Indranil Dutta

Consonant-f0 interaction in Malayalam stops: Automatic or controlled?

Voicing is known to perturb f0 in the following vowel (House and Fairbanks, 1953). This lowering has been attributed to biomechanical constraints (Svantesson and House, 2006). On the basis of evidence from languages where the [voice] feature does not manifest itself phonetically, yet is accompanied by f0 lowering, it has been argued that f0 lowering serves to enhance the phonological [voice] contrast. We present results from an acoustic study of f0 perturbation in Malayalam, which exhibits predictable voicing in lexical items from its Dravidian stratum. Underlying voiceless stops only appear initially, and are voiced intervocalically. Malayalam also contains a large Sanskrit stratum lexicon, where phonetic voicing is tied to its stratal affiliation. We obtained a significant effect of laryngeal setting on z-score normalized f0 at vowel onset. These results suggest voicing is significant in determining the onset f0 of the vowel, regardless of the phonological status of voicing in a language We conclude tha...

Evidence for active cavity expansion through advanced lingual place of constriction in voiced geminates

Geminate voicing requires both active and passive cavity expansion in order to delay the equalization of subglottal and supraglottal pressure (Ohala 1983, Ham 2001). Studies report differences in stop closure duration (Lahiri & Hankamer 1988) and pre-geminate vowel duration as cues to the laryngeal contrast. Following Schroeder (1967), Mermelstein (1967), and Iskarous (2010), we present results from formant transitions in the VC:V context and the anti-symmetric Fourier component coefficient, an = -2 (Fn-fn)/fn, to partially recover the area function of the vocal tract behind the geminate constriction. Voiced geminates exhibit significantly higher and positive coefficient values than their voiceless counterparts which implies that measured Fn values for voiced geminates are lower than their neutral tract formant values. This finding indicates that voiced geminates have a more anterior constriction compared to their voiceless counterparts. Without ruling out the role of passive cavity expansion through the cheek muscles, we conclude that in order to delay the pressure equalization the lingual constriction in voiced geminates is maintained at an anterior location compared to voiceless geminates in Bangla. Results from the slopes of first order F2 locus-equations are also presented to measure the differences in coarticulatory resistance between the voiced and voiceless geminates.Geminate voicing requires both active and passive cavity expansion in order to delay the equalization of subglottal and supraglottal pressure (Ohala 1983, Ham 2001). Studies report differences in stop closure duration (Lahiri & Hankamer 1988) and pre-geminate vowel duration as cues to the laryngeal contrast. Following Schroeder (1967), Mermelstein (1967), and Iskarous (2010), we present results from formant transitions in the VC:V context and the anti-symmetric Fourier component coefficient, an = -2 (Fn-fn)/fn, to partially recover the area function of the vocal tract behind the geminate constriction. Voiced geminates exhibit significantly higher and positive coefficient values than their voiceless counterparts which implies that measured Fn values for voiced geminates are lower than their neutral tract formant values. This finding indicates that voiced geminates have a more anterior constriction compared to their voiceless counterparts. Without ruling out the role of passive cavity expansion through the ...

Acoustic correlates of the four-way laryngeal contrast in Marathi stops

As many of its geographical and typological neighbors, Marathi demonstrates a four-way laryngeal contrast in stop consonants, which includes plain voiced and voiceless stops as well as aspirated voiceless stops and typologically rare aspirated (breathy) voiced stops. The present study investigates the acoustic correlates of the laryngeal contrast in word-initial velar stops in Marathi. Thirty three native speakers of Marathi were recorded in Mumbai, Maharashtra, India. Positive and negative Voice Onset Time (VOT), fundamental frequency at the onset of voicing (onset f0), and spectral intensity as a measure of breathiness during the following vowel were investigated as correlates of laryngeal distinctions. Preliminary results show that all four stop categories are well-distinguished via VOT and onset f0 measures. Initial spectral intensity measurements suggest that breathiness in Marathi stops is largely contained in the release and does not extend far into the following vowel, unlike what has been found in Hindi (Dutta, 2007). The results will be discussed with respect to the typology of voicing contrast across languages, specifically the relevance of VOT distinctions and the use of onset f0, and compared to relevant findings for typologically similar languages.As many of its geographical and typological neighbors, Marathi demonstrates a four-way laryngeal contrast in stop consonants, which includes plain voiced and voiceless stops as well as aspirated voiceless stops and typologically rare aspirated (breathy) voiced stops. The present study investigates the acoustic correlates of the laryngeal contrast in word-initial velar stops in Marathi. Thirty three native speakers of Marathi were recorded in Mumbai, Maharashtra, India. Positive and negative Voice Onset Time (VOT), fundamental frequency at the onset of voicing (onset f0), and spectral intensity as a measure of breathiness during the following vowel were investigated as correlates of laryngeal distinctions. Preliminary results show that all four stop categories are well-distinguished via VOT and onset f0 measures. Initial spectral intensity measurements suggest that breathiness in Marathi stops is largely contained in the release and does not extend far into the following vowel, unlike what has been found i...

Onset f0 as a correlate of voicing in Marathi

Onset f0 (the fundamental frequency at the onset of vowel immediately following the consonant) has been shown to correlate with voicing in a variety of languages with two-way voicing contrasts in stop consonants, such as the voice-type contrast (between prevoiced and voiceless unaspirated stops) or the aspiration contrast (between voiceless unaspirated and voiceless aspirated stops). Onset f0 is typically higher after phonologically voiceless than after voiced stops. With some notable exceptions, onset f0 data from languages with three-way and four-way voicing contrasts is relatively scarce. The present study explores onset f0 as a correlate of voicing in Marathi, an Indo-Aryan language with a four-way voicing contrast which includes both aspirated and unaspirated voiced and voiceless stops. Results demonstrate that onset f0 covaries with both laryngeal voicing and aspiration in Marathi: f0 is significantly lower for both sets of voiced than voiceless stops. Onset f0 is also significantly lower for both s...

Tonal alignment in Deori

This paper reports on the results from an experiment on tone in Deori, a language spoken by about 20,000 people in Assam (India). Data from 10 speakers where the target word bearing the tonal contrast appeared in the sentence medial position is presented. Time-normalized pitch of different words shows that words may have a lexically specified high or low tone. A high tone may contrast with a low tone, but its phonetic implementation of rise or fall in a disyllabic word depends on whether the syllable on which the contrast appears is initial or final. A tonal contrast on the first syllable leads to a falling contour, but when the contrastive tone appears on the second syllable of a disyllabic word then the tonal contour is falling. Exceptions to this pattern appear in closed disyllables where a steep rise in either the low or high tone is not observed. A high or a low tone may also contrast with a word which is not specified with any tone, in which case there is no rise or fall. Statistical analyses show t...

Coarticulation and contrast in static and dynamic models of second formant trajectories

Real (Stevens et al., 1966) and virtual F2 locus (Sussman et al., 1991) measures are presented for Malayalam lingual plosives. We show that in distinguishing voiceless coronals (dental, alveolar, and retroflex) in VCːV sequences, F2 onsets derived from first-order locus equations (LEs) show only partial delineation of the contrast. The dental-alveolar contrast is effectively maintained, but retroflex and alveolar stops show no significant difference in F2 onset. Following Lindblom and Sussman’s (2012) examination of LEs as a measure of relative coarticulatory resistance, we report F2 slopes for the three coronal stops in VC and CV transitions to assess the implications of this metric in Malayalam. Our findings on the ordering of slope values from steepest to flattest did not follow predictions based on expectations of relative articulatory complexity; namely, alveolars generated a flatter slope than retroflexes, despite Dart and Nihalani’s (1999) demonstration that the retroflex gesture is more complex wi...

Consonant-f0 interaction under predictable voicing: f0 lowering due to phonetic factors

We report on consonant-f0 interactions in Malayalam. Crosslinguistically, voicing lowers f0 in the following vowel (House and Fairbanks 1953, Hombert 1978, Clements 2002, Moreton 2006). While this lowering has been attributed to physiological and phonetic factors (Stevens 1998, Atkinson 1978, and Honda 2004), Ohde (1984), Svantesson and House (2006), and Keating (1984) have argued that f0 lowering serves to maintain a phonological contrast between voiced and voiceless segments. Voicing in Malayalam is predictable; voiceless stops appear initially, and voiced intervocalically. We report on data from 6 native speakers. 3 repetitions of each word in a frame sentence were recorded. Since Malayalam words only appear with voiceless initial stops, we contrast these with nasals from corresponding places of articulation. We examine time-normalized f0 perturbation due to voicing in Malayalam for all places of articulation for both initial and medial contexts. Our findings lend support to the physiological and phonetic account of f0 lowering. Malayalam exhibits a pattern of f0 lowering following voicing and raising during the production of voiceless segments. In spite of Malayalam voicing being predictable, the f0 perturbation in the following vowel follows the cross-linguistic pattern of lowering following voiced segments. This finding dove-tails with the results from earlier studies that take physiological factors to be responsible for the lowering of f0 following voiced segments.

Enhancement of laryngeal features under segmental and prosodic conditioning

Evidence in support of enhancement features (Keyser and Stevens, 2006) is presented from an acoustic study of laryngeal contrasts in Hindi. Segmental contrasts where defining features and their corresponding acoustic outcomes are attenuated are said to be made acoustically salient by enhancing gestural features (Stevens and Keyser, 2010). In this study, the four-way laryngeal contrasts in Hindi are examined under varying prosodic and segmental contexts. In segmental contexts where a defining distinctive feature (slack vocal folds) is attenuated in its acoustic manifestation, namely, closure duration (CD), an enhancing gesture (spread glottis) (Avery and Idsardi, 2001) is added to increase the saliency of contrasts between voiced aspirated stops (VAS) and voiced stops (VS). The acoustic consequence of this enhancement is f0 lowering in the following vowel. Similar results are obtained under weak prosodic conditions where both f0 lowering and increase in spectral tilt result from the enhancing gesture (spre...

Louder is longer: Amplitude conditioned lengthening in diasporic Siraiki

Based on a longitudinal study of three generations of diasporic Siraiki speakers in India, it is shown that first generation (Gen1) implosives develop into lengthened plosives, intervocalically, in the subsequent two generations (Gen2/Gen3). Historically, Prakrit geminates developed into Siraiki and Sindhi implosives. Ohala (1992a) observes that maintaining voicing during intervocalic geminates is facilitated by increased oral cavity volume, achieved by lowered larynx (LL). The consequence of this cavity enlargement is listener misperception of the accompanying cues that lead to the genesis of implosives in the intervocalic position in Sindhi (Ohala, 1993). Closure voicing durations, root mean square (rms) amplitude during closure, and f0 following release are measured in this study. It is observed that in contexts where the Gen1 speakers produce increased rms amplitude at the release of implosives, the subsequent Gen2/Gen3 produce lengthened plosives. Increased peak-amplitude has been shown to be a weak ...