Beruflich Dokumente
Kultur Dokumente
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to International Journal of
American Linguistics.
http://www.jstor.org
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
A REVISED PROPOSAL OF PROTO-TUKANOAN CONSONANTS
AND TUKANOAN FAMILY CLASSIFICATION 1
Thiago Chacon
Universidade Católica de Brasília
1 This article is the result of several years of comparative research on Tukanoan languages.
Special thanks go to Aryon Rodrigues, who provided me the opportunity to begin comparative
research on the Tukanoan family while I was still an undergraduate student at the University
of Brasilia; to Lyle Campbell, Bob Blust, Patience Epps, and several other professors and col-
leagues, who helped me with supervision and comments on earlier versions; and foremost to
Kristine Stenzel, Elsa Gomez-Imbert, and Denny Moore, who reviewed the paper for IJAL and
helped me with the organization and content of the final version.
2 There has never been any systematic attempt by researchers to distinguish between language
and dialect in Tukanoan languages, a task that is made especially difficult because of the tendency
to correlate language and ethnicity in the Tukanoan language-speaking groups. Some Eastern
Tukanoan languages might well be classified by an outsider as dialects of each other, given the
limited phonological and lexical distinctions between them (e.g., tan, ret, and yah, or bas and
edu). However, language is a very important marker of social identity of all groups in the region,
and speakers are highly aware of linguistic distinctions and pay attention to small details that
justify treating “dialects” as distinct languages. This paper focuses on genetic classification across
major subgroups in the Tukanoan family rather than on the specific relationships of dialects or
between very closely related languages. It cannot (due to space and data limitations) provide
data for each specific language/dialect listed in (1) above; however, throughout the preparation
of this paper, I have observed that dialects or very closely related languages share fundamental
phonological features that, on the one hand, support this reconstruction and classification of
languages in the family and, on the other, are sufficient to continue classifying them as distinct.
275
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
276 international journal of american linguistics
languages and their ISO codes, as well as information about whether the
language is still spoken († = extinct), is given in (1). 3
(1) Languages of the Tukanoan family
Arapaso† (ara) (ISO code: arj); Bará (bar) (ISO code: bao);
Barasano (bas) (ISO code: bsn); Desano (des) (ISO code: des);
Eduria (edu) (ISO code: bsn); Karapana (kar) (ISO code: cbc);
Koreguahe (kor) (ISO code: coe); Kubeo (kub) (ISO code: cub);
Kueretu† (kue) (no current ISO code); Maihɨki (mai) (ISO code:
ore); Makaguahe† (mag) (ISO code: mcl); Makuna (mak) (ISO code:
myy); Miriti-Tapuya† (mir) (ISO code: mmv); Pirá-Tapuya (pir)
(ISO code: pir); Pisamira (pis) (no current ISO code); Retuarã (ret)
(ISO code: tnc); Sekoya (sek) (ISO code: sey); Siona (sio) (ISO
code: snn); Siriano (sir) (ISO code: sri); Tama† (tam) (ISO code:
ten); Tanimuka (tan) (ISO code: tnc); Tatuyo (tat) (ISO code: tav);
Tetéte† (tet) (ISO code: teb); Tukano (tuk) (ISO code: tuo); Tuyuka
(tuy) (ISO code: tue); Wanano (wan) (ISO code: gvc); Yahuna†
(yah) (ISO code: ynu); Yupua† (yup) (no current ISO code); Yuruti
(yur) (ISO code: yui)
Tukanoan languages are found in four different South American countries—
Brazil, Colombia, Ecuador, and Peru—and are widely distributed throughout
the Northwest Amazon, from the Vaupes and Ayari Rivers (western affluents
of the Rio Negro), through the Apaporis, Caquetá/Japurá, Putumayo / Iça, and
Napo Rivers, all northern tributaries of the Solimões/Marañón River. They
are surrounded by Cariban, Nadahup (Makuan), Arawakan, Witotoan, Boran,
and Tupian languages, as represented in the map in figure 1.
The first attempted classification of Tukanoan languages was by Brinton
(1891; 1892), who grouped certain Tukanoan languages, along with Betoi
and Japu, into a single Betoya stock. This classification was challenged by
Beuchat and Rivet (1911), who identified Betoi as a Chibchan language. The
term “Tukano / Tucano” has been used to identify the languages presented in
(1) above since then.
The most widely accepted classification of Tukanoan languages was pro-
posed by Mason (1950), building on Beuchat and Rivet (1911) but incorpo-
rating additional linguistic, geographic, and ethnographic information. He
Thus, although I may use a token to represent a sound change shared by dialects or closely related
languages, I always acknowledge to which dialect/language the token belongs.
3 There are many alternate names of the languages listed in (1), especially given the com-
mon practice in South America for groups to be known by heteronyms, which are very often
derogatory, while still keeping their autonyms, which have gained recognition by outsiders over
the past few decades. The following are the most common and/or recent names found in the
literature: Bara ~ Waimaha, Southern Barasano; Eduria ~ Taiwano; Kubeo ~ Pamie; Pira-Tapuyo
~ Wa’ikhana; Pisamira ~ Pápiwa; Maihɨki ~ Koto, Orejón, Maihuna; Retuarã ~ Letuama; Tukano
~ Ye’pa Masa; Siona ~ Piojé, Zeona; Wanano (Guanano) ~ Kotiria; Yuruti ~ Wahiara.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 277
divided the Tukanoan family into two major branches: Eastern Tukanoan (et),
comprising Tuyuka, Bará, Pirá-Tapuya, Wanano, Tukano, Yurutí, Karapana,
Tatuyo, Kubeo, Makuna, Barasano, Miriti-Tapuya, Desano, Siriano, Let-
uama, Tanimuka, Koretu, and Yauna—languages spoken in the Apaporis and
Vaupes River Basins of the Brazilian–Colombian border region; and Western
Tukanoan (wt), comprising Maihikɨ [Orehon], Koreguahe, Tama, Sekoya,
Siona, Makaguahe, and Teté—languages spoken in western Colombia, eastern
Ecuador, and southeastern Peru, on the Caquetá, Putumayo, and Napo Rivers.
Mason’s (1950) classification was challenged by Waltz and Wheeler (1972),
who proposed a third—Middle Tukanoan—branch that actually included just
one language, Kubeo, representing a “midpoint” between wt and et languages.
Waltz and Wheeler’s (1972) reformulation was based on comparison of 278
words from 16 Tukanoan languages, gathered in the early years of the Summer
Institute of Linguistics’ (SIL) missionary work in Colombia. Their classifica-
tion of the family utilized lexicostatistical methods, though they presented
some phonological correspondences and a few inconsistent proto-sound re-
constructions. Only one wt language, Siona, was represented.
Many SIL publications, catalogs of South American languages (e.g., Tovar
and Tovar 1984), and Ramirez (1997b) followed Waltz and Wheeler’s analysis.
However, the “Middle Tukanoan” branch has remained loosely defined. Barnes
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
278 international journal of american linguistics
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 279
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
280 international journal of american linguistics
4 An exception to voiced segments having nasal allophones is found in Maihɨki /g/ accord-
ing to more recent descriptions of the language (Lev Michael, personal communication, 2012),
although Vellie, Brend, and Powlison (1976) describe /g/ having the allophone [ŋ] in nasal
environments. Likewise, /r/ in kor is also never nasalized (see Cook and Criswell 1993). /h/ is
the only voiceless segment reported to be nasalized, in wan (Stenzel 2013), while it is transparent
for nasal harmony in wt languages.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 281
5 The following references that have not been directly cited in this paper are important sources
of data on Tukanoan languages and cultural groups: Alemán M., Lopez H., and Miller (2000)
on Desanol; Cabalzar (2008) on Tuyuka; Hugh-Jones (1979) on Barasano; Jackson (1983) on
Bará; Metzger (2000) on Karapana; Morse, Salser, and Salser (1999) on Kubeo; Piaguaje et al.
(1992) on Sekoya; Smothermon and Smothermon (1993) on Makuna; and Waltz, Simmons de
Jones, and de Waltz (2007) on Wanano.
6 The data in Waltz and Wheeler (1972) was much more limited than the data available for the
present study and they did not consider independent historical sources such as Koch-Grünberg’s
word lists (1913;1914). Another shortcoming in their study, likely due to the lack of overall data
on the family at the time, was their use of correspondences from a limited number of languages
as the basis for their reconstruction of proto-sounds.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
282 international journal of american linguistics
of labialized velar stops (cf. kw and k’w), but I did find some indications and
note in Appendix A that these are very likely reconstructions. Appendix B,
published online only, gives a full representation of all correspondences upon
which the reconstruction of proto-sounds is based.
2.3. Classification of the Tukanoan family. The classification in fig-
ure 3 revises earlier classifications and includes a number of Tukanoan
languages not included in more recent literature. It proposes a more precise
and inclusive classification of languages into subgroups, so that in addition
to the Western and Eastern division, it classifies the et languages into three
subgroups: Southern-et, Western-et, and Eastern-et (the latter further di-
vided into Branch-I and Branch-II). A map of the distribution of the major
branches is given in figure 4.
2.4. On contact and genetic relations within the Tukanoan fam-
ily. One of the greatest challenges in subgrouping closely related lan-
guages that have maintained close contact after their diversification is the
choice of criteria used to disentangle retentions from the common proto-
language, shared innovations, and contact-induced changes.
In this study, subgroup classification is primarily based on shared phono-
logical innovations. Cognate percentages are used only secondarily, for three
reasons: (i) shared lexical items can be retentions, which cannot be used
to propose genetic subgroupings (Campbell 2004); (ii) lexical borrowing
among et languages might be more common than we currently recognize,
based on comparative studies between Tukanoan and Arawakan languages
(see Sorensen 1967 and Aikhenvald 2002); (iii) speakers of closely related
et languages may be more attuned to phonological traits than lexical ones.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 283
Point (i) above is perhaps the main issue behind the classification of kub as
Middle/Central Tukanoan in Waltz and Wheeler (1972). Waltz and Wheeler
gave priority to lexicostatistics, and the Kubeo lexicon shows partial simi-
larities to both Western and Eastern Tukanoan languages. Chacon (2013; see
also 2012) shows that Kubeo has retained several lexical archaisms from
Proto-Tukanoan as well as certain grammatical and phonological traits. This is
because Kubeo shares fewer of the areal features found in et languages from
the Papurí and Pirá-Paraná Rivers, a fact also directly related to point (ii).
Point (ii) refers to cases where languages (A) and (B) share fundamental
sound changes but are less alike with respect to vocabulary. 7 Whenever this
contradiction is found, it is possible to find a third language (C) that shares
fewer fundamental sound changes with (A) and (B) but shares more vocabu-
lary with (A), for instance. Normally we find that (C) and (A) have an intense
intermarrying relationship, and their lexical similarities can be attributed to
contact.
7 By “fundamental sound changes,” I mean changes that are both more characteristic of a
consistent sound change shared genetically by distinct languages and that can also be argued to
have occurred earlier than other, more contingent changes. “Non-fundamental sound changes”
are those that can easily occur independently in distinct languages or that represent sporadic but
shared changes based on chance or areal trends.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
284 international journal of american linguistics
A few cases illustrating point (ii) are given below. In each case, language
(C) shares more basic vocabulary with language (A), based on the lexical fig-
ures in Ramirez (1997b:15) and those reported in Waltz and Wheeler (1972),
although languages (A) and (B) are more closely related to each other than to
(C), based on phonological innovations that are the basis for the classifica-
tion in figure 3.
bará (A), tukano (B), and tuyuka (C)
desano (A), makuna (B), and tukano (C)
pirá-tapuyo (A), tuyuka (B), and tukano (C)
This follows from the fact that long-term intermarriage practices between
speakers of different languages can engender several types of linguistic con-
vergence, from lexical borrowings to shared phonological and structural traits.
While the sharing of structural features is pervasive in the region and native
speakers are less attuned to it than they are to other types of phonological and
lexical diffusion (see Sorensen 1967 and Aikhenvald 2001a; 2002), we can
suppose that prolonged contact among closely related Tukanoan languages
might produce different results than contact between Tukanoan and non-
Tukanoan languages. This is due both to closer structural and phonological
similarities among Tukanoan languages and to more intense and long-term
multilateral, inter-community linguistic exogamy among et groups than with
groups speaking unrelated languages.
A related factor with a role similar to that of linguistic exogamy is the is-
sue of linguistic sub-areas in the region (i.e., smaller, denser linguistic areas
within a larger, more inclusive one). Linguistic exogamy is, in fact, one of the
social bounds defining local linguistic sub-areas in the Vaupes, but geography
is a more general factor, since marriage alliances and other social relations
not dependent on marriage are always stronger between neighboring groups
(see also Stenzel 2005). 8
In the three sets below, languages (B) and (C) are spoken by groups in
geographically contiguous areas and show greater lexical similarity than do
languages (A) and (B), which actually have a closer genetic relationship based
on phonological innovations.
wanano (A), tuyuka (B), and bará (C)
tukano (A), tatuyo (B), and tuyuka (C)
kubeo (A), barasano (B), and karapana (C)
8 The notion of “linguistic area” needs to be better defined for this region, which would allow
us to understand areal effects within more general geographic zones (e.g., the entire Northwest
Amazon) and those pertaining to particular sub-areas (e.g., the Pirá-Paraná, the Papuri-Vaupes,
the Pirá-Paraná and upper sections of the Papuri, the Tiquie and Vaupes, etc.)
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 285
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
286 international journal of american linguistics
tan kub des bas tuk wan kar kue mai kor sek sio
p [ɸ] p p h p ph p h h h h h root-initial
p p p h p p p h h h h h root-medial
tan kub des bas tuk wan kar kue mai kor sek sio
t t t t t th t t t th t t root-initial
t t
d/_ṽ d d t t t t n/_ṽ t th t t root-medial
Fig. 6.—*t reflexes in root-initial and medial positions. “v” stands for vowel.
9 The Kubeo exception /d/ in ‘grass’ seems to be conditioned by a sporadic change where
pre-Kubeo *taja [taða] dropped medial /j/ [ð] and voiced word-initial /t/.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 287
tan kub des bas tuk wan kar kue mai kor sek sio
k k k k k kh k k k kh k k root-initial
k k g k k k k k k kh k k root-medial
∅/_ṽ tʃ/e,i_ kj/e_
tan kub des bas tuk wan kar kue mai kor sek sio
t t t t t t t t t t t t *tt
tʃ/_i
k k k k k k k k k k k k *kk
kanoan languages. This can explain the shape of pea ‘firewood’ in des. In this language, the
word for ‘father’ is pagɨ, but it is often pronounced as paɨ or simply pɨ, showing the tendency
for /g/ deletion. See 3.1.3 below for more on /g/ in Tukanoan languages.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
288 international journal of american linguistics
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 289
tan kub des bas tuk wan kar kue mai kor sek sio
b b b b b b b b / p ʔb p p p’ root-initial
v’p v’p v’p
v’b/_ṽ b b b v’b/_ṽ v’b/_ṽ p h/∅ h h h h root-medial
tan kub des bas tuk wan kar kue mai kor sek sio
r d d r d d r ? ʔd r d ɖ root-initial
t v’t v’t t/l
v’r/_ṽ d d d v’d/_ṽ v’d/_ṽ t n/_ṽ t th t t root-medial
*p’ in root-initial position are a *p’ > b change in et languages, *p’ > ʔb
(a pre-glottalized bilabial stop) 11 in mai, retention of *p’ in sio, and *p’ >
p in kor and sek.
kue displays a very interesting pattern of variation of <p> and <b> in Koch-
Grünberg’s (1914) orthography. It is reasonable to think that the actual seg-
ment represented by either orthographic symbol was “ambiguous” in terms of
voicing in Koch-Grünberg’s perception. In his transcriptions of et languages
that have true voiced stops, <b> is consistently used. This ambiguity in kue
is possibly linked to the phonetic realization of laryngealized stops in kue
more than a century ago, which may represent a stage in which a distinct
reflex of *p’ was retained.
In root-medial position, *p’ merged with *p in all wt languages (note
figures 5 and 9)—hence, the h reflexes (and those where h > ∅). Merger of
laryngealized and plain stops in root-medial position is recurrent in other
reflexes of *C’ for wt (see figures 10 and 11). There was an unconditional
*p’ > b change in kub, des, sir, yup, bas, mak, and edu; in tan, ret, yah, tuk,
wan, and pir, *p’ laryngealized the preceding vowel: *p’ > v’_, while in kar,
tat, bar, tuy, and yur, vowels did not become laryngealized. In the latter two
groups of et languages, *p’ > b [m] before a nasalized vowel.
remarkable that every word analyzed as pre-glottalized by Velie (1975) and Velie, Brend, and
Powlison (1976) clearly matches the cognates I reconstruct as having *C’. On the other hand,
non-pre-glottalized voiced stops in these sources are reconstructed to different proto-sounds,
especially *w > b. More recently, Lev Michael (personal communication, 2011), who is conduct-
ing a detailed phonological investigation of the language, told me that some of the contrasts
analyzed by Velie (1975) and Velie, Brend, and Powlison (1976) no longer exist in the language.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
290 international journal of american linguistics
Consider now the reflexes of *t’ (fig. 10), whose cognates in Appendix A
are traira fish, heavy, wife, and woman for root-initial position, and excre-
ment, aracu fish, three, tobacco, and wait for root-medial position.
All Tukanoan languages changed *t’ > d or r in root-initial position, though
in some, such as yah (Koch-Grünberg 1913) and sio (Wheeler 1987a), d has a
retroflex realization. r in kor is realized as a retroflex tap (Cook and Criswell
1993), and Metzger and Metzger (1973:116) state that kar has a retroflex
allophone of r. These realizations suggest that there were two stages in the
evolution of root-initial *t’: first, **t’ > *ɖ and, later, *ɖ changed to the sounds
currently represented in the correspondences above.
In root-medial position, wt languages merged *t’ with *t, except for kue,
whose reflexes present a number of inconsistencies. The pattern for et lan-
guages is an exact parallel to that of *p’: some languages changed *t’ > d
(kub, des, sir, yup, bas, edu, and mak); some retained t in oral contexts and
changed *t’ > d in nasal contexts, with accompanying laryngealization of
the preceding vowel (tuk, wan, pir, tan, ret, and yah); and others merged
*t’ with *t in oral contexts and changed *t’ > d in nasal contexts (kar, tat,
bar, tuy, and yur).
The phoneme r in kor has an interesting feature that can be explained by a
*t’ > r change. r is the only “sonorant” phoneme in the language that blocks
nasal harmony, while other sonorant phonemes are targets for nasal harmony
(see Cook and Criswell 1993). It seems reasonable to postulate that the reason
for the exceptional pattern of r is because this is the only sonorant segment
historically derived from a non-sonorant sound. Similar facts are found in des
as well (see discussion of examples in figure 12).
kar and tat are special with respect to the reflexes of *t’ because there
is evidence that they first underwent the change *t’ > d (with [d] surfacing
root-initially and [r] root-medially). Later, in non-nasal contexts only, d > r,
while d merged with *n in nasal contexts. The d > r change was possibly an
areal influence from bas and mak, both of which changed *t’ and *n to r. 12
Several examples support this hypothesis: in oral roots, kar r corresponds to
bas r, e.g., kar ria : bas ria ‘river’; whereas in most nasal roots, the initial
alveolar segments are distinct, such as in kar ~diti : bas ~riti ‘charcoal’.
Two objections could be raised for this reconstruction of *t’. First, given
the regular reflexes of *t’ in root-initial position, one could ask: why not
postulate *d or *r instead of *t’? The answer would be that not only does *t’
represent a sound that is consistent with the natural class of laryngealized stops
in Proto-Tukanoan, but reflexes of *t’ resemble those of *p’ in root-medial
12 There seems to be an /r/ isogloss, which is present in most Tukanoan languages of the
Pirá-Paraná and Apaporis Rivers in root-initial position, contrasting with /d/ in et in the Vaupes
sub-area.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 291
tan kub des bas tuk wan kar kue mai kor sek sio
∅ k g g ∅ k ∅ k g/∅ k k k’ root-initial
k ∅ g g v’k v’k k k k kh k k root-medial
13 Indeed, Cook and Criswell (1993) mention that /t/ is very rare in Koreguaje, while Wheeler
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
292 international journal of american linguistics
Similarly to the reflexes of *t’ and *p’, in tuk, wan, and pir, *k’ > v’k in
root-middle position, with laryngealization of the vowel preceding *C’. tan,
ret, yah, kar, and tat (which changed *k’ > ∅ root-initially) and yur, pis, and
tuy (which changed *k’ > k root-initially) merged *k’ with *k root-medially.
The observed similarities between reflexes of *k’ and other *C’ reflexes
support the present reconstruction, though what is actually more interesting
are the types of irregularities found in the reflexes of *k’. For instance, mai has
been analyzed as having pre-glottalized stops ʔb and ʔd (reflexes of *p’ and
*t’), while g is not reported to be pre-glottalized (see Velie 1975 and Velie,
Brend, and Powlison 1976). Also, among et languages there is a split between
those that voiced *k’ root-initially, those that dropped it in a possible **k’ >
*g > ∅ cline, and those that apparently merged *k’ with *k. These changes
involve languages that we have reason to classify as belonging to different
subgroups, indicating that not all languages of each genetic subgroup under-
went the same changes.
Considering what might have caused *k’ reflexes to deviate so strongly
from the more regular patterns observed for other types of *C’, I postulate
three factors that were separately or jointly involved:
(i) Longer retention of *k’: It is likely that *k’ was retained longer than
other *C’ in et languages; this could be responsible for the greater diversity
of reflexes, since each smaller genetic subgroup would have changed *k’
independently from the others. Maddieson (1984:40) states the implicational
hierarchy where if a language has g, then it is very likely to have d and b; thus
it seems that the *k’ > g change took place after those of *p’ > b and *t’ > d. 14
(ii) Structural patterns: Paradigmatic factors related to the structure of
segment inventories from different proto-languages in the evolution of the
Tukanoan family also framed the particular evolution of *k’. It is likely that
these factors were responsible for both the longer retention of *k’ and for its
different reflexes: *k’ > g, **k’ > *g > ∅, and *k’ > k. Maddieson (1984:40,
118) discusses how voicing is more favored in bilabial stops than in velar
stops, which explains why g is often absent from series of voiced plosives.
On the other hand, ejective stops (which, within the glottal class, share many
phonetic similarities with laryngealized stops) show the reverse tendency:
velar ejectives kʔ are more common than bilabial ejectives pʔ. 15
(iii) Contact: Contact could also have played a role in the evolution of *k’.
First, there is evidence for contact-induced changes related to *k’ reflexes
in languages from different Tukanoan subgroups (especially tat influencing
kar). Second, Northern Arawakan languages—which lack g—could have
14 There are only three counterexamples to this generalization in Maddieson’s (1984) data-
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 293
-p’i /
-bi -~bi -~bi -~bi -~witat -pi / -hi -ʔphi third-person masculine
influenced the reflexes of *k’ in different ways: (a) inducing the dropping
of *g (as perhaps in tuk, bar, tat root-initially, kub root-medially, and as
Aikhenvald 2002 suggests for the influence of Yukuna [Arawakan] on Re
tuarã); (b) preventing the *k’ > g change from taking place or causing a further
**k’ > *g > k change root-initially (as in kub, wan, pir, tuy, yur, and pis). 16
It is unnecessary and quite likely impossible to make a more precise assess-
ment of the correlation of these factors: all three are sound explanations for
the challenging reflexes of *k’.
The evolution of *p’, *t’, and *k’ in bound morphemes is also very il-
luminating for the history of these sounds. Consider the correspondences
shown in figure 12.
While *p’ merged with *p (> h) root-medially in wt languages, it did not
undergo the same change in bound morphemes (alternations of these forms
are discussed in 3.3.3). Also, des -bɨ ‘non-third-person animate’ is opaque
to nasal spreading from a preceding nasal morpheme. This fact can only be
explained synchronically by underlying specifications, since almost all mor-
phemes with initial voiced stops can be nasalized in the language. If it is true
that this morpheme is derived from *-p’ɨ, then we may postulate that when
the initial consonant became voiced, the morpheme still remained opaque to
nasal spreading (since *p’ was opaque to nasal spreading in Proto-Tukanoan;
cf. the behavior of p’ [< *p’] in sio).
Another interesting pattern is the process whereby *p’ > b > w in most
bound forms, especially in et languages. It is nevertheless intriguing that
not all forms display this change. One hypothesis is that the change only af-
fected morphemes with a certain boundness status: i.e., suffixes, being more
closely bound than clitics, underwent the full change, while clitics retained
a more conservative realization of *p’. This could explain why the form for
third-person masculine in sio, which has clitic status, still retains the initial
laryngealized stop (see Wheeler 1987a). However, further investigation of
this hypothesis is still necessary.
I turn now to the reflexes of *t’ (fig. 13) and k’ (fig. 14) in bound morphemes.
16 Contact between Arawak and Tukanoan languages was particularly strong (at least in
historical times) in the cases of kub and wan with Baniwa-Curripaco; ret and tan with Yukuna;
and tuk with Tariana.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
294 international journal of american linguistics
tan kub des bas tuk wan kar mai sek kor sio gloss
-de -re -de / case/oblique/
-re -de -de -re -de -de -re -de [-te] [-the] -ʔthe referential
-do -do /
— -dõ -do -ro -do -do -ro -do [-to] — -ʔtho part of a whole
tan kub des bas tuk wan kar mai sek kor sio gloss
-k’o
-ko -ko -go -go -(g)o -ko -o -o -o -o [-ʔkho] feminine
Reflexes of *t’ in bound morphemes resemble those observed for the same
sound in root-initial position. It is interestingly, though, that in languages such
as bas, which has a three-way alveolar contrast (d, t, and r), the bound mor-
phemes above have initial r rather than d. This shows a parallel change of *t’
in bound morphemes and in root-initial position, in contrast to its change in
root-medial position, which yielded d. In languages with a two-way alveolar
contrast, all bound morphemes have d.
et languages that changed *k’ > k root-initially, such as kub and wan, exhibit
the same reflex in bound morphemes; likewise, et languages displaying an
initial *k’ > g change in roots, such as bas and des, show the same change in
bound morphemes. In tuk, tat, bar, and kar, the *k’ > g > ∅ change postulated
for root-initial position is also attested in bound morphemes (although in tuk,
g in bound morphemes is still attested in formal speech styles; see Ramirez
1997b). The facts relating to wt languages are more complex, but it appears
that *k’ > ∅ in bound forms in most of these languages, with the exception
of sio, where it was retained.
3.3.1. Evidence against the reconstruction of voiced stops. Both Waltz
and Wheeler (1972) and Malone (1986) proposed the reconstruction of *b,
*d, and *g to account for the correspondences here reconstructed as a la-
ryngealized series. I turn now to evidence supporting non-reconstruction of
voiced stops.
First, in the previous section, we observed that laryngealization of the vowel
preceding *C’ is a recurrent pattern in certain et languages and that languages
lacking this laryngealization are also the ones in which glottal stop *ʔ merged
with ∅. There is no way to account for these systematic differences through
the reconstruction of voiced stops.
Additionally, nasalization triggered the root-medial changes of *p’ > b and
*t’ > d/r in et languages that did not undergo the same changes in non-nasal
contexts—in which they reflected a voiceless realization of *C’. wt languages,
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 295
on the other hand, retained voiceless stops in all contexts root-medially (in-
cluding in nasal contexts). Nasalization is cross-linguistically an environment
that can trigger lenition processes, especially those involving voicing, as im-
plied in the evolution of *C’. It thus seems unlikely that nasalization would
trigger devoicing of *b, *d, and *g in those languages.
The same reasoning can be applied to languages that changed *C’ > Cvoiced
root-initially but have voiceless stops in the correspondences of root-medial
*C’. Postulating a voiced series would require an ad hoc explanation for reten-
tion of voiced *b, *d, and *g root-initially and their devoicing root-medially.
Moreover, the retention of laryngealized stops in sio is totally unmotivated
if voiced stops are their assumed diachronic source.
We also observe some intricate assimilation and dissimilation processes
in words in which root-initial *C’ is followed by a plain stop *C with the
same point of articulation, as in the cognates fishing net, spider, to break, to
squeeze, and to swell. These words show diverse resolutions for the sequence
*p’VpV, and such irregularities are taken as evidence of a conflict of differ-
ent laryngeal settings in the articulation of these words, exacerbated by the
fact of both stops having the same point of articulation. Such processes are
common in languages in which stops contrast based on laryngeal features of
[constricted glottis] and [spread glottis] (see Fallon 2002).
Finally, it is worth entertaining the possibility that what I have been treat-
ing as a laryngeal feature associated with stops may actually come from
other sources, such as underlying laryngealized vowels or a low tone. Much
comparative investigation on tones still needs to be done, so it is a bit early
to reject or accept this possibility. Regarding laryngealized vowels, in Chacon
(forthcoming), I list several arguments against their reconstruction, especially
the following ones: (i) no Tukanoan languages currently have laryngealized
vowels phonemically; (ii) there is never a laryngeal trace in an environment
surrounding a proto-nasal stop *m and *n (only surrounding reflexes of *C’);
(iii) the evidence for laryngealized phonemic vowels is so weak that one
would have to assume that they existed in pre-Proto-Tukanoan, making such a
proposal even less concrete. In any case, at this point, I think we have several
reasons to treat laryngealization as a feature of segments when it comes to
the analysis of reflexes of *C’.
3.3.2. Phonetic and phonological basis of *C’ diachronic evolu-
tion. There is ample evidence that the *C’ series was composed of pre-
laryngealized stops presenting creaky voicing. Creaky voice involves vibration
of the vocal folds (in contrast to plain voiceless stops) and is a midpoint
between full closure of the glottis, as for ejectives, and full aperture of the
glottis, as for voiced stops (Ladefoged and Maddieson 1996). This partial
closure can explain voiceless reflexes of *C’, while partial aperture can ex-
plain the voiced reflexes. Moreover, the laryngealized feature can elucidate
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
296 international journal of american linguistics
17 tuk, wan, and pir changed *k’ > v’k in root-medial position, with accompanying laryn-
gealization of the vowel preceding *C’, as noted in the reflexes of *t’ and *p’. The fact that tuk
root-initial *k’ underwent a different change indicates that laryngealization of vowels preceding
*C’ is a retention from Proto-Eastern-Tukanoan, rather than a more recent innovation.
18 When creaky voice articulation overlaps with the full production of the stop, the acoustic
distinction between creaky voice and modal voice is reduced to different degrees of increas-
ing amplitude (as in creaky voice) or decreasing amplitude (as in modal voice) of vocal folds’
vibration pulses, and to the regularity of the pulses of vibration: in creaky voice, pulses are more
irregular in shape and there are possibly interleaving peaks of smaller duration in the vibration
pulses, whereas in modal voice, pulses are more regular in shape (see Ladefoged and Maddieson’s
1996:54 discussion of creaky voice vs. modal voice in Fula).
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 297
There is evidence to suggest that *C’ reflexes are the result of two (or more)
distinct changes, possibly occurring at separate points in time. For example,
while all wt languages merged *C’ with *C in root-medial position, in root-
initial position the reflexes of *C’ are distinct in each language. On the other
hand, while all et languages have voiced reflexes of *C’ root-initially, reflexes
of *C’ root-medially tend to be more diverse.
It is interesting that in languages where *C’ > Cvoiced both root-medially
and root-initially (bas, mak, yup, des, sir, and kub), the change also resulted
in loss of the creaky voice effect on the preceding vowel. Indeed, bas, mak,
and kub also completely dropped *ʔ (see 3.5 for an analysis of *ʔ), while
there is substantial variation within des and sir regarding retentions of *ʔ
(for des, see Kaye 1970 and Miller 1999; for sir, see Criswell and Bandrup
2000). It is moreover curious that Gomez-Imbert (personal communication,
2009) and Kristine Stenzel (personal communication, 2012) report that voiced
stops (reflexes of *C’) in bas and wan respectively are frequently realized
as implosives.
In ret, tan, and yah, *C’ reflexes root-initially are consistent with the et
pattern but vary root-medially: vowels preceding *p’ are always laryngeal-
ized; vowels are only laryngealized preceding *t’ in a nasal environment; and
preceding *k’ vowels are never laryngealized. Also *p’ and *t’ changed to b
and d respectively in nasal contexts.
In mai, *C’ may have changed to a voiced pre-glottalized stop in root-initial
position (see Velie 1976; but also see n. 11) but merged with plain voice-
less stops root-internally following the pattern in other wt languages. In the
process, the laryngeal trace on the preceding vowel was also lost, completing
the merger where there had once been a contrast. Reflexes of root-initial *p’
and *t’ are voiced (and pre-glottalized, thus contrasting with d and b from
different diachronic origins). Pre-glottalization is also attested in words that
historically had *C’ and later became nasalized but still retained the glottal
feature as [ʔm] and [ʔn]. In mai, reflexes of *m and *n are not glottalized,
further suggesting retention of a contrast between *C’ and other C’s. In wt
sio, *C’ was retained in root-initial position, where it laryngealizes the sur-
rounding vowels, which are realized with “little strength and voicing [that]
is minimally perceptible” (Wheeler 1987a:85).
Finally, changes in bound morphemes also seem to have different phono-
logical conditioning from changes occurring in root-initial position at the
stem level, even though in both cases, the C is preceded by a vowel. For
instance, in tuk, des, and wan, there is pre-aspiration of plain voiceless stops
root-internally, while no pre-aspiration occurs before a suffix beginning with
a voiceless stop. This suggests that root-medial changes and those affecting
bound morphemes were conditioned at two distinct phonological levels. Still,
more research on bound morphemes needs to be done, especially given the
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
298 international journal of american linguistics
fact that Proto-Tukanoan appears to have been less agglutinating and had
fewer inflectional forms than its daughter languages.
3.3.3. Morphophonemic alternations involving *C’ reflexes. Through-
out the Tukanoan family there are morphophonemic alternations that shed
some light on the grammatical past of Proto-Tukanoan and on the early con-
trast of *C’ and *C. In general, such morphophonemic processes involve the
alternation of a lenis form, such as g, with a fortis form, such as k or kh, which
can be reconstructed as alternations of *C’ and *C. These alternations involve
rules that are on the frontier between synchrony and diachrony.
In wt languages, suffixes beginning with a stop that is a reflex of *C’ ap-
pear to undergo morphophonemic alternations triggered by the presence vs.
absence of stress. In sio, Wheeler (1987a) states that in unstressed position,
affixes such as -k’o ‘copula’ or -b’ɨ ‘rounded classifier’ surface as [go] and
[bɨ] respectively, while in stressed position they undergo fortition, surfacing
as [kho] and [phɨ], as in (4a) and (4b) respectively. Also, before these stressed
affixes, a glottal stop is epenthesized (adapted from Wheeler 1987a): 19
(4a) saí-maʔì-k’o-k’ò > [saímaʔìgoʔkho]
go-neg-nmz.fem-cop
‘the one that is not going’
(4b) s’iá-t’a-b̀ -t’ɨ > [s’iáraʔph̀ rɨ]
egg-dim-cl.round-dim
‘small egg’
Stressed syllables are a typical environment in which consonantal strength-
ening occurs. Plain stops or aspirated stops can be considered more fortis
than laryngealized or voiced stops (see Fallon 2002 and Lavoie 2001). The
epenthetic glottal stop in sio, according to Wheeler, occurs only in these
alternations before a stop. From a diachronic perspective, one may think of
the glottal as a residual feature of *C’, which is now manifested in the coda
position of the preceding syllable, while the stop is strengthened by aspiration.
In 3.3, we saw that while *C’ merged with *C root-medially in wt lan-
guages, it was retained root-initially and in grammatical morphemes in sio.
In light of these alternations, and considering that stress falls on the second
syllable of roots (assigned left to right) in sio and sek, it is possible that the
merger of *C’ with *C in C2 position in (C1)VC2V roots was triggered by
regular stress placement on the second syllable.
kor has an alternation similar to that observed in sio , even though kor
does not have similar laryngealized stops (see Cook and Criswell 1993). For
instance, -pi ‘agent’ alternates with [hi], while -re ‘complement’ alternates
with [-ʔthe], the glottal ʔ in the latter form also being epenthetic. Its absence
19 An acute accent marks primary stress and a grave accent marks secondary stress.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 299
20 The morpheme be(ti) ‘negation’ can surface alternately as [be] or [beti]; see Gomez-
Imbert (2004:60).
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
300 international journal of american linguistics
the alternate forms h, t, and k are reflexes of *p, *t, and *k. In tat, w (in most
affixes), r, and ∅ are reflexes of *p’, *t’, and *k’ respectively, while p, t, and
k are the alternate reflexes of *p, *t, and *k.
Ramirez (1997b) describes processes in tuk that are very similar to the
alternation in tat. There are some interesting differences, however, such as
the aspiration of all alternate (fortis) forms: w (< *p’) alternates with [ph]; g
or ∅ (< *k’) alternates with [kh]; d [ɾ] (< *t’) alternates with [th].
The reconstruction of *C’ and *C systems in Proto-Tukanoan provides an
elegant explanation for all these alternations in et and wt languages. More-
over, given the similarity of processes observed in both wt and et languages,
these systems constitute a valuable piece of information about the morpho-
phonemics of Proto-Tukanoan.
3.3.4. Regional consideration with respect to *C’. We should note
that languages with glottalized stops often present themselves in clusters
correlating to a specific linguistic area. It is hard to know if this was the case
for Proto-Tukanoan, since there is still much to learn about the history of the
Northwest Amazon.
It is much more common to find full series of glottalized stops in languages
of the Andes than in those of the Amazon, though throughout Amazonia there
are numerous languages with some kind of glottal stop, such as implosives
(in Karajá [Rodrigues 1999], Proto-Arawan [Dixon 1999], Koaia [Aihken-
vald and Dixon 1999], Sabané, Nambikwaran [Araújo 2004]); ejectives (in
Cahuapanan [Wise 1999], Trumai, Wari’, Itonama, Movima [Aikhenvald and
Dixon 1999]); and laryngealized stops (in Wapishana (Cariban) [Ladefoged
and Maddieson 1996], Hup [Epps 2008]).
Rodrigues (2007) reconstructs a full series of glottalized stops in Proto-Tupí.
It is interesting that the correspondences he finds to support his reconstructions
are very similar to the ones presented here for Proto-Tukanoan laryngealized
stops. For instance, *p’ in Proto-Tupí is based on correspondences of p’ : p
: w : b : ∅; *t’ is based on correspondences of t : l : d : s : h : ∅; *k’ is based
on correspondences of k : ʔ : ∅ (Rodrigues 2007:172, 198).
Hup, a Nadahup (also referred to as Makuan) language of the Vaupes, is
described by Epps (2008) as having a complex series of laryngealized conso-
nants. The phonetic properties of these stops make the “following vowel . . .
consistently laryngealized; in other words, pronounced with ‘creaky voice’”
(Epps 2008:89). This post-laryngealization represents the crucial difference
between Hup and Proto-Tukanoan laryngealized stops. It is possible that Hup
laryngealized consonants are historically derived from a cluster of voiceless
stop plus a glottal stop, as Epps (2008:107) indicates. This might explain why
Hup has a full series of laryngealized consonants, including laryngealized
palatal glides, which are rare cross-linguistically.
One could wonder whether Proto-Tukanoan laryngealized stops might also
be derived from a similar cluster. Unfortunately, the data does not provide clear
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 301
21 In the preparation for this paper and in different publications, I have hesitated about the
exact reconstruction of *tj, *tj’, and *c sounds. Sometimes I have thought them to be alveo-
palatal affricates: *ts, *ts’, and *tʃ. However, *tj, *tj’, and *c sounds were chosen because they
do more justice to the range of existing reflexes and the assimilation processes in the history of
Tukanoan languages.
22 Kubeo currently has a phoneme /tʃ/, though it is a late development (Chacon 2012).
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
302 international journal of american linguistics
Context tan kub des yup bas tuk bar tat wan kar pis kue mai kor sek sio
*j #_ j j j j j j j j j j j j j j j j
V_V]STEM j j j j j s h h s s tʃ j j j j j
~V_V]STEM j j j j j j j j j j j j j j j j
Vlow_Vlow c
*s #_ h h s s c s h h s h h s s s s s
V_V h h s s c s h h s s tʃ s s s s s
tʃ/_i
*tj #_ h h j ts j j j j j j j t s s s s
s/_e s/_e tʃ/ _ i
V_V]stem h h s ts c s h h s s tʃ t s s s s
*c #_ h h s tʃ c s h h s h h t s s s tʃ
*tj’ #_ j h j j j j j j j j j j d j s’ s’
/ #_ i r h d d r d d r d r d r [l] j j s’ s’
(iii) *tj root-initially: cold, insect sp., peccary, tree. *tj root-medially:
capybara, hot, people, you all.
(iv) *c root-initially: dove, navel, to gather/collect. (No
correspondences root-medially.)
(v) *tj’ root-initially: black, blood, to chew, distant, egg, face, poison,
river, tongue/liver, to chew.
The sound with the most regular reflexes is *j, which, nevertheless, merged
with *tj, *s, and *c in root-internal position and non-nasal contexts in tuk,
bar, tat, wan, pir, kar, pis, tuy, and yur.
*s, *tj, and *c have very similar reflexes in many Tukanoan languages. The
data supporting the reconstruction of *tj are the correspondences of s in sio,
sek, kor, mai : t in kue : ts in yup : and j in des, bas, tuk, bar, tat, wan, pir,
kar, pis, tuy, and yur. In most cases, *tj merged with *s and *c in root-medial
position. The stop *c, on the other hand, was retained in sio, bas, pis, and
yup, and in kue *c has the reflex t (the same as for *tj, indicating a merger
of *c and *tj with *t). Otherwise, *c seems to have merged with *s in most
languages. For its part, *s was retained or changed to h in most languages
and merged with another *c in pis, bas, and in mai preceding i.
The key reflexes of *tj’ are an alveolar stop d or tap r, palatal glide j, the
unique s’ (~ <z>) sound in sio and sek, and kub h, the invariable reflex of Proto-
Tukanoan coronal stops and fricatives in this language. In sio and sek, the
contrast of *tj and *tj’ was retained (though indirectly as s vs. s’ respectively).
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 303
tan kub des bas/edu/ tuk wan kar kue mai kor sek sio gloss
mak
{ru’pa} {doba} {doa} ruhiedu duhi duhi ruí — (jũi) (jũ’i) (jũi) (jũ’i) to sit
~dahe hoewe ~dasi race dase dasa rase — jãse jãse jãse ~dahe toucan
riho hihe disi rice ɨ’se dɨse rise / risepu (jo) (ji’obo) (jɨ’o) (ji’obo) mouth
di(t)siyup jise
rupú hipo dipu riho dɨpo dapu rɨpo — ~tʃio ~siho ~sio ~s’iho head
In mai, *tj’ > d and kor *tj’ merged with *j. This merger also occurred in
most et languages, which also changed *tj > j. Still, *tj’ has different reflexes
before i in non-nasal environments, likely linked to its alveopalatal articula-
tion. The most curious fact is that it inverts the correspondences between
most et languages and mai. While mai had d and most et languages had j as
reflexes of *tj’ in the “elsewhere” environment, *tj’ preceding i in mai has j
where most et languages have d.
Before I analyze in greater detail the phonetics and phonology of this
class of sounds in Proto-Tukanoan, it is important to consider a few more
problematic cases. First, figure 17 shows words that present problematic corre-
spondences or possible suggestions for the reconstruction of different sounds.
head and lake are the only words in figure 17 with cognates across major
branches of the family; nevertheless, while sio and et present correspondences
similar to *tj’ reflexes, mai, kor, and sek have correspondences based on *tj,
*c, or *s. This may be an indication of areal convergence of these three wt
languages. The remaining words in figure 17 do not constitute a correspon-
dence set across the major subgroups of the family. 23
There are two other correspondence sets that are very intriguing because
they show subtle differences from what has been shown so far and could foster
an alternate explanation for how coronal consonants evolved. Consider the
words in figures 18 and 19.
23 The problem with the other correspondence sets is that some words are not cognates and
others are borrowings. For instance, ‘to sit’ is clearly a case of different cognates among et and
wt languages, given that there are no correspondences between wt /j/ and et /d/ or /r/, while the
correspondence between the medial /h/ in some et and /ʔ/ in wt is also not attested. One could
force an interpretation of the root for ‘to sit’ patterning with ‘toucan’. However, ‘toucan’ is very
likely a borrowing from an Arawakan language, e.g., yásene in Tariana, tʃaase in Piapoco, and
yase in Yukuna (see Huber and Reed 1992 and Aikhenvald 2001b). The word for ‘mouth’ is
clearly composed of two cognate sets.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
304 international journal of american linguistics
kub des yup bas tuk wan kar mai kor gloss
— pája sasa haya pa’sa pa’sa pasa — — float
tan kub des yup bas tuk wan kar sek kor gloss
~haha ~haha ~sada ~sasa ~yaca ~sada ~sada ~yasa sã’sa sã’sa kingfisher
The correspondence set in figure 18 is different from the most regular cases
of *j in word-medial position given in figure 16, because bas has j rather than
c between two low vowels. These are the only two examples found for this
pattern. One could account for this set by proposing a distinct proto-sound,
such as a laryngealized *j’ 24 or a combination of glottal stop plus *j. The
latter seems preferable because the stress pattern in the des form conforms to
reflexes of glottal stops in this language (see Miller 1999). 25
Finally, the data in figure 19 could be interpreted as reflexes of *tj’ in
root-medial position. This reconstruction is favored in particular for the d
([ɾ] ~ [n]) reflexes, along with s and h. However, caution is advised: first, it
is likely that some sort of assimilation and metathesis between the two stops
happened in the history of this word; second, the word could also have been
subject to borrowing and/or some sort of sound symbolism. These facts should
be kept in mind for future studies, when additional information can be taken
into consideration.
3.4.1. Phonetics and phonology of the diachronic evolution of coronal
consonants. In this section, I assess allophonic realizations of sibilants,
alveopalatal, and palatal sounds in Tukanoan languages, further supporting
the idea that Tukanoan languages retained a contrast of coronal consonants
until very recently.
The sounds *s, *c, and *tj present several cases of merger in different
Tukanoan languages. The phoneme c in bas (Gomez-Imbert 2004) is the
reflex of *s, *c, and instances of *tj in root-medial position. The Barasana
and Eduria dictionary (Barasana Literacy Committee 2009) acknowledges
24 Although laryngealized glides are not very common, there are some cases in different areas
of the world. Indeed, Hup, a neighboring Nadahup (also known by the derogatory “Makuan”
name) language, also has a laryngealized glide (Epps 2008).
25 This is evidence that the reconstruction of glottal stop (see 3.7) is still only partial, because
it takes into account only glottal stops occurring between two vowels, while the words in figure
18 indicate that the glottal stop can also occur before a glide.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 305
that in bas c can be realized either as [s], in general, or [tʃ] before u: ~cuka
‘baby’ is pronounced [tʃũkã]. The phoneme s also has a [ts] allophone in sek
(Johnson and Peeke 1962:89), which merged *tj with *s, and also in kar
(Metzger and Metzger 1973:128), which merged *s and *c. For pis, which
merged *s with *c, Gonzales de Pérez (2000) analyzes tʃ as a phoneme with
different realizations, the most common being [tʃ], [ts], and [s], depending
on still unclear phonological contexts and also depending on the speaker and
sociolinguistic variation (Gonzales de Pérez 2000). 26 For mai, which merged
*c and *tj with *s, Velie, Brend, and Powlison (1976:15) state that s is in free
variation with the voiceless dental fricative [θ], the latter being more restricted
to speakers with less contact with Spanish speakers, and that s has either [tʃ]
or [ʃ] allophones before i.
Interestingly, Koch-Grünberg’s (1914) transcriptions show that des, wan,
pir, pis, tuy, and the languages of the Pirá-Paraná River systematically had
[ts] and [tʃ] (<ts>, <tx>, and <tš> in Koch-Grünberg’s orthography) as the
equivalents for what has been analyzed as s more recently for most of these
languages. From his data, it is unclear whether these sounds were phonemes
or allophones of a single sound. Since Koch-Grünberg was a speaker of
German—which has contrasts of s, z, ts, and tʃ—it is very likely that his
transcriptions are phonetically accurate.
As an areal consideration on variation in the realization of alveolar and
palatal stops, it is also worth mentioning the realization of c in Hup. Epps
(2008) shows that Hup has a complex set of allophonic realization for c, such
as [ʃ], [tj], and [tʃ], and also presents debuccalization of fricatives, such as the
realization of ç as [jh] in syllable-final position. Debuccalization of fricatives
and affricates also occurred in the history of kub, ret, tan, yah, bar, tat, kar,
and pis in the Tukanoan family.
Reflexes of *j show that this sound also had overlapping realizations with
*s, *c, *tj, and *tj’. Most recent analyses of Tukanoan languages describe j
with very complex allophonic realizations, ranging from a palatalized alveolar
[dj], to the palatal affricate [ʤ], the glide [j], and the post-alveolar approximant
[ð̞] (see also Gonzales de Pérez and Rodrigues de Montes 2000). In tan, the
contrast of [ts] (as the allophone of s) with j is neutralized in word-initial posi-
tion in the speech of elders (Eraso 1999). If we assume that Proto-Tukanoan
*j also had a set of complex allophonic realizations, it is likely that the more
frontal and alveolar allophones of *j would inevitably overlap with *tj and *tj’
and the more posterior realization would overlap with *c and allophones of *s.
26 Recently, Katherine Bolaños (personal communication, 2009), after conducting fieldwork
with the lone Pisamira family that still uses the language on a daily basis, reported their more
frequent use of [s] rather than [tʃ], while among people who do not speak the language on a
regular basis, [tʃ] has been preserved. This is an interesting sociolinguistic situation, showing
that in Pisamira, /s/ is the more recent variant displacing [tʃ].
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
306 international journal of american linguistics
It is notable that in bas and mak, *j > c only between low vowels, while kub
has an allophonic realization of j as [ð̞] between low vowels (Chacon 2012).
It moreover appears that a dental or alveolar realization of j occurs over an
extensive linguistic area, from the Orinoco Basin to the Caquetá/Japurá River
and all northern tributaries of the Amazon in the same region. Ramirez (2001)
indicates that in Arawakan languages from the Japurá–Colombia divison, *j
has several different alveolar and palatal reflexes, such as j : dz : ts : tʃ : d :
ʤ . These are also attested in several Cariban languages, such as Taulepang,
Makushi (Migliazza 1967), and possibly Dekuana (or Ye’kwana/Ye’kuana),
and in the Yanomaman family, where *j in root-initial position was retained
in Yanomam and Yanomamɨ but changed to tʃ in Yanam (Ninam) and to ts
in Sanɨma (Migliazza 1972).
At some point in the history of different Tukanoan languages, the realiza-
tions of *tj and *tj’ overlapped with *s, *c, *j, *t’, and *t. The sound *tj has
reflexes realized with frication in all Tukanoan languages, which shows that
it probably had ts as one of its allophones before subsequent independent
changes. In wt languages, kue changed intermediate ts independently to s
before e and to t elsewhere and the remaining languages merged ts with *s.
*tj’ was retained until a late period and changed independently in all wt
languages. sio and sek changed *tj’ > s’ in a change sort of parallel to *tj > s.
In et languages, the pattern is much more complex, suggesting late re-
tentions of *tj word-initially and *tj’ elsewhere. In many languages, there
seems to have existed a late merger between *tj’ and *tj with *j. Exceptions
to this late merger are tan, ret, and yah where *tj > h, yup where *tj > ts
(and s before e), and kub where *tj and *tj’ > h. Because these languages are
geographic outliers with respect to the et language that merged *tj and *tj’
with *j, and because we know that yup (which did not undergo that merger)
is very closely related to des and sir (which have undergone the merger), 27
we have independent evidence to believe that the late changes involving *tj
and *tj’ in et are partially a result of areal diffusion.
Although the data are still too complex to allow more solid conclusions, it
seems likely that in et *tj > ts and *tj’ > dz. In a more recent period, *j merged
with ts in word-medial position in eastern-et languages. In word-initial po-
sition, contrast of ts, dz, and j was problematic, as well as that between dz
and reflexes of *t’ before i, which ended up causing several distinct types of
shared, independent, or contact-induced changes.
3.4.2. Morphophonemic alternations involving coronals. There are
synchronic morphophonological processes involving s or c and j in et lan-
guages that also support the complex relations between *j and *tj and *c
and *s.
27 By the time of Koch-Grünberg’s travels in the region, yup was being spoken in an entirely
different geographic area from its most closely related languages, des and sir.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 307
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
308 international journal of american linguistics
tan kub des bas tuk wan kar kue mai kor sek sio
w w > ∅ /_ṽ w W w w w w/u b/u w w w root-initial
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 309
28 It is likely that Proto-Tukanoan had nasalization as a feature of the entire syllable, as is
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
310 international journal of american linguistics
such as *n, *m, *w, and *j (creating the natural class of voiced stops). Hence,
nasalization could assimilate and spread to any voiced segment, merging **m
with *b < **p’, **n with *d < **t’. In sio and kor, however, this merger never
happened (in sek only d (< *t’) merged with *n).
Because Proto-Tukanoan had a consonantal system lacking a natural class
of voiced stops, it is likely that nasal spreading in the proto-language was
similar to the pattern found synchronically in sio and kor. Nasality was trig-
gered by nasal vowels and nasal stops. Syllables with nasal vowels would be
partially nasalized with a non-sonorant stop (*p’ã > [p’ã]) and completely
nasalized with a sonorant consonant (*/jã/ > [ɲã]). Nasal spreading was al-
lowed through sonorant segments and blocked by *C’ and *C.
3.7. Glottal sounds. The two glottal sounds reconstructed for Proto-
Tukanoan, *h and *ʔ, show systematic reflexes only in root-medial position.
The cognates that support the reconstruction of *h are big, pacu fish, heron,
and macaw, while black ink, bone, fish, i, buriti palm, and path support the
reconstruction of *ʔ.
Cross-linguistically, we find that h is often an unstable segment, a tendency
that holds true in Tukanoan languages. wt languages, as well as kub, tan, ret,
kar, tat, yur, tuy, bar, and pis, merged h with ∅. This may have caused a pull
chain in the long term, since in most of these languages *p and/or coronal
fricatives and affricates later changed to h.
We find similar variation in the reflexes of *ʔ, which was dropped in kub,
bas, mak, kue, kar, tat, bar, tuy, yur, pis, and mai (the only wt language
to drop this sound). Among these, kub, kar, tat, bar, tuy, yur, pis, and mai
also dropped *h.
3.7.1. Different sources of ʔ in Tukanoan. The cognates illustrating
the retention of *ʔ represent the instances of ʔ having the widest distribution
across the Tukanoan family. These are considered reflexes of the true Proto-
Tukanoan glottal stop, occurring in words that all had the invariable structure
*(C)VʔV. In descriptions of Tukanoan languages, we nevertheless find other
cases of ʔ that cannot be reconstructed to the proto-language.
The first case is that of words in certain et languages with the template
(C1)VʔC2V, reflexes of *C’ in root-medial position that left a laryngealized
trace on the preceding vowel (see 3.3 above). This laryngeal trace cannot be
reconstructed to wt languages except as reflexes of *C’. Conversely, there
are instances of ʔ in (C1)VʔC2V words in wt that cannot be reconstructed to
Proto-Tukanoan, such as the correspondences bas wese ‘garden’ : tuk wese
‘garden’ : sek we’se ‘outside’ : kor we’se ‘outside’.
Thus, sounds analyzed as ʔ in present Tukanoan languages have a hetero-
geneous history; tone and morphophonological boundaries are likely to be
additional sources of ʔ. It also has distinct phonetic properties: in kor it is
associated with high pitch, in sio and sek it is sometimes associated with stress,
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 311
and in some et languages it is associated with low tone and creaky voice on
the vowel. In languages that merged it with ∅, such as mai (see Wheeler 1992)
and bas, it is realized as low tone, while in kub and des it is realized as stress
on the preceding syllable. 29
Gomez-Imbert (2011) proposes the glottal stop in et languages as a default
consonant which appears in sequences of vowels (VV) to restore the canonical
CV syllable template. This is an interesting proposal, which seems to be the
case for the glottal stop *ʔ reconstructed for Proto-Tukanoan. Stenzel (2007)
analyzes a “glottal suprasegment” for wan, which is a nice synchronic solution
for the analytical problems imposed by (C)VʔCV words. The glottal supra-
segmental is restricted to the first mora (vowel) of words and is manifested at
least in some realizations as a low tone and creaky voice (additional acoustic
information can be found in Stenzel and Demolin [forthcoming]). Ramirez
(1997b) refers to a “glottal tone” for tuk and states that an epenthetic glottal
stop [ʔ] can occur at the beginning of words that start with a vowel, another
recognized “default” function of ʔ. In kue, there are some interesting cases
of h epenthesized at the beginning of words, such as wasp, snake, and pacu.
Finally, the role of *ʔ as a default consonant for defining syllable boundar-
ies can be indirectly observed in kar and tat words, where *ʔ > ∅ but other
types of sounds developed as default consonants, usually homorganic with
the articulation of the preceding vowel, as in the words tat õwa : tuk õ’a
‘bone’, tat pɨga : tuk pɨ’a ‘two’. 30
29 Although this study does not deal with prosody in Proto-Tukanoan, a comparative list of
tones and stress is being prepared and preliminary comparison corroborates the above observations.
30 Thus, cases of bas ∅ : tat g after /ɨ/ do not constitute a dropping of /g/ in bas but an
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
312 international journal of american linguistics
might have been in contact with these languages and influenced a p > h change in Tukanoan
languages. (p > h or ɸ occurred in several languages of the Guiana branch of the Cariban family,
of which Karihona is a member. Also, the Karihona ethnonym suggests a debucalization of a
bilabial stop: *karipona > karihona.) For this paper, the evidence is inconclusive.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 313
The earliest split within the wt group was the separation of kue. In fact,
Koch-Grünberg and Tastevin encountered this group on the lower Apaporis
River, evidence that there was also a large geographic separation between
kue and the other wt languages.
kue has unique sound changes in comparison to all other Tukanoan lan-
guages, such as t reflexes of both *tj and *c, and h in word-initial position
corresponding to ∅ in other Tukanoan languages. kue shares some changes
with et languages, notably the reflexes of *tj’, and nasal reflexes of *t’ and
*t in root-medial position. All the changes common to kue and et languages
are additionally shared with ret, tan, and yah, while other fundamental et
sound changes that are absent in ret, tan, and yah are not found in kue.
Koch-Grünberg (2005 [1909]) clearly points out the close contact between
kue and the et languages spoken along the Apaporis and Pirá-Parana Rivers,
especially tan, ret, and yah, so this is another likely case of diffusion of
changes between et and wt languages.
A second split in wt was the separation of mai. Independent changes in mai
include reflexes of *C’ > Cvoiced (and pre-glottalized) stops, *s > tʃ / _i, *tj’
> j or d, *w > b, and *ʔ > low tone (see Wheeler 1992).
The third split was the divergence of kor from sek and sio, which is evident
in the different reflexes of *tj’ and *C’, the development of an aspirated series
as reflexes of *C in kor, and a retroflex r as a reflex of *t’.
4.3. Diversification within et languages. Before the et languages
split into two major sub-branches, they all shared the following fundamen-
tal sound changes: *C’ > Cvoiced root-initially, *p’ and *t’ > b and d root-
medially in nasal contexts. *k’ was likely retained until more recently. Nasal
stops *m and *n had already merged with b and d, and at this point it is
likely that in all et languages, vowels preceding *C’ were laryngealized. In
addition, it is likely that *s, *c, *h, and *ʔ were still present, and *tj’ and
*tj might have evolved into *ts and *dz, giving the following phonological
inventory for Proto-et:
(7) Proto-et consonant inventory
p t ts c k h
b d dz k’ ʔ
s
w j
There are two major sound changes that mark the separation of the two
sub-branches of et: in eastern-et languages *j > s / V_V in all languages. On
the other hand, western-et languages share the fundamental unconditioned
change of *p’,*t’, and *k’ > b, d, and g root-medially. The fact that tan, ret,
and yah do not share any of these specific changes implies that they compose
a different sub-branch of et languages, here referred to as southern-et.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
314 international journal of american linguistics
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 315
32 The creation myths of many (but not all) et groups involve migration from a place in
the Southeast.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
316 international journal of american linguistics
33 The continua proposed here are to be understood as related more to the history of the
languages involved than to the history of the people that speak those languages. It is important
to keep in mind that ethnohistory is distinct from linguistic history; languages and ethnicity
do not necessarily correlate. What is more significant for the Tukanoan case is that an ethnic
group “A” might nowadays speak a language very similar to the language of another group “B.”
Nevertheless, the ethnohistories of groups “A” and “B” might indicate that they have distinct
descent lines, not supporting the linguistic evidence that their language are closely related.
What the ethonohistories may not tell us about are the linguistic processes of convergence and
language shift that lead two ethnically distinct groups to speak the same or similar languages.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 317
In contrast, the divergence of des/sir and bas/mak could have been strength-
ened because of their positioning in distinct linguistic sub-areas. While des and
sir have closer bonds to the Vaupes groups, bas and mak are deeply embedded
in the Pirá-Paraná sub-area. Consequently, they have incorporated different
areal features. For instance, des has pre-aspiration of voiceless stops as reflexes
of geminate-stop simplification, while bas and mak geminated all medial stops.
bas and mak have the isogloss r, which was diffused to tat and kar—two
of the other eastern-et languages in the Pirá-Paraná sub-area—while des
and sir lack it altogether. To explain why tan, ret, and yah also share the r
isogloss, along with kue, we certainly need a more adequate areal investiga-
tion, incorporating the idea of networks, beyond the idea of linguistic areas.
Lack of contact also led to cases of linguistic divergence, manifested by
retention of Proto-Tukanoan features and relatively independent evolution in
several phonological and grammatical respects. This is the case for kub and
tan and ret and yah, for instance, but it can be observed in wan as well. This
language has many changes independent from other eastern-et languages
that can be accounted for by independent evolution as well as contact with
Arawakan-speaking groups (see Waltz 2002, Stenzel 2013, and Stenzel and
Gomez-Imbert 2009). As a result, wan is the most distinct language in the
eastern-et sub-branch.
A final point of discussion of the prehistory of et languages involves con-
tact with other linguistic families. The Vaupes area has at least three major
linguistic families: Tukanoan, Nadahup (or Makuan), and Arawakan. The
study of contact between Nadahup and Tukanoan languages is just begin-
ning, but so far it has mostly focused on Tukanoan influence on Nadahup
languages (see Epps 2005; 2007). Regarding Tukanoan–Arawakan contact,
Aikhenvald (2002) outlines two types of contact relations—one leading to
indirect, multilateral diffusion of linguistic traits and the other to unilateral
diffusion of linguistic traits from tuk to Tariana (Arawakan).
It is likely that there was a period of contact between Tukanoan and Ar-
awakan languages before the arrival of the Tariana in the Vaupes area. A
case in point is kub, which has an older substratum of Arawakan influence
(see Goldman 2004 and Chacon 2013). The exact impact of this substratum
still needs to be assessed in kub and other Tukanoan languages, but it was
perhaps this period of influence that led to the emergence of relative cultural
homogeneity between Arawakan and Tukanoan language groups. During this
period of contact, the two linguistic families coexisted in the same area, but
the Arawakan groups dominated most of the Vaupes River and the Tukanoan
groups lived in a smaller area than they presently inhabit (see Wright 2005).
This period coincided with the arrival of the Tariana (Arawakan) and their
gradual domination of locations along the Vaupes River after warfare with
Tukanoan groups.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
318 international journal of american linguistics
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 319
REFERENCES
Aikhenvald, Alexandra Y. 2001a. Language awareness and correct speech among the Tariana
of Northwestern Amazonia. Anthropological Linguistics 43, no. 4:411–30.
. 2001b. DicionárioTariana–Portuguêse–Português–Tariana. Boletim do Museu Paraense
Emílio Goeldi: Série Antropológica, vol. 17, no. 1. Belém do Pará: Museu Paraense Emílio
Goeldi.
. 2002. Language Contact in Amazonia. Oxford: Oxford University Press.
Aikhenvald, Alexandra Y., and R. M. W. Dixon. 1999. Other small families and isolates. The
Amazonian Languages, ed. R. M. W. Dixon and Alexandra Y. Aikhenvald, pp. 341–68. Cam-
bridge: Cambridge University Press.
Alemán M., Tulio; Reinaldo López H.; and Marion Miller. 2000. Wirã ya, peamasa ya wereri-
turi: Desano–Español, diccionario bilingüe de 896 palabras. Bogotá: Editorial Alberto Lleras
Camargo.
Araújo, Gabriel Anutes de. 2004. A Grammar of Sabanê, a Nambikwaran Language. Utrecht:
LOT.
Ardila, Olga. 1998. Aspectos fonológicos de las lenguas tucano-orientales: Una visión compara-
tiva. Forma y Función, vol. 11. Bogotá: Universidad Nacional de Colombia.
Barasana Literacy Committee. 2009. Diccionario bilingüe: Eduria and Barasana–Español, Es-
pañol–Eduria and Barasana, comp. Paula S. Jones and Wendell H. Jones. Bogotá: Editorial
Fundación para el Desarrollo de los Pueblos Marginados.
Barnes, Janet. 1980. La reconstrucción de algunas formas del proto tucano-barasano-tuyuca.
Artículos en Lingüística y Campos Afines 8:37–66.
. 1999. Tucano. The Amazonian Languages, ed. R. M. W. Dixon and Alexandra Y.
Aikhenvald, pp. 207–26. Cambridge: Cambridge University Press.
Bellier, Iréne. 1994. Los Maihuna: Tucano Occidentales. Guia Etnográfico del Alto-Amazonas,
vol. 1, ed. Fernando Santos and Frederica Barclay. Quito: Ulpiano Páez.
Beuchat, Henri, and Paul Rivet. 1911. La famille Betoya ou tucano. Memoir de la Societé
Linguistique de Paris 17:117–36, 162–90.
Brinton, Daniel G. 1891. The American Race: A Linguistic Classification and Ethnographic
Description of the Native Tribes of North and South America. New York.
. 1892. Studies in South American Native Languages. Proceedings of the American
Philosophical Society, vol. 30. Philadelphia.
Campbell, Lyle. 2004. Historical Linguistics: An Introduction. 2nd ed. Cambridge, Mass.: The
M.I.T. Press.
Cabalzar, Aloisio. 2008. Filhos da Cobra de Pedra: Organização Social e Tarjetórioas Tuyukas
no rio Tiquié. São Paulo: Fundação Editora da Unesp.
Chacon, Thiago. 2009. Nasality in Tukanoan: Synchrony and diachrony. Paper presented at the
Annual Meeting of the Society for the Study of the Indigenous Languages of the Americas,
San Francisco.
. 2012. The phonology and morphology of Kubeo: The documentation, theory and de-
scription of an Amazonian language. Ph.D. dissertation, University of Hawai’i.
. 2013. Kubeo: Linguistic and cultural interactions in the Upper Rio Negro. Upper Rio
Negro: Cultural and Linguistic Interaction in Northwestern Amazonia, ed. Patience Epps and
Kristine Stenzel, pp. 403–40. Rio de Janeiro: Museu Nacional and Museu do Índio Funai.
. Forthcoming. The reconstruction of laryngealization in Proto-Tukanoan. Laryngeal
Features in the Americas, ed. Leo Wetzels. Leiden: Brill.
Cook, Dorothy M., and Linda L. Criswell. 1993. El idioma koreguaje (Tucano occidental).
Bogotá: Instituto Lingüístico de Verano.
Criswell, Linda, and Beverly Brandrup. 2000. Un bosquejo fonológico y gramatical del si-
riano. Lenguas indígenas de Colombia: Una visión descriptiva, ed. María Stella González de
Pérez and María Luisa Rodríguez de Montes, pp. 395–417. Bogotá: Instituto Caro y Cuervo.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
320 international journal of american linguistics
Dixon, R. M. W. 1999. Arawá. The Amazonian Languages, ed. R. M. W. Dixon and Alexandra
Y. Aikhenvald, pp. 292–306. Cambridge: Cambridge University Press.
Dupont, Carlos M. 1988. Armonía nasal en la lengua koreguaje. Cuadernos de lingüística his-
pánica, pp. 105–25. Tunja: Universidad Pedagógica y Tecnológica de Colombia.
Ehret, Christopher. 1976. Linguistic evidence and its correlation with archaelogy. World
Archaelogy 8:5–18.
Epps, Patience. 2005. Areal diffusion and the development of evidentiality: Evidence from Hup.
Studies in Language 29:617–50.
. 2007. Birth of a noun classification system: The case of Hup. Language Endangerment
and Endangered Languages: Linguistic and Anthropological Studies with Special Emphasis
on the Languages and Cultures of the Andean–Amazonian Border Area, ed. Leo Wetzels,
pp. 107–28. Leiden: Research School of Asian, African, and Amerindian Studies (CNWS),
Leiden University.
. 2008. A Grammar of Hup. Berlin: Mouton de Gruyter.
Eraso, N. 1999. Cambios tonales en la palabra verbal. Congreso de lingüistica amerindia y cri
olla, vol. 6, no. 1, ed. J. Landaburu and T. Rojas Curieux, pp. 72–95. Bogotá: Editorial Artes.
Fallon, Paul D. 2002. The Synchronic and Diachronic Phonology of Ejectives. New York:
Routledge.
Franchetto, Bruna, and Elsa Gomez-Imbert. 2003. Review of The Amazonian Languages e dited
by R. M. W. Dixon and Alexandra Y. Aikhenvald. IJAL 69:232–38.
Goldman, Irving. 2004. Cubeo Hehenawa Religious Thought: Metaphysics of a Northwestern
Amazonian People. New York: Columbia University Press.
Gomez-Imbert, Elsa. 1993. Problemas en torno a la comparación de las lenguas tucano-
orientales. Estado actual de la clasificación de las lenguas indígenas de Colombia, ed. Maria
Luisa R odríguez de Montes, pp. 235–67. Bogotá: Instituto Caro y Cuervo.
. 1999. Variétés tonales sur fond d’exogamie linguistique. Cahiers de Grammaire,
vol. 24, pp. 67–94. Toulouse: ERSS, Université Toulouse–Le Mirail.
. 2004. Fonología de dos idiomas Tukano orientales. Amerindia, vol. 29. Paris: CELIA.
. 2011. La famille tukano. Dictionnaire des langues du monde, pp. 1454–60. Paris:
Presses Universitaires de France.
Gomez-Imbert, Elsa, and Stephen Hugh-Jones. 2000. Introducción al estudio de las lenguas del
Piraparaná (Vaupes). Lenguas indígenas de Colombia: Una visión descriptiva, ed. Maria
Stella González de Pérez and Maria Luisa Rodríguez de Montes, pp. 321–56. Bogotá: Insti-
tuto Caro y Cuervo.
Gonzalez de Pérez, María Stella. 2000. Bases para el estudio de la lengua pisamira. Lenguas
indígenas de Colombia: Una visíon descriptiva, ed. Maria Stella González de Pérez and Maria
Luisa Rodríguez de Montes, pp. 373–94. Bogotá: Instituto Caro y Cuervo.
González de Pérez, Maria Stella, and Maria Luisa Rodríguez de Montes, eds. 2000. Lenguas
indígenas de Colombia: Una visión descriptiva. Bogotá: Instituto Caro y Cuervo.
Huber, Randal Q., and Robert B. Reed, comps. 1992. Vocabulario comparativo: Palabras selec-
tas de lenguas indígenas de Colombia. Bogotá: Instituto Lingüístico de Verano.
Hugh-Jones, Christine. 1979. From the Milk River: Spatial and Temporal Processes in the North-
west Amazon. Cambridge: Cambridge University Press.
Jackson, Jean. 1983. The Fish People: Linguistic Exogamy and Tukanoan Identity in Northwest
Amazonia. Cambridge: Cambridge University Press.
Johnson, Orville, and Stephen Levinson. 1990. Gramática secoya. Cuadernos Etnolinguísticos,
vol. 11. Quito: Instituto Linguístico de Verano.
Johnson, Orville, and Catherine Peeke. 1962. Phonemic units in the Sekoya word. Studies in
Ecuadorian Indian Languages, ed. Catherine Peeke and Benjamin Elson, pp. 78–95. Norman,
Okla.: Summer Institute of Linguistics.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
tukanoan family classification 321
Kaye, Jonathan. 1970. The Desano verb: Problems in semantics, syntax and phonology. Ph.D.
dissertation, Columbia University.
Koch-Grünberg, Theodor. 1913. Betoya-Sprachen Nordwestbrasiliens und der angrenzenden
Gebiete. Anthropos 8.
. 1914. Betoya-Sprachen Nordwestbrasiliens und der angrenzenden Gebiete. Anthropos 9.
. 2005 [1909]. Dois Anos entre os Indígenas: Viagens no noroeste do Brasil [Two years
among the Indians: Travels in the Brazilian Northwest] (1903/1905). Manaus: EDUA and
FSDB.
Ladefoged, Peter, and Ian Maddieson. 1996. The Sounds of the World’s Languages. Oxford:
Blackwell.
Lavoie, Lisa M. 2001. Consonant Strength: Phonological Patterns and Phonetic Manifestations.
New York: Garland.
Maddieson, Ian. 1984. Patterns of Sounds. Cambridge: Cambridge University Press.
Malone, T. 1986. Proto-Tucanoan and Tucanoan genetic relationship. Ms., Instituto Linguístico
de Verano, Colombia.
Mason, J. A. 1950. The languages of South American Indians. Bureau of American Ethnology
Bulletin 143:157–317.
Metzger, Lois, and Ronald G. Metzger. 1973. Fonología del Carapana. Sistemas fonológicos de
idiomas colombianos, vol. 2. Bogotá: Ministerio de Gobierno. <http://www.sil.org/americas/
colombia/pubs/abstract.asp?id=11733>.
Metzger, Ronald G., comp. 2000. Marĩ yaye mena Carapana, yaia yaye mena Español macãrĩcã
tuti = Carapana–Español, diccionario de 1000 palabras. Bogotá: Editorial Alberta Lleras
Camargo.
Michael, Lev; Christine Beier; Stephanie Farmer; Gregory Finley; and John Sylak. 2011.
Diccionario Bilingüe Máíjɨ́ki–Castellano y Castellano–Máíjɨ́ki (versión agosto 2011).
Ms., Máíhɨ́ki Project.
Migliazza, Ernesto. 1967. Grupos lingüísticos do território federal de Roraima. Atas do simpósio
sobre a biota Amazônica, vol. 2, Antropologia, pp.153–73. Belém do Pará : Museu Paraense
Emilio Goeldi.
. 1972. Yanomama grammar and intelligibility. Ph.D. dissertation, Indiana University,
Bloomington.
Miller, Marion. 1999. Desano Grammar: Studies in the Languages of Colombia. Dallas: Sum-
mer Institute of Linguistics and the University of Texas at Arlington.
Morse, Nancy L.; Jay K. Salser; and Neva Salser, comps. 1999. Diccionario ilustrado bilingüe:
Cubeo–Español, Español–Cubeo. Bogotá: Editorial Alberto Lleras Camargo.
Piaguaje, Ramón; Elías Piaguaje; Orville E. Johnson; and Mary Johnson, comps. 1992. Vocabu-
lario Secoya. Quito: Instituto Lingüístico de Verano.
Ramirez, Henri. 1997a. A fala Tukano dos Ye’pâ-Masa: Vol. 2: Dicionário. Manaus: Inspetoria
Salesiana Missionaria da Amazônia, CEDEM.
. 1997b. A fala Tukano dos Ye’pâ-Masa. Vol. 1: Gramática. Manaus: Inspetoria Salesi-
ana Missionaria da Amazônia, CEDEM.
. 2001. Línguas Arawak da Amazônia Setentrional: Comparação e descrição. Manaus:
Editora da Universidade do Amazonas.
Rodrigues, Aryon D. 1999. Macro-Jê. The Amazonian Languages, ed. R. M. W. Dixon and Al-
exandra Y. Aikhenvald, pp. 165–206. Cambridge: Cambridge University Press.
. 2007. Proto-Tupí. Línguas e Culturas Tupí, ed. Ana Cabral, Suelly Arruda Câmara,
and Aryon Dall’Igna Rodrigues, vol. 1, pp. 167–203. Campinas: Editora Curt Nimuendajú.
Smothermon, Jeffrey R., and Josephine H. Smothermon, comps. 1993. Masa ye, gawa ye rãca
ãmara tuti: Macuna–Español, diccionario de 850 palabras. Bogotá: Editorial Alberto Lleras
Camargo.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions
322 international journal of american linguistics
Sorensen, Arthur Peter, Jr. 1967. Multilingualism in the Northwest Amazon. American An-
thropologist 69:670–84.
Stenzel, Kristine 2005. Multilingualism: Northwest Amazonia revisited. Paper presented at
the Second Congress on Indigenous Languages of Latin America (CILLA), Austin, Texas.
. 2007. Glottalization and other suprasegmental features in Wanano. IJAL 73:331–66.
. 2013. A Reference Grammar of Kotiria (Wanano). Lincoln: University of Nebraska
Press.
Stenzel, Kristine, and Didier Demolin. Forthcoming. Traços Laringais em Kotiria e Wa’ikhana
(Tukano Oriental). Fonologia: Teorias e perspectivas. Porto Alegre: EdiPUCRS.
Stenzel, Kristine, and Elsa Gomez-Imbert. 2009. Contato Lingüístico e Mudança Lingüística no
Noroeste Amazônico: O caso do Kotiria (Wanano). Revista da ABRALIN 8:71–100.
Tovar, Antonio, and Consuelo Larrueca de Tovar. 1984. Catálogo de las lenguas de la América
del Sur. Madrid: Editorial Gredos.
Velie, Daniel. 1975. Bosquejo de la fonología y gramática del idioma Orejón (coto). Datos Etno-
Lingüísticos, no. 10. Lima: Instituto Lingüístico de Verano.
Velie, Daniel, and Virginia Velie. 1981. Vocabulario Orejón. Lima: Ministerio de Educación
and Instituto Lingüístico de Verano.
Velie, Daniel; Ruth Brend; and Esther Powlison. 1976. Fonología del Orejón. Datos Etno-
linguísticos: Colección de los Archivos del ILV 2008. Lima, Peru: Instituto Linguistico de
Verano. < http://www.sil.org/americas/peru/ show_work.asp?id= 52278>.
Waltz, Nathan E. 2002. Innovations in Wanano (Eastern Tucanoan) when compared to Pirata-
puyo. IJAL 68:157–215.
Waltz, Nathan E., and Alva Wheeler. 1972. Proto-Tucanoan. Comparative Studies in Amerin-
dian Languages, ed. Esther Matteson, pp. 19–49. The Hague: Mouton.
Waltz, Nathan E.; Paula Simmons de Jones; and Carolyn de Waltz, eds. 2007. Diccionario
bilingüe: Wanano o Guanano–Español, Español–Wanano o Guanano. Bogotá: Editorial Fun-
dación para el Desarrollo de los Pueblos Marginados.
Wheeler, Alva. 1987a. El pueblo siona del río Putumayo, Colombia. Vol. 1: Etnología,
gramática, textos. Bogotá: Instituto Linguistico de Verano.
. 1987b. El pueblo siona del río Putumayo, Colombia. Vol. 2: Diccionario. Bogotá:
Instituto Linguístico de Verano.
. 1992. Comparaciones lingüisticas en el grupo Tucano Occidental. Estudios Compara-
tivos Proto Tucano, ed. Stephen H. Levinsohn, pp. 17–53. Bogotá: Alberto Lleras Camargo.
. 2000. La lengua Siona. Lenguas indígenas de Colombia: Una visión descriptiva, ed.
María Stella González de Pérez and María Luisa Rodríguez de Montes, pp. 181–98. Bogotá:
Instituto Caro y Cuervo.
Wise, Mary Ruth. Small language families and isolates in Peru. The Amazonian Languages,
ed. R. M. W. Dixon and Alexandra Y. Aikhenvald, pp. 307–40. Cambridge: Cambridge Uni-
versity Press.
Wright, Robin. 2005. História Indígena e do Indigenismo no Alto Rio Negro. Campinas, São
Paulo: Mercado de Letras/Instituto Socioambiental.
This content downloaded from 129.82.28.144 on Thu, 31 Dec 2015 12:54:47 UTC
All use subject to JSTOR Terms and Conditions