Evidence of Gemination in Persian (Phonetic and Phonological Study of Lexical and Post-Lexical Geminates)
17 論文
Evidence of Gemination in Persian (Phonetic and phonological study of lexical and post-lexical geminates)
Mahjoub ZIRAK and Peter M. SKAER
Graduate School of Integrated Arts and Sciences, Hiroshima University
Abstract the perceptual distance between the phonemic The purpose of this study is to examine the categories. phonetic interpretation of geminate contrast and the articulatory differences between the lexical and 1. Introduction post-lexical geminates in Persian. In terms of two geminate types, the findings indicated that word- This paper presents fi ndings on the articulatory boundary geminates display the same temporal conditioning of lexical and post lexical geminate values as lexical geminates, however, unlike lexical consonants in Standard Contemporary Persian. The geminate consonants, the phonetic implementation phonetic and phonological distinctions between the of adjacent identical consonants in word-boundary different types of geminates have been studied in geminates are not allocated the feature [+tense]. many of the world’s languages, as well as studying Such phonologically difference affected the vowels the differences between singleton and geminate preceding the two geminate types. Results from consonants. Phonological research describes analyzing the qualitatively short and long vowels two different types of geminates with differing preceding the word-boundary geminates showed structures. Lexical geminates are phonemic and separate distributions for consonant sequences represent as single melodic units associated with affected by different vowel types in which the two timing slots (Leben 1980), whereas post tenseness of the vowel // and the duration of lexical geminates are formed by concatenation of consonant sequences would result in considerable two identical consonants across a word boundary interaction. Contrary to RMS amplitude, formant or by total assimilation of a segment which takes frequencies as a robust secondary cue, could the identity of the adjacent segment at a word contribute to the perception of the vowel and internal morpheme boundary (McCarthy 1986; consonant discrimination in two types of geminates Hayes 1986; Lahiri and Hankamer 1988; Ridouane in Persian. These results demonstrate that temporal 2010; Oh and Redford 2011). The representations compensation is maintained with the interaction of singletons, lexical geminates, geminates created between the preceding vowels and consonants in by concatenation of two identical stops across a two geminate types. Duration as a primary correlate morpheme boundary and a geminate derived by would be enhanced by Formant Frequency values total assimilation are illustrated in the Table 1. as an additional acoustic correlate and increases 18 Mahjoub ZIRAK and Peter M. SKAER
Table1. The representations of singleton and three types of geminate consonants singleton Lexical geminate Concatenated geminate Assimilated geminate
X X X X X X X
k k k k d k
To some extent, morphological boundaries preceding them (e.g. Italian: Espisito and di impact the phonetic manifestation of geminates. Bendetto 1999; Pickett et al. 1999; Arabic: Al- Within phonetic implementation, lexical geminates Tamimi 2004; Malayalam: Local and Simpson are phonetically long segments which contrast 1999; English, German, Spanish and French: with their singleton counterparts as short segments Delattre’s 1971; etc.). Several studies provided (Ladefoged and Madieson 1996), whereas the post evidence that relative duration plays a critical lexical geminates are not contrastive in length. role in distinguishing lexical and post lexical Some phonetic studies have shown that generally geminates, where the absolute consonant duration lexical and post lexical geminates are phonetically is indistinguishable. Oh and Redford (2011), identical, as in Bengali and Turkish (Lahiri and by taking an example from Tashlhiyt Berber Hankamer 1988), Lavantine Arabic (Miller (Ridouane 2010), state that the preceding vowel 1987), Sardinian (Ladd and Scobbie 2003) and duration differences for lexical and post lexical Tashlhiyt Berber (Ridouane 2010). This is because geminates should have translated into relative of the phonological representation of geminates duration differences between the two types of which captures the similarity in the number of geminates, since Ridouane found that there is timing units lexical and post lexical geminates no difference in absolute consonant duration of are associated to. Based on this phonological lexical and post lexical geminates. Unlike those representation, the difference between lexical and studies reviewed above which attempted to defi ne post lexical geminates will not be manifested in the relative duration differences of contrasting absolute timing differences at the phonetic level. categories, Hansen (2004) stated that the C/V ratio is not a clear discriminator between singleton 1.1. Relative duration and geminates in Persian as has been observed in Italian and Icelandic. He claimed that this is partly Although absolute duration of a consonant is due to the fact that in Persian the vowel preceding usually the primary acoustic correlate of geminate the geminate tends to be longer. Regardless of the contrasts, this indistinguishable cue is always representation of geminates and many languages supported by relative duration. Relative duration, in which gemination has been characterized often expressed in duration ratios between two or as a phonological manifestation, phonetic more segments, presents a more reliable criterion implementation of gemination is described as for defining the segmental duration in signaling having an extremely limited domain. This study phonemic contrast (Idemaru and Guion 2010). examines the way these abstract phonological More substantial differences of relative duration representations are refl ected in the phonetic details manifest themselves in the duration of preceding of speech production, and the way they are related vowels. Acoustic evidence has been offered to the distinct behavior of the different types of suggesting that there is a temporal compensation geminates (namely, Ambiguity and Inalterability). relationship between geminates and the vowels We argue that Persian provides an interesting data Evidence of Gemination in Persian 19 set which represents this explanation. due to the fact that in Persian the vowel preceding the geminate tends to be longer. He considered 1.2. Current study the syllable duration as a more useful parameter for understanding the distinction between the Persian has twenty-three consonantal and six categories. As Hansen (2004) only analyzed vocalic phonemes (University of Victoria Phonetic the duration of two stop consonants without Database, UPVD 1999). Persian maintains a considering the different durations of 3 vowels phonological contrast between geminate and preceding them, it is not clear how the ratio of other singleton consonants. For example [], consonants to identical preceding vowels duration ‘classified’ contrasts with [], ‘mysterious’. will play out. As all consonant types in Persian can Persian has three types of geminates represented be geminate, consonants to an identical preceding as lexical geminates, geminates created by vowel ratio may need to be specifi ed for managing concatenation of two identical stops across a the data of singletons and geminates. While the morpheme boundary and a geminate derived by fi ndings by Hansen (2004) in Persian geminates are total assimilation. Thurgood (1993) states that reporting from word-internal geminates’ data, one cross-linguistically geminates tend to occur in wonders if it is possible to place the effect in more two basic environments; intervocalically and contexts as well as word-boundary geminates. after short stressed vowels. In Persian, lexical The present analysis focused on the production geminates can occur only in word medial of Persian word internal lexical geminates, which positions and must be preceded by short vowels. are all only found in Arabic loan words, and word- Interesting observations have been made by Iranian boundary post-lexical geminates to investigate researchers regarding geminates: Mahootian the hypothesis that word internal geminates may (1997) states that for native (non-Arabic) words differ from those that arise across word boundary “the geminates in these words are often, if not (i.e. concatenated geminates). Word internal usually, reduced.” Deyhim (2000) transcribed the geminates were compared to phrases with word pronunciations of 16 speakers of Tehrani Persian boundary geminates and to word with singletons and whereas geminated stop consonants in words to distinguish not only the geminate/singleton of Arabic origin are pronounced as geminates contrast, but also to establish whether durational by all informants, the geminate stops in native and non-durational properties would distinguish words are pronounced by some of the informants geminates from different boundary types. as singletons. For the durational contrasts on Some lexical geminates of Arabic loanwords geminate stops in Persian, Hansen (2004) provided in Persian are from the Active Participle derived acoustic evidence suggesting that Persian geminate from Arabic verb forms like ‘teacher’; stops are clearly distinguished from singleton stops the infinitive form of the verb like in production when speaking rate is taken into ‘thought’ (Mitchell 1990). In the case of post account. He examined closure duration, preceding lexical geminates in Persian our study considered vowel duration, closure to preceding vowel ratio word boundary concatenated geminates. Persian and the utterance duration and showed that the concatenated geminates in word boundaries are ratio of geminate to singleton closure consistently represented underlyingly as two timing slots each decreased with increased speaking rate. Hansen associated with a single melodic unit (e.g. /b+b (2004) verified that the C/V ratio is not a clear ndi/→ (Table 2). discriminator between geminate and singletons 20 Mahjoub ZIRAK and Peter M. SKAER
Table 2. Phonetic distribution of word internal and word-boundary geminates in Persian. Word-internal Word-boundary δ δ δ δ δ δ
x x x x x x x x x x x x x x x x x
m o d b b b n d i [moddb] [bbndi]
1.2.1. Consonant types ratios variability and interfere with the production of post lexical geminates. One issue that received relatively less attention in previous literature is the comparison of different 1.2.2. Vowel types manners of geminates in Persian. Other languages that have been studied in this light include Italian Preceding vowel duration was also found (Payne 2005), Cypriot Greek (Arvantini and to be significantly affected by gemination. The Tserdanelis 2000), Buginese, Madurese and interaction between preceding vowel duration and Toba Batak (Cohn et al. 1999). All consonant gemination has been reported in many languages phonemes in Persian can be geminate and contrast (Maddieson 1985). Languages vary in whether phonemically with their singleton counterparts. geminates shorten or lengthen the preceding In the present study, ten consonant types [k, b, vowels. In some languages there are no substantial t, d, s, z, l, r, n, m] were examined in the two differences in preceding vowel duration between geminate type contexts. This subset of consonants singleton and geminates; for example Egyptian was chosen to explore the effect of some certain Arabic (Norlin 1987) and Hungarian (Ham 2001). parameters including place of articulation (bilabial, In Cypriot Greek, there is a slight tendency toward labiodental, coronal); manner of articulation shortening before geminates, but this tendency (stop, fricative, nasal, lateral); and, for the stops, is not very consistent (Arvantini and Tserdanelis phonological voicing (voiced vs. voiceless), to 2000). This shortening is generally explained by keep the dataset to a manageable size. The aim syllable structure differences between singletons was to compare the different manner of consonant and geminates: The vowel is longer in an open geminate types in lexical and post lexical situations syllable (V.CV) and shorter in a closed syllable and examine whether intrinsic durational properties (VC.CV). Ridouane (2010) pointed out that of consonants condition the way the consonant this explanation can probably account for the is geminated and in particular whether these shortening observed in intervocalic position where differences result in asymmetries in the duration the first part of the geminate closes the syllable. ratios contrast (see also Payne 2005). Studies have The Modern Persian vowel system is generally found that the distinctiveness of duration contrasts considered to be a quality-based system in which does depend on consonant manner. For example quantity is not contrastive; e.g. /dir/, ‘late’ contrasts the predicted relationship between sonority and with /del/, ‘heart’ and /dur/, ‘far’ contrasts with geminate markedness is supported by Podesva’s /dor/, ‘round’. A qualitative distinction (tenseness) (2000) survey of 52 languages. It was hypothesized categorizes the Persian vowels into two groups: that articulatory factors would result in durational long vowels /, i, u/ with the feature [+tense], Evidence of Gemination in Persian 21 versus short vowels /, e, o/ with the feature a more complete picture of the articulatory [-tense] (Rouhani Rahbar 2009). According conditioning of geminate consonant duration and to several studies on Persian vowel systems, contrast. In particular, it seeks to establish how far the length of Persian vowels changes based temporal properties like consonant types and local on the structure they occur in (Samareh 1992; ratios such as consonant/vowel preceding duration Toosarvandi 2004). Samareh (1977) considered ratio shape variability in gemination types and two functionally different groups of Persian vowels undermine the contrast between the two types. A with respect to the possible following consonant. further aim is to explore whether non-durational The short vowels /, e, o/ can occur before all properties also distinguish lexical geminates combinations, while the long vowels /, i, u/ have from post lexical, as possibly evident from RMS a very limited occurrence preceding consonant amplitude during the release of consonants clusters. This context constriction is also the same and formant analysis of the preceding vowel’s for geminate consonants in Persian. Qualitatively structure. The current study also focuses on the long vowels cannot occur before geminate influential role of relational timing of consonant consonants. As the preceding vowel duration was to preceding vowel duration to accurately classify found to be significantly affected by geminates, lexical and post lexical geminate productions. We this study carried out a phonetic analysis of two predict that there is more stability in higher-order qualitatively short // and long // Persian vowels acoustic properties, namely the relative duration as the preceding vowels of lexical and post lexical of consonant to vowel duration, than the absolute geminates to investigate whether they have the duration of the consonants. To test this, we same behavior as seen in other languages in which conducted an acoustic study to examine absolute vowels are significantly shorter before lexical consonant duration and various local ratios as well geminates than concatenated word boundary as consonant to preceding vowel duration ratios geminates (Tashlhiyt Berber; Ridouane 2010). It (C:V1). The categorization power of duration is hypothesized that temporal and non-temporal ratios versus absolute consonant duration was parameters significantly affected by gemination also considered. Finally we conclude with a brief and word boundary concatenated geminates in discussion of some non-durational analyses as well Persian may shorten the long vowels preceding as RMS amplitude, phonation differences in the them, contrary to some investigations in other consonants and the quality and quantity effects on languages. This different behavior is due to the the preceding vowel by analyzing the Formants. special characteristics of the Persian vowel system and their manifestations in different contexts, 2. Methods that is, qualitatively long vowels could occur 2.1. Participants before concatenated geminates contrary to lexical geminates. Eight native speakers of Persian (4 male, 4 female) who reported normal hearing, participated. 1.3. Aims of the present study All were native speakers of Persian and were born of Persian speaking parents. The mean age ±SD This study seeks to identify the two types of of the participants was 31.2 ±4.7 ranging from 25 geminates across different boundaries, namely, to 40 years old. None of them reported any speech word-internal (lexical) and word-boundary disorder. (concatenated) geminates in Persian and provide 22 Mahjoub ZIRAK and Peter M. SKAER
2.2. Stimuli word-boundary geminates and their singleton counterparts, the disyllabic words of the form
The stimuli were 32 Persian and Arabic loan CV1C1.C2VC were chosen. The syllables C2VC of words, presented in Tables 3 and 4. All test words items were all stressed. To test the vowel duration, are stressed on the last syllable, because stress the preceding vowel of V1 in the lexical words of falls on the last syllable of Persian nouns. Three the form CVC1V1C2C2VC and also their singleton factors, namely, absolute consonant durations, counterparts of the form C1V1C2VC covered the the duration of the preceding vowel and the low front //. Although it may have been ideal to local ratio of consonant to preceding vowel vary the vowel context of the preceding position, duration were examined in lexical and word- only /-/ is used for the vowel context in order boundary concatenated geminate contexts and to obtain a manageable number of test words. compared with their singleton counterparts. To Two different vowel contexts were chosen for measure the duration, we followed the standard word boundary geminates of the form CV1C1. criteria of segmentation. Following Arvaniti and C1VC in which they differed in vowel preceding
Tserdanelis (2000) and Ridouane (2010), separate of V1 namely, low front // versus low back // measurements were taken for closure duration and (short vs. long vowel). Regarding the comparison release duration values for voiceless stops, VOT, between word-boundary geminates and their and for voiced stops, to highlight the difference in singleton counterparts, the disyllabic words of the gemination effect and the two geminate types for form CV1C1.C2VC with the vowel context of /C these two series of stops. For analyzing the target C1/ and /CC1/ were chosen. It is worth noting that consonants in lexical geminates, the trisyllabic V1 of all test words which preceded singleton and
Arabic loanwords of the form CVC1V1C2C2VC geminate consonants were unstressed.
([mofl) were chosen in which C2 cover the Mapping the difference between the two ten consonant types [k, b, t, d, s, z, l, r, n, m] and geminate types, and the effect of two vowel the last syllables, C2VC, were stressed. The tokens types on word-boundary geminates, additional of concatenated word boundary geminates were measurements, namely non-temporal indices limited to four stop consonants [b, d, c, t] and the were obtained. It is worth noting that the longer test words were disyllabic of the form CV1C1.C1VC. duration of the closure for geminates may lead to
The last syllable C1VC of items were all stressed higher amplitude upon a release of a long plosive similar to the first set of items. In constructing (Hankamer et al. 1989). RMS amplitude of the the test stimuli to evaluate the gemination type target consonants in word-internal and word- differences, lexical geminates which contained boundary geminates was measured in the VOT and only the same four stop consonants of [b, d, c, t] release duration of stop consonants. Differences were selected. Word internal and word boundary between the two geminate types are not restricted geminates were compared also with their singleton only to duration. To test the consonantal resonance counterparts to provide a more complete picture and vowel qualities in word-boundary and lexical of the articulatory conditioning of geminate types geminates, F1 and F2 of the two qualitatively compared to singletons. To this aim, the singleton short and long vowels preceding the word- word test counterparts for lexical geminates boundary geminate consonants and the singleton were selected from disyllabic words of the form counterparts, F1 and F2 of the target consonants in
C1V1C2VC. The syllables C2VC of items were word-boundary, F1 and F2 of the vowel preceding all stressed. Regarding the comparison between the lexical geminates and F1 and F2 of the target Evidence of Gemination in Persian 23 lexical geminate consonants were measured at amplitude and Formant analysis. The subset of mid-point. Constructing the manageable data set, non-temporal parameters consist of lexical and lexical and word-boundary geminates containing word-boundary geminate consonants and the the stop consonants were selected for analyzing vowels preceding them. The 8 subjects were asked the formants. The materials collected for analyzing to produce each token 3 times, yielding 768 tokens temporal parameters was also subject to the RMS in total.
Table 3. List of stimuli with the target singleton/lexical geminates
Table 4. List of stimuli with the target singleton/word-boundary concatenated geminates
2.3. Procedure before recording began. Unfamiliar words were defi ned and produced by the experimenter for the The 32 words shown in Table 3 and 4 were speaker. All recordings were made in an isolated imbedded in similar carrier phrase used in Hansen situation using a Shure SM10A Microphone and a (2004): Zoom H4 digital recorder. The recorded utterances Ali goft ke ____ naboud. were saved in digital format for later analysis. Ali said that ____not-was. Ali said that (the)_____wasn’t there. 2.4. Measurements Participants read the sentences from 3 randomized lists of 32 words, yielding a total of The spoken phrases were displayed in Praat 768 words collected for analysis. Each speaker (Boersma and Weenik 2007). Measurement criteria was given several minutes to look over the stimuli were established after visual inspection of both the 24 Mahjoub ZIRAK and Peter M. SKAER spectrogram and the waveform. Acoustic duration matched singletons and the vowels preceding of the pre-consonant vowel and consonant duration them. The differences between the two kinds of for fricative and nasals and closure duration for geminates regarding the duration of their preceding stops were considered as temporal parameters. vowels and the consonant durations were also Segmentation procedures described below analyzed by using a two-way ANOVA. Separate followed those outlined in prior studies (Lahiri two-way ANOVAs were run on measurements for and Hankamer 1988; Hankamer et al. 1989; Payne non-temporal parameters, namely RMS amplitude 2005). The duration of stop consonants which and formant analysis for each condition. An alpha is defined as the closure duration and the VOT, level of 0.05 was set as the level of significance. was measured from the offset of the preceding SPSS 20.0 statistical software was used for all of vowel from the last complete periodic cycle in the descriptive and analytic statistics. the waveform to the onset of the first complete periodic cycle in the waveform of the following 3. Results vowel. Similar to previous studies (Cohn et al. 3.1. Singleton versus lexical geminates 1999; Payne 2005; Idemaru and Guion 2008), VOT was included as part of the consonant duration The fi rst part of analysis investigated duration which enabled a compatible measurement for the differences of lexical geminates and their singleton durations of voiced and voiceless stops. Fricative counterparts in terms of absolute and relative consonants was taken to be the period of time duration of the target consonants. Consonant elapsed between two points of the left edge of duration is supple according to consonant types frication noise to the onset of the following vowel, and geminate types. Table 5 reflects the mean defined as the first complete periodic cycle. The durations of singleton and lexical geminate tokens sonorant consonants were measured from the across consonant types vary from 63.87 ms (SD = sudden reduction in amplitude and lowering of 2.034) to 149.143 ms (SD = 9.325) for singletons, formants to the onset of the following vowel. The and from 162.541ms (SD = 7.601) to 246.351ms vowel duration was measured from the onset of (SD = 13.602) for geminates. Figure 1 shows the first complete periodic cycle to the offset of mean duration across consonant types. As might the last complete periodic cycle in the waveform be expected, voiceless consonants are longer than with reference to the visible energy of the second their voiced counterparts. Measurements show the formant (F2) of a time-locked spectrogram. longer average of 1.1 times for geminates and 1.3 Advanced statistical methods were used in order times for singletons. As a preliminary screening, to consider the main effects of all factors as well a 2 x 10 (Geminacy x Consonant Type) GLM as factor by factor interactions. The General linear univariate analysis of variance (ANOVA) was run. Model (GLM) univariate procedures which provide As predicted, the difference between consonant analysis of variance were considered to be an types was highly signifi cant [F(9, 460) = 1467.338, appropriate model in this study. Two separate two- p < 0.000]. The same test revealed that differences way univariate ANOVAs were utilized to assess between Geminacy, namely singleton versus lexical differences of consonant types and preceding geminate consonants were signifi cant [F(1,460) = vowel durations between lexical geminates and 41504.79, p < 0.000]. There were also a signifi cant their singleton counterparts. Separate two-way interaction between consonant type and Geminacy ANOVAs were conducted for the analysis of the [F(9, 460) = 248.372, p < 0.000]. The factor duration of word-boundary geminates and the consonant type tested the effect of 10 consonant Evidence of Gemination in Persian 25 types regarding the manner of articulation and and 162.541 ms for the matched geminates. The found that the geminate counterparts of intrinsically factor quantity tested the effect of singleton and short singletons are relatively short; it also applies geminate consonant on duration. Along with many to long consonants. The longest durations were other study results on various languages (Esposito obtained for voiceless stop /k/ for which the mean and Di Benedetto 1999; Lahiri and Hankamer durations of 149.143 ms and 246.351 ms were 1988; Redouane 2010) consonant durations show measured for singleton and geminate respectively. the most robust correlations that distinguishes The shortest duration were measured for the lateral singletons from geminates in Persian. /l/ with the durations of 63.87ms for singleton
Table 5: Mean duration (ms) and Standard Deviation of the target consonants in singleton/lexical geminates (N=480) tkbdszmnlr 144.168 149.143 115.321 96.327 134.13 123.357 85.915 71.661 63.872 65.789 singleton (10.561) (9.325) (8.361) (4.012) (11.025) (12.305) (9.203) (3.201) (2.034) (4.102) 244.961 246.351 221.584 198.698 194.552 178.699 205.694 203.669 162.541 165.989 geminate (11.817) (13.602) (20.301) (8.604) (5.302) (10.551) (14.506) (8.642) (7.601) (9.632)
Figure 1: Mean duration (ms) of the singleton and lexical geminate consonant types
Table 6 sums up the effect of gemination on are significantly affected by gemination, except the release duration of voiced and voiceless stops the VOT duration for voiceless stops which had as well as their closure duration, considered as the the same duration for singleton and geminates. temporal parameters. The table shows the mean The GLM univariate analysis of variance indicated duration (ms) and standard deviation of the tokens that closure duration of stops were highly affected across speakers, repetitions and consonant types. by gemination [F(1,184) = 33587.19, p < 0.000]. Release duration values are presented separately to The significant difference between consonant highlight the differences in gemination effect for types was also predicted [F(3,184) = 687.658, voiced and voiceless stops. As Figure 2 illustrates p < 0.000]. There was no significant interaction the durational differences, all temporal parameters between consonant types and the geminacy [F(3, 26 Mahjoub ZIRAK and Peter M. SKAER
184) = 5.903, p = 0.01]. Signifi cant differences in <0.000]; while the same test revealed that there are the release duration according to the presence or no significant differences between singleton and absence of gemination were confi rmed by one-way geminate voiceless stops in case of VOT [F(1,92) = ANOVA only for voiced stops [F(1,92) = 464.76, p 0.9838, p = 0.3238].
Table 6: Mean duration (ms) and Standard Deviation of 4 stop consonants across the temporal parameters (Cld = Closure duration, Rld = Release duration of voiced stops, VOT = VOT of voiceless stops) (N=192). t Cld t VOT k Cld k VOT b Cld b Rld d Cld d Rld 65.766 78.402 74.911 74.232 100.213 15.112 79.936 16.391 singleton (8.521) (9.565) (8.022) (6.325) (13.057) (2.042) (4.407) (2.385) 167.409 77.552 173.919 72.432 197.247 24.335 177.462 21.236 geminate (10.817) (8.264) (11.254) (5.287) (11.452) (3.898) (7.567) (4.603)
Figure 2: Effect of gemination on the temporal parameters
As the Table 7 and Figure 3 illustrate, [F(9,460) = 20.604, P <0.000]. Despite this, the preceding vowel duration //, is affected by duration differences between the vowels preceding gemination. The GLM univariate test indicated the singleton and geminate consonants were on that there was a significant difference of vowel average 14.294 ms and only the duration difference duration due to gemination [F(1,460) = 1493.75, P for /k/ (24.659 ms) and /d/ (33.079 ms) were <0.000]. Contrary to Hansen (2004) who claimed longer than the average. Considering the maximum that in Persian, the preceding vowel is lengthened (107.161 ms) and minimum (57.882 ms) average before geminates, our data revealed that vowels difference between the duration of singleton and tended to be shorter before geminates but the effect geminate consonants mentioned in Table 5 and was not consistent across consonant types. The Figure 1, enhance the possibility that the effect same test revealed that the effect of consonant of gemination on relative duration ratio to be types was also significant [F(9,460) = 53.112, p considered as a more robust perceptual cue than the <0.000]. The interaction between the consonant duration of the absolute preceding vowel duration types and gemination contrast was also signifi cant alone. Evidence of Gemination in Persian 27
Table 7: Mean duration (ms) and Standard Deviation of the vowels preceding the target singleton/lexical geminate consonants (N=480)
tkbdszmnlr 116.168 137.243 126.547 136.333 118.456 125.255 113.771 116.939 124.438 132.466 singleton (8.803) (10.829) (16.252) (8.022) (20.336) (9.668) (14.597) (16.332) (20.124) (21.005) 105.888 112.584 116.351 103.254 106.332 108.522 99.2102 99.251 108.471 115.441 geminate (9.355) (11.692) (4.394) (9.632) (12.374) (7.548) (3.205) (3.602) (9.302) (8.201)
Figure 2: Mean Duration (ms) of the vowels preceding the target consonants
The consonant to vowel preceding ratio of of consonants to their vowels preceding them each consonant type across the function of lexical [F(1,140) = 724.572, p <0.000]. The consonant geminates and the matched singletons are shown type articulation also affected the consonant in Table 8 and Figure 3. The results revealed that to vowel preceding ratio [F(9,140) = 22.097 p gemination contrast was highly affected the ratio <0.000].
Table 8: Vowels preceding/Consonant Duration Ratios as a function of singleton/lexical geminate consonants (N=160) C/V1 t k b d s z m n l r
singleton 1.241 1.086 0.911 0.706 1.132 0.984 0.755 0.612 0.513 0.496
geminate 2.313 2.188 1.904 1.924 1.829 1.646 2.073 2.052 1.498 1.437 28 Mahjoub ZIRAK and Peter M. SKAER
Figure 3: Vowels preceding/Consonant Duration Ratios as a function of singleton/lexical geminate consonants
Figure 4: Scatter plot of consonant duration (ms) vs. preceding vowel duration (ms). The circles are singletons and the triangles are geminates, colored in case of different consonant types (N=160).
The effect of consonant to preceding vowel them. Thus, the data presented in the plot shows ratio (C/V1) is clearly indicated in the scatter that the difference between lexical geminates and plot in Figure 4. As predicted, vowels preceding the matched singletons can be discriminated with singleton consonants were longer than the geminate almost all of the C/V1 ratios in Persian. consonants which shorten the vowels preceding Evidence of Gemination in Persian 29
3.2. Lexical versus word-boundary concatenated of variance ANOVA revealed that there were geminates no high significant differences between them [F(1,184) = 6.867, p = 0.0314]. Although the The phonetic characteristics of differences effect of consonant types was significant in each between geminate types in Persian have not been geminate type [F(3,184) = 182.233, p <0.0001], significantly investigated as in other languages. their interaction showed no significant difference Table 9 and Figure 5 show the duration of stop [F(3,184) = 1.225, p = 0.302]. The effects of consonant properties affected by gemination in geminate type on VOT [F(1,92) = 0.093, p = lexical and word-boundary geminates; namely, 0.835], and release duration [F(1,92) = 0.789, p closure duration (Cld), VOT and release duration = 0.376] were not significant. Thus, in case of of the four voiceless and voiced stops /b, k, t, the consonant duration, geminates produced the d/. Geminate consonants were identical for both same released duration in two different boundary types of geminates to control the comparison types. It is possible that the small difference found between them. Results show that the trend is still between the duration of the closure duration may longer for the closure duration of word-boundary be attributed to the pause inserted by the speakers geminates than the lexical ones, with the difference in word-boundary concatenated geminates. of the average of 5 ms, but the univariate analysis
Table 9: Mean Duration (ms) and Standard Deviation of Consonants in two types of geminates. (Cld = Closure duration, Rld = Release duration of voiced stops, VOT = VOT of voiceless stops) (N=192) t Cld t VOT k Cld k VOT b Cld b Rld d Cld d Rld Word- 172.229 78.583 177.418 73.255 205.697 23.634 183.927 20.655 boundary (7.854) (5.142) (8.996) (2.175) (11.012) (1.281) (8.692) (6.201) 167.409 77.552 173.919 72.432 197.247 24.335 177.462 21.236 Lexical (10.817) (7.128) (11.254) (5.964) (11.452) (3.162) (7.567) (2.038)
Figure 5: Duration (ms) of the target consonants in the function of word-boundary and lexical geminates 30 Mahjoub ZIRAK and Peter M. SKAER
Looking at the results found in case of their matched singletons [F(1,190) = 181.330, p vowels preceding the two types of geminates, <0.000], there were no high statistical differences contrary to what was found in Bengali by Lahiri between the vowel duration preceding the word- and Hankamer (1988), significant differences boundary geminate consonants and the singleton were observed in the preceding vowel durations counterparts [F(1,190) = 7.636, p = 0.06]. However between word-boundary and lexical geminates in the analysis of word-boundary geminates revealed Persian [F(1,190) = 28.882, p <0.000]. In order the same temporal values as lexical geminates, to obtain a thoroughly controlled comparison, the their effects on the preceding vowels displayed vowels preceding the two geminate types were that lexical geminates always affected the duration identical (//). Measurements of the preceding of the preceding vowels by shortening them while vowel durations displayed that these segments word-boundary geminates did not. It is supposed are signifi cantly shorter before lexical than word- that the different behavior is due to the fact that boundary geminates; the same results were also word-boundary geminates are represented as two obtained by Ridouane (2010) on Tashlhiyt Berber. adjacent identical consonants, each associated with Results illustrated in Table 10 and Figure 6 one melodic position, which is not sufficient to indicated that while there was a highly signifi cant manifest as a lexical geminate with doubly linked difference between the duration values of the segments. vowels preceding the lexical geminates and
Table 10: Duration (ms) and Standard Deviation of the vowel preceding the target consonants in the function of word-boundary and lexical geminates and their matched singletons (N=382)
Word-boundary Lexical 122.032 129.072 singleton (12.958) (10.904) 119.198 109.518 geminate (10.225) (6.745)
Figure 6: Duration (ms) of the vowel preceding the target consonants in the function of word-boundary and lexical geminates and their matched singletons Evidence of Gemination in Persian 31
The conventional view is that the qualitatively vowel duration was also affected by the consonant long vowels could not occur before lexical type differences and the same test revealed the geminates due to the phonotactic constraints in signifi cant differences of vowel differences in each Persian. However, as with the word-boundary context [F(3,184) = 103.646, p<0.000], but the geminate articulations, long vowels precede the interaction between the vowel types and consonant geminates as well as short vowels. Investigating types were not significant [F(3,184) = 2.364, p the results in Table 11 and Figure 7, and based on = 0.779]. As predicted there were no significant the results from the GLM univariate analysis of differences between the VOTs [F(1,92) = 0.735, p variance, closure durations following the low back = 0.393] and the release durations [F(1,92) = 0.876, vowel // were shorter than when they follow the p = 0.346] following the two qualitatively short low front // [F(1,184) = 48.939, p <0.000] (mean and long vowels. difference = 10.014 ms). It is expected that the
Table 11: Mean duration (ms) and Standard Deviation of 4 stop consonants in the context of two different vowel types (Cld = Closure duration, Rld = Release duration of voiced stops, VOT = VOT of voiceless stops) (N=192) t Cld t VOT k Cld k VOT b Cld b Rld d Cld d Rld 161.255 80.367 168.375 74.059 193.311 24.095 172.894 20.158 (9.102) (8.321) (8.641) (6.325) (12.302) (3.521) (9.321) (5.336) 172.229 78.083 177.418 73.452 205.083 23.173 183.961 21.153 (7.845) (4.014) (8.901) (5.366) (10.605) (2.005) (8.692) (3.102)
Figure 7: Duration (ms) of the word-boundary concatenated geminates following the qualitatively short and long vowels
Given the relative classification power of of the short and long vowel durations preceding the two qualitative vowel types, along with their the word-boundary geminates and their matched affection on the consonants following them, singletons identified almost the same duration the expected results were obtained from the differences for // occurring before singleton and investigated data (Table 12). The acoustic study word-boundary geminates [F(1,190) = 7.636, p = 32 Mahjoub ZIRAK and Peter M. SKAER
0.06], whereas the significant difference between than before their matched singletons. So comparing the duration of low back vowel // preceding the results in Figure 7 and Figure 8, as long as singleton and word-boundary geminates [F(1,190) the long vowel shortened the closure duration of = 474.450, P <0.000] results in the phonotactic the consonants following it, the word-boundary patterning in which long vowels associated to two geminates also had the same effect on long vowels timing slots in the melodic tier tend to phonetically preceding them. articulate shorter before word-boundary geminates
Table 12: Duration (ms) and Standard Deviation of the qualitatively short and long vowels preceding the word-boundary geminate consonants and their matched singletons (N=384) 164.048 122.032 singleton (5.987) (12.958) 122.729 119.198 geminate (8.694) (10.225)
Figure 8: Duration (ms) of the qualitatively short and long vowels preceding the word-boundary geminate consonants and their matched singletons
3.2.1. Non-temporal analysis of lexical/word- variance and the results displayed in Figure 9 boundary geminates show the significant difference between the RMS Since the release duration of the target stop amplitude of the four different consonant types consonants could become an important cue to [F(3,56) = 170.571, p <0.000], while a small realize a difference between geminate types, the signifi cant difference was found between whether RMS amplitude of the bursts were taken from the geminates were lexical or word-boundary types the VOTs and release durations of the target [F(1,56) = 5.0298, p = 0.028]. The small signifi cant consonants. The GLM univariate analysis of difference of RMS amplitude which occurred in Evidence of Gemination in Persian 33 the two geminate types results in the differences of the RMS amplitude in the different contexts of observed in the RMS amplitude of voiceless stops geminate types. The RMS of /dd/ was not affected within which the VOT of /kk/ showed a higher by either of the geminate types [F(1,30) = 0.621, p RMS amplitude in word-boundary geminate than = 0.436]. The geminate /bb/ had an average of 0.255 in lexical [F(1,30) = 12.787, p = 0.001]. The higher RMS in the lexical context than in the word- same results were found for /tt/ which displayed boundary type, but this small amount didn’t result the higher RMS amplitude in word-boundary in a signifi cant difference of RMS between the two geminates [F(1,30) = 79.423, p <0.0001]. The geminate types [F(1,30) = 0.942, p = 0.339]. voiced stops didn’t show any signifi cant differences
Figure 9: Mean normalized RMS amplitude for voiced and voiceless stop consonants in lexical and word-boundary geminates (N=63)
Figure 10: Mean normalized RMS amplitude for voiced and voiceless stop consonants preceding the qualitatively short // and long // vowels in word-boundary geminates 34 Mahjoub ZIRAK and Peter M. SKAER
Figure 10 indicated the results observed for the compare the differences of the word boundary stop RMS amplitude of the consonant types following geminates following the two different vowels and the two different vowel types, namely low front to compare the results with lexical stop geminates. // and low back // in word-boundary geminates. As predicted, results did show a significantly The RMS amplitude was not affected much by the higher F1 [F(1,94) = 2033.199, p <0.000] and quality of the vowel types preceding them except F2 [F(1,94) = 7474.427, p <0.000] of the vowel for /tt/ [F(1,30) = 51.901, p <0.000] which had // than the vowel // before word-boundary a significantly higher RMS amplitude before the geminates. Results also displayed the effect of short vowel than before the long one. There was word-boundary geminates on the vowels preceding also a small signifi cant difference in RMS of /dd/ them compared with a singleton context, in which [F(1,30) = 4.984, p = 0.033], and /kk/ [F(1,30) F1 of both vowels // and // was higher than = 4.129, p = 0.05], but /bb/ [F(1,30) = 2.462, p = before singletons [F(1,94) = 238.143, p<0.000], 0.127] displayed no signifi cant differences in RMS [F(1,94) = 365.360, p<0.000]. While F2 of the amplitude. As expected, word-boundary geminates vowel // preceding geminates was higher (mean preceding with the long vowel // tend to be difference = 124.970) than in singletons [F(1,94) = produced by lower RMS amplitude compared to 155.279, p<0.000], results did show the lower F2 their counterparts preceding by short vowels, but for // preceding geminates than (mean difference the difference is not suffi cient enough to serve as = -112.805) when it occurred before singletons additional indices. [F(1,94) = 619.497, p<0.000]. Considering the We have shown that the phonetic effectiveness of two different vowel types on the implementation of geminate types and the matched word-boundary stop geminates following them, singletons are not restricted to duration alone. high statistically significant differences were In Figure 11, the scatter plot presented clearly observed (Figure 12) where F1 [F(1,94) = 210.815, F1-F2 space formant plots illustrates the gestural p <0.000] and F2 [F(1,94) = 858.726, p <0.000] differences between the low back // and low front were both higher for geminates following // than // preceding 4 voiced and voiceless stops in word- when following //. The results clearly displayed boundary geminates with their matched singletons. the effect of the vowels on the word-boundary Two separate one-way ANOVA were performed geminate stops following it, where identical to compare the formant differences of each geminate stops were produced with different phonologically short and long vowel preceding formants when following different vowels in terms word-boundary stop geminates and their matched of phonetic shapes. singletons. Two separate tests were performed to Evidence of Gemination in Persian 35
Figure 11: F1-F2 plots of the short // and long // vowels before singleton and word-boundary geminate stops (where C = /b, d, k, t/)
Figure 12: F1-F2 plots of the C.C in word-boundary geminates following two vowel types, (where C.C = /b, d, k, t/)
As Figure 13 shows, F1 values were lower 2212.844, p <0.000] and F2 is higher for lexical for vowel // preceding lexical geminates geminates [F(1,94) = 2713.453, p <0.000]. As (mean difference= -50.23) compared to when expected the consonant resonance in word- it preceded word-boundary geminates [F(1,94) boundary and lexical geminates shown in Figure = 301.416, p <0.000], while F2 is higher (mean 14, display the difference of the two types of difference= 117.704) when preceding lexical geminates in which lexical geminates have clearer geminates [F(1,94) = 1139.694, p <0.000]. Figure resonance (more palatalized) than their word- 14 illustrates the same results for F1 and F2 of the boundary counterparts as they were articulated stop consonants in two geminate types where F1 with higher F2 and lower F1. Considering the is higher for word-boundary geminates [F(1,94) = effect of geminate types to vowel qualities, the 36 Mahjoub ZIRAK and Peter M. SKAER results indicate that the vowels preceding the specifi cally here, the low front vowel // in CC lexical geminates are more peripheral in quality (lexical) was produced fronter than that in C.C than those preceding the word-boundary types; (word-boundary).
Figure 13: F1-F2 plots of the short vowel // preceding word-boundary and lexical geminates (where CC = /b, d, k, t/)
Figure 14: F1-F2 plots of the CC in word-boundary and lexical geminates following // (where CC = /b, d, k, t/)
4. Discussion analyzing the data from absolute and relational durations of intervocalic consonants and the vowels The specific purpose of this study was preceding them has shown that geminate and to assess the difference between two types of singleton duration, and the contrast between them, geminates, namely, lexical and word-boundary vary according to consonant type. In this case each concatenated geminates in Persian. Firstly, consonant appears to have an inbuilt relationship Evidence of Gemination in Persian 37 between geminate and singleton duration and the the trend of 5 ms longer in the closure duration result of the geminate function show the longest of word-boundary than lexical geminates which duration for /k/ and the shortest one for /l/. Our data might be interpreted as a pause, which speakers demonstrated that in stop consonants the difference almost always inserted for word-boundaries. between singleton and geminate stops lies only This result for Persian geminates is the same as in closure duration used as a robust correlate for the findings which Oh and Redford (2012) found gemination contrast. Although there is a signifi cant in word-boundary fake geminates in English in difference between the release duration of voiced careful speech. They found that in careful speech, stops in singleton and geminates, the lack of speakers also occasionally inserted a pause at a effect of geminacy contrast on VOT in Persian morpheme boundary and between fi rst and second is similar to many other languages including syllable in words with singletons, pause duration Labanese Arabic (Ham 2001), Bengali (Hankamer in word-boundary geminates was longer than in et al. 1989) and Berber (Ridouane 2010), whereas word-internal geminates or singletons. In contrast Arvantini and Tserdanelis (2000) observed a longer to the clear findings through temporal analysis, VOT for geminates in Cypriot Greek. Regarding non-durational boundary cues may not be reliable the vowels preceding the target consonants, it enough for producing useful generalizations. should be noted that although preceding vowels are Contrary to Ridouane (2010) who found that shortened before geminates, the absolute duration lexical geminates were produced with higher RMS differences between vowels preceding singleton amplitude compared to concatenated geminates in and geminates compared to the closure duration Tashlhiyt Berber, our data investigated in Persian (14.294 ms vs. minimum 57.882 ms), didn’t suggested that, due to the lack of the possibility display a reliable difference. These fi ndings suggest to predict the RMS differences in two geminate that durational differences take place in relative types, RMS amplitude doesn’t appear to be a rather than absolute duration in that the perception good candidate for a consistent secondary cue to of gemination contrast is mainly achieved through discriminate the consonant behavior in the two the relative duration for each of the singleton and types of geminates. Unlike the RMS amplitude geminate consonants and the surrounding vowels. differences as an insufficient secondary cue to The results from the acoustic study of enhance the durational correlate, formant analysis geminate types suggest that the same sound may of consonants in the two geminate types and the be implemented with different lengths in different vowels preceding them may serve as a robust cue contexts. The findings show that word-boundary and can be found throughout our data. Our fi ndings geminates display the same temporal values as suggest that lexical stop geminates have clearer lexical geminates, all being produced with the resonance than word-boundary geminates. In same closure duration, VOT and release duration addition, vowel // preceding a lexical geminate for stops. Although the absolute duration of is relatively more front than the vowel preceding lexical and word-boundary geminates were similar the word-boundary geminates. Thus the lexical and substantially different from their singleton geminate effect with lower F1 and higher F2 values durations, once pre-consonantal vowel length was on the vowel preceding them and the result is taken into account, lexical and word-boundary relatively the same formant value for vowels. This geminates differed and only the vowel // result also may be the case for word-boundary stop preceding the lexical geminates were longer than geminates where the vowel // before adjacent before singletons. Furthermore, the results display consonants with higher F1 and lower F2 may be 38 Mahjoub ZIRAK and Peter M. SKAER implemented with the same Formant values. Thus following //, this difference is not significant the findings in formant analysis of two geminate enough to enhance the primary correlate. The types suggest that the effect of geminacy with small difference obtained may be dependent on respect to the more palatalization form is present the tenseness of the vowel // which affects the at the underlying representation of geminate types following consonant sequence to be implemented and therefore also present at the stage of gestural with lower RMS amplitude. planning. The fact that effects are stronger for Interestingly, long vowels which affect the lexical geminates providing the lower F1 than closure duration of the following consonant word-boundaries indicates that non-durational sequence can also be shortened by the effectiveness indices are more robust cues for gemination of word-boundary geminates, compared to their contrast and that the geminate types are gesturally matched singletons. In contrast, the same durations different. were interpreted for both the short vowel // Although the configuration of a two timing preceding word-boundary geminates and their slots representation for word-boundary geminates singleton counterparts. The findings that // manifest phonetically identical characteristics as behaved differently from // in the same context lexical geminates, our results strongly suggest could indicate that this different behavior is due that word-boundary geminates are presented to the phonologically active feature in Persian as consonant sequences. The finding that these namely, [±tense] as a distinctive feature for the geminates precede by vowels with the same quality contrast, which is strongly accounted for duration as pre-singletons suggest that the by categorizing the Persian vowels into two groups phonetic implementation rule whereby doubly based on phonotactic patterning such as restrictions linked segments are assigned the feature [+tense] on following consonants and versification and (Ridouane 2010), will not apply to these two raising vowel harmony observed in this language. adjacent identical consonants since the structural Considering the categorization of vowels /, i, u/ description of this rule implies that the melodic vs. /, e, o/, all vowels which are considered as tier be linked to two timing positions, thus there long in quantity-based accounts in Persian has the is no geminate tenseness to shorten the preceding feature [+tense] in the qualitative feature system of vowel. This opinion is consistent with the fi nding vowels. Vowels with the feature [+tense] represent of the pause speakers inserted in word-boundary with a melodic tier associated to two timing geminates. Clearly, both findings argue for the positions unlike [-tense] vowel which associates to fact that having identical consonant sequences is one timing position. Thus the syllable containing not suffi cient to consider it as a lexical geminate. a vowel with the feature [+tense] which is already In terms of this phonological behavior, closure bimoraic, couldn’t precede the lexical geminates as duration of the stops following the qualitatively a doubly linked segment with the feature [-tense] long vowel //, which occur only before word- (Foley 1977; Churma 1988), whereas the allowance boundary geminates, were shorter than when they of occurring before word-boundary geminates is follow the short vowel //. Unlike durational due to the failure of tenseness in the concatenated boundaries as a consistent primary correlate, the segments in a word-boundary (Keyser and Stevens non-temporal parameter of RMS amplitude was not 2006, Ridouane 2010). sufficient enough to serve as a robust secondary The findings that the duration of the vowel cue. Although word-boundary geminates following // was significantly shorter compared to its // have been produced with lower RMS than singleton counterparts and similar to the duration Evidence of Gemination in Persian 39 of the vowel // before word-boundary geminates than they occur before the vowel //. So such a (122.729 ms) are more opaque. It may reflect dark resonance (absence of palatalization) pattern in the quality of // with the feature [+tense] of word-boundary geminate implemented after which has the restrictions for occurring before the vowel // would justify the reciprocal effect of consonant sequences. Although the word-boundary the feature [tense] of the vowel and the adjacent geminates didn’t contain the feature [tense], consonant duration of word-boundary geminates. the virtual similarity of the adjacent consonant To summarize, the patterns of phonetic shape duration between word-boundary and lexical involving the durational and spectral variation geminates, results in shortening the duration of in vowels in word-boundary geminates which the [+tense] vowel // so that the syllable’s weight affect the resonance of the consonant sequences is maintained. This similarity may be because following them, are also a robust secondary cue Persian speakers are sensitive to the relative along with duration and are found in this part of duration of consonant sequences to the preceding our dataset. vowels and their perceptions from [+tense] vowels results in shortening them. Another possibility 5. Conclusion is that the observed surface differences are due to the difference in underlying representation of The purpose of this study was to examine gestures of word-boundary geminates and vowel the phonetic characteristics of lexical and word- types preceding them. It is worth noting that, boundary geminates considering the effectiveness the underlying gesture for vowels // and // of two qualitatively short and long vowels before singleton consonants is spatially different preceding them in Persian. We first investigated from when they occur before word-boundary the phonetic interpretation of geminatation geminates. However our fi ndings suggest that these contrast in Persian, with geminate and singleton differences did not display the similar phonetic consonant duration, the effect of consonant types shape for vowels before word-boundary and and the interaction of the preceding vowel duration lexical geminates. In this interpretation, the vowel and the local ratio of C/V1. The non-temporal // preceding the word-boundary geminates with parameters, namely RMS amplitude and the high F1 and F2 is implemented more front and formant frequencies were also examined to explore higher than the vowel // which refl ects the gesture their effect on distinguishing the different geminate of closer and less front before word-boundary types. The general conclusion of this study is that geminates. In other words, considering the effect of the primary correlate distinguishing singletons the vowel on the following consonant sequences, from geminates is duration. But gemination the word-boundary geminates with the vowel // contrast strongly relies on relative rather than the preceding them are produced more palatalized like absolute duration of consonants and the vowels lexical geminates, but with higher F1 value than preceding them due to the small absolute durational the lexical ones. This implementation is in contrast differences. The effect of geminacy on consonant with the fact of more front vowels before lexical type also was demonstrated by the longest duration geminates which shows the different gesture of for /k/ and the shortest one for /l/. Although in stop word-boundary geminates compared to lexical consonants the geminatation contrast lies only in ones. However, consonant sequences following the closure duration as a robust correlate, VOT and long vowel // which have been affected by the release duration weren’t affected by gemination. vowel, are relatively produced with less clearance In terms of two geminate types, the fi ndings show 40 Mahjoub ZIRAK and Peter M. SKAER that word-boundary geminates display the same of intervocalic singleton and geminate sonorants in temporal values as lexical geminates, all being Jordanian Arabic. Alarabiyya 29. 37-52. produced with the same closure duration. However Arvantini, A. and G. Tserdanelis, 2000. On the phonetics of these two geminate types differed only in the case geminates: Evidence from Cypriot Greek. Proceedings that while lexical geminates shortened the vowel of the International Conference on Spoken language // preceding them word-boundary geminates did Processing, Beiging. 23-27. not. Furthermore, the results show the trend of 5 Boersma, P. and D. Weenik. 2007. Praat: Doing phonetics by ms longer in the closure duration of word-boundary computer (version 4.6.12) [computer program]. http:// than lexical geminates which might be interpreted www.praat/org/ as a pause where speakers almost always inserted Churma, D. 1988. On “on geminates”. Ms., Suny-Buffalo. when speaking carefully. This result with the Cohn, A., W. Ham, and R. Podesva 1999. The phonetic finding of the pause speakers inserted in word- realization of singleton-geminate contrasts in three boundary geminates consistently enough for the languages of Indonesia. Preceedings of the 14th fact that the phonetic implementation rule whereby International Congress of Phonetic Sciences. 587-590. doubly linked segments are assigned the feature Delattre, P. 1971. Consonant gemination in four languages: [+tense], will not apply for these two adjacent An acoustic, perceptual and radiographic study. IRAL 9. identical consonants. The reciprocal effectiveness Part I. 31-52; Part II. 97-113. of the qualitatively long vowel // and the Deyhim, G. 2000. Farhang-i Avayi-I Farsi [Persian consonant sequences following it in word-boundary pronunciation dictionary]. Tehran: Farhang Moaser geminates can be represented by the phonologically Publisher. active feature [+tense] of the vowel //. This Esposito, A. and M. Benedetto. 1999. Acoustical and means that Persian speakers can take advantage perceptual study of gemination in Italian stops. Journal of relative duration of consonant sequences of of the Accoustical Society of America 106. 2051-2062. the preceding vowels in their perceptions from Foley, J. 1977. Foundations of Phonological Theory. [+tense] vowels and shorten them. The study also Cambridge University Press. presented evidence that RMS amplitude, contrary Ham, W. 2001. Phonetic and phonological aspects of to durational differences as a robust primary geminate timing. New York: Routledge. correlate, didn’t appear to be a good candidate for Hankamer, J., A. Lahiri, and J. Koreman. 1989. Perception a consistent secondary cue to discriminate between of consonant length: Voiceless stops in Turkish and the two vowel types and consonant behavior in the Bengali. Journal of Phonetics 17. 283-298. two types of geminates. In contrast, the distinction Hansen, B. 2004. Production of Persian Geminate Stops: between the underlying gesture of consonants Effect of varying speaking rate. Proceedings of the 2003 in the geminate types and two vowel types in Texas Linguistics Society Conference, Agwuele, Warren word-boundary geminates displayed that formant and S.H. Park (eds.). 86-95. Somerville: Cascadilla frequencies could serve as robust secondary cue Press. to distinguish the geminate types and interpret the Hayes, B. 1986. Inalterability in CV Phonology. Language manifestation of tense articulation of vowel types 62. 321-351. in Persian. Idemaru, K. and S. Guion. 2008. Acoustic covariants of length contrast in Japanese stops. JIP A 38. 167-186. 6. References Idemaru, K. and S. Guion. 2010. Relational timing in the production and perception of Japanese singleton and Al-Tamimi, F. 2004. An experimental phonetic study geminate stops. Phonetica 67. 25-46. Evidence of Gemination in Persian 41
Keyser, S. and K. Stevens, 2006. Enhancement and overlap Representation of Fake Geminates in English. Journal in the speech chain Language 82(1). 33-63. of Phonetics 40(1). 82-91. Ladd, R. and J. Scobbie. 2003. External Sandhi as gestural Payne, E. 2005. Phonetic Variation in Italian Consonant overlap counter evidence from Sardinian. In Local J. Gemination. Journal of the International Phonetic R. Ogden, and R. Temple, (eds.), Papers in Laboratory Association 35(2). 153-181. Phonology VI: Phonetic Interpretation.164-182. Pickett, R., S. Blumstein, and M. Burton. 1999. Effect of Cambridge University Press. speaking rate on the singleton/geminate consonant Ladefoged, P. and I. Maddieson. 1996. The Sounds of the contrast in Italian. Phonetica 56. 135-157. World’s Languages. Oxford: Blackwell publishers. Pind, J. 1986. The Perception of quantity in Icelandic. Lahiri, A. and J. Hankamer. 1988. The Timing of Geminate Phonetica 43. 116-139. Consonants. Journal of Phonetics 16. 327-338. Podesva, R. 2000. Constraints on geminates in Burmese and Leben, W. 1980. A Metrical Analysis of Length. Linguistic Selayarese. In Proceedings of West Coast Conference Inquiry 11. 497-510. on Formal Linguistics 19, Bilerey-Mosier, and B Local, J. and A. Simpson. 1999. Phonetic implementation of Lillehaugen (eds.) 343-356. Somerville: Cascadilla geminates in Malayalam nouns. Proceedings of the 14th Press. International Congress of Phonetic Sciences1. 595-598. Ridouane, R. 2010. Gemination at the Junction of Phonetics Maddieson, I. 1985. Phonetic Cues to Syllabification. In and Phonology. Laboratory Phonology X.61-90. Fromkin, V. (ed.). Phonetic Linguistics: Essays in Rohany Rahbar, E. 2009. On Contrast in the Persian Vowel Honor of Peter Ladefoged. NewYork: Academic Press. System. Toronto Working Papers in Linguistics 31. 203-221. Samareh, Y. 1977. The arrangement of fegmental phonemes Mahootian, S. 1997. Persian (Descriptive Grammars). in Farsi. Tehran: University of Tehran Press. London: Routledge. Samareh, Y. 1992. Avashenasi-e Zaban-e Farsi [Persian McCarthy, J. 1986. OCP effects: Gemination and Language Phonetics]. Tehran: University Publication Antigemination. Linguistic Inquiry 17. 207-263. Center. Miller, A. 1987. Phonetic Characteristics of Levantine Thurgood, G. 1993. Geminates: A Cross-Linguistic Arabic Geminates with Differing Morpheme and Examination. In Nevis, J., G. McMnamin and Syllable Structures. Ohio State Papers from the G. Thurgood (eds.), Papers in honor of Fredrick Linguistics Laboratory 36. 120-140. Brengleman on the occasion of the twenty fifth Mitchell, T. 1990. Pronouncing Arabic 1. Oxford: Oxford anniversary of the department of Linguistics. 129-139. University Press. Frenso: California State University. Norlin, K. 1987. A phonetic study of emphasis and vowels Toosarvandi, M. 2004. Vowel Length in Modern Farsi. in Egyptian Arabic. Lund University Department of Journal of Royal Asiatic Society 14. 241-251. Linguistics Working Papers 30.1-119. UPVD (University of Victoria Phonetic Database) 1999. Oh. G. and M. Redford. 2012. The Production and Phonetic Version 4, Kay Elemetrics corp. N.J.