mehri ejectives. Arabic translation of each item was also provided. For each recording, the speaker was shown the verbal root, asked to conjugate it in perf. 3ms and imperf. 3ms and to produce the items with a natural speaking rate while maintaining the same tempo throughout the session. A total of 397 tokens were included in the acoustical analysis (8 verbal forms × 2 positions × 5 speakers × 5 repetitions − 3 tokens). Three tokens were excluded from analysis because they were mispronounced (2 repetitions of [θ’əbuːr] and one repetition of [θəbuːr]). Acoustic data were segmented and labeled using Praat (Boersma and Weenink 2012) by an expert in phonetics and were checked afterwards by the two authors. Praat was also used for acoustic analyses. Fundamental frequency, intensity, center of gravity were measured automatically at midpoint, while formants were measured at 1/3, 1/2, and 2/3 of the vowel. All measurements made use of the standard parameters implemented in Praat. Different durational and non-durational values were taken from the acoustic records for each item (No measurements were taken from /uː/ because of the diphthongization that affects this vowel in the ejective context and from the vowel /a/ in /iʃabla/). Non-durational values included:

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access on ejective fricatives in omani mehri 147

figure 2 The temporal intervals measured: V = preceding vowel duration, C = pre-frication closure interval, F = frication noise, L = post-frication glottal lag. Token [iθ’uːbər] “to blame”, for speaker s2.

1. Number of fricatives with pre-frication (near-)silent intervals 2. Number of fricatives with post-frication (near-)silent intervals, 3. Effect of ejectivity on the intensity of frication noise, 4. Effect of ejectivity on the center of gravity (CoG) of frication noise, 5. Effect of ejectivity on pitch at the onset of preceding and following vowels, and 6. Effects of ejectivity on formant structure of preceding and following vowels.

Temporal measurements, illustrated in Figure 2, included:

1. Preceding vowel duration, 2. Pre-frication closure duration, 3. Frication noise duration, 4. Total fricative duration, and 5. Post-frication glottal lag duration.

Statistical analysis was computed with the r package (version 3.2.1). Levels of significance were established through anovas using the acoustic parameter as a dependent factor. Speaker, airstream (ejective vs. non-ejective) and place of articulation were simultaneously used as independent factors. Initial and intervocalic fricatives were analyzed separately.

4 Results

4.1 Presence/absence of silence Counts of ejective fricatives with silent lags depending on word position and place of articulation are given in Table 3. While no silent lags occurred dur-

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access 148 ridouane and gendrot table 3 Counts of ejective fricatives with pre-frication and post- frication silent intervals, grouped by place of articulation. Symbols as in Figure 2.

Initial Intervocalic Total for L

LCL

Alveolar 13/25 0/25 9/25 22/50 Interdental 17/23 8/25 5/25 22/48 Lateral 13/25 7/25 8/25 21/50 Palatoalveolar 17/25 0/25 5/25 22/50 Total 60/98 15/100 27/100 87/198 ing the production of pulmonic fricatives, pre- and post-frication silent intervals were observed for ejective fricatives. Regarding pre-frication silence, only interdental /θ’/ and lateral /ɬ’/ manifested pre-release closure in intervocalic position, the two other fricatives are never affricated. Although Table 3 doesn’t report counts of pre-frication silent lags for ejectives in word-initial position, closer examination of the acoustic signals and spectrograms shows that ejective fricatives in this position can also be produced with affrication. Affrication translates into a more or less visible release burst before frication noise as Fig- ure 3 shows. We intended not to report counts of pre-frication silent closures in word-initial position, as these can’t be evaluated in a systematic and reliable way based on acoustic data only. Electropalatographic or aerodynamic data would be a more reliable means for accounting for the affrication of ejectives in word-initial position. Post-frication glottal lags are rather common during the production of Mehri ejective fricatives. Word-initial ejectives (61%) are much more frequently produced with such lags than word-intervocalic ones (27%). Differences were also observed depending on place of articulation: again interdental /θ’/ and lateral /ɬ’/ displayed more silent lags than alveolar /s’/ and palatoalveolar /ʃ’/. These results show that contra to previous reports all fricatives, including interdentals, can display post-frication silent intervals characteristics of ejectivity. It is important to note, however, that important differences among native speakers were observed, as shown in Table 4: s3 produced the largest amount of silent lags while s5 and s1 produced the least, s4 and s2 lay in between.

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access on ejective fricatives in omani mehri 149

figure 3 Pre- and post-frication lags for [θ’əbuːr] “to blame” and [ʃ’əruːm] “to slap” for speaker s2

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access 150 ridouane and gendrot table 4 Counts of ejective fricatives with pre-frication and post-frication silence intervals for the 5 subjects, grouped by place of articulation

[s’] [θ’] [ɬ’] [ʃ’] Total s1 0 6 2 0 8 s2 8 7 5 5 25 s3 10 10 10 10 40 s4 4 6 7 5 23 s5 0 1 3 2 6 Total 22 30 28 22 102

table 5 Effect of ejectivity on f0, CoG and intensity, grouped by word position

Position Airstream f0 CoG Intensity

Initial Ejective 142.88 5036 57.96 Pulmonic 137.71 4266 54.08 P-Value 0.89 < 0,0001 0.002 Inter. Ejective 121.12 4721 59 Pulmonic 120.84 3874 58.4 P-Value 0.93 < 0.0001 0.21

4.2 Intensity, CoG and pitch Results for intensity, CoG and pitch are summarized inTable 5. Regarding pitch, results show that the presence or absence of ejectivity has no significant effect on f0 of the following vowels (for initial fricatives) and on f0 of the preceding vowel (for intervocalic fricatives). Frication noise intensity varies for initial fricatives (see also Figure 4), with significantly higher intensity for ejectives. No such differences were observed, however, in intervocalic position. Ejective fricatives were well differentiated from their pulmonic counterparts by the CoG measurement: it is significantly higher for ejectives than for their pulmonic counterparts (see also Figure 5 for subject s3). As expected, important differences were also observed between different places of articulation. CoG was for example highest for alveolars, which fits with the previsions men-

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access on ejective fricatives in omani mehri 151

figure 4 Illustration of frication intensity for tokens [səbuːħ] “to fix” and [s’əbuːr] “to support” for speaker s3

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access 152 ridouane and gendrot

figure 5 Bar graph for the CoG measurements for speaker s3 in the word-initial position tioned in the literature and further validates our CoG measurements. All pair- wise comparisons of fricatives depending on place of articulation were significant at the p<.0001 level, except for the pair palatoalveolar vs. lateral (p = 0.25).

4.3 Effects on vowel formants As briefly outlined above, several authors have argued that Mehri ejective fricatives are generally realized with an emphatic or dorsopharyngealized con- figuration, akin to what occurs in Arabic languages. Dorsopharyngealization induces important qualitative effects on adjacent vowels, manifested by a raising of f1—albeit generally slight—and a large drop of f2 of the adjacent vowels. The produced f1-f2 proximity was shown to be a highly reliable acoustic cue to dorsopharyngealization in different Arabic varieties (Obrecht 1961, Yeou 2001). One would expect similar acoustic effects of Mehri ejective fricatives on neigh- boring vowels if they were dorsopharyngealized. We measured f1 and f2 of the schwa vowel following initial fricatives and f1 and f2 of the vowel /i/ preceding intervocalic fricatives. In intervocalic position, vowel f1 is higher in the context of ejective fricatives (340.29 Hz, sd=39.09) compared to their pulmonic counterparts (310.55 Hz, sd=28.99). It is important to note, however, that the size of this difference, though significant at the p<.0001 level, is not as important as the one reported in emphatic consonants in Arabic (Yeou 2001). No significant differences between ejective and pulmonic contexts were found as far as vowel f2 is concerned (p=0.22). For word-initial context, significant differences were observed both for vowel f1 and f2. Schwa vowel has a significantly higher f1 and significantly lower f2 following ejective fricatives (f1: 455.94 Hz, sd=50.41; f2: 1200.56 Hz, sd=141.10) compared to their plain counterparts (f1: 392.07, sd=38.60; f2: 1417.49, sd=168.97). In this position, ejective fricatives display expected acoustic outcomes of a dorsopharyngealized articulation. This is illustrated in Fig- ure 6.

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access on ejective fricatives in omani mehri 153

figure 6 Illustration of vowel formants for tokens [ɬəbuːr] “to measure” and [ɬ’əbuːl] “to drink” for speaker s2

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access 154 ridouane and gendrot

4.4 Temporal values Ejectivity has no significant effect on preceding vowel duration (p=0.51). The vowel /i/ has virtually the same duration preceding an ejective fricative (87.82 ms, sd=26.29) and its pulmonic counterpart (90.26 ms, sd=28,68). Two sets of measurements were applied to fricative duration for both initial and intervocalic positions. In set1 we measured only frication duration by subtracting the duration of any silent intervals. In set2 we measured the entire fricative duration including silent intervals, if any. For intervocalic position, set1 measurements yield significant differences between ejectives and their plain counterparts (p=0.005). As expected, frication noise for ejectives (87.9 ms, sd=33.03) was shorter than for their plain counterparts (105.8 ms, sd=19.19). No differences were observed, however, when total fricative duration was measured (107.48, sd=21.12 for ejectives, and 105.8, sd=19.19 for their plain counterparts). Absence of significant differences was also observed for initial fricatives (122.39 ms, sd=21.12 for ejectives, and 123.88 ms, sd=19.19 for their plain counterparts) as far as set2 is concerned (p=0.38). Again, as in set1 measurements for intervocalic position, the duration of frication noise is shorter in word- initial ejective fricatives than in plain fricatives at the p<.0001 level (101.94 ms, sd=27.30 for ejectives and 123.88 ms, sd=19.19 for their plain counterparts). Alveolars have the longest duration of this interval and interdentals the short- est. Palatoalveolars and laterals lay in between. The mean duration of the post-frication silent interval in initial position (25.43, sd=12.04) is only slightly longer than in intervocalic position (20.63, sd=8.4). Virtually the same temporal values were obtained for pre-frication noise interval for intervocalic ejectives (24.3, sd=8.8). These durational values are closer to those reported for ejective /s’/ in Amharic (Demolin 2002) and Tigrinya (Shosted and Rose 2011), when compared to the longer lags reported for Tlingit ejectives (Maddieson et al. 2001).

5 Discussion and conclusion

The thread of this research has been to determine how Mehri native speakers implement ejectivity in fricatives given its aerodynamic incompatibility with a continuing flow of air. We recorded 5 native speakers producing different types of pulmonic and ejective fricatives in word-initial and -intervocalic positions. A common acoustic cue to ejectivity generally observed in world languages is the silent lag following closure release. This silent gap, due to glottal closure, has been reported for ejectives, e.g. in Tigrinya (Kingston 1985), Ingush (Warner 1996), Waima’a (Stevens and Hajek 2004), Georgian (Vicenik, 2010),

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access on ejective fricatives in omani mehri 155 and Maya Yucatec (Bayraktar and Ridouane 2016). Fre Woldu (1985) showed that this period of post-closure silence is the most reliable cue in distinguish- ing pulmonic and ejective stops. Our results show that Mehri ejective fricatives, regardless of their place of articulation, also display post-frication silent lags, though with some inter-speaker variability. This is an important result since it provides evidence that fricatives, including interdentals, display a telltale characteristic of ejectivity, contra to previous reports (e.g. Watson and Bellem 2010). Important differences were observed depending on word position: word-initial ejectives (61%) are much more frequently produced with post-frication closure lags than word-intervocalic ones (27%).We come back to this prosodically induced difference below, when we consider the other cues available in these two word positions. In addition to post-frication silent intervals, Mehri ejective fricatives can also be produced with a silent gap preceding frication noise. This closure lag is only produced by some subjects and occurred in only 15% of the tokens in intervocalic position, much less than the 80% reported for Tigrinya for example (Shosted and Rose 2011). As reported above, there is some debate as to the source of this silent period. Maddieson et al. (2001) who observed the same phenomenon on Tlingit interpreted this as evidence that formation of a glottal closure occurs before the onset of frication in the production of ejective fricatives and that these segments should be described as real ejectives produced with glottal closure throughout the production of the frication. Shosted and Rose (2011), on the other hand, argued that the source of this silent gap is oral rather than glottal. They interpret this silence as linguopalatal closure on the basis of comparative electropalatographic and aerodynamic evidence in the related language Amharic (Demolin 2002, 2004). In the absence of articulatory data on the production of Mehri ejective fricatives, we leave the question whether the source of pre-frication silence is oral or glottal undecided. Accord- ing to Shosted and Rose (2011), affrication of the ejective /s’/ in Tigrinya is a strategy used to satisfy the aeroacoustic demands of ejectivity. The changing of the manner of articulation of /s’/ in Tigrinya has no consequence on the consonantal inventory of the language, since Tigrinya has no homorganic affricate to pair with /s’/. The same is true for Mehri, as the language lacks phone- mic affricates. Surprisingly, although changing Mehri fricatives to affricatives is not expected to have any effect on the phonological system of the language, ejective alveolars and palatoalveolars are never affricated in intervocalic position. Ejective fricatives display other characteristics of an ejective articulation. Intensity and CoG, for example, are higher for ejective fricatives compared to their non-ejective counterparts. The higher CoG of the noise distribution of

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access 156 ridouane and gendrot ejective fricatives suggests that they are produced with a narrower constriction compared to the pulmonic fricatives. This narrower constriction has the effect of increasing the intraoral air pressure while reducing the airflow through the constriction. This reduced amount of the airflow for ejectives is compati- ble with their shorter frication noise duration. Indeed, similar to what has been reported in other languages (Tlingit, Tigrinya, Kabardian), Mehri ejective fricatives have significantly shorter frication noise duration than their pulmonic counterparts. But this temporal difference fades away when total fricative duration is measured. Similarly, ejectivity has no significant effect on preceding vowel duration. This result differs from what has been reported for Tigrinya, for example, where vowels were shown to have a significantly shorter duration when preceding ejectives. Shorter vowel duration in Tigrinya may be interpreted as a strategy to compensate for the shorter fricative duration of ejective fricatives. In Mehri, no such temporal compensation is needed since ejective fricatives have the same total duration as their pulmonic counterparts. Taken together the results reported so far provide important evidence that Mehri fricatives display characteristics of an ejective articulation. There is one aspect of our results however which may be interpreted not as a correlate of ejectivity but rather of dorsopharyngealization. This concerns the effect of ejectivity on the formant structure of adjacent vowels. Similar to what has been observed for dorsopharyngealized consonants in Arabic, Mehri ejective fricatives, mainly in word-initial position, induce raising of f1 combined with a drop of f2. A large drop of f2 has been shown to be the main perceptual cue to dorsopharyngealization in Arabic (Obrecht 1968). According to Yeou (1995), even a relatively small f2 lowering, similar to what is reported in our results, may be salient enough to perceive dorsopharyngealization when f1 is high enough. The exact articulatory mechanism responsible for this acoustic effect on vowel quality is not clear, however. Lowering of f2 after ejective fricatives can be attributed to a pharyngeal constriction. But it can also be the consequence of a simultaneous depression of the palatine dorsum (Ali and Daniloff 1981, Zeroual et al. 2007). The fact that Mehri ejectives don’t induce substantial raising of f1 is another argument showing that the pharyngeal constriction during the production of these fricatives is not narrow enough. As already stated, the effect of ejectivity in fricatives reflects important acoustic variability depending mainly on word position. Unlike in initial position, intervocalic ejective fricatives display less cues to ejectivity: they are less intense, induce less effect on vowel f2, and, most importantly, are produced with less post-frication lags. These prosodically induced acoustic variations might be perceptually motivated. In intervocalic position, induced vowel diphthongization is perceptually salient enough to allow speakers recover the con-

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access on ejective fricatives in omani mehri 157 trast. The poor acoustic cues to ejectivity may be a diagnostic of the ongoing neutralization of this laryngeal contrast for fricatives. The fact that initial ejective fricatives display acoustic correlates of dorsopharyngealization could be interpreted as yet another strategy to further marginalize ejectivity in fricatives. Serving as an enhancing feature, dorsopharyngealization helps increase the perceptual distance between ejectives and their pulmonic counterparts. As several scholars have already argued, one can reasonably attribute this feature to the strong influence of Arabic, a consequence of the long-lasting contact between the two languages. Massive borrowing of Arabic vocabularies, none of which has an ejective fricative, should decrease the frequency of occurrence of these segments in the lexicon, which will inevitably decrease the frequency of occurrence of ejective fricatives in Mehri speech. Future work should bring attention to ejective fricatives as areally rare sounds, and as a specificity of the language that may be particularly fragile in a context of endan- germent.

References

Ali, L.H. & Daniloff, R.G. (1972). A contrastive cinefluorographic investigation of the articulation of emphatic-non emphatic cognate consonants. Studia Linguistica 26, 81–105. Bayraktar, E. & Ridouane, R. (2016). Le vot des éjectives: le cas du maya yucatèque. Actes des 29e Journées d’Etudes sur la Parole. Paris. Beck, D. (2006). The emergence of ejective fricatives in Upper Necaxa Totonac. In R. Kirchner (ed.), University of Alberta Working Papers in Linguistics 1, 1–18. Bendjaballah, S., & Ridouane, R. (2016). Word-final geminates in a Modern South Ara- bian language: Phonetics and Phonology. 15th Conference of Laboratory Phonology (LabPhon 15). Cornell University, Ithaca, ny (usa). Bendjaballah, S., & Ségeral, P. (2014). The phonology of ‘idle glottis’ consonants in the Mehri of Oman (Modern South Arabian). Journal of Semitic Studies 59 (1), 161–204. Boersma, P., & Weenink, D. (2012). Praat: doing phonetics by computer. http://www .praat.org/. Corriente, F. (2003). On the degree of kinship between Arabic and Northwest Semitic. Proceedings of aida 5, Cadix, 187–194. Demolin, D. (2002). The search for primitives in phonology and the explanation of sound patterns: The contribution of fieldwork studies. In Gussenhoven, C., Warner, N. (eds.), Laboratory phonology 7, The Hague: Walter de Gruyter, 455–514. Demolin, D. (2004). Acoustic and aerodynamic characteristics of ejectives in Amharic. Journal of the Acoustical Society of America 115(5), 2610.

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access 158 ridouane and gendrot

Esling, J.H. (1996). Pharyngeal consonants and the aryepiglottic sphincter. Journal of the International Phonetic Association 26: 65–88. FreWoldu, K. (1985). The perception and production of Tigrinya stops (Reports from Uppsala University Department of Linguistics 13). Uppsala: Uppsala University. Gordon, M., & Appelbaum, A. (2006). Phonetic structures of Turkish Kabardian. Jour- nal of the International Phonetic Association 36(2), 159–186. Hetzron, R. (1974). La division des langues sémitiques. In Caquot, A. & D. Cohen (Ed), Actes du 1e congrès internationale de la linguistique sémitique et chamito-sémitique, Paris, 16–19 juillet, 1969, La Haye & Paris: Mouton & Co., pp. 181–194. Hetzron, R. (ed.) (1997). The Semitic languages. London: Routledge. Jahn, A. (1902). Die Mehri-Sprache in Südarabien: Texte und Wörterbuch. Wien: Hölder. Jahn, A. (1905). Grammatik der Mehri-Sprache in Südarabien. Wien: Hölder. Johnstone, T.M. (1975). The Modern South Arabian languages. Afroasiatic Linguistics 1(5), 93–121 / Monographic Journals of the Near East, 1–29. Kingston, J. (1985). The phonetics and phonology of the timing of oral and glottal events. Ph.D. dissertation, University of California, Berkeley. Lonnet, A. (1993). Quelques résultats en linguistique sudarabique moderne. Quaderni di Studi Arabici 11, 37–82. Lonnet, A., & Simeone-Senelle, M.C. (1997). La phonologie des langues sudarabiques modernes. In: Kaye, A. (ed.), Phonologies in Asia and Africa 1, Winona-Lake: Einsen- brauns, 337–372. Maddieson, I. (1997). Combining frication and glottal constriction: Two solutions to a dilemma. Journal of the Acoustical Society of America 102(5), 3135. Maddieson, I. (1998). Why make life hard? Resolutions to problems of rare and difficult sound types. Proceedings of the Annual Meeting of the Berkeley Linguistic Society 24, 367–380. Maddieson, I. (2013). Glottolized consonants. In: Dryer, M.S., Haspelmath, M. (eds.),The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolu- tionary Anthropology. (Available online at http://wals.info/chapter/7, accessed on 2015-01-31). Maddieson, I., Smith, C.L., & Bessell, N. (2001). Aspects of the phonetics of Tlingit. Anthropological Linguistics 43 (2), 135–176. Moisik, S.R. (2008). A three-dimensional model of the larynx and the laryngeal con- strictor mechanism: Visually synthesizing pharyngeal and epiglottal articulations observed in laryngoscopy. ma thesis, University of Victoria. Müller, D.H. (1909). Südarabische Expedition. Band ix. Mehri- und Ḥaḍrami-Texte gesammelt im Jahre 1902 in Gischin von Dr. Wilhelm Hein, bearbeitet und heraus- gegeben von Dav. Heinr. Müller. Wien: Hölder. Obrecht, D. (1961). Effects of the Second Format in the Perception of Verlarization in Lebanese Arabic. Ph.D. Dissertation, University of Pennsylvania.

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access on ejective fricatives in omani mehri 159

Rubin, A. (2010). The Mehri language of Oman. Leiden: Brill. Shosted, R.K., & Rose, S. (2011). Affricating ejective fricatives: The case of Tigrinya. Journal of the International Phonetic Association 41(1), 41–65. Sima, A. (2009). Mehri-Texte aus der jemenitischen Sharqīyah: Transkribiert unter Mitwirkung von ʕAskari Saʕd Ḥugayrān. Edited, annotated and introduced by J.C.E. Watson & W. Arnold. Wiesbaden: Harrassowitz. Simeone-Senelle, M.-C. (1997). The Modern South Arabian languages. In R. Hetzron (ed.), The Semitic languages, 378–423. London: Routledge. Stevens, M. & Hajek, J. (2004). A preliminary investigation of some acoustic characteristics of ejectives in Waima’a: vot and closure duration. Proceedings of the tenth Aus- tralian International Conference on Speech Science and Technology, 277–282. S. Cas- sidy, F. Cox, R. Mannell & S. Palethorpe (eds.), Sydney: assta. Vicenik, C. (2010). An acoustic study of Georgian stop consonants. Journal of The International Phonetic Association 40, 59–92. Warner, N. (1996). Acoustic Characteristics of Ejectives in Ingush. Proceedings of the International Conference on Spoken Language Processing, Philadelphia, 1525–1528. Watson, J. (2011). Glottalisation and neutralisation in Yemeni Arabic and Mehri: An acoustic study. In: Heselwood, B., Hassan, Z.M. (eds.), Arabic Instrumental Phonetics, Amsterdam: Benjamins, 436–452. Watson, J. (2012). The Structure of Mehri. Wiesbaden: Harrassowitz. Watson, J., & Bellem, A. (2010). A detective story: Emphatics in Mehri. Proceedings of the Seminar for Arabian Studies 40, 345–356. Yeou, M. (2001). Pharyngealization in Arabic: Modelling, acoustic analysis, airflow and perception. Revue de La Faculté des Lettres El Jadida 6, 51–70. Zeroual, C., Hoole, P., Fuchs, S. & Esling, J.H. (2007). ema study of the coronal emphatic and non-emphatic plosive consonants of Moroccan Arabic. Proceedings of the xvith International Conference on Phonetic Sciences, Saarbrücken, 397–340.

Brill’s Journal of Afroasiatic Languages and Linguistics 9 (2017) 139–159 Downloaded from Brill.com09/30/2021 08:44:52PM via free access