You are on page 1of 18

U. Alberta Working Papers in Linguistics, vol. 1, 2006.

Robert Kirchner (ed.)

THE EMERGENCE OF EJECTIVE FRICATIVES


IN UPPER NECAXA TOTONAC1
David Beck, University of Alberta
dbeck@ualberta.ca

Upper Necaxa Totonac (a.k.a. Patla-Chicontla) has an unusual phonemic inventory in that
it has a series of three ejective fricatives (including the rare /˚’/), but no ejective stops.
This paper examines the diachronic path that has lead to the current situation and dis-
cusses some of the phonetic and phonological characteristics.

Upper Necaxa Totonac (UNT) is a member of the isolate Totonac-Tepehua linguistic family
spoken by some 3,000 people living in and around three villages—Patla, Chicontla, and Caca-
huatlan—in the Necaxa River Valley in the northeastern part of the state of Puebla, Mexico.2
Phonologically, Upper Necaxa differs from its sister languages in a number of ways, the most
striking of which for the linguist is the presence of three ejective fricatives—/s’/, /ß’/, and /¬’/.
These segments are all the more unusual from a typological point of view in that UNT lacks an
accompanying set of ejective stops. Such a consonantal inventory is, to my knowledge, unprece-
dented in the world’s languages, yet examination of the historical-reconstructive data shows that
this highly unique situation emerged as a result of a sequence of rather mundane phonological

1
I’d like to thank David Bennett, Robert Kirchner, Ryan Klint, Peter Ladefoged, Paulette Levy, and Terry Nearey
for discussing this topic with me, Ian Catford for providing me with some much needed background on the problem,
and Jim Watters for generously sharing his Tepehua database. Thanks also to Ryan Klint for doing the dirty work
collecting the instrumental data (and for taking his turn hauling the heavy machinery back and forth from the field),
to Robert Kirchner for lending us the air-flow measurement equipment, and to Yvonne Lam for walking me through
the statistical analysis. Support for this research was provided by the Social Sciences and Humanities Research
Council of Canada, the Organization of American States, the Wenner-Gren Foundation, the University of Toronto,
and the University of Alberta. My heartfelt thanks go out to the people of Patla and Chicontla, Mexico, for their
warmth and hospitality, and to my consultants on this topic, Alvaro Barragán Alvarez, Roberto Barragán Alvarez,
Rosendo Melo Márquez, and Porfirio Sampayo Macín for their indulgence. Lah'alhu:matzá.
2
According to the Ethnologue database (Grimes 2004), Upper Necaxa (Ethnologue’s Patla-Chicontla [TOT]) is
also spoken in the neighbouring village of Tecpatlán. However, none of the sound changes described here apply to
the Tecpatlán variant of Totonac, which in my own experience has proven to resemble, both lexically and phonol-
ogically, the variant spoken in Zihuateutla, which Ethnologue groups with the Totonac of Apapantilla (Ethnologue’s
Xicotepec de Juárez [TOO]). Reference is also made to an unnamed fifth location, most likely the village of San
Pedro Tlalontongo just downriver from Chicontla. The linguistic situation here is less clear, given the moribund
state of the San Pedro variant and the difficulty in finding native speakers who are not intimately familiar with (and
able to speak in) the Chicontla dialect of UNT. Chicontla speakers, however, consistently report that San Pedro To-
tonac is “different” and cite lexical distinctions that group it with Filomena Mata ([TLP]) and other variants to the
south and east. Whether this warrants excluding San Pedro Totonac from UNT—or whether the sound changes de-
scribed here apply to this variant (most Chicontla speakers say they do not)—remains an open question.
2 U. Alberta Working Papers in Linguistics, vol. 1, 2006.

changes—specifically, by a sequence of historical sound-shifts changing Proto-Totonacan *q to


*÷ followed by a shift from Pre-UNT fricative (F) + /÷/ clusters to modern UNT ejective frica-
tives (*F÷ > F’).

In this paper, I will examine the UNT ejective fricative first from a historical and then from a
phonological perspective. In section 1, I explore the diachronic evidence for the proposed sound
changes by comparing UNT forms to data from other languages in the family, beginning in sec-
tion 1.1 with evidence for the shift from the uvular to the glottal stop (*q > *÷), and then turning
in section 1.2 to the subsequent change from consonant-cluster to ejectivized fricative. Having
established the diachronic path for the creation of these segments, I will turn in section 2 to the
equally important topic of their phonetic properties, with an eye towards establishing that the
process creating them is indeed a diachronic rather than a synchronic one, and that ejective frica-
tives in modern UNT are in fact distinct segments from F÷ clusters. In section 2.1, I will present
evidence to this effect drawn from an instrumental study including measurements of air-flow and
intra-oral pressure, and, in section 2.2, I discuss some of the phonological and phonotactic char-
acteristics of the UNT ejective fricatives.

1. ORIGINS OF UNT EJECTIVE FRICATIVES

The key to understanding the diachronic development of ejective fricatives in Upper Necaxa lies
in an examination of certain of the sound changes that have taken place in UNT to set it apart
from its congener languages. The distinctive phonological features of UNT are readily apparent
by comparing its full consonantal inventory, shown in Table 1, with that of a more typical To-
tonacan language, Papantla, shown in Table 2:

Table 1. Upper Necaxa consonants


bilabial alveolar lateral-alveolar post-alveolar velar glottal
stop p t k ÷
affricate ts tß
s ¬ ß
fricative x
s’ ¬’ ß’
approximate w l j
nasal m n Ñ

Table 2. Papantla consonants (based on Levy 1987)


bilabial alveolar lateral-alveolar post-alveolar velar uvular glottal
stop p t k q
affricate ts t¬ tß
fricative s ¬ ß h
approximate w l j
nasal m n N
Beck, The emergence of ejective fricatives in Upper Necaxa Totonac 3

One of the changes that sets UNT apart from all of the other languages of the family is the ab-
sence of the lateral affricate, /t¬/, which has neutralized with the voiceless lateral fricative, /¬/.
More significant for the purposes of this paper, however, is the loss of the uvular consonants, /q/
and [N] (an allophone of /n/ associated with /q/).3

In place of the pan-Totonacan /q/, UNT has a phonemic glottal stop whose distribution closely
parallels that of other stop-consonants. In other Totonacan languages, [÷] is either the phonetic
realization of the laryngealization of vowels (Levy 1987: 59), or a consonant phoneme with con-
strained or specialized distribution (word-final in nominals and as an epenthetic onset to underly-
ingly vowel-initial syllables—MacKay 1999; McQuown 1990). As shown both by comparative
reconstruction and by the synchronic behaviour of /÷/, the absence of /q/ and the presence of /÷/
in UNT are the result of a diachronic process leniting uvular stops to glottal stops (section 1.1).
This historical lenition of *q to ÷ also seems to explain the presence of the three ejective frica-
tives: comparison of UNT words containing ejective fricatives with cognates in other languages
suggests that these are the result of a diachronic sequence of sound changes beginning with the
initial shift of Proto-Totonacan *q > ÷, followed by the collapse of fricative plus glottal stop se-
quences to ejective fricatives, *F÷ > F’ (section 1.2).

1.1 Cognates of Upper Necaxa /÷/

The first step in the diachronic sequence leading to the emergence of the ejective fricatives in
UNT is the replacement of the voiceless uvular stop, /q/, with a glottal stop, /÷/. This change is
thorough-going (no words in UNT have /q/) and is recognized as a dialect marker by speakers in
the Necaxa Valley and surrounding regions. Obvious cognates of Upper Necaxa words with /÷/
are found in other Totonac-Tepehua languages. A few of these are given in Table 3 on the fol-
lowing page.4 As the data in Table 3 shows, the shift *q > ÷ occurred in all contexts where *q
was permitted by Totonacan phonotactics—between vowels (a, b), word-initially (c), word-
finally (d), and preceding non-plosive consonants (e, j). In Pisa Flores Tepehua, the same shift
seems to have occurred in some words between vowels (d – f) and, in a few cases, word-initially
(c).5 Evidence for the origin of UNT /÷/ in *q can also be seen in the assimilatory effects it has on
/i/ and /u/, lowering these to /e/ and /o/, respectively. UNT not only lacks stems with /i/ and /u/
adjacent to /÷/, but also has a morphophonemic process lowering high vowels in the presence of

3
The UNT inventory is also distinguished from that of Papantla by the presence of /x/ as opposed to /h/ (which is
the sound found also in Misantla—MacKay 1999—and Coatepec—McQuown 1990). This distinction, however,
seems to be shared with Apapantilla and other Totonacs spoken in the region of Xicotepec de Juárez, and may pro-
vide a basis for the recognition of a “Northern” Totonacan group (as may the—possibly related—phonemic status of
the vowels /e/ and /o/, see section 1.1 below). Although this group has enjoyed some currency in the literature—as,
for example, in MacKay (1999), MacKay & Trechsel (2004), and Beck (2000)—it is based more on impressionistic
evidence than hard reconstruction. P. Levy (p.c.) suggests that the origin of the Northern Totonac grouping can be
traced back to the S.I.L. lexicographer H. P. Aschmann.
4
Language data in this and subsequent tables are: UNT (Beck 2001, 2004); Apapantilla—Reid and Bishop (1974);
Papantla (Aschmann 1973); Zapotitlan (Aschmann 1983); Tepehua (Watters n.d.a, n.d.b).
5
According to an anonymous reviewer, the shift from /q/ to /÷/ is complete in Pisa Flores Tepehua and nearly com-
plete in Huehuetla, where /q/ persists only in the speech of the oldest speakers.

3
4 U. Alberta Working Papers in Linguistics, vol. 1, 2006.

Table 3. Cognates of Upper Necaxa /÷/


UNT Apapantilla Papantla Zapotitlan Tepehua*
a) ‘far’ ma÷át maÓqát máqat maqát Tl máqati
b) ‘sky’ aÓ÷apún aÓqapu…n aÓqapu…n aÓqapu…n Tl ÷aqapu…-
Tl qáhin
c) ‘turtle’ ÷ayán qayan qayin qayan PF ÷ahín
H q’ahín
Tl aqtßóÓqni
d) ‘freshwater shrimp’ a÷atßó÷ aÓqatßøq aqátßoq aÓqatßoÓqa
PF a÷atßóqni
e) ‘to bloom’ pa÷¬- paq¬- paq¬- paq¬- —
Tl ÷aqtsoqóqni
f) ‘knee’ tso÷ósni tsoqosnî tsoqosni — PF tso÷ótni
H tsoq’oti
Tl paqatßuh
g) ‘wing’ pe÷én peqen paqan paqan PF pa÷atßuh
H peqsti¬
h) ‘West Indian Elm’ a…÷é…ßtu a…qé…ßtu aÓqaßti — Tl ÷aqayßtah
i) ‘explode’ paÑ- panq- panq- panq- Tl paq-
PF mó…qßnu…
j) ‘owl’ mo…Ñßú móÓnqßuÓ monqßnú móÓnqßnuÓ
H mo…qßnu…
*Tl = Tlachichilco; PF = Pisa Flores; H = Huehuetla
/÷/, as in le…÷amá…n ‘toy’ (from li…- ‘INSTRUMENT’ + ÷amá…n ‘to play with/at’).6 Lowering of /i/
and /o/ by /q/ is attested in other Totonacan languages, although /e/ and /o/ are treated as allo-
phones of /i/ and /u/ in Papantla (Levy 1987), Coatepec (McQuown 1990), and Misantla
(MacKay 1999). In Apapantilla, on the other hand, Reid (1991) treats /e/ and /o/ as phonemic,
and the same seems to be appropriate—at least impressionistically—in other languages of the
Northern Sierra, more potential evidence for the proposed Northern grouping of Totonacan lan-
guages (see fn. 3 above). Most occurrences of the mid-vowels in UNT do seem to have their ori-
gins in lowering processes triggered by /÷/ or by the velar fricative /x/, though there are a number
of words (e.g. tsewaníÔ ‘pretty’) where it seems to be necessary to treat /e/ and (to a lesser extent)
/o/ as phonemes in their own right.

Another uvular-like phonological property that UNT /÷/ has is that it triggers the backing of /n/
to /Ñ/, as can be seen by the correspondences in Table 3 (i) and (j) between UNT /Ñ/ and general
Totonacan /nq/ (phonetically [Nq]—Levy 1987: 53; MacKay 1999: 58). In Upper Necaxa, /Ñ/ is
generally an allophone of /n/ found either in a syllabic coda within a stem adjacent to an audible
glottal stop (e.g. putséÑ÷e ‘black’, spinéÑ÷e ‘red’), or in similar environments created by mor-
phophonemic processes (e.g. kiÑ÷ayán ‘my turtle’ from kin- ‘my’ and ÷ayán ‘turtle’, paÑ- ‘to ex-
plode’ > páÑ÷a ‘it explodes’ vs. páÑliÔ ‘it exploded’). There are a few instances of UNT /Ñ/ in co-
das inside of stems (corresponding to /nq/ in sister languages) without an audible /÷/ (e.g. aÓ÷óÑßa

6
Note that this process seems to be on its way out with younger speakers, many of whom will spontaneously pro-
duce forms such as li…÷amá …n or even li…amá…n, though they do so inconsistently and seem to use the innovative
forms as free variants of the forms with the lowered vowel. Speakers of the Patla dialect are less consistent about
lowering vowels in these contexts than are speakers from Chicontla.
Beck, The emergence of ejective fricatives in Upper Necaxa Totonac 5

‘she braids her hair’).7 The fact that the historical *nq cluster is realized as a velar, rather than a
uvular, nasal likely indicates that these segments are reflexes of a historical nasal-backing proc-
ess rather than the result of a synchronic assimilatory rule (which would result in [N] rather than
[Ñ]). The realization of the backed nasal as a velar seems to be part and parcel of the wholesale
elimination of uvular segments from the inventory, the *N > /Ñ/ shift and the shift from *q to /÷/
effectively removing the uvular place of articulation from modern UNT phonology.

Taken together with the comparative evidence, these phonological effects of UNT /÷/—vowel-
lowering and the backing of /n/—point strongly to the historical source of the Upper Necaxa
glottal stop being Proto-Northern Totonac *q. This represents a fairly standard process of leni-
tion attested in a variety of languages around the world (e.g., Cairene Arabic, which has replaced
Classical Arabic /q/ with /÷/—Mitchell 1990: 54). As noted above, the same process also seems
to be found in Tepehua, although how thorough-going this change has been will have to await
further documentation.

1.2 Cognates of Upper Necaxa F’

From a typological point of view, the most remarkable thing about the Upper Necaxa phonemic
inventory is the presence of a series of three ejective fricative phonemes in the absence of any
ejective stops.8 As shown in Table 4 on the next page, UNT /s’/, /ß ’ /, and /¬’/ correspond to /sq/,
/ßq/, and /¬q/ sequences in other Totonac-Tepehua languages, linking their presence to the same
process of lenition that gave rise to a phonemic /÷/ from historical *q. Although the comparative
data here is somewhat scarcer than that for the plain *q > ÷ shift, these and a dozen or so addi-
tional examples of cognates from two or more Totonacan languages show a clear distributional
parallel between ejective fricatives (F’) in UNT and Fq clusters in sister languages. It is also
suggestive that UNT F’ has the same vowel-lowering properties as /q/ (b, e, f, g), further evi-
dence of their historical source in *Fq clusters, which are known to lower vowels in other To-
tonacan languages.

Unlike the lenition of *q to ÷, the shift from *F÷ to F’ does not seem to be a widely-attested
one cross-linguistically, although “fusion” processes that produce ejective stops from stop + glot-
tal stop (T÷) sequences seem to be relatively frequent, both synchronically and diachronically
(Fallon 2002). Synchronic phonological fusion creating ejective stops is reported in Kashaya
(Buckley 1994), Bella Coola (Nater 1984), Yurok (Robins 1958), Athapaskan (Rice 1994), and

7
Interestingly, nasals seem to be transparent to the vowel-lowering effect of /q/ and /÷/, as shown in Table 3(j), ex-
plaining why no UNT roots containing /Ñ/ have high vowels in adjacent syllables.
8
However, ejective stops have been reported in two other language in the family. According to García Ramos
(1979), there are ejective stops in Papantla, although Paulette Levy (p.c.) expresses some skepticism on the issue
and, impressionistically, I have not been able to hear any strongly ejectivized sounds either in Levy’s field record-
ings or in my one or two brief exposures to speakers of this variety of Totonac. Watters (1988) also posits ejective
stops for Tlachichilco Tepehua, although the origins of these (like the ejectives proposed for Papantla by García
Ramos) appear to be sequences of a stop followed by a short laryngealized vowel. In UNT such sequences can give
rise to very weakly ejectivized velar stops in certain environments with some speakers (e.g., [wak÷áÓ¬] from wakáÓ¬
‘be high’).

5
6 U. Alberta Working Papers in Linguistics, vol. 1, 2006.

Table 4. Cognates of Upper Necaxa F’

UNT Apapantilla Papantla Zapotitlan Tepehua


Tl. s÷at’a*
a) ‘small; baby’ s’áÓtaÓ sqáÓtaÓ sqaÓtaÓ sqaÓtaÓ
H. sq’at’a
b) ‘lamp’† pu…maÓ÷s’ót pu…maqsqo pu…sqon pu…maqsqo —
c) ‘corn husk’ ß’a…m ßqa…m ßqam ßqa…m Tl. ßqa…n
d) ‘to yawn’ ta÷e¬ß’apáÓ÷- taqa¬ßqaÓpaÓ:ta aqßqaÓpa…tnan taßqapa…ta —
e) ‘iguana’ ß’o…lúluÓ ßqoÓluÓluÓ ßqo…lúlu ßqo…luÓluÓ Tl. ßqu…lú¬
¬’oyú… ¬qoyunún ¬qoyú ¬qoyunan
f) ‘burn, cook’‡
‘burn trash’ ‘burn’ ‘cook’ ‘bake in kiln’
qa¬qaÓwi
g) ‘gums’ ÷e¬’éwiÔ qa¬qeÓwiÔ — —
‘jawbone’
h) ‘crooked’ ¬’awíliÔ ¬qaÓwiÔliÔ ¬qaÓwili — —
*Tl = Tlachichilco; H = Huehuetla

Also Misantla squnah ‘warm’ (MacKay 1999: 65).

Also Misantla ¬quyunan ‘s/he fires clay’ (MacKay 1999: 66).

Klamath (Barker 1964).9 All of these languages also have underlyingly ejective phonemes. Syn-
chronic fusion processes that create ejective stops and ejective fricatives are found in Barbareño
Chumash (Beeler 1976) and Tsou (Ladefoged & Zeitoun 1993).

Evidence for diachronic fusion of obstruent – glottal stop sequences is harder to come by,
though oblique reference to C÷ sequences being the historical source of ejective consonants is
made from time to time in the literature (e.g. Greenberg 1970). Fusion is posited as the origin of
ejective stops in Washo (Jacobsen 1977), Nitinaht and Makah (Jacobsen 1969), Alaskan Haida
(Levine 1981), Southern Pomo (McLendon 1973, 1976), Caddo (Chafe 1976), Hausa (Skinner
1975), and Circassian (Colarusso 1989). According to Blust (1980), glottalized consonants in
Yapese (including ejective stops) seem to have arisen from a process collapsing *CVC syllables
first to *CC clusters and then to C’. Of particular interest here are some examples involving *tVq
sequences:

(1) *bu(r)teq > buut’


*mataqu > mat’aaw
*taqi > t’aay
(Blust 1980: 152)

Presumably, these forms would have passed through a *tq > *t÷ > t’ progression analogous to
that proposed here for Pre-Upper Necaxan *Fq—*Fq > *F÷ > F’. The fact that this process did
not give rise to ejective stops in UNT as it did in Yapese follows from Totonacan phonotactics,

9
Fallon (2002) also reports other types of synchronic phonological processes creating ejectives from plain obstru-
ents in Acoma, Coeur d’Alene, Colville, Cowichan, Huamelultec Chontal, Jacaltec, Kalispel, Makah, Mazahua
Otomí, Nez Perce, Nootka, Okanagan, Pame, Takelma, Tzetzal, Tzutujil, Zuni, Aaari, Rutul, and Mingrelian.
Beck, The emergence of ejective fricatives in Upper Necaxa Totonac 7

which rule out stop-stop (TT) clusters (McQuown 1990: 63 – 64; Levy 1987: 83 – 87; MacKay
1999: 63 – 67), removing the input (*Tq) for a process creating ejective stops.

2. PHONETIC AND PHONOLOGICAL PROPERTIES OF UNT EJECTIVE FRICATIVES

Having established the diachronic path for the development of the segments transcribed in Upper
Necaxa Totonac as ejective fricatives, in this section I will turn to their phonetic and
phonological properties in order to show that they are, in fact, underlyingly ejective segments
rather than F÷ clusters or F÷ clusters “collapsed” by a synchronic process of fusion. Although the
ejective fricatives are fairly distinctive on an impressionistic level, sounding “sharper” than ordi-
nary fricatives and being accompanied by visible glottal-raising, reliably identifying the seg-
ments instrumentally is a somewhat more difficult task. There is relatively little literature avail-
able on the phonetic properties of ejective fricatives—not surprisingly, given the rarity of these
segments in the world’s languages. What literature there is (e.g., Maddieson, Bessell, and Smith
1996; Maddieson 1998) suggest that the most significant features of ejectivization of fricatives
can be found in measurements of peak oral air flow and intra-oral pressure. With this in mind,
the author and his research team took an air-flow mask to the field in the summers of 2003 and
2004 and made measurements of two adult male speakers of Upper Necaxa Totonac reading
from prepared lists of forms (seven tokens each of 3 words containing /¬’/, 3 words containing
/s’/, four words containing /ß’/, and 5 words containing /ß÷/ clusters). The data was then examined
by the author using MacQuirer 6.0 airflow analysis software and Praat 4.3, and statistics were
run on the data using SPSS. The results of these analyses, presented in section 2.1, show that
measurements of the timing of the peak in intra-oral pressure and the time differential between
this peak and the peak in oral air flow, as well as the overall duration of frication, clearly distin-
guish ejective fricatives both from plain fricatives in F÷ clusters and from plain fricatives fol-
lowed by vowels. In addition, comparison of the duration of the hiatus between the end of frica-
tion and the onset of voicing in the following vowel for ejective fricatives and F÷ clusters also
revealed statistically significant differences. Following the discussion of the instrumental data, I
take up the phonological and phonotactic properties of the ejective fricatives in section 2.2, out-
lining some issues raised by their special phonotactics.

2.1 Phonetic properties of UNT ejective fricatives

Because ejective (glottalic egressive) segments are created using a pulse of air driven by the
abrupt raising of the closed glottis (Ladefoged 2001), one of their principle aerodynamic charac-
teristics should be a peak in intra-oral pressure near the beginning of the articulation. For ejective
stops, Maddieson (1998) shows that /p’/ in Yapese has a sharp peak in intra-oral pressure imme-
diately before the release of the consonant and the beginning of oral air flow, and that the intra-
oral pressure for this segment is considerably higher than for non-ejectivized segments. High in-
tra-oral air pressure is also reported by Maddieson, Bessell, and Smith (1996) for ejective frica-
tives in Tlingit; these authors also note that ejective fricatives in Tlingit are followed by a period
of near-silence which set them off from the beginning of the next segment. In addition, Maddi-
eson, Bessell, and Smith (1996) and Maddieson (1998) both suggest that ejective fricatives

7
8 U. Alberta Working Papers in Linguistics, vol. 1, 2006.

should be distinguishable from plain fricatives in terms of their overall duration, pulmonic frica-
tives being sustainable over a longer period than fricatives created by glottal-raising.

For the purposes of this paper, measurements of peak intra-oral pressure, the timing of peak in-
tra-oral pressure and peak oral air flow, and total duration of frication were made on multiple
tokens of one example each of an ejective /ß’/, an /ß/ in /ß÷/ clusters, and a plain /ß/ in a contrastive
environment (between the vowels /i/ and /a/).10 Because both F’ and F÷ clusters are followed by
an audible hiatus, measurements were taken of the gap between the end of frication and the VOT
of the following vowel for the words containing these sounds as well. These measurements show
that on three of the four counts—timing of intra-oral pressure, timing of peak oral air flow, and
duration—the ejective fricatives were distinct from both fricatives in F÷ clusters and from plain
fricatives. Ejective fricatives were also shown to be distinct from F÷ clusters in terms of the
length of hiatus preceding VOT. The peak value of intra-oral pressure was found not to be sig-
nificant (in all three segment types peak values ranged between 7 and 9 cm H20).

The typical phonetic properties of ejective fricatives found in our study are shown in Figure 1:

Figure 1. miß’á…m ‘your cornhusks’

This diagram shows the cumulative results of spectral analysis and measurements of airflow and
intra-oral pressure for one token of the word miß’á…m ‘your (mi-) cornhusks (ß’a…m)’. In terms of
air flow, the fricative shows a peak 22 ms after the onset of the fricative. The intra-oral pressure
also spikes at the beginning of the fricative, but peaks about 10 ms before the peak in air flow (8
cm H2O at 12 ms). The total duration of the fricative is 112 ms (this is the shortest recorded to-
ken, the mean duration of ejectives in the sample is 143 ms), and there appears to be very little
hiatus (0.5 ms) between the end of frication and the onset of voicing in the following vowel.

10
This was only done for /ß’/ because this is the only of the three ejectives for which tokens of the corresponding
F÷ cluster was collected. We plan to collect some tokens of /¬÷/ at some point in the future, but /s÷/ is unattested,
there being no productive s-final prefixes.
Beck, The emergence of ejective fricatives in Upper Necaxa Totonac 9

These features are quite distinct from those of an /ß÷/ cluster, illustrated in Figure 2, a recording
of iß÷awátßa ‘his/her (iß-) son (÷awátßa)’:

Figure 2. iß÷awátßa ‘his son’

Here we see the air flow peak 16 ms after the beginning of the fricative, considerably sooner than
for the ejective. The intra-oral pressure, on the other hand, peaks at a higher level (9 cm H20) at
28 ms, 12 ms after the peak in air flow. The total duration of the fricative is 90 ms, less than that
of the ejective, and it is followed by a hiatus of 4 ms.

Although no deliberate effort was made to collect airflow data on tokens of plain fricatives,
one of the elicitation forms did contain a non-ejective /ß/ followed by a vowel, as in Figure 3:

Figure 3. ißaÓ÷aß’ó¬ ‘his/her ear’

9
10 U. Alberta Working Papers in Linguistics, vol. 1, 2006.

In the plain fricative, the air flow shows a similar pattern to that of the fricative in the /ß÷/ cluster,
peaking 18 ms after the beginning of frication; the intra-oral pressure peaks 11 ms later (7 cm
H2O at 29 ms), showing the same relative timing of air-flow peak and intra-oral pressure as the
fricative shown in Figure 2. The total duration of frication is 104 ms and it is followed by a short
hiatus (7 ms) due to following laryngealized vowel. The duration of the plain fricative here con-
trasts with the duration of the ejective fricative in the same wordform (140 ms) and that in Figure
1 (112 ms), but is comparable to the duration of the fricative shown in Figure 2 (90 ms).

As mentioned above, there are three significant measurements in this data that seem to distin-
guish the ejective fricatives from fricatives in /ß÷/ clusters and plain fricatives—timing of the
peak in intra-oral air pressure from the beginning of frication (IOP), time difference between the
peak in air flow and IOP (AF-IO), and duration of the segment (D). These are summarized in
Table 5 for the recorded tokens from one speaker of miß’á…m, iß÷awátßa, and ißaÓ÷aß’ó¬:11

Table 5. Means for IOP, AF-IO, and D

N Mean Std. Deviation Std. Error


ejective 7 19.86 9.737 3.680
IOP (ms) cluster 6 34.33 8.189 3.343
plain 6 33.17 8.060 3.291
ejective 7 4.71 4.386 1.658
AF-IO (ms) cluster 6 -13.00 6.723 2.745
plain 6 -5.67 7.941 3.242
ejective 7 143.00 24.617 9.304
D (ms) cluster 6 101.50 7.342 2.997
plain 6 96.00 14.128 5.768

The means in this table show clearly that the ejective fricatives differ on all counts from the clus-
ters and the plain fricatives, and that the latter, as expected, pattern together. This observation is
confirmed by performing One-way ANOVA tests on these samples, giving the results summa-
rized in Table 6:

Table 6. One-way ANOVA for IOP, AF-IO, and D scores

Sum of Squares df Mean Square F Sig.


Between Groups 857.397 2 428.699
IOP Within Groups 1229.024 16 76.814 5.581 .014
Total 2086.421 18
Between Groups 1033.764 2 516.882
AF-IO Within Groups 656.762 16 41.048 12.592 .001
11
Total 1690.526 18
Note that although
Betweenseven tokens were recorded
Groups of each word, only
8747.447 2 six each were used for the cluster and the
4373.724
plain fricative. The first recorded token of iß÷awátßa from this speaker was mispronounced as [iß’awátßa], while the
D token of the
first Within Groupsdid not show a discrete
plain fricative 4903.500 16 in airflow
measurable peak 306.469 14.271 .000
during frication.
Total 13650.947 18
Beck, The emergence of ejective fricatives in Upper Necaxa Totonac 11

According to these tests, the difference in IOP among all groups is significant (F(2,16)=5.581,
p=.014, main effect for IOP among all three groups), as are both the difference in AF-IO among
all groups (F(2,16)=12.592, p=.001, main effect for AF-IO among all three groups) and the dif-
ference in D among all groups (F(2,16)=14.271, p=.000, main effect for D among all three
groups). In order to identify which group or groups was the source of the differences found by
the One-way ANOVA test, a Tukey post-hoc test was carried out to provide for pair-wise com-
parison among groups, as shown in Table 7:

Table 7. Tukey post-hoc test for IOP, AF-IO, and D

(I) (J) Mean Difference (I-J) Std. Error Sig.


cluster -14.48* 4.876 .023
ejective
IOP plain -13.31* 4.876 .037
cluster plain 1.17 5.060 .971
cluster 17.71* 3.564 .000
ejective
AF-IO plain 10.38* 3.564 .026
cluster plain -7.33 3.699 .149
cluster 41.50* 9.740 .002
ejective
D plain 47.00* 9.740 .001
cluster plain 5.50 10.107 .851
*p < .05

Pair-wise post-hoc comparisons using a Tukey test revealed that the difference in IOP is signifi-
cant between ejectives and clusters (p=.023) and between ejectives and plain fricatives (p=.037),
but not between clusters and plain fricatives (p=.971). Similarly, significant differences in AF-IO
were found between ejectives and clusters (p=.000) and between ejectives and plain fricatives
(p=.026), but not between clusters and plain fricatives (p=.149). The same was found for D,
where there were significant differences between ejectives and clusters (p=.002) and between
ejectives and plain fricatives (p=.001), but not between clusters and plain fricatives (p=.851). As
with the previous tests, these results clearly show that the ejectives are readily distinguishable on
all three measures from plain fricative and fricatives in /ß÷/ clusters, whereas the fricatives in
clusters and plain fricatives are indistinguishable on these counts.12

Measurements of the hiatus (H) between the end of frication and the onset of voicing for the
following vowel were also made for the ejectives and the clusters. The plain fricatives were ex-
cluded from this test because in the sample they are followed by a short layrngealized vowel, /aÓ/,

12
Because of the small sample sizes here, the data were also checked with a non-parametric Mann-Whitney U-test,
which confirmed the statistical significance of all comparisons between ejectives and clusters (p=.022, p=.000,
p=.000) and ejectives and plain fricatives (p=.045, p=.010, p=.000), as well as the non-significance of the distinction
between clusters and plain fricatives (p=.872, p=.198, p=.630).

11
12 U. Alberta Working Papers in Linguistics, vol. 1, 2006.

which is frequently realized with a brief initial glottal closure ([÷aÓ] or [÷a], depending on the envi-
ronment). The means for the sample are shown in Table 8:13

Table 8. Means for H

N Mean Std. Deviation Std. Error


ejective 7 .943 .6024 .2277
H (ms)
cluster 5 3.44 1.7516 .7833

As the table shows, the H associated with the cluster (3.44 ms) is considerably shorter than that
associated with the ejective fricative (.943 ms). Running an independent samples T-test on this
data shows that this difference is statistically significant (t(10)=-3.548, p=.005). These results
seem to be in line with the analysis that the hiatus following the ejective fricative is the result of
articulatory phonetic effects (an interruption of airflow at the point where the closed glottis
reaches its maximum height), whereas the hiatus in the clusters is the realization of a phonemic
glottal stop.

The fact that the measurements relating to intra-oral pressure, IOP and AF-IO, distinguish ejec-
tives from non-ejective fricatives seems to follow rather naturally from the differing source of
the pressure in the oral cavity. In the case of ejectives, this pressure would come from the raising
of the closed glottis, which presumably would be timed with the raising of the tongue to create a
high-pressure region in the oral cavity in order to initiate frication. The intra-oral pressure would
build fairly quickly and peak at the beginning of the articulation, while the air flow resulting
from it would lag behind. Intra-oral pressure in the plain fricatives, on the other hand, would be
the result of the constriction created behind the tongue as it is raised towards the roof of the
mouth. Frication would begin before the point of maximum constriction, delaying the peak in
intra-oral pressure relative to the onset of frication.

On the other hand, the fact that the third measure, duration of frication, plays out in the way it
does is a bit counter-intuitive. As Maddieson (1998) notes, one might normally expect ejectiv-
ized fricatives to be shorter than plain ones due to the limited time it is possible to raise the glot-
tis and maintain glottal closure. In this data, however, it seems that ejective fricatives are signifi-
cantly longer than the non-ejectives both in clusters and when followed by a vowel. What is es-
pecially surprising about this data, however, is not the relatively long duration of the ejective
fricatives (mean 143 ms—cf. the mean of 148 ms reported by Maddieson, Bessell, and Smith
1996 for all ejective fricatives in Tlingit), but the relatively short mean duration of the plain
fricatives (102 ms in clusters and 96 ms followed by vowels—again, cf. the mean duration of
pulmonic fricatives in Tlingit, 222 ms). This fact helps to explain another discrepancy between
the data presented here and the Tlingit data reported in Maddieson, Bessell, and Smith (1996)
having to do with the contour of the peak in intra-oral pressure. In Tlingit ejective fricatives, in-
tra-oral pressure peaks and then falls abruptly, whereas the intra-oral pressure in pulmonic frica-

13
Note that a further token was removed from the cluster sample due to an unusually long hiatus (90 ms), probably
due to speaker hesitation.
Beck, The emergence of ejective fricatives in Upper Necaxa Totonac 13

tives shows a flattened peak and remains elevated for a relatively longer period during frication.
In UNT, on the other hand, there seems to be relatively little difference between ejectives and
plain fricatives in the way intra-oral pressure drops following the peak, a fact which can be ac-
counted for by the relatively short duration of the plain fricatives, which precludes an extended
period of intra-oral pressure.14 At the present, however, I have no explanation for the relatively
short duration of pulmonic fricatives itself; however, resolving this issue seems crucially to hinge
on examination of data from plain fricatives in a wider variety of environments, and will have to
be left for future study.

2.2 The phonology and phonotactics of UNT ejective fricatives

Although the phonetic evidence for the presence of ejective fricatives in UNT is quite clear, de-
finitive phonological evidence—that is, argumentation based on phonotactics or phonological
processes—is somewhat harder to come by. Indeed, when considered from the point of view of
phonotactics, the distribution of the ejective fricatives more closely resembles that of FT clusters
than it does the distribution of other fricatives. The search for phonological evidence for the dis-
tinctiveness of F’ from F÷ in the modern language is further complicated by the dearth of con-
trastive environments for the two types of sounds: the diachronic fusion process has eliminated
all instances of F÷ from roots, and there is only a limited number of environments where F÷ clus-
ters are created by morphological processes. Because these all involve prefixation, and word-
level stress in UNT falls at the end of the word, stress-assignment rules offer no clues as to the
effects of syllabification processes on F’ vs F÷ clusters, although impressionistic evidence drawn
from careful speech (produced for a linguist struggling to transcribe words containing unfamiliar
sounds) does distinguish the two cases.

The canonical syllable structure of UNT is CV(C), with the possibility of having clusters of up
to two consonants in onset and coda position.15 The permissible patterns of consonant groupings
in onsets and codas are shown in Figure 4 on the next page. Although there are a few vowel-
initial prefixes and lexical items, syllables in UNT preferentially have an onset which may con-

14
Another difference between the UNT data and Maddieson, Bessell, and Smith’s (1996) data from Tlingit con-
cerns the differences in peak values of intra-oral pressure between plain and ejective fricatives. In Tlingit, the dis-
tinction is a marked one, with ejective fricatives having almost double the value for intra-oral pressure than that
measured for pulmonic fricatives (17.0 cm H2O mean for /s’/ vs 8.6 cm H2O for /s/). In contrast, the peaks in intra-
oral pressure in UNT ejectives in the sample examined for this study seem closer in value to those reported for pul-
monic fricatives in Tlingit and, as noted above, did not show any significant statistical differences from intra-oral
pressure in UNT clusters and plain fricatives. This may mean that UNT ejective fricatives are much more “weakly”
ejectivized than are ejective fricatives in Tlingit, though it would be preferable to make direct comparisons of identi-
cal segments between the two languages (Maddieson, Bessell, and Smith base their data on measurements of /s’/ and
/¬’/, whereas the UNT measurements are of /ß’/). It would also be useful to draw on a larger sample of tokens and
speakers from the UNT data (the present study includes 7 tokens from a single speaker), particularly given that there
was a great deal of variation reported between individual speakers in the Tlingit study. This will have to be left for
future investigation.
15
Much of the discussion of UNT phonotactics here is drawn from Kirchner and Varelas (2002), which contains a
survey of the possible consonant clusters and their distributions, although in their paper the authors have organized
the data in terms of word-initial, word-final, and medial positions rather than in terms of syllable structure as I have
done for the present discussion.

13
14 U. Alberta Working Papers in Linguistics, vol. 1, 2006.

C V
C† C= consonant
F= fricative

FC* TF* N = nasal


T = stop

TF* NF* *not F’, not ÷


†not F’

Figure 4. UNT syllable structure

sist of a single consonant, a fricative + consonant cluster, or a stop + fricative cluster. Codas can
be empty (not shown) or can consist of a single syllable, a stop + fricative cluster, or a nasal +
fricative cluster. The major difference between this template and syllable structure in Papantla
(Levy 1987) and Coatepec (McQuown 1990) is that both of these other languages allow NkF and
NqF clusters in coda position, whereas UNT has only NF clusters.16 This difference seems to
stem from the fact that historical Nq clusters in UNT have become simply /Ñ/ unless the original
glottal stop is preserved by a syllable boundary (i.e., by appearing in the onset of the following
syllable), reducing *NqF sequences simply to ÑF. In *NkF clusters, the nasal would have assimi-
lated to the following velar place of articulation, and modern UNT seems, not unreasonably, to
have neutralized the distinction between syllables with ÑF and ÑkF in coda position.

What is most striking about the template in Figure 4 from the point of view of the ejective
fricatives is that their distribution is restricted to one position, syllabic onset, and they can not be
combined with any other consonant in this position. This distribution exactly parallels the distri-
bution of FC clusters. Of course, given the relatively recent nature of the shift from *F÷ to mod-
ern F’, it hardly seems surprising that the distribution of the modern segments reflects that of
their historical source. Indeed, if the syllable structure of pre-UNT was the same as that shown
for the present-day language in Figure 5, the absence of ejective fricatives in any position but
syllabic onset is entirely predictable. In order for an ejective fricative in modern UNT to appear
as part of an FC onset cluster, it would have to have had its origin in a historical F÷C onset clus-
ter, something ruled out by the template. Similarly, for an ejective to appear in a TF’ onset clus-
ter, the source would have had to have been an illegal TF÷ onset. As far as codas go, UNT
phonotactics would appear to rule out clusters ending in anything but a fricative—once again
removing any potential diachronic source for an ejective fricative in any kind of coda.

Of course, it might also be possible to argue that the segments analyzed as ejective fricatives in
this paper are, in fact, the creations of a process of synchronic, rather than diachronic, fusion, and
that in modern UNT they remain underlyingly sequences of a fricative and a glottal stop. This

16
Levy (1987: 84) also reports kFC (and /ktr/) clusters in onset position in Papantla, but notes that these occur only
on words prefixed with the first-person subject morpheme ik- (reduced to [k-] in casual speech). The same is also
true of UNT, where the initial /k/ of such clusters tends to re-syllabify (often across a word-boundary) leftward if
preceded by a vowel.
Beck, The emergence of ejective fricatives in Upper Necaxa Totonac 15

argument is made even more tempting by the fact that, in roots, there are no phonemic contrasts
between F’ and F÷, the latter having been eliminated by the fusion process in question. The main
difficulty with this position is, of course, that it seems at odds with the phonetic data presented in
section 2.1, which show that there is a distinction maintained between underlying F’ and F÷ se-
quences created by morphological processes. While there are relatively few processes that create
such sequences (there are only three productive prefixes that end in fricatives—iß- ‘his/her’, ÷e¬-
‘mouth (interior)’, and ki¬- ‘mouth (exterior)’), the fact that they exist and that a rather subtle
phonetic distinction is maintained in these cases argues strongly for the recognition of ejective
fricatives as distinctive segments in the UNT phonological inventory.

An additional phonological argument for treating the ejectives as being underlyingly single
segments comes from syllabification. Because of the UNT preference for CVC syllable structure,
FC clusters in the onset of a non-initial syllable will be re-syllabified leftward when preceded by
a syllable without a coda:

(2) místu ‘cat’ [«mis.tu]


pá¬ka ‘griddle’ [«pa¬.ka]
taßtú ‘leave’ [taß.«tu]

This holds true for mono-morphemic roots such as místu and pá¬ka, as well as for complex stems
such as taßtú, formed from the inchoative prefix ta- and the bound stative root -ßtu. Here, the re-
syllabification takes place even across a morpheme boundary. This contrasts very clearly with
the behaviour of ejectives preceded by a vowel-final prefix, as shown in (3):

(3) tatu…s’á… [ta.tu….«s’a…]


ta–tu…–s’á…
INCHOATIVE–foot–split
‘split one’s foot open’

In this example, the verb stem s’á… ‘split open’ takes a vowel-final partonymic prefix tu…-; the
root-initial ejective fricative, however, does not re-syllabify as /ß.÷/, as we might expect if it were
underlyingly an F÷ cluster. As shown by the example of taßtú in (2), the morpheme boundary can
not be invoked to block cluster re-syllabification. This is an important point because in order to
maintain the position that the ejectives are created by a synchronic process of fusion from under-
lying clusters, it would be necessary to rely upon a morpheme boundary or some other morpho-
phonemic process to block fusion in those cases where F÷ clusters are created by affixation. In
these situations, as we saw in section 2.1, the distinction between F’ and F÷ clusters is main-
tained, and any attempt to posit a specific rule to block fusion in just these environments seems
highly stipulative, particularly given the lack of any positive phonological evidence in favour of
treating the ejectives as underlying clusters. In the final analysis, it seems that it is preferable to
treat the ejectives fricatives as underlyingly just that—ejective fricatives—and to account for
their restricted distribution in terms of their historical origins.

15
16 U. Alberta Working Papers in Linguistics, vol. 1, 2006.

3. CONCLUSION

This paper has presented comparative and phonetic evidence for the existence of a series of three
ejective fricatives—/s’/, /ß’/, and /¬’/—in Upper Necaxa Totonac. Measurement of duration, air-
flow, and intra-oral pressure confirm that these segments are ejectivized and that they are distinct
from plain fricatives and from clusters of fricatives followed by glottal stops; measurements of
the duration of the hiatus following ejective fricatives as compared to the hiatus following the
fricative in F÷ clusters is consistent with the analysis that the hiatus in the second of these cases
is the realization of a phonemic glottal stop. Although the distribution of the ejective fricatives
remains largely that of the F÷ clusters that gave rise to them, impressionistic evidence from syl-
labification gives some support to the notion that they are indeed single segments in the modern
language.

The comparative evidence points to Upper Necaxa F’ arising through a fairly clear diachronic
sequence:

Fq > F÷ > F’

Each step in the chain, the lenition of *q to *÷ and the collapse of *F÷ to F’, is in itself unre-
markable, as is the type of consonantal inventory we would want to posit for Proto-Totonacan.
The absence of ejective stops from the UNT phonemic inventory can be attributed to a simple
Proto-Totonacan phonotactic restriction—a constraint against *Tq clusters—that is still found in
most if not all the extant daughter languages and which is in and of itself completely run-of-the-
mill. Nevertheless, in Upper Necaxa Totonac, the net result of the application of two relatively
ordinary phonological processes to an unremarkable phonological inventory has resulted in a
phonemic system that is, to my knowledge, unattested in any other language of the world.

REFERENCES

ARANA, EVANGELINA. 1953. Reconstrucción del prototonaco. In Ignacio Bernal (ed.), Huastecos,
totonacos y sus vecinos. Revista mexicana de estudios antropológicos no. 23, pp. 123 – 30.
ASCHMANN, HERMAN P. 1946. Totonaco phonemes. International Journal of American Linguis-
tics 12, 34 – 43.
———1973. Diccionario totonaco de Papantla. Mexico: Summer Institute of Linguistics.
———1983. Vocabulario totonaco de la Sierra. Mexico: Summer Institute of Linguistics.
BARKER, MUHAMMAD ABD-AL RAHMAN. 1964. Klamath grammar. University of California Pa-
pers in Linguistics no. 32. Berkeley: University of California Press.
BECK, DAVID. 2000. The syntax, semantics, and typology of adjectives in Upper Necaxa Toto-
nac. Linguistic Typology 4, 213 – 50.
———2001. Primer vocabulario del idioma totonaco del Río Necaxa. Edmonton: University of
Alberta.
———2004. Upper Necaxa Totonac. Munich: LINCOM Europa.
Beck, The emergence of ejective fricatives in Upper Necaxa Totonac 17

BEELER, MADISON S. 1976. Barbareño Chumash: A farrago. In Margaret Langdon and Shirley
Silver (ed.), Hokan Studies: Papers from the First Conference on Hokan languages, pp. 251
– 69. Janua linguarum, series practica, no. 181. The Hague: Mouton.
BLUST, ROBERT. 1980. More on the origins of glottalic consonants. Lingua 52, 125 – 56.
BUCKLEY, EUGENE. 1992. Theoretical aspects of Kashaya phonology and morphology. Disserta-
tions in Linguistics. Stanford: CSLI Publications.
CHAFE, WALLACE L. 1976. The Caddoan, Iroquoian, and Souian languages. Trends in Linguis-
tics: State-of-the-Art Reports, no. 3. The Hague: Mouton.
COLARUSSO, JOHN. 1989. Proto-Northwest Caucasian (or how to crack a very hard nut). In How-
ard I. Aronson (ed.), The non-Slavic languages of the USSR, Linguistic Studies, pp. 20 – 55.
Chicago: Chicago Linguistic Society.
FALLON, PAUL D. 2002. The synchronic and diachronic phonology of ejectives. Routledge Out-
standing Dissertations in Linguistics Series. New York: Routledge.
GARCÍA RAMOS, CRESCENCIO. 1979. Fonología del totonaco del Tajín, Veracruz. Cuadernos An-
tropológicos (Veracruz) 2, 133 – 76.
GARCÍA ROJAS, BLANCA. 1978. Dialectología de la zona totonaco-tepehua. Honours thesis: Na-
tional School of Anthropology and History, Mexico.
GREENBERG, JOSEPH. 1970. Generalizations concerning glottalic consonants. International Jour-
nal of American Linguistics 36, 123 – 45.
GRIMES, BARBARA F. 2004. Ethnologue: Languages of the World. On-line database
(http://ww.ethnologue .com). Summer Institute of Linguistics.
ICHON, ALAIN. 1990 [l973]. La religión de los totonacos de la Sierra. Mexico City: Instituto Na-
cional Indigenista.
JACOBSEN, WILLIAM H., JR. 1969. Origin of the Nootka pharyngeals. International Journal of
American Linguistics 35, 125 – 53.
———1977. A glimpse of the Pre-Washo pronominal system. In Kenneth Whistler, Robert D.
van Valin, Jr., Chris Chiarello, Jeri J. Jaeger, Miriam Petruck, Henry Thompson, Ronya
Javkin, and Anthony Woodbury (eds.), Berkeley Linguistics Society no. 3, pp. 55 – 73. Ber-
keley: Berkeley Linguistics Society.
KIRCHNER, ROBERT, and ELENI VARELAS. 2002. A cue-based approach to the initial and final
phonotactics of Upper Necaxa Totonac. In Leora Bar-el, Linda Watt and I. Wilson (eds.),
Proceedings of WSCLA 7: The Workshop on Structure and Constituency in the Languages of
the Americas, pp. 89 – 102. Vancouver: University of British Columbia Working Papers in
Linguistics.
LADEFOGED, PETER. 2001. A course in phonetics. New York: Harcourt Brace.
LADEFOGED, PETER, and ELIZABETH ZEITOUN. 1993. Pulmonic ingressive phones do not occur in
Tsou. Journal of the International Phonetic Association 23, 13 - 15.
LEVINE, ROBERT D. 1981. Review of the Haida Dictionary, by Erma Lawrence et al. Interna-
tional Journal of American Linguistics 47, 354 - 8.
LEVY, PAULETTE. 1987. Fonología del totonaco de Papantla, Veracruz. Mexico: Universidad
Nacional Autónoma de México.
MACKAY, CAROLYN J. 1999. A grammar of Misantla Totonac. Salt Lake City: University of
Utah Press.

17
18 U. Alberta Working Papers in Linguistics, vol. 1, 2006.

MADDIESON, IAN. 1998. Why make life hard?—resolutions to problems of rare and difficult
sound types. UCLA Working Papers in Phonetics 96, 106 – 18.
MADDIESON, IAN, NICOLA BESSELL, and CAROLINE LAWS SMITH. 1996. A preliminary report on
the phonetics of Tlingit. UCLA Working Papers in Phonetics 93, 125 – 48.
MCLENDON, SALLY. 1973. Proto Pomo. University of California Papers in Linguistics no. 71.
Berkeley: University of California Press.
———1976. The Proto-Pomo pronominal system. In Margaret Langdon and Shirley Silver
(eds.), Hokan Studies: Papers from the First Conference on Hokan languages. Janua lingua-
rum, series practica, no. 181, pp. 29 – 54. The Hague: Mouton.
MCQUOWN, NORMAN A. 1977. La familia lingüística totonacana. Chicago: University of Chi-
cago Library.
———1990. Grámatica de la lengua totonaca. Mexico: Universidad Nacional Autónoma de
México.
MITCHELL, T. F. 1990. Pronouncing Arabic: I. Oxford: Clarendon Press.
NATER, HANK F. 1984. The Bella Coola language. Ottawa: National Museum of Man.
REID, AILEEN A., and RUTH G. BISHOP. 1974. Diccionario totonaco de Xicotepec de Juárez.
Mexico: Summer Institute of Linguistics.
RICE, KEREN. 1994. Laryngeal features in Athapaskan languages. Phonology 11, 107 – 47.
ROBINS, ROBERT H. 1958. The Yurok language: Grammar, text, lexicon. University of California
Papers in Linguistics no. 15. Berkeley: University of California Press.
SKINNER, NEIL. 1971. /ts/ and /k’/ in Hausa. Anthropological Linguistics 133, 301 – 10.
TROIANI, DUNA. 2004. Aperçu grammatical du totonaque de Huehuetla, Puebla, Mexique. Mu-
nich: LINCOM Europa.
WATTERS, JAMES. n.d.a Una base de datos para la reconstrucción del idioma protototonaco.
Ms., Summer Institute of Linguistics.
———n.d.b. Lexical database for Tlachichilco Tepehua. Shoebox database file.

You might also like