Plosives, affricates, and ejective plosives and affricates

Plosives are distinguished at four places of articulation: bilabial, dental, velar, and glottal. Affricates are postalveolar; the aspirated postalveolar affricate ʧʰ is very rare, only occurring in a few words (Collord Reference Collord1968: 2). According to Newman (Reference Newman1944: 14), the aspirated alveolar plosive in Yokuts (represented as /ṭ/ in Newman) is the source of both the alveolar fricative /s/ and the aspirated postalveolar /ʧʰ/ in Chukchansi. The dental and velar plosives are often accompanied by slight affrication, especially the aspirated plosives and ejectives: [tΘ tΘș kx kxș]. This affrication is never as long as that of the postalveolar affricates [ʧ ʧʰ ʧș]. Perceptually, the frication noise in dental plosives is always dental [Θ], while the frication noise in velar plosives is sometimes retracted, sounding uvular [qχ qχș] (see Figures 8 and 3, 5, respectively). In fact, velar plosives in general have a retracted quality, as noticed by an anonymous reviewer. The third author, a native speaker of Chukchansi, has the impression that before front and central vowels, the velar plosives in Chukchansi are slightly retracted compared to English velar plosives (e.g. comparing Chukchansi /kʰaʔjuʔ/ [k̠ʰaʔjuʔ] ‘coyote’ (nom) with English /kʰatn̩/ ‘cotton’). However, before back vowels, the author feels no difference between velar plosives in Chukchansi and in English. In the absence of articulatory data, which is not available, we cannot be certain about the precise place of articulation of the dental and velar plosives or the postalveolar affricates.

Figure 3

A spectrogram and waveform of /kʰaʔjuʔ/ ‘coyote’ (nom), illustrating the aspirated velar plosive with a 98 ms VOT.

Figure 4

A spectrogram and waveform of /kajis/ ‘good’ (nom), illustrating the unaspirated velar plosive with a 19 ms VOT.

Plosives and affricates in Chukchansi (and other Yokuts languages) do not have contrastive voicing (e.g. Newman Reference Newman1944, Kroeber Reference Kroeber1963, Collord Reference Collord1968). The unaspirated bilabial plosive may, however, appear voiced word-initially and intervocalically, i.e. in onset position. For example, see /܀ajin̰/ [怂ajin̰] ‘acorn’ (nom) in the wordlist (1) above and /܀aː܀as/ [怂aː怂as] ‘potato’ (nom) in Figure 7. Plosives and affricates, except for the glottal plosive, instead show a three-way contrast between voiceless aspirated (Figure 3), voiceless unaspirated (Figure 4), and ejective (Figure 5). Figures 3–5 show this three-way contrast at the velar place of articulation: /k kʰ kș/. For the case of the aspirated velar plosive /kʰ/ in /kʰaʔjuʔ/ ‘coyote’ (nom) in Figure 3, the duration of the aspiration is 98 ms. In the case of the unaspirated velar plosive /k/ in /kajis/ ‘good’ (nom) in Figure 4, a brief release burst of 19 ms is present, followed immediately by the periodic waveform of the vowel.

In these two cases, the contrasting unaspirated and aspirated plosives are in onset position. The aspiration contrast is neutralized when these plosives occur in the coda, where only ejective plosives are typically released. We do not find any consistent differences in the release of coda plosives based on the manner of the following onset consonant.

In word-final coda position, Chukchansi exclusively has aspirated, not unaspirated plosives at the phonetic level: [pʰ tʰ kʰ]. Chukchansi words that end morphophonemically in unaspirated plosives /p t k/ include bare roots in the unmarked nominative case and particles without suffixes (though word-final [k] does appear as an allomorph of the imperative suffix /-k(a)/). In six such words illustrated in (2)–(4), the final plosives are realized with aspiration: [pʰ tʰ kʰ].

Comparing the words above ending in the unaspirated phonemes /p t k/, as in (2)–(4), with words ending in the aspirated phonemes /pʰ tʰ kʰ/ in (5) and (6), the contrast is neutralized. Both unaspirated and aspirated plosives are released word-finally with aspiration (contrasting with ejectives, which have a glottal release following the oral release).

The aspirated release of the word-final consonants in (2)–(4) may be related to their position at the end of the utterance. Note that word-final vowels in many of the recordings are also accompanied by aspiration. There may be a final glottal abduction gesture resulting in aspiration of our word-final tokens, as suggested by a reviewer, though we have not determined whether its domain is the word, phrase or utterance. The word-final consonants in (2)–(4) are written as unaspirated in the orthographic column due to related morphological forms in which these consonants occur in onset position and are clearly unaspirated, like [k] in [ʔåkitʰ] (7). The contrast between the two forms of /ʔaːk/ with [kʰ] in (4a) and [k] in (7) shows the alternation between aspirated and unaspirated allophones of the phoneme /k/ (the [i] in the surface form is due to phonotactics, as discussed below in the section ‘Syllable structure and vowel alternation’).

The facts in (2)–(4) and (5)–(6) suggest that aspiration is not contrastive in coda position. No aspiration contrast for postalveolar affricates /ʧ ʧʰ/ is observed in coda position.

Ejectives form another series of plosives and affricates in Chukchansi. In previous literature of Yokuts languages, this type of plosive is referred to as ‘glottalized’ (Newman Reference Newman1944, Kroeber Reference Kroeber1963, Collord Reference Collord1968, Gamble Reference Gamble1978). Figure 5 illustrates a case where a velar ejective /kș/ occurs in word-initial position, in /kșajaʃ/ ‘wild carrot’ (nom).

Figure 5

A spectrogram and waveform of /kșajaʃ/ ‘wild carrot’ (nom), illustrating the velar ejective with a VOT of 148 ms.

The supralaryngeal closure of the ejective is followed by the release burst of the compressed air from the oral cavity. The release burst is often followed by further aperiodic noise that indicates affrication (see /tșojoʃ/ [t^Θșojoʃ] ‘arrow’ (nom) in Figure 8 as well). As an anonymous reviewer points out, this noise cannot be aspiration, which requires an open glottis, though the glottis is still closed at this phase of the ejective. The affrication noise is followed by a period of silence representing the length of time it takes from the release of the obstruction in the oral cavity to the release of the glottalic obstruction. The initial portion of the vowel following the release of the glottalic obstruction is often laryngealized, as shown in Figure 5 (see also /pșonoʃ/ ‘hand’ (nom) in Figure 18, though compare /tșojoʃ/ ‘arrow’ (nom) in Figure 8 without vowel laryngealization). For the case in Figure 5, the entire VOT from release of oral closure to beginning of the (laryngealized) vowel is 148 ms.

The intense burst, long VOT and long total duration of Chukchansi ejectives such as /kș/ in Figure 5 are characteristics of ‘fortis’ ejectives (as opposed to ‘lenis’), in the sense of Kingston (Reference Kingston1985). See also the division between ‘strong’ vs. ‘weak’ in Bird (Reference Bird2002), ‘stiff’ vs. ‘slack’ in Kingston (Reference Kingston, Sharon and Keren2005), and ‘complex’ vs. ‘simplex’ in McDonough & Wood (Reference McDonough and Valerie2008), Chukchansi ejectives being ‘strong’, ‘stiff’, and ‘complex’, respectively. Ejectives in Chukchansi are thus similar to the strong ejectives found in many Athabaskan and Salishan languages, including Navajo and Montana Salish (e.g. Lindau Reference Lindau1984, McDonough & Ladefoged Reference McDonough and Peter1993, Flemming, Ladefoged & Thomason Reference Flemming, Ladefoged and Thomason1994, McDonough & Wood Reference McDonough and Valerie2008, among others), as opposed to those in Hausa, Quiché, and many languages of the Caucasus (Lindau Reference Lindau1984, Kingston Reference Kingston1985, Vicenik Reference Vicenik2010). However, unlike typical ‘fortis’ or ‘stiff’ ejectives in Kingston’s (Reference Kingston1985, Reference Kingston, Sharon and Keren2005) typology, vowels following ejectives in Chukchansi often begin with laryngealization and low amplitude, like ‘lenis’ or ‘slack’ ejectives. This finding is in line with the proposals in Wright, Hargus & Davis (Reference Wright, Hargus and Davis2002) and Vicenik (Reference Vicenik2010) that a simple, two-way classification of ejectives does not account for their different phonetic characteristics.

Figure 6 shows the mean values for Voice Onset Time (VOT) in ms for each of the plosive and affricate phonemes. VOT was calculated from the release of full oral closure to the onset of the voiced vowel. VOT also includes the fricative release portion of postalveolar affricates as well as any fricative release portions of dental and velar plosives. These values are taken from 226 plosive or affricate tokens in either word-initial or intervocalic position and in either stressed or unstressed position from 63 words recorded for this study in addition to the wordlist at the beginning of this section. These words were read three times each. Due to the small number of tokens recorded, the different environments of stress and position were combined; this does not seem to have influenced the aggregated values in Figure 6.

Figure 6

Mean VOT (in ms) for plosives and affricates. Error bars show standard deviation grouped by laryngeal state (unaspirated, aspirated, ejective). Numbers show number of tokens.

The VOT mean values in Figure 6 show the typical crosslinguistic pattern for the plosives where bilabials have the shortest VOT and velars the longest (Maddieson Reference Maddieson, John and Hardcastle1997). The longer measurement for postalveolar affricates is also typical, as it takes into account the fricative release component of the sound. The fricative portion of postalveolar affricates is always longer than the fricative portion in dental and velar plosives and ejectives that have affrication.

The mean VOT measurement (−29 ms) for the unaspirated bilabial plosive combines both voiced tokens [b] and unvoiced tokens [p]. This is responsible for the high standard deviation (41 ms) of the unaspirated bilabial plosive in Table 1. Splitting up the 34 total tokens of the unaspirated bilabial plosive /p/, 22 tokens are voiced [b] (mean VOT: −55 ms, SD: 25 ms), and 12 tokens are unvoiced [p] (mean VOT: 20 ms, SD: 7 ms). Eighteen of the 22 of voiced [b] tokens are in word-initial position, while the remaining four voiced [b] tokens are intervocalic. Eight of the 12 unvoiced [p] tokens are in word-initial position, while four of the unvoiced [p] tokens are intervocalic. These VOT values of the voiced [b] and voiceless [p] tokens combine to give the negative mean VOT of the 34 total tokens of /p/ (−29 ms; SD: 41 ms).

Figure 7

A spectrogram and waveform of /paːpas/ [baːbas] ‘potato’ (nom), illustrating pre-voiced word-initial and fully voiced intervocalic tokens [b] of the unaspirated bilabial plosive /p/; the word-initial plosive has a negative VOT of 54 ms.

In Figure 7, both the word-initial and intervocalic tokens of the unaspirated bilabial plosive /p/ in /paːpas/ ‘potato’ (nom) are voiced [b], as illustrated by their periodic waveforms and the voicing bars. The word-initial token of /p/, realized as [b], in Figure 7 has a negative VOT of −54 ms. Compare this case to the one in Figure 4, with the unaspirated velar plosive /k/ in /kajis/ ‘good’ (nom), which is realized as unvoiced [k] and has a positive VOT of 19 ms.

One finding particular to Chukchansi is shown in Figure 6: velar and especially dental ejectives have a much higher mean VOT than their aspirated counterparts (about 2.25:1 and 1.75:1, respectively), but bilabial ejectives only have somewhat higher mean VOT (1.5:1) and postalveolar ejectives slightly shorter (0.85:1) than their aspirated counterparts. The greater relative length of dental and velar ejectives may be due to the appearance of long, aperiodic noise with steady amplitude in many tokens. This noise indicates affrication, as confirmed by perception and noted by multiple reviewers.

This release pattern contrasts with that of ‘strong’ or ‘stiff’ ejectives in many other languages, where the rapid release of oral closure while the glottis remains closed yields a burst with high initial amplitude that quickly lessens (Lindau Reference Lindau1984, Kingston Reference Kingston1985). While affrication also occurs in non-ejective dental and velar plosives (as noted by multiple reviewers), the fricated portion in the non-ejective plosives comprises most of the VOT, following the burst of the plosive release and preceding the adjacent vowel. In ejective plosives, on the other hand, the fricated portion only makes up some of the VOT, which includes the silent period before the onset of the vowel as well. Consequently, the VOT of dental and velar ejectives is longer than the VOT of aspirated dental and velar plosives. In addition, the VOT of ejectives is similar for dental and velar ejectives, while aspirated plosives have the typically crosslinguistic pattern where velars have longer VOT than dentals (Maddieson Reference Maddieson, John and Hardcastle1997), as pointed out by an anonymous reviewer. Lastly, the Chukchansi data follow the cross-linguistic tendency for ejective and aspirated plosives to be differentiated by VOT (among other properties), as shown by Cho & Ladefoged (Reference Cho and Ladefoged1999).

Figure 8 shows the affricated release burst of the dental ejective token /tș/ in /tșojoʃ/ [t^Θșojoʃ] ‘arrowș (nom). As shown, the amplitude of the oral release of the ejective begins low and slowly increases before lessening into the silent period, which is then followed by the vowel /o/. In Figure 8, the affricated release burst phase lasts 58 ms, while the silent phase lasts 66 ms, adding up to the whole VOT of 124 ms. Compare Figure 5, with the velar ejective /kș/ in /kșajaʃ/ ‘wild carrot’ (nom), which has a different release burst pattern (high initial amplitude, rapidly reduced). We are unsure whether these different release burst patterns are systematic or not, or whether they are an epiphenomenon of the place of articulation of the fricative release, as suggested by an anonymous reviewer.

Figure 8

A spectrogram and waveform of /tșojoʃ/ [t^Θșojoʃ] ‘arrow’ (nom), illustrating the dental ejective [t^Θș] with affricated release phase of 58 ms and a silent phase of 66 ms.

While velar and dental ejectives often have the release burst pattern shown in Figures 5 and 8, respectively, bilabial ejectives never do. Similar to bilabial ejectives, bilabial plosives are never affricated. In addition, some tokens of the bilabial ejective [pș] are closer to ‘lenis’ or ‘slack’ ejectives in the sense of Kingston (Reference Kingston1985, Reference Kingston, Sharon and Keren2005), as opposed to dental and velar ejectives [tș kș], which are always clearly ‘fortis’ or ‘stiff’. While bilabial ejectives always have a release burst and a silent period, as in ‘fortis’ ejectives, these phases are shorter in duration than those in dental and velar ejectives. This difference in duration may be related to the affricated release burst in the latter but not the former ejectives, and it may also contribute to the longer VOT measurements for dental and velar ejectives vis-à-vis their aspirated counterparts vs. in bilabials.

Figure 9 shows the non-affricated release burst of the bilabial ejective token /pș/ in /pșaːja/ ‘child’ (acc). In Figure 9, the non-affricated release burst phase lasts 22 ms, while the silent phase lasts 46 ms, adding up to the whole VOT of 68 ms.

Figure 9

A spectrogram and waveform of /pșaːja/ ‘child’ (acc), illustrating the bilabial ejective [pș] with a non-affricated release phase of 22 ms and a silent phase of 46 ms.

The three ejective tokens in Figures 5, 8, and 9 illustrate the variable pronunciation of ejectives in Chukchansi. Similar to ejectives in Witsuwit’en (Wright et al. Reference Wright, Hargus and Davis2002, Hargus Reference Hargus2007) and Hul’q’umi’num’ (Percival Reference Percival2019), different ejective tokens in Chukchansi may contrast in the intensity of the oral release burst, though this appears to correlate with place of articulation. Ejective tokens also contrast in whether the following vowel starts with laryngealization: Figure 5 shows heavy laryngealization and Figure 8 shows none, with Figure 9 intermediate. The difference in vowel laryngealization does not appear to be correlated with the difference in burst intensity or duration of VOT, again supporting Wright et al.’s (Reference Wright, Hargus and Davis2002) and Vicenik’s (Reference Vicenik2010) arguments against a simple ‘fortis/stiff’ vs. ‘lenis/slack’ contrast in ejectives.

Fricatives

Fricatives in Chukchansi are distinguished at four places of articulation: alveolar, post-alveolar, velar and glottal. All fricatives in Chukchansi are voiceless in all positions. The alveolar and post-alveolar fricatives are both sibilants. The alveolar sibilant often has an impressionistic retroflex quality. Compare (8) [χoːwiʂ], which is audibly post-alveolar or retroflex, with (9) [soχ], which is audibly alveolar. The velar fricative often sounds retracted, as noted by reviewers, especially next to the back vowels /u ů o o̊/. This retraction is audible in (8) [χoːwiʂ] and (9) [soχ] (see also Collord Reference Collord1968: 2–3).

Due to a lack of articulatory data, which is unavailable, we cannot be certain about the precise place of articulation of the fricatives. Moreover, we have not found any discernable pattern for the variation in tokens of /s/; the presence of alveolar [s] vs. post-alveolar [ʃ] or retroflex [ʂ] tokens is not consistently correlated with syllable position or adjacent vowels. For some words with /s/, we hear the alveolar [s] token in some utterances and the post-alveolar [ʃ] or retroflex [ʂ] token in other utterances of the same words.

Sonorants

Nasals and approximants contrast modally voiced and laryngealized segments in Chukchansi. The laryngealized or glottalized sonorants are post-glottalized, not pre-glottalized (see Ladefoged & Maddieson Reference Ladefoged and Ian1996, Howe & Pulleyblank Reference Howe and Douglas2001, Bird et al. Reference Bird, Caldecott, Campbell, Gick and Shaw2008 for pre- vs. post-glottalization). This contrast is phonotactically limited: while modally voiced sonorants can surface in any position, laryngealized sonorants only surface after a vowel. Moreover, sonorants that are underlyingly laryngealized are typically articulated with modal voice when intervocalic. Most often, laryngealization only shows up acoustically in coda position, at the end of the sonorant segment. In onset position, then, sonorants almost always have modal voice, as explained below.

The phonotactic restriction on laryngealized sonorants drives alternations between modally voiced and laryngealized segments on the surface, as with the root /tal̰w̰/ ‘trip someone’ with the two underlyingly laryngealized sonorants /l̰/ and /w̰/ (10). In (10a), [l̰] is in the coda and retains laryngealization, while [w] is in the onset and is modally voiced. In (10b), this is reversed: laryngealized coda [w̰] and modally voiced onset [l].

Laryngealized sonorants may surface in intervocalic position in order to keep a lexical contrast. For example, the roots /saw/ ‘scream’ (11a) and /saw̰/ ‘water’ in (11b) only differ in laryngealization of the labial-velar approximant.

When the approximant is intervocalic, the laryngealized /w̰/ in (11b) often surfaces with laryngealization in the middle of the segment in order to preserve the contrast with plain /w/ in (11a). The third author, a native speaker, syllabifies intervocalic laryngealized sonorants as in (11b): the first syllable ends with a laryngealized sonorant, and the second syllable begins with the same sonorant but modally voiced.

In Figure 11, /w̰/ in /saw̰itʰ/ ‘watered’ (11b) shows reduced amplitude and pitch and increased duration in comparison to /w/ in /sawitʰ/ ‘screamed’ (Figure 10, example (11a)). The laryngealization or creaky voicing of /w̰/, visible in Figure 11, is strongest in the middle of the segment but spreads rightward into the following vowel. We consider this sonorant to be mid-glottalized, but with the laryngeal gesture persisting to the end of the oral (labial and tongue body) gestures. Mid-glottalization is consistent with the third author’s native speaker intuitions about syllabification in (11b), with the strongest laryngealization flanked by the oral articulation of the sonorant on either side. Similar variation in the timing of the laryngeal gesture in laryngealized sonorants also occurs in St’át’imcets (Bird Reference Bird2011).

There is diachronic evidence that Chukchansi has reanalyzed some intervocalic laryngealized sonorants as a [ʔ]+sonorant sequence. For example, cognates of /kʰaʔju/ ‘coyote’ in other Yokuts varieties have a laryngealized sonorant: /kʰaj̃u/ (Newman Reference Newman1944, Gamble Reference Gamble2018). These Yokuts varieties retain the glottalization contrast intervocalically; it is likely that the intervocalic [ʔ.j] sequence in (12) is the Chukchansi reflex of intervocalic /j̃/.

As previously shown for [kʰaʔ.juʔ] in Figure 3, laryngealization occurs before modal voicing of the sonorant, yielding the pre-glottalized sequence [ʔ.j]; though, as an anonymous reviewer notes, there is overlap between the glottal gesture and the oral (palatal) gesture. This timing contrasts with that of laryngealized sonorants in general in Chukchansi, which are mid-glottalized or post-glottalized, not pre-glottalized. In Figure 11, showing [saw̰.witʰ] ‘watered’ (rec.pst), and Figures 14 and 15 further below, showing [ʔan̰mi] ‘while leaning’ and [saw̰mi] ‘while watering’, laryngealization occurs toward the middle (Figure 11) or the end (Figures 14 and 15) of the sonorants.

Figure 10

A spectrogram and waveform of /sawitʰ/ ‘screamed’ (rec.pst), illustrating the intensity (red solid line) and the pitch (blue dotted line) of the modally voiced labial-velar approximant in intervocalic position.

Figure 11

A spectrogram and waveform of /saw̰itʰ/ ‘watered’ (rec.pst), illustrating the intensity (red solid line) and the pitch (blue dotted line) of the laryngealized labial-velar approximant in intervocalic position.

In word-medial position when the sonorant precedes another consonant, i.e. in coda position, the contrast between modally voiced and laryngealized sonorants can be observed as a contrast in pitch and intensity (in addition to the presence of a glottal gesture in the latter, which shows up in the spectrum as silence and may be perceived as a glottal stop). Glottalized sonorants tend to have a sharp decrease in pitch and intensity from the preceding vowel, while plain sonorants show either a slight decrease in pitch and intensity or none at all. Examples of the plain alveolar nasal /n/ and plain labial-velar approximant /w/ preceding another consonant are shown in Figures 12 and 13. In the spectrogram of the plain nasal in /tʰanmi/ ‘while going’ (Figure 12), the intensity of the signal (red solid line) decreases slightly during the alveolar nasal into the following bilabial nasal. There is no sharp change in the pitch level (blue dotted line) either into, during, or out of the alveolar nasal. In the spectrogram of /sawmi/ ‘while screaming’ (Figure 12), there are small decreases in both pitch and intensity during the labial-velar approximant, which continue to decrease following this segment.

Figure 12

A spectrogram and waveform of /tʰanmi/ ‘while going’, illustrating the intensity (red solid line) and the pitch (blue dotted line) of the modally voiced alveolar nasal in coda position.

Figure 13

A spectrogram and waveform of /sawmi/ ‘while screaming’, illustrating the intensity (red solid line) and the pitch (blue dotted line) of the modally voiced labial-velar approximant in coda position.

Figure 14

A spectrogram and waveform of /ʔan̰mi/ ‘while leaning’, illustrating the intensity (red solid line) and the pitch (blue dotted line) of the laryngealized alveolar nasal in coda position.

Figure 15

A spectrogram and waveform of /saw̰mi/ ‘while watering’, illustrating the intensity (red solid line) and the pitch (blue dotted line) of the laryngealized labial-velar approximant in coda position.

When a laryngealized nasal or approximant is in coda position, the decrease in amplitude, intensity and pitch level is dramatic compared to the cases of the modally voiced sonorant. The intensity level for the laryngealized portions of the laryngealized alveolar nasal in /ʔan̰mi/ ‘while leaning’ (Figure 14) and the laryngealized labial-velar approximant in /saw̰mi/ ‘while watering’ (Figure 15) drops below the bottom threshold of 40 dB on the y-axis. The amplitude of the laryngealized portion of the laryngealized sonorants is also much reduced relative to the modally voiced sonorants in Figures 12 and 13. Due to the cessation of voicing in the laryngealized portion of the segments in Figures 14 and 15, f0 is absent in the middle of this portion. In both figures, intensity and amplitude levels rise sharply and pitch returns after the laryngealized sonorant ends and the following modally voiced sonorant begins. In Figure 14, the pitch drops slightly at the beginning of modal voicing on [m] in [ʔan̰mi], though we are not certain whether this is due to the previous glottal constriction having raised f0 beforehand, as suggested by a reviewer, or because of a general downdrift of pitch toward the end of a word spoken in isolation.

In the laryngealized portions of the laryngealized sonorants in Figures 14 and 15, there is a period of silence, especially noticeable in Figure 15. The brief period of silence in the laryngealized sonorants in these cases is typically perceived as a glottal stop, as also noted in Collord (Reference Collord1968: 4), e.g. /ʔan̰mi/ is perceived as [ʔanʔmi]. In other tokens of laryngealized sonorants, there is no full glottal closure or silence, but creaky voicing instead. For example, in [saw̰itʰ] in Figure 11, the laryngealized sonorant /w̰/ has creaky voicing but not full glottal closure (see also Bird Reference Bird2011 for St’át’imcets). The same variation occurs with intervocalic glottal stops /ʔ/, which are sometimes realized as a full closure [ʔ] and other times as creaky voice on the neighboring vowels.

Vowel F1 and F2

In addition to being distinguished by duration, short and long vowels in Chukchansi are also distinguished in the vowel space. Impressionistically, long vowels in Chukchansi are more peripheral than short vowels. Table 1 shows the mean values and standard deviations of the first two formants for the ten Chukchansi vowels. The measurements were based on vowels in the penultimate (stressed) position of two- or three-syllable words and in an open syllable.Footnote 2 Based on the recording of one native speaker, these words were produced in isolation, repeated two or three times, with a total of 247 tokens. Consonant environment was not controlled for, to allow for a balanced number of tokens. The wordlist includes both words recorded for this study and additional words that were necessary to obtain enough tokens of all the vowels. Formant measurements were made at the midpoint of the vowel in Praat (version 6.0.41; Boersma & Weenink Reference Boersma and Weenink2018), using a modified script from Christian DiCanio (http://www.acsu.buffalo.edu/∼cdicanio/scripts.html).

Table 1

Mean F1/F2 values (in Hz), Standard Deviations (in parentheses) and number of tokens of Chukchansi vowels.

The F1 mean values in Table 1 show that high and mid short vowels tend to have higher values (and are thus lower in the vowel space) compared to their long counterparts. The F1 mean value for /a/, on the other hand, is lower than for /aː/, suggesting that the short vowel is higher in the vowel space than its long counterpart for the low vowel. The F2 mean values for /i e/ are lower than for /iː eː/, but they are higher for /o a/ than for /oː aː/. This suggests that, with the possible exception of /u uː/, long vowels are slightly more peripheral than short vowels, as cross-linguistically expected (e.g. Johnson & Martin Reference Johnson and Jack2001 for Creek, Maddieson, Smith & Bessell Reference Maddieson, Smith and Bessell2001 for Tlingit, Hirata & Tsukada 2009 for Japanese).Footnote 3 Figure 16 shows the vowel space of the ten vowels in Chukchansi, with the IPA symbol at the mean and the ellipses giving one standard deviation.

Figure 16

Chukchansi vowel space, with ellipses indicating one standard deviation away from the mean. The figure was created using the phonR package in R (McCloy Reference McCloy2016).

As shown in Figure 16 below, there is quite a bit of overlap in the acoustic space between a short vowel and its long counterpart. The degree of overlap is greater for /u uː/ than for the other vowel pairs. Among the long vowels, there is a slight overlap between the front vowels /iː eː/ and between the back vowels /uː oː/. The long vowel /aː/ does not overlap with the other long vowels. Among the short vowels, a small amount of overlap occurs among adjacent vowels, with the largest overlap between /a o/.

Vowel length

All vowels in Chukchansi have contrastive length, as illustrated in Figures 17–20. The contrasted vowels are in an open penultimate syllable, thus bearing the primary stress, as described below in the section ‘Prosodic Features’. They show the relative duration of short and long vowels. The vowels are preceded by the plosives /t pș tʰ/ or the nasal /n/ and followed by the modally voiced alveolar sonorants /n l/. Vowel duration was measured from the beginning of high-amplitude, periodic waveforms to the end of the high-amplitude in the waveform. The waveforms in Figures 17–20 show that the distinction in amplitude clearly shows the boundary between vowels and sonorants. Figures 17 and 18 illustrate the short vowels /e/ in /tenel̰/ and /o/ in /pșonoʃ/.

Figures 19 and 20 show the long vowel counterparts: /eː/ in /tʰeːlij̃/ and /oː/ in /noːnipș/. The duration difference for short vs. long vowels in the cases presented in Figures 17–20 is greater than a 1:2.0 ratio. In the case of /tenel̰/ vs. /tʰeːlij̃/, the ratio is 1:2.6. In the case of /pșonoʃ/ vs. /noːnipș/, it is 1:2.2.

Table 2

Mean duration values (in ms), Standard Deviations (in parentheses), number of tokens and duration ratios of Chukchansi vowels.

^aMartin’s (2011) study is the only other existing acoustic investigation of Chukchansi vowels. His study is based on two speakers, one of whom is the third author in the present study. He finds the duration ratio for short vs. long vowels to be 1:1.5 for mid front and high vowels, and about 1:2.0 for mid back and low vowels.

Figure 17

A spectrogram and waveform of /tenel̰/ ‘hole in rock’ (nom), illustrating the mid short vowel /e/ in open penult with a 102 ms duration.

Figure 18

A spectrogram and waveform of /pșonoʃ/ ‘hand’ (nom), illustrating the mid short vowel /o/ in open penult with a 98 ms duration.

Duration measurements of short and long vowels are presented in Table 2. The duration measurements were based on the same vowels from which the formant measurements were made.

In addition to the mean duration values and standard deviations, Table 2 shows the number of tokens analyzed and the duration ratios of the vowels with contrastive length. The results indicate that for all vowels, the duration ratios for short vs. long vowels are at least 1:2.0. The smallest duration ratio is observed for the mid back vowels /o oː/ and the greatest duration ratio for the low vowels /a aː/.

Figure 19

A spectrogram and waveform of /tʰeːlij̃/ ‘tooth’ (nom), illustrating the mid long vowel /eː/ in open penult with a 268 ms duration.

Figure 20

A spectrogram and waveform of /noːnipș/ ‘nine’ (nom), illustrating the mid long vowel /oː/ in open penult with a 215 ms duration.