Introduction

The supplementary materials provide some additional information on the !Xóõ corpus, illustrate some cases of excluded lexical items and notes for future research, and includes additional figures showing the timecourse of the acoustic measures as they vary by phonation type and by lexical item.

The corpus

The files used in this study comes from freely-accessible recordings made by Tony Traill and Peter Ladefoged in the summer of 1979 near Lokolane, Botswana.

The files can be acccessed on the UCLA Phonetics Lab Archive. Sound files analyzed in this study include “nmn_word-list_1979_5”, “nmn_word-list_1979_6”, “nmn_word-list_1979_7”, “nmn_word-list_1979_8”, “nmn_word-list_1979_12”, and “nmn_word-list_1979_14”. Some of the words in these recordings were excluded; see details in paper and following tabs in this section.

Wordlist

The words included in the main paper are shown below. I’ve also included differences in the spelling, glosses and transcriptions between the Archive and Traill’s dictionary. (In the paper, harsh vowels are transcribed using a superscript voiced epiglottal trill; I use a voiceless epiglottal trill here because the voiced epiglottal trill has no unicode value.)

Phonation Transcription Spelling (Traill 1994) Spelling (Archive) Gloss (Traill 1994) Gloss (Archive) Tone (Traill 1994) Tone (Archive)
Breathy ǃā̤o ǃāho ǃāhoh slope slope M M
Breathy ǃā̤le ǃāhle ǃāhle stand on climb up M M
Breathy ǃà̤la ǃàhla ǃàhla Peeling-bark Ochna species of tree L L
Modal ᶢǂàa gǂàa ǂɡàa exploit refuse to help L L
Modal ǂába ǂába ǂabá peg peg H H
Modal ǁáa ǁáa ǁáa Camel Thorn tree Camelthorn tree H H
Creaky ᶢǃᵡâˀje ɡǃxâ’je ɡǃxâ’aje udder udder F F
Creaky ᶢǀàˀje ɡǀà’je ǀɡà’aje bend bend L L
Pharyngealized qâˤa qâ̰a qà̰a long ago long ago F L
Pharyngealized táˤi tá̰i ta̰í far far away H H
Pharyngealized ǂqàˤn ǂqà̰n ǂqà̰n smooth smoothe L L
Breathy-creaky ǀà̤ˀje ǀàh’je ǀāh’aje wait for wait for him/it L M
Breathy-creaky ᶢǀā̤ˀbe ɡǀāh’be ǀāh’abe hide hide it M M
Breathy-creaky ᵑǀā̤ˀbe ǀnāh’be ǃnāhabe surround it surround it M M
Harsh ˀᵑǃàᵸo ’ǃnà̰ho ˀǃnà̰o base base L L
Harsh ˀᵑǃàᵸle ’ǃnà̰hle ˀǃnà̰le lower it lower it L L
Harsh ǂɢàᵸli ɢǂà̰hli nǂɢà̰hi Bladethorn tree species of tree L L

Voiced ejective clicks?

One of the words in these recordings, ǃgā’a ‘lower’ (per transcriptions in Archive) or ‘low’ (in Traill’s 1994 dictionary), is rather consistently realized with a glottal stop after the release of the click. The following vowel is modal. Sometimes however, there is a weak vowel following the release of the quick and prior to the glottal closure; see sample spectrograms below. This word is perhaps best transcribed as [ᶢǃˀāa] with a voiced post-glottalized click, or as [ᶢǃ’āa] with a voiced ejective click. Such sounds are documented for West !Xóõ (Naumann, 2016) but not in Traill’s work on East !Xóõ.

Voiceless nasal click

Another word, ǁn̥â’am ‘damp’ was ultimately excluded from the dataset because the word was often produced with breathy release of the click (see two upper panels in spectrogram below). In addition, in some tokens the breathy release of the click was followed by pharyngealization or harsh voice. See two lower panels in the figure below.

Pharyngealized-creaky vowels

In this dataset, there was one lexical item with a pharyngealized-creaky vowel: ǃn̥ɑ̰̂’ɑm ‘evade’ (‘avoid’ in Traill’s dictionary).Some sample spectrograms of the pharyngealized-creaky token are shown below.

Many of the tokens appear indistinguishable from pharyngealized vowels, which also show irregular creaky voicing, as shown below (repeated from Figure 2 in paper). It is possible that this word is in fact pharyngealized, instead of pharyngealized-creaky. But it is equally possible that the type of creaky voice differs between pharyngealized vs. pharyngealized-creaky vowels, as was found for pharyngealized vs. creaky vowels.

Acoustic measures

H1*-H2*

Averaged across all lexical items

By phonation and lexical item

HNR \(<\) 500 Hz

Averaged across all lexical items

By phonation and lexical item

HNR \(<\) 3500 Hz

Averaged across all lexical items

By phonation and lexical item

Strength of Excitation

Averaged across all lexical items

By phonation and lexical item

f0

Averaged across all lexical items

By phonation and lexical item

By tone and lexical item

Tone values are based off of entries in Traill’s (1994) !Xóõ dictionary. See discussion in paper about the possibility of collapsing the four categories to H (tones ‘H’ and ‘F’ in figure) and L (tones ‘L’ and ‘M’ in figure).

In fact, the f0 onsets appear to cluster in two groups: one averaging around 175 Hz; the other between 130-150 Hz. (Midpoints are problematic for harsh and pharyngealized vowels, which show sharp f0 dips.) This might lend support to a two-toned analysis.

An additional issue: The f0 onset for qâ̰a /qâˤa/ ‘long ago’ looks like those of other pharyngealized low tones, though it is labeled with a falling tone. Note that this word is transcribed/spelled as qà̰a with a low tone in the Archive; see the Wordlist above. This suggests a typo in Traill’s dictionary.

F1

Averaged across all lexical items

By phonation and lexical item

Other measures

F2

Duration

Pharyngealized vowels are longest, followed by harsh vowels. Breathy and creaky vowels are shortest.

H1*-A1*

H1*-A1* is not included in the main paper because it is a spectral tilt measure that does not appear in Kreiman et al.’s (2014) voice model. Nonetheless, it is frequently used to distinguish phonation types. Interestingly, according to this measure harsh voice looks more similar to breathy voice throughout its duration, and creaky voice has the lowest spectral tilt in the first third, as expected. See discussion in Garellek (to appear) The phonetics of voice on theoretical and practical issues with using this measure in addition to spectral tilt measures in Kreiman et al.’s model.

Terminology

The acoustic results of this study have implications for the terms used to describe acoustically the phonation contrasts in East !Xóõ. I defer to Traill as to how best to describe the contrasts in terms of phonological behavior, but discuss some phonotactic issues with respect to the combinations of phonation types below.

Based on frequency of use in current studies of voice quality/phonation, I favor the terms breathy, modal, creaky, pharyngealized and harsh as terms used to describe the phonation types in East !Xóõ. These describe quality of the phonation types using current and well-known terms. Though these terms are not always articulatorily explanative, this is probably an advantage for the language, given the spatial and temporal variability in production.

Breathy

Breathy vowels in the language have also been called ‘murmured’ (Ladefoged 1983), but this term has in recent decades fallen out of date among linguists researching breathy voice.

Creaky

Creaky vowels could also be called ‘glottalized’ (as is used, along with ‘tense’, for West !Xóõ, e.g. in Naumann 2016) or ‘laryngealized’ (as used for East !Xóõ by Ladefoged 1983). It should be noted, however, that both these terms are somewhat problematic. Articulatorily, glottalization (constriction of the vocal folds) is found for pharyngealized and harsh vowels as well as creaky ones, and `laryngealization’ is general enough articulatorily to encompass pharyngealized and harsh voice, perhaps even breathy voice (since breathy voice involves the larynx). ‘Tense’ voice is often used for constricted but high-pitched voicing, as discussed by Keating et al. (2015).

Another possible term that could be used for creaky vowels is ‘checked’, i.e. a vowel that is followed by a glottal stop. Many creaky vowels are in fact followed by sustained vocal fold constriction (a phonetic glottal stop), as Traill (1985) himself remarked on. The glottal stop, whether realized with sustained closure or as creaky voice, can account for the presence of creaky voice on creaky vowels, which tends to be phased towards the middle and end of the vowel. However, if one adopts the term ‘checked’ for creaky vowels, there are phonotactic considerations that come into play when considering whether the checked quality is due to a glottal stop, which is a segment in the language: if all creaky vowels are analyzed as modal vowels followed by a glottal stop, then this changes the phonotactic rules of East !Xóõ (see Traill 1985).

I would also argue against using terms that describe temporal characteristics, at least for East !Xóõ, because of temporal instability of the voice quality. For instance, while many of the creaky vowels end in a glottal stop (supported their being called as ‘checked’), not all do. Some are more ‘rearticulated’, involving maximal constriction in the middle of the vowel, and others have creaky voice throughout. So, the temporal and spatial vagueness of a term like ‘creaky’ is advantageous for East !Xóõ.

The vagueness of a term like ‘crealy’ is also advantageous when doing cross-dialectal and cross-linguistic comparisons. The creaky (`glottalized, interrupted’) vowels in Ju|’hoansi are creakiest in the middle of the vowel (Snyman 1977a), but cross-linguistic comparisons for languages like !Xóõ and Ju|’hoansi are potentially made more complicated when different terms are used to describe phonological properties that are likely similar.

Pharyngealized

These vowels have been called ‘pressed’ and ‘voiced strident’ in notes by Ladefoged and Traill (see Archive notes: http://archive.phonetics.ucla.edu/Language/NMN/nmn_word-list_1979_06.jpg; http://archive.phonetics.ucla.edu/Language/NMN/nmn_word-list_1979_15.jpg). ‘Pressed’ was used in work on Ju|‘hoansi (Snyman 1977b); these vowels are called ’epiglottalized’ by Miller (2007). ‘Epiglottalized’ would be an apt term for pharyngealized vowels in !Xóõ, given that their acoustic characteristics suggest that they have epilaryngeal constriction.

Bonny Sands pointed out to me that in Ju|‘hoansi and N|uu, the pharyngealized/epiglottalized vowels tend to be more fricated/trilled than in !Xóõ, in which pharyngealized vowel can have an epiglottal stop (see spectrograms above) but generally don’t have aryepiglottic frication/trilling. This suggests to me that in Ju|’hoansi and N|uu, these epiglottalized vowels might best be described as ’harsh’, akin to harsh vowels in !Xóõ, which also show aryepiglottic frication/trilling.

Harsh

Harsh vowels (phonologically breathy-pharyngealized, per Traill) in !Xóõ are usually called ‘strident’ (Traill 1985, Ladefoged & Maddieson 1996, Naumann 2016). It is my personal opinion that this term can be done away with in favor of ‘harsh’. Although Traill (1986) explicitly claims that the strident vowels are not harsh (following Laver 1980’s taxonomy), this is because there is no evidence for ventricular fold vibration during the strident vowels. But in more current research, ‘harsh’ voice refers to a rough, growling voice quality that involves epilaryngeal constriction (see discussion in Moisik 2013), just like the so-called ‘strident’ vowels in !Xóõ.

Combinations

It is still unclear whether we need to assume, as Traill (1985, 1994) does, that the phonation types can combine with each other to form breathy-creaky (breathy-glottalized), breathy-pharyngealized (strident), pharyngealized-creaky (pharyngealized-glottalized), and breathy-pharyngealized-creaky (strident-glottalized) vowels. Naumann (2016) says these combinations cannot occur in West !Xóõ, and consequently treats strident vowels as a separate phonological feature.

If East !Xóõ were analyzed as having a similar system (with modal, creaky, breathy, pharyngealized, and harsh vowels), then perhaps what differentiates the two dialects is more about phonotactics: namely, that East !Xóõ allows non-modal phonation types to be followed by a word-medial glottal stop. This is how Naumann (2016) tries to reconcile the differences between the two dialects, and he claims that some (but not all) words with combinations of a phonation type with creaky voice could be reanalyzed as having one phonation type followed by a glottal stop.

If this is indeed the case, this would change both the phonotactic and phonation analyses of the language. See also discussion in Footnote 2 of the paper about the ‘harsh-creaky’ words in Traill’s dictionary that are marked for breathy voice, pharyngealization, and creaky voice. I echo Naumann’s desire for more cross-dialectal and dialect-internal analyses of the phonotactics of the language, in particular with regards to word-medial consonants.