4.2.1 Anterior constrictions across click types

As shown in table 2, Nǀuu contrasts five click types – four that have a coronal anterior constriction, and one that has a bilabial anterior constriction. Miller et al. (Reference Miller, Brugman and Sands2007) and Sands et al. (Reference Sands, Brugman, Exter, Namaseb and Miller2007) provide palatographic and linguographic data on the four coronal types. Results show that the dental click has a laminal dental articulation, with a fairly broad area of contact. The central and lateral alveolar clicks are clearly alveolar in Nǀuu, and not postalveolar as in IsiXhosa (Sands Reference Sands1991), IsiZulu (Doke Reference Doke1923) and Khoekhoe (Beach Reference Beach1938). Tongue contact primarily involves the tongue tip, as has been reported for ǃXóõ (Traill Reference Traill1985). The palatal click is characterized by an extremely wide contact area between the alveolar ridge and the palatal region. We suspect, based on our viewing of ultrasound movies of this click in Nǀuu and EPG data for this click in Khoekhoe, that the long laminal contact seen in palatograms is due to dynamic movement of the anterior constriction during production of the palatal click, rather than a long laminal closure at the time of the release.

The bilabial click in Nǀuu has been assessed through video images of the lips during its production (Miller et al. Reference Miller, Brugman and Sands2007). Dynamic changes in lip posture show compression and release over the initial 20 ms interval of production, followed by puckering of the lips over the remaining 20 ms of the segment's duration. There is also very salient rounding coarticulation on following vowels in Nǀuu, and words like ‘meat’ are realized as both [⊙oe] and [⊙əe] by the same speaker in the same context. We have also documented a variant production of the bilabial click by Speaker HK that involves contact between the lower lip and the upper teeth. This variant has also been reported in ǃXóõ (Ladefoged & Maddieson Reference Ladefoged and Maddieson1996: 251). Overall, our investigation of anterior place in Nǀuu clicks has shown that clicks exhibit individual phonetic variation similar to that documented for pulmonic consonants.

Figure 2 provides waveforms of words that begin with each of the five lingual stops, as well as bark-scaled LPC spectra computed over 10 ms windows. For the spectra, 30 ms windows were extracted with the center of the window aligned at the left edge of the burst. A barkfilter was then created using a 10 ms window, with the bandwidth of each filter being one bark. The waveforms show that the bilabial, dental and lateral alveolar clicks are noisy, while the central alveolar and palatal clicks are abrupt, much like those reported for ǃXóõ (Ladefoged & Traill Reference Ladefoged and Traill1994, Ladefoged & Maddieson Reference Ladefoged and Maddieson1996). That is, the bursts of the central alveolar and palatal clicks exhibit a sharp, intense transient, with little turbulent noise, while the bursts in bilabial, dental and lateral alveolar clicks are noisy, making it difficult to isolate the transient. The differences in noisiness correlate with durational differences: noisy clicks are longer than abrupt ones.

Figure 2

Representative waveforms of bursts extracted from the frame sentence [nɑ kɑ _____] ‘I say _____’ for the lingual stops in: (a) [⊙oɑᵪe] ‘daughter’, (b) [ǀɑɑᵪe] ‘niece’, (c) [!ɑmɑ] ‘kidney’, (d) [ǁɑɑᵪe] ‘sister’ and (e) [ǂəusi] ‘tsama melon’. Normalized bark-scaled LPC spectra of bursts in words: (a) [⊙uⁿuⁿ] ‘son’, (b) [ǀuⁿuⁿ] ‘boil’, (c) [!uu] ‘acacia’, (d) [ǁuu] ‘grasshopper’ and (e) [ǂuuke] ‘fly’ (Speaker KE).

Ladefoged & Maddieson (Reference Ladefoged and Maddieson1996) attribute both the shorter duration and the dominance of the transient in the abrupt clicks to a faster anterior release. It is interesting to note that the clicks in most languages pattern like those in Nǀuu with respect to their abruptness, but Fulop et al. (Reference Fulop, Ladefoged, Liu and Vossen2003) report that palatal clicks in the Bantu language Shiyeyi tend to be longer and noisier, like the dental clicks, and that lateral clicks are often sharp like the central alveolar clicks. We expect that cross-linguistic studies will show more complexity of this type, as found for pulmonic coronal stops in European languages (e.g. Dart Reference Dart1998).

It is important to note that there is no indication in these waveforms of a pulmonic burst between the click burst and the vowel. Descriptions of clicks often claim that the release of the posterior constriction is a pulmonic stop (e.g. Ladefoged & Traill Reference Ladefoged and Traill1994), but it is crucial to our subsequent argument that this is not the case. We sometimes find low-intensity events between the lingual burst and the vowel onset, and these events may well correspond to the release of the posterior constriction, but they are low amplitude and generally imperceptible, especially given their proximity to the high-amplitude click burst. Traill (Reference Traill1985: 125–126), in fact, makes this same observation for ǃXóõ. He cites Beach's (Reference Beach1938: 82) comment that the posterior closure in Khoekhoe clicks can be released ‘practically silently’ and maintains that the same is true of ǃXóõ. This observation will be important in our discussion of the distinction between lingual and linguo-pulmonic clicks in section 4.4, and we maintain that it is the norm rather than the exception in click production.

The click burst spectra in figure 2 (see also Miller, Brugman & Sands Reference Sands, Miller, Brugman, Payne and Peña2007) provide acoustic evidence of both anterior and posterior place differences among the different clicks. The overall energy distribution patterns in these bursts are consistent with those reported for ǃXóõ (Ladefoged & Traill Reference Ladefoged and Traill1994, Ladefoged & Maddieson Reference Ladefoged and Maddieson1996), in that the dental and palatal bursts have more energy at higher frequencies, the central and lateral alveolar clicks have more energy at lower frequencies and the bilabial click spectra are relatively diffuse. Moreover, we are able to consistently identify two peaks in the burst spectra of the four coronal click types, as well as about half of the bilabial click tokens. The identified peaks in these spectra are marked with vertical lines in figure 2. Though we have not found significant differences among the lower peaks (P1), the higher peak (P2) does distinguish all five click types (Miller, Brugman & Sands Reference Sands, Miller, Brugman, Payne and Peña2007).

The spectral peaks can be modeled with an adaptation of a proposal by Stevens (Reference Stevens1999: 186–187), who models the peaks in abrupt clicks with a short tube. He claims that the whole tube in front of the posterior constriction contributes the spectral peaks in click bursts because the anterior constriction is released within a few milliseconds. Kagaya (Reference Kagaya1978), on the other hand, attributes P1 to the cavity in front of the constriction, as it is in pulmonic egressive stops (Fant Reference Fant1960). We attribute P1 to the entire cavity volume in front of the posterior constriction following Stevens (Reference Stevens1999: 186–187). Although Stevens does not model the higher peak seen in figure 2, we attribute this higher frequency peak to the volume of the lingual cavity between the two constrictions in the coronal clicks. Since we propose that P2 relates to the volume of the lingual cavity, we discuss these values here. P1 values are discussed in section 4.2.2 below, since they provide data about the posterior place of articulation.

The P2 frequency values for the tokens in figure 2 are: 20 bark for [⊙], 17 bark for [ǀ], 15 bark for [ǂ], 14 bark for [ǃ] and 12 bark for [ǁ]. These are comparable to the averages reported for three speakers in Miller, Brugman & Sands (Reference Sands, Miller, Brugman, Payne and Peña2007). The values suggest that among the coronal clicks, the dental has the smallest lingual cavity, followed by the palatal, and the two alveolar clicks. Finally, the high-frequency P2 peak found in about half our tokens of the bilabial click may be attributable to a simultaneous dental constriction in these productions. Traill (Reference Traill1985) reports such a variant in ǃXóõ, and the fact that only half the tokens are characterized by such a peak makes this interpretation likely.

4.2.2 Posterior constrictions across click types

We now turn to our description of the posterior places of articulation in the five Nǀuu click types ([⊙ ǀ ǃ ǁ ǂ]). At least since Doke (Reference Doke1923), the phonetic literature has assumed that all basic click types are produced with a velar posterior place of articulation. Miller, Namaseb & Iskarous (Reference Miller2007) have, however, shown that the palatal and postalveolar clicks in Khoekhoe in fact have different posterior places of articulation, and we will see that the same is true in Nǀuu (Miller et al. Reference Miller, Brugman, Howell and Sands2006). Our first goal here is to provide evidence supporting our claim that the posterior constrictions in Nǀuu alveolar [ǃ] and palatal [ǂ] clicks are qualitatively different from velar stops. Previous descriptions of clicks have focused on the location of the front part of the posterior constriction, which is usually described as velar and equated with [k], but we will show that the shape and dynamics of the tongue root in clicks are actually very different from [k] and from each other. We attribute these differences to the muscular activity involved in click production and constraints on overall tongue shape. We will begin with ultrasound data that show different posterior places of articulation for [ǃ] and [ǂ], and move on to spectral data from the linguo-pulmonic clicks that suggest that the same pattern obtains in these segments, as well.

We begin with our ultrasound data, which allow us to pinpoint the part of the tongue used in the production of clicks, as well as the location of the release relative to velar and uvular consonants. Our discussion in this section distinguishes between three posterior parts of the tongue, namely the dorsum, the upper tongue root and the tongue root proper (the lower part of the tongue root). Generally, the tongue dorsum is at, or in front of, the posterior constriction during click closures, while the upper tongue root and the tongue root proper are behind it. By upper tongue root, we mean the part of the tongue that is at the interface between the oral and pharyngeal cavities in rest position, and by the tongue root proper, we mean the part in the upper oropharynx. We also need to distinguish between the raising of the back of the tongue that occurs in high back vowels and is primarily a vocalic gesture (Esling Reference Esling2005) and the retraction of the tongue root into the pharyngeal region that is involved in alveolar click production and is a consonantal gesture. We show that the raising of the back of the tongue into the uvulo-pharyngeal region found in palatal clicks is in part articulated in the back of the uvular region and is very similar to the production of [u], which follows the clicks in this study. The tongue root proper forms the upper part of the laryngeal constrictor mechanism described by Esling (Reference Esling2005) so in terms of his model, the alveolar click involves the laryngeal constrictor mechanism. However, Edmondson et al. (Reference Edmondson, Padayodi, Hassan and Esling2007) show that languages such as Yi, Akan and Kabiye ‘can use constriction to achieve significant pharyngeal resonator reduction without changing phonation to any great extent’ (p. 2068). The production of Nǀuu alveolar clicks is likely similar to the production of vowels in these languages, then, in using tongue root retraction, and probable concomitant aryepiglottic fold constriction, without changing phonation, as there are no voice quality cues associated with plain voiceless click consonants. Further, constriction of the tongue root proper that is involved in alveolar click production in Nǀuu, is the likely lingual accompaniment to aryepiglottic sphinctering found in epiglottalized vowels in the language (Traill Reference Traill, Singer and Lundy1986; Esling Reference Esling1996, Reference Esling2005; Hess Reference Hess1998). The epiglottalized vowels in Nǀuu involve non-modal phonation similar to the [+ constricted] vowels in Bai and Somali (Edmondson et al. Reference Edmondson, Padayodi, Hassan and Esling2007), which is similar for most speakers to that found in Juǀ’hoansi (Miller-Ockhuizen Reference Miller-Ockhuizen2003, Miller Reference Miller2007). We expect that raising of the upper tongue root involved in palatal click production, would not accompany aryepiglottic sphinctering, as this does not involve the laryngeal sphincter mechanism.

When interpreting tongue traces in this section, it is important to remember that they show the position of the tongue relative to the ultrasound probe, which was positioned beneath the jaw. Raising and lowering of the tongue with respect to the jaw will generally result in raising and lowering with respect to the (hard) palate, but this technology underestimates displacement that results from jaw movement. In addition, ultrasound is an imprecise technology for imaging the tip of the tongue. Ultrasound depends on the transmission of sound waves through tissue, and the sublingual cavity associated with a raised tongue tip allows only poor transmission of these waves. We were not, therefore, able to trace the tip of the tongue reliably in the alveolar clicks. We do, however, know from Traill's x-ray recordings of ǃXóõ that the tongue tip in these segments is pointed upward. We can also deduce the apical gesture in the alveolar click from the extremely concave tongue body shape that we see in our ultrasound videos. We also observe that the tongue root gesture used in the alveolar click involves a convex shape, which corroborates Esling's (Reference Esling2005) claim that tongue body and root shapes are independent.

Figure 3 provides representative ultrasound tongue traces from frames that correspond to the closure and release portions of the alveolar [ǃ] and palatal [ǂ] clicks, as well as traces of the pulmonic velar and uvular stops from the frame sentence. These plots make the differences in tongue body shape for these two click types quite clear. Looking first at the top panel, we see small differences in the location of the upper tongue root (i.e. the back sides of the constrictions) in the velar stop [k], the uvular stop [q] and the posterior constriction in the central alveolar lingual stop [ǃ]. Specifically, the upper tongue root in [k] is most advanced, that in [q] is intermediate and that in [ǃ] is most retracted. The differences may in part be due to vowel context. Our impression from viewing the dynamics of this segment in the source ultrasound movies, as well as Traill's x-ray recordings of similar segments in ǃXóõ, is that the same is true of the tongue root proper. One difference between velar and uvular pulmonic stops cross-linguistically is that uvular segments are characterized by tongue root retraction. Hess (Reference Hess1998), for instance, reproduces x-ray traces from Tunisian and Iraqi Arabic that show a much more retracted upper and lower tongue root in [q] than [k]. The retraction of the tongue root proper involved in the production of [ǃ] supports its analysis as an alveolar-uvular segment.

Figure 3

Tongue traces of the closures and releases of the central alveolar [!] (top) and palatal [ǂ] (bottom) click types, as well as velar and uvular pulmonic stop closures in the frame sentence [nɑ kɑ _____ nɑ kɑ qo^ʢnɑ^ʢiⁿ] ‘I say _____, I say famished’. (Note that ‘=’ is used for ‘ǂ’ in the plot.)

Turning to the posterior constriction in the palatal click, we see that the upper tongue root is raised into the back of the uvular region, hence its description as palatal-uvular in table 2. The constriction is technically uvulo-pharyngeal, with contact along the uvula, which is raised in the production of this click. There cannot be a complete pharyngeal constriction, nor can the velum be completely raised in nasal varieties of these clicks, because such raising would block nasal airflow. We do, however, see pronounced retraction in the videos. We know that the upper tongue root is raised and the tongue dorsum is retracted into the upper pharynx, presumably causing bunching. This is akin to the raising gesture that Esling (Reference Esling2005) finds in high back vowels, and differs from the retraction gesture typically found in uvular pulmonic consonants. We surmise the raising of the upper tongue root from hyoid movement, which is visible as movement of the hyoid shadow in the ultrasound recordings. In the palatal click, the upper tongue root makes a constriction in the back of the uvular region. Constrictions can often be seen in ultrasound because the soft palate becomes visible in the image when the tongue touches it. That is, the uvulo-pharyngeal constriction found in the palatal click is not a retraction gesture as found in epiglottals, but is rather a raising gesture like that Esling (Reference Esling2005) describes in the production of high back vowels. The similarity between the [u] gesture that follows the clicks in our data and the palatal click posterior tongue gesture is quite striking. We are not aware of post-velar pulmonic obstruents with such high retracted constrictions, though Ladefoged & Maddieson (Reference Ladefoged and Maddieson1996: 169–170) describe a somewhat similar constriction in the Danish approximant [ʁ], which we might transcribe as []. It might be that the back uvular component of the posterior gesture in the palatal click is more similar to this more open constriction in the Danish approximant. There may also be an upper pharyngeal more open constriction in the palatal click. We cannot ascertain the degree of constriction with our current ultrasound recordings, as we have not been able to trace the palate, but it is clear that the back part of the posterior lingual gesture found in the palatal clicks is a more open gesture, rather than a more constricted consonantal gesture.

In addition to the clear differences in overall tongue shape, our ultrasound recordings also show differences in the timing of the anterior and posterior releases of the central alveolar [ǃ] and palatal [ǂ] click types. In recordings of the central alveolar click, we sometimes see the tongue tip moving downward from the apical articulation which suggests the uncurling of an extreme concave tongue shape. The deep concave tongue body shape indicates a high volume cavity between the two constrictions. In general, the front of the tongue in this click seems to move down faster than the back of the tongue, which is probably related to the extreme jaw lowering involved in its production. In the palatal click, on the other hand, the two constrictions seem to be lowered more simultaneously, as is suggested from the successive frames traced in the bottom panel of figure 3. The cavity between the constrictions is much shallower in this click. These differences in tongue shape and the resultant differences in cavity size account for the differences in the high spectral peak, P2, shown in figure 2 and discussed in section 4.2.1 above.

Our impression from viewing the ultrasound video is that the relationship between the anterior constriction location, the posterior constriction location and resulting cavity volume are highly constrained by the tongue musculature. The tongue muscles are interconnected, and the tongue itself can be divided into four main extrinsic muscles (Zemlin Reference Zemlin1968, Harris, Vatikiotis-Bateson & Alfonso Reference Harris, Vatikiotis-Bateson, Alfonso, Ohala, Nearey, Derwing, Hodge and Wiebe1992, Honda Reference Honda1996), which are divided into groups of compatible vs. incompatible pairs. We suspect that there is muscular incompatibility between apical alveolar and back uvular articulations, the latter of which involve raising the back of the tongue, as well as between laminal dental or laminal pre-palatal anterior constrictions and uvular posterior constrictions, which involve consonantal tongue root retraction. In the palatal click, the tongue body is raised upwards and forwards by anterior genioglossus muscle activity, which goes hand-in-hand with the advancement and raising of the upper tongue root accomplished through the compatible styloglossus muscle activity found in high back vowels (Esling Reference Esling2005). These movements are surmised from the swinging action of the hyoid bone that can be deduced from the movement of its shadow, which we have observed in every token of [ǂ] we have viewed in both Nǀuu and Khoekhoe. Additionally, we surmise that the alveolar click is produced using a smaller degree of styloglossus activity which pulls the tongue dorsum upwards, as well as a larger degree of hyoglossus activity that pulls the tongue root backward.

We turn now to the acoustic evidence for posterior place of articulation differences among the coronal clicks. Recall that we attribute the lower peak in the spectrum (P1) to the cavity in front of the posterior constriction following Stevens (Reference Stevens1999). The P1 values for the dental and palatal clicks are higher than those in the central and lateral alveolar clicks, suggesting a smaller cavity in front of the posterior constriction for the dental and palatal clicks. We attribute the higher peaks found for the dental and palatal clicks to their long, flat anterior and posterior constrictions, which lead to a smaller cavity volume. P1 in alveolar clicks is lower than P1 in the palatal and dental clicks because their apical anterior constrictions and narrow posterior constrictions allow a more concave tongue shape and, consequently, a larger cavity volume. We also see subtle P1 differences in the central alveolar and lateral alveolar clicks. Specifically, P1 is lower for the central alveolar click than for the lateral alveolar click, which we take to indicate a larger cavity in front of the posterior constriction for [ǃ] than [ǁ]. This observation is consistent with the slightly more laminal anterior constriction in [ǁ] (Sandset al. Reference Sands, Brugman, Exter, Namaseb and Miller2007), which leads to a slightly less concave tongue shape and slightly smaller cavity.

We now turn to acoustic data that corroborate our ultrasound findings, specifically to evidence provided by linguo-pulmonic stops (i.e. [⊙⊃q ǀ⊃q ǃ⊃q ǁ⊃q ǂ⊃q]). We will see below that these stops are characterized by two bursts, the first lingual and the second pulmonic, and we argue in section 4.4 below that lingual and linguo-pulmonic segments with the same click type (e.g. [ǃ] and [ǃ⊃q]) have the same posterior place of articulation. There are, however, differences in the posterior place of articulation across the different linguo-pulmonic stop types, just as we saw in the ultrasound traces of [ǃ] and [ǂ]. The second, pulmonic burst in these segments provides information about their posterior places of articulation. Representative spectra of such bursts are provided in figure 4, where we see that the energy distribution in the bilabial and alveolar linguo-pulmonic stops is very similar to that in the uvular pulmonic stop. All four exhibit a sharply falling spectrum between 0 and 10 bark. Pulmonic bursts in dental and palatal linguo-pulmonic stops, on the other hand, exhibit energy throughout this range. Thus, these spectra provide clear evidence that the pulmonic portion of the bilabial and alveolar linguo-pulmonic stops differs from the pulmonic portion of their dental and palatal counterparts. These data are incompatible with an analysis that represents the pulmonic release of all linguo-pulmonic stops as a uvular stop, whether as part of the same segment or as a separate segment. Moreover, they support our contention that the posterior constrictions vary with click type.

Figure 4

Spectra for posterior bursts in linguo-pulmonic stops extracted from words in the frame sentence [nɑ kɑ _____] ‘I say _____’: (a) [q'ui] ‘good’, (b) [⊙⊃qʰuiɑ] ‘sweat’, (c) [ǀ⊃quu] ‘tobacco’, (d) [!⊃qui] ‘ashes’, (e) [ǁ⊃quu] ‘urine’ and (f) [ǂ⊃quu] ‘neck’ (Speaker GS).

Taken together, these data show that the posterior closures in clicks involve post-velar constrictions that are qualitatively different from [k]. We, therefore, propose a terminological shift from velaric airstream mechanism to the articulatorily more accurate lingual airstream mechanism (Miller, Namaseb & Iskarous Reference Miller2007).Footnote ⁵ The use of velaric as a descriptor for clicks dates to Beach (Reference Beach1938: 74), whose definition of click encompasses all segments with ingressive airflow. That is, he distinguishes between velaric, glottalic and pulmonic clicks, which in current terminology would be clicks, implosives and pulmonic ingressives. In a system that defines clicks in terms of ‘the inner boundary of the chamber wherein the air is rarefied’, the term velaric makes sense, but as the name of an airstream it is problematic. The other two airstream mechanisms have names that reflect the anatomical source of the airflow – the lungs in the case of the pulmonic airstream and the glottis in the case of the glottalic airstream. By extension, the term velaric suggests that this airstream is somehow initiated by the velum or that it involves a velar stop, and this is clearly not the case. Rather, the tongue is used to create a negative pressure cavity, the anterior release of which initiates the ingressive flow of air. For this reason, we adopt the term lingual airstream mechanism in describing these segments.