PRE-ORALIZED NASAL CODAS IN MAMAINDÊ And the Oral Vowel Enhancement Proposal

Full text

PRE-ORALIZED NASAL CODAS IN MAMAINDÊ
And$the$Oral$Vowel$Enhancement$Proposal$
david m eberhard
sil
!"#$%&'(#!%")
)
The various Nambikwara nations gained short lived notoriety in the 1940s thru the writings of the
French anthropologist Claude Levi Strauss, who dubbed them the ‘earth people’, for their characteristic aversion
to sleeping anywhere other than on the ground. These ‘earth peoples’ constitute a loose-knit cluster of bands,
native to an area covering some 50,000 square kilometers in northwest Brazil, not far from the Bolivia border.
When considered altogether, they number some 1200 individuals. However, each Nambikwara band considers
itself quite independent from the others, with its own individual identity and language. Due to distinct variables
of social contact, these small languages find themselves at various stages along the road to endangerment and
extinction.
In the midst of this rather fragile family is Mamaindê, spoken by some 250 individuals. It can be
classified as a member of the Guapore cluster of the Northern Nambikwara branch of the Nambikwara language
family. The Mamaindê live in 4 communities, which are having increased contact with the outside world,
resulting in a high rate of bilingualism with Portuguese. A few of the younger children are only passive speakers
of the language, signaling that the highly endangered stage has already begun.
As far as its phonology is concerned, Mamaindê has proved to be quite fascinating. Some of the major
points of interest include 4 contrastive vowel qualities (oral, nasal, creaky voice, and nasal & creaky voice -
yielding 16 vowel phonemes); it also has an intricate place feature spreading process from vowels to consonants
(requiring a new look at feature geometry); a tone system with 2 lexical tones, tone plateauing and floating tones;
a stress system that is predicated on syllable weight, syllable position, and 4 lexical strata; and a coda position
which routinely employs biphasic oral/nasal segments.

)
%*+,(#!-,.)%/)#0!.).#'&1)
It is this last point which will be the focus of our study today. The behavior of the nasal coda is one of the more
salient features of the Mamaindê sound system. This behavior results in the formation of three biphasic
oral/nasal segments in syllable final positions: labial [bm], coronal [dn], and dorsal [gŋ]. It will be shown that in
the Mamaindê language these complex codas require a phonology which treats their primary articulation as nasal
[dn, bm, and gŋ] rather than oral [dn, bm, gŋ], and thus the process involved is one of pre-oralization as opposed to
post-nasalization. In this presentation, my objective will be to explore three alternatives for the analysis of this
pre-oralization process, first from a phonological (autosegmental) perspective, and then two analyses from a
phonetic implementation approach. This paper will show that one of the phonetic alternatives is the more
satisfactory of the three. It will also demonstrate that adopting a specific set of parameters proposed by Wetzels
(2008) is helpful in predicting the type of phonetic enhancement found in this language.
)
&!.#$!*'#!%")%/)%$234"2.23)*!502.!().,67,"#.)
Biphasic oral/nasal segments (as well as their counterpart nasal/oral segments) have been documented in a large
number of languages and language families around the world, many of these endangered. They seem to be
concentrated in four general areas; South America, Africa, Australia, and Austronesia (Ladefoged, 1993:165-166;
Wetzels, 2009:1; see also Rosendall, 1992, and Mills, 1984 for data on the African languages Gwari and Senoufo
respectively). However, other parts of the world are not without them. Chen and Clumeck give examples of
post-stopped nasals in Cantonese (1975:128). In Borneo, the Bonggi language (Boutin 1993, 2000), as well as the
Malayic and Land Dayak language families (Blust 1997; Court 1972; Scott; 1964) are reported to have this feature,
which is also commonly referred to in the literature as pre- or post-nasalization, or, when looked at another way,
pre-oralization, denasalization, or pre-plosion.

Oral-nasal contours are also quite prevalent among Amazonian languages. Wetzels (2009), whose recent
study focuses on nasal/oral contour consonants in South America, gives examples from eight Amazonian
languages with this feature: Yuhup, Dâw, Wari, Wansajot, Kaingang, Maxacali, Barasana, and Mebengokre. In
Amazonia, the language families which have been most documented in terms of oral/nasal contour segments
appear to be the Macro-Je family (specifically Apinaye, Kaingang, Maxacali, and Xokleng) and the Maku family
(Yuhup and Dâw). To this list the Nambikwara family must also be added (specifically Southern Nambikwara
and all the Northern Nambikwara lects, namely Mamaindê, Negarotê, Latundê and Lakondê), as it becomes
evident that oral/nasal contours occupy a prominent place in most of the phonologies of Nambikwara languages
outside of Sabanê.
In the Nambikwara languages, biphasic oral/nasal segments are the result of oralization, being output
variants of simple nasals. These languages also limit the oralization of nasal stops to the coda position, and thus
only the pre-oralized nasals are present. Post-oralized nasals, such as those found in Kaingang and Maxacali, are
not a Nambikwara phenomenon, these being more characteristic of Macro-Je languages (See Wetzels 2009). Of
the three pre-oralized nasal forms found in the family - [bm], [dn] and [gŋ] - Southern Nambikwara (Kroeker,
2001:79), Mamaindê (Eberhard, 2003a), and Negarotê1 possess all three, while Latundê and Lakondê (Telles,
2002:56-57) have only the labial [bm] and coronal [dn] varieties, apparently lacking the dorsal counterpart.2
Antunes (2004) does not mention any pre-oralized nasal variants for Sabanê. Mamaindê pre-oralized nasals,
however, can be distinguished from all the other Nambikwara biphasic segments by the unique manner in which
they acquire their place features (we will touch on this later - see Eberhard, 2009).
)
)
)
)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))
8)#9:)":;<=>?@)A<?<)B>C:D) E=>C) CF) >GH) 7<C<IHA@)A<?<J<D:K)#9:D:)?G>) L:B?D) <=:) M:=F)DICIL<=)<HA)?9:=:)<=:)<)
HNCJ:=)>E)":;<=>?@)LIMIH;)GI?9IH)?9:)7<C<IHA@)B>CCNHI?F)G9>)B>H?IHN:)?>)ND:)D>C:)":;<=>?@)E>=CDK)
O)#:LL:D)POQQORSTUSVW)H>?:D)?9<?)?9:)X=:UD?>XX:A)H<D<LDY)G9IL:)B>CC>H)IH)3<Z>HA@Y)<=:)[NI?:)=<=:)IH)3<?NHA@K)

#0,)7272!"&\)*!502.!()%$234"2.23).,67,"#.)
)
We will start our discussion of the MD biphasic oral/nasal segments by defining their phonological status. Here
is a sampling of the abundant examples found in the language, organized according to the variant involved:
!"#$%!
/kadeunta-latʰa-wa/ [gadeubmdalatʰwa] it is alive)
/waun-latʰa-wa/ [waubmlatʰwa] it is red)
/taun-tu/ [daubmdu] tail)
/leunʔ-tu/ [leubmʔdu] tapir)
)"&ŋ]!
/sin-tu/ [sigŋdu] meat)
/jalinʔja̰-tu/ [jaligŋʔja̰ɾu] puberty flute music)
/walinʔ-tu/ [waligŋʔdu] anteater: type ‘mirim’)
/wainsi-tu/ [waigŋtʃiɾu] medicine)
"'(%!
/jalakwatun-tu/ [jalakwadudndu] howler monkey)
/ʔon-latʰa-wa/ [ʔodnlatʰwa] he is lazy
/nahon-sa̰-tu/ [nahodnsa̰ɾu] sweet beverage)
/nakajanʔ-tu/ [nagajadnʔdu] person/indian)
!
While these sequences appear at first glance to be the result of intrusive stops, I intend to show that
there is no epenthesis occurring here at all – it is simply a case of gestural timing. In articulatory terms, the
opening of the velum is simply delayed until after the tongue has assumed its place of articulation for the nasal.
One of the basic questions being asked in this study is whether this timing delay is phonological or phonetic in
character.
First I give distributional evidence that these biphasic coda segments in Mamaindê are in fact simple
nasals underlyingly. The forms listed here show biphasic segments alternating with simple nasal stops. The
conditioning environment has to do with syllable position and the quality of the preceding vowel. When the
simple nasal consonant is in the coda position and preceded by a tautosyllabic oral vowel, it is systematically
realized as a biphasic nasal.

PQ8W /alain-sihtaʔ/ [a.laigŋ.sih.r̥aʔ])
cross.over-!"K%$&,$K#%K&. in order for him/her to cross over
)
PQOW /naʔtun-hĩʔ/ [naʔ.dudn.hĩʔ])
to.be.full-#0,"K&. he/she was full, then…)
)
)
PQ]W /kateun-kʰatoʔ/ [ka.deubm.kʰə.ɾəʔ] )
be.alive-#0,"K.. he/she was still alive, then…)
In the intervocalic position, however, or after nasal vowels, only the simple nasal appears. Note that the addition
of the first person marker /a/ in the first three forms forces the resyllabification of the nasal from coda to onset,
and the biphasic variant is lost.
PQ^W /alain-a-sihtaʔ/ [a.lai.na.sih.ɾ̥aʔ] )
cross.over-.8U!"K%$&,$K#%K&.) ) in order for me to cross over
)
PQSW /naʔtun-a-hĩʔ/ [naʔ.duː.na.hĩʔ] )
to.be.full-.8U#0,"K&. I was full, then…
PQTW /kateun-a-hĩʔ/ [ka.deu.na.hĩʔ] )
be.alive-.8U#0,"K&. I was (still) alive, then…
(07) /wanũn-sa/ [wanũnsa]
good-speech good speech

(08) /na-ʔmĩn/ [naʔmĩn]
3Poss-skin his skin
(09) /hãn-latha-wa/ [hãnlathwa]
white-S3-Decl it is white
Oral stops, on the other hand, are never found alternating with these contour segments. This
distribution pattern, then, provides sufficient evidence for establishing the simple nasal consonant as the lexical
form for each of these biphasic nasals.
Additional support comes from the phonetic data. In most tokens during natural speech, the nasal phase
of biphasic stops is from 10% to 100% longer than the oral phase (range of oral phase: from .042ms - .090ms /
range of nasal phase: from .081ms - .110ms). But it is in careful speech, when the syllables involved receive added
emphasis, that the difference between the phases is even more pronounced, the extra length invariably falling on
the nasal phase, and never on the oral phase. This results in tokens such as the one shown here for the word
/ankalathawa/ ‘it is precious’, where the duration of the nasal murmur can be up to 4 times as long as the oral
phase (oral phase: 118ms - .230ms /nasal phase: .231ms - .447ms - ).

AN$ EXAMPLE$ OF$ THE$ FORM$ /ANKALATHAWA/$‘IT$IS$ PRECIOUS’,$[

AGNGALATHWA],$ WITH$THE$ SYLLABLE$ CONTAINING$ THE$ BIPHASIC$
STOP$ BEING$ STRESSED$ AND$ EMPHASIZED.$HERE$ THE$ ORAL$ PHASE$ IS$ .118MS$ WHILE$ THE$ NASAL$ PHASE$ IS$ .447MS.$$THIS$ IS$
CHARACTERISTIC$OF$THESE$EMPHATIC$FORMS,$WHERE$THE$NASAL$PHASE$CAN$BE$UP$TO$4$TIMES$AS$LONG$AS$THE$ORAL$PHASE.$
This particular recording was extracted from a text by Joaquim Mamaindê, who is answering my
question of why they practice the female puberty rite – and the emphasis he places on his answer is telling not
only phonologically but also culturally. He basically had a one word answer – /ankalathawa/ - ‘because it is
precious’ - with emphatic lengthening on the stressed syllable, and as you will notice in this spectogram, the extra
length falls directly on the nasal portion of the biphasic segment.
There have not been any examples found where the opposite is true – with the oral phase being
significantly longer than the nasal phase. So phonetics supports phonology in giving primary status to the nasal
portion of these segments.3
)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))
])In other Amazonian languages, such as Dâw, where these biphasic segments have been analyzed as simple nasals at the
phonemic level, the nasal phase has also been measured as being significantly longer than the oral phase. S. Martins (2004: 47)
cites V. Martins (1994) who measured the nasal phase of Dâw as being from 30% to 50% longer than the oral phase.
)

The last task in laying the groundwork for what follows is to determine whether the nasal VOWELS
are contrastive or derived in this language. As we have mentioned, the pre-oralized codas always occur after oral
vowels, and never after nasal vowels. Although that might lead us to believe that nasal vowels are derived, there
is still plenty of data which shows that nasal vowels must be seen as contrastive in current Mamaindê speech.
P8QW /han-latʰa-wa/  [hadnlatʰwa] it wiggles)
P88W /hãn-latʰa-wa/  [hãnlatʰwa] it is white)
P8OW /nahĩʔ/  [nahĩʔ] then…)
P8]W /nahiktu/  [nahiktu] his hand)
P8^W /hũtu/  [hũɾu] savannah wolf- (lobo guara))
P8SW /huktu/  [huktu] bow)
P8TW /han-aʔ-wa/  [haːnaʔwa] it does not wiggle)
P8VW /hãn-aʔ-wa/  [hãːnaʔwa] it is not white)
)
Note that the difference between the last two forms here can only be attributed to a contrastive
difference in vowel quality.
The two conditions, then, to this pre-oralization process in Mamaindé are the orality of the nucleus, and
the coda position of the nasal consonant.
There are several plausible ways to account for this oralization process. The first one is phonological in
nature, appealing to autosegmental spreading rules, while the last two will be phonetic, appealing to articulatory
gestures.
)
)
)

%$23!_2#!%"R)2")2'#%.,67,"#23)2((%'"#))
)
In autosegmental terms, this process could be viewed as a simple case of oral or [-nasal] spreading. When a nasal
occurs in a coda position preceded by an oral vowel, the orality (or [-nasal] feature) from the vowel spreads to the
nasal.
ORAL$SPREADING$RULE$
[-nasal] [+nasal]
V C
σ
Notice that after the spreading of the [-nasal] feature, instead of suffering delinking, the nasal coda
retains its association to [+nasal]. This results in a biphasic nasal segment, one which is doubly associated to a [-
nasal] and a [+nasal] feature.
ORAL$SPREADING$DEMONSTRATED$IN$THE$ROOT$FORM$$/SUN/$‐$‘HIT’$!$
)
[-nasal] [+nasal]
s u n
σ
output: [sudn]
)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))
^)#9:)>XX>DI?:)>E)?9ID)DX=:<AIH;)B>NLA)<LD>)>BBN=))U)E>=):`<CXL:Y)?9:)D9<=IH;)>E)abH<D<Lc)?9=>N;9>N?)?9:)=9FC:)P<D)IH)
$‘hard’).* However,* since* both* the* vowel* and* the* coda* in* ) <=:) NHA:=LFIH;LF) H<D<L) ?>) J:;IH) GI?9) IH)
7<C<IHA@Y)I?)G>NLA)J:)ICX>DDIJL:)?>)A:?:=CIH:)G9:?9:=)?9:)D9<=IH;)>E)<)abH<D<Lc)E:<?N=:)IH)?9:D:)B<D:D)ID)AN:)?>)
?9:)%(5)>=)?>)H<D<L)DX=:<AIH;K))))

In this account however, the terminal features of the Soft Palate node would need to be binary, [+nasal]
and [-nasal], since there must be some [oral] or [-nasal] feature to spread to the coda. While not a standard
approach, the spreading of the [-nasal] feature is not a new idea. A similar analysis of oralized nasals has already
been suggested by Anderson (1974:272-74) for certain Macro-Ge languages such as Maxakali, Kaingang, and
Apinaye. Oral spreading in general has also been allowed for by Wetzels and Goedemans (2008:335) in their
development of the NASDAT database for nasal harmony systems. And studies continue to emerge which
utilize the [-nasal] feature (for instance, Kaiser’s (2008:290) analysis of nasal spreading in Paraguayan Guarani, and
Silva’s (2008:304) account of nasal harmony in Desano).
However, this use of binary nasality does not minimize the arguments to the contrary, which see this as
an unfortunate result in light of the fact that feature theory continues to move further in the direction of unary
features. Strong arguments against binary features for nasality have been offered by Piggot, 1992, Kenstowicz,
1994:492, 504; and Steriade, 1995, among others.5 Therefore, if possible, an alternative way of handling the
formation of these biphasic segments is preferred, one which does NOT resort to the oral spreading approach.
)
)
)
)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))
S)5I;;>?)P8ddOWY)>H)?9:)>?9:=)9<HAY)BL<ICD)?9<?)aUH<D<Lc)DX=:<AIH;)ID)H>?)H:B:DD<=FK)0:)<H<LFe:D)H<D<L)DFD?:CD)GI?9)
J>?9)DICXL:)H<D<LD)<HA) X=:H<D<LIe:A) D:;C:H?D) <D)9<MIH;)H>)H<D<L)B>HD>H<H?D)<?) <LLY) JN?) DICXLF) aUB>H?c)D:;C:H?D)
NHDX:BIEI:A)E>=) H<D<LI?FK) !H) ?9:)B>H?:`?) >E) H<D<L) M>G:LDY)?9:F) ?<Z:) >H)?9:) H<D<L) [N<LI?F) >E) ?9:)<Af<B:H?)M>G:LK)!H)
X=:H<D<LIe:A) D:;C:H?DY) ?9:I=) H<D<LI?F) ID) >HLF) <) X9:H>C:H>H) >E) ?9:) <=?IBNL<?>=F) <XX<=<?NDK) *N?) IH) 7<C<IHA@) 9ID)
<H<LFDID)ID)H>?)<A:[N<?:Y)E>=)I?)ID)[NI?:)BL:<=)?9<?)?9:)H<D<L)[N<LI?F)>E)?9:)H<D<L)B>HD>H<H?D)IH)?9ID)L<H;N<;:)B<HH>?)
J:)A:=IM:A) E=>C)?9:)H<D<LI?F)>E)<HF)<Af<B:H?)H<D<L)M>G:LY)DIHB:)C<HF)>E)?9:C)>BBN=)IH)?9:)B>H?:`?)>E)>=<L)M>G:LD))
P/‘hisUbrother’).* If* an* intervocalic* coda* in* a* particular* morpheme* is* treated* as* underlyingly* nasal*
P/‘hitUNEG’),* then* when* that* same* C>=X9:C:) >BBN=D) IH) <)AIEE:=:H?)B>H?:`?Y) ?9:) NHA:=LFIH;) H<?N=:) >E)
?9<?) B>A<) CND?) D?ILL) J:) B>HDIA:=:A) <D) H<D<L) ) P//$ ‘* hitU.]K5$.K&,(3’WK) #9:D:) H<D<L) B>A<D) <=:) ?9:=:E>=:)
NHA:=LFIH;) H<D<L) B>HD>H<H?D) G9IB9)J:B>C:) L:DD) H<D<L) G9:H) ?9:F) >BBN=)IH) DX:BIEIB) :HMI=>HC:H?DK) g9:H) ?9:F) <=:)
IH?:=M>B<LIBY)?9:F) <=:) =:<LIe:A)<D)DICXL:) H<D<LDY) <HA)G9:H) E>LL>G:A) JF) <)B>HD>H<H?) ?9:F) <=:)=:<LIe:A)<D) B>H?>N=)
X=:U>=<LIe:A)H<D<LDK).?:=I<A:)P8ddSR8^dW)<LD>)<=;N:D) D?=>H;LF) E>=)<)C>H>M<L:H?)H<D<L)E:<?N=:)PD::)<LD>) h:HD?>GIBe)
8dd^R^dOWK)

%$23!_2#!%"R)O)50%",#!()2((%'"#.)
)
I will now turn to several phonetic accounts of these contour segments in other languages, where the formation
of these forms is seen as a phonetic behavior, instead of a phonological process. Both of these will make use of
Keyser and Stevens (2006) idea of a language specific Phonetic Enhancement Component, where secondary
articulatory gestures add cues to the acoustic signal in order to enhance or make more prominent specific
phonological contrasts which are in danger of losing their saliency.
(%&2)532(,),"02"(,7,"#).#$2#,61)
We will first consider Keyser and Steven’s phonetic approach to biphasic nasals in several Southern
Central Australian languages. In these languages (Arabana, Aranda, and WaNgaNuru as well as in Olgolo, a
North Queensland language), according to Keyser and Stevens, there are up to 6 distinctive places of articulation
for these biphasic coda nasals (unfortunately the authors did not provide us with any specific data).
The oral phase of these contour segments is homorganic with the nasal phase, and nasality is
(apparently) not distinctive on vowels. In the light of this information, Keyser and Stevens view the biphasic
segments as a result of a phonetic enhancement of the place of articulation of the coda. Specifically, the addition
of an oral phase to the coda allows the vowel formants to be unperturbed by nasalization, thus enabling the
formant transitions to be distinctive, highlighting the contrast found in the place features of the following coda. I
will label this the CPE, or the Coda Place Enhancement strategy.
The Coda Place Enhancement strategy, however, runs into problems with the Mamaindê data. Note
that in the list of Mamaindê forms provided, the labial biphasic form [bm] is found only following the high back
rounded vowel of a diphthong. The dorsal forms are encountered only after the high front vowel, and the
coronal forms are found only following /a/, /u/, or /o/.
"#$%!)*++*,-(&!./0.!*1!.20.3!
/kadeunta-latʰa-wa/ [gadeubmdalatʰwa] it is alive)
/waun-latʰa-wa/ [waubmlatʰwa] it is red)
/taun-tu/ [daubmdu] tail)
/leunʔ-tu/ [leubmʔdu] tapir

)"&ŋ]))*++*,-(&!.-.!
/sin-tu/ [sigŋdu] meat)
/jalinʔja̰-tu/ [jaligŋʔja̰ɾu] puberty flute music)
/walinʔ-tu/ [waligŋʔdu] anteater: type ‘mirim’)
/wainsi-tu/ [waigŋtʃiɾu] medicine)
"'(%))*++*,-(&!./.4!.*.4!/('!.0.!
/jalakwatun-tu/ [jalakwadudndu] howler monkey)
/ʔon-latʰa-wa/ [ʔodnlatʰwa] he is lazy
/nahon-sa̰-tu/ [nahodnsa̰ɾu] sweet beverage)
/nakajanʔ-tu/ [nagajadnʔdu] person/indian)
It becomes obvious then that the place features of the codas are derived from or in some way contingent
upon the previous vowel. While this much is clear, the specific details of this vowel place feature spreading
would actually require another paper, and we obviously don’t have time for that here. Suffice it to say that by
the use of some basic underspecification, as well as a feature geometry structure which combines elements of
both the Clements and the Halle/Sagey models, the place features of the Mamaindê coda are entirely predictable.
While not the main purpose of this study, I will quickly step thru the highlights of an analysis which
can predict the place features of these segments. The first step in predicting the place features of the Mamaindê
coda is to adopt a feature geometry model that is a combination of both the Halle and the Clements models –
Hallian by viewing all vowels as Dorsal, and Clement-like in separating the height node from the place node.
Within the Dorsal node, we can use [+/-back] to differentiate the front and back vowels. By viewing the front
vowels as [-back] Dorsals, it facilitates their connection to a pre-velar consonant (and all of the Dorsal codas in
MD are actually pre-velars), which would also be characterized as a [-back] Dorsal.6
)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))
T)2L?9>N;9)H>?)DX:BIEI:A)IH)?9:)A<?<)X=:D:H?:AY):M:=F)IHD?<HB:)>E)4Z4)E>LL>GIH;)4I4)ID)=:<LIe:A)<D)/k̯/.)

7272!"&\)-%g,3)/,2#'$,)6,%7,#$1)
) ) ))-)
)
) ) ->B) ) )
) )
) ) -5L<B:) ) 2X:=?N=:)
)
3<JI<L) ) &>=D<L) ) ab4U9I;9cV)
)
) ) ab4UJ<BZc)
)
Secondly, we need to underspecify the vowel features in such a way that only the Dorsal, Labial, and
Height features are needed (this includes an even further underspecification of the second place vowel of
diphthongs to where these only have 2 features each).
7272!"&\)-%g,3)/,2#'$,.)Pg!#0)&!50#0%"6.W))
))))))))))))))))))))
&>=D<L)
3<JI<L)
2X:=?N=:)
)
aJ<BZc)
)
a9I;9c)
I)
5!
!
6!
:)
5!
!
5!
<)
6!
!
5!
>)
6!
7!
5!
N)
6!
7!
6!
-N)
!
7!
6!
-I)
5!
!
6!
)
Once Mamaindê feature organization has been worked out, the actual assimilation process is simply a
matter of spreading the VPlace node from a high vowel to an underspecified coda. Note that while the height of
the vowels is a required part of the environment, the height feature is separate from Place and does not spread.
)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))
Vg:)B>NLA)fND?)<D):<DILF)ND:)(L:C:H?D)E:<?N=:D)E>=)M>G:L)9:I;9?)a>X:H8cY)a>X:HOc)9:=:K)!)<C)>X?IH;)E>=)?9:)E:<?N=:)
a9I;9c)DIHB:)I?)ID)C>=:)B>CC>HLF)ND:AK)

"#$%&'()'%*+,'(-#,'%./01(,+$'(2%+*3-'14'0*%$56(
! ! ! ! !σ!
!
! ! ! 8! ! 9!
!
!
! ! !!!!!!!!!!8*:/+-:! !!!!!!!!!!!!!9;+/:2!
!
!
! <=21>012!
!
!
"6?-&?%!!!!!!!!!!!8;+/:2! ! !
))
While there are many details to the above process which are not addressed in full here, the important
thing to note is that the place features of the Mamaindê coda are predictable. This means that the Coda Place
Enhancement strategy is not worthy of any more consideration in the case of the Mamaindê biphasics, since the
place features of the Mamaindê coda nasals are not contrastive but derived, and thus not in any need of
enhancement.
-%!(,&).#%5),"02"(,7,"#)%$))%$23)-%g,3),"02"(,7,"#)
Two more phonetic enhancement approaches to biphasic nasals are offered by Wetzels (2008), in a study of
oral/nasal contour segments in Amazonian languages. Here Wetzels cites examples from Kaingang, Maxacali,
Dâw, and Wansajot, some of which had previously been analyzed as requiring the spreading of the [-nasal].
Wetzels builds on the general idea of phonetic enhancement introduced by Keyser and Stevens, as well as on
earlier observations made by Steriade (1993) regarding the need to maintain contrastive nasality on vowels in
several of these Amazonian languages.
Depending on the language, biphasic segments could be analyzed as being either underlyingly nasal, or
underlyingly oral. Wetzels proposes that both of these cases can be accounted for by way of phonetic
)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))
i)!)<C)H>?)IHBLNAIH;)<)BL<DDIB<L);:H:=<?IM:)=NL:)9:=:)DIHB:)I?)G>NLA)J:)[NI?:)B>CXL:`)<HA)H>?)9:LXENL)>=)IHDI;9?ENL)IH)
<HF)G<FK))

enhancement, whether it involves a nasal phase being added to an oral Consonant, or an oral phase being added
to a nasal Consonant. In the former case, where biphasic segments are underlyingly oral, Wetzels contends that
languages may add a nasal phase to these segments for the purpose of enhancing the voicing of the coda
consonant. This of course would be useful in languages which exhibit a contrast between voiced and voiceless
stops, and in environments where that contrast is in danger of losing its perceptual saliency (in syllable coda
position, word finally, or in the dorsal place of articulation). This type of enhancement is labeled by Wetzels
(2008:9-11) as Voiced Stop Enhancement (VSE).
In the latter case, where these contour segments are underlyingly nasal, an oral phase may be added to a
nasal consonant in order to make the orality of oral vowels more prominent preceding nasal codas. This is useful
in languages where a primary contrast already exists between oral and nasal vowels and in contexts where this
contrast might be in danger of being neutralized (such as before nasal codas). This type of enhancement is
referred to by Wetzels as Oral Vowel Enhancement (OVE).
@ABCADE!AFG<F9AHAFB!;<I<HABAIE!
Wetzels (2008:11) goes on to outline seven parameters to predict whether such oral/nasal contour sequences are
the byproduct of Voiced Stop Enhancement or Oral Vowel Enhancement. These parameters have been
simplified for our purposes below:
WETZELS’$PARAMETERS$FOR$PHONETIC$ENHANCEMENT$OF$BIPHASIC$ORAL/NASAL$SEGMENTS:$
A.) /OralV/~/NasalV/: Does the language have contrastive oral/nasal vowels?
If not, the OVE (Oral Vowel Enhancement) is excluded.
B.) P/B~M: Does the language exhibit a contrast between nasal and non-nasal obstruents? 9
If so, VSE (Voiced Stop Enhancement) is excluded. This is due to an expectation that enhancement
features will not be used as primary contrastive features in a language. Specifically, one would not
expect (partial) nasalization to be used as an enhancement feature for voiced obstruents in a system
that already uses the nasal feature as a primary feature to distinguish nasal consonants from non-
nasal consonants.
)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))
d)#9:)G>=AIH;)>E)?9ID)X<=<C:?:=)9<D)J::H)B9<H;:A)E=>C)?9:)>=I;IH<L)?>)J:??:=)EI?)?9:)7<C<IHA@)A<?<K)

C.) g>d>b: Does the language display a preference in the place of articulation in regards to the biphasic
stops, such that dorsals > (are preferred over) coronals > (are preferred over) labials?
If so, then the VSE is predicted. This is because the necessity for secondary gestures to enhance the
primary [voice] feature is greatest for dorsal stops, for which the area between the glottis and the
point of constriction is smallest, while it is less urgent for alveolar and coronal stops and least
urgent for labial stops.
D.) [Vbm] vs [mbV]: Are the contour nasal stops found in the coda as opposed to the onset? If so, OVE is
predicted.
E.) [mb]/[bm] ~ [b]: Do the biphasic segments alternate with voiced stops? If so, then VSE is predicted.
F.) [mb]/[bm] ~ [m]: Do the biphasic segments alternate with nasal stops? If so, then OVE is predicted.10
G.) [NasalV] [NasalC]: Are nasal consonants restricted to syllables with nasal vowels? If so, then OVE is excluded.
@ABCADE!;<I<HABAIE!<;;DJAK!BL!H<H<JFKM!
When we apply the above parameters to Mamaindê, we find that all but one of the seven Mamaindê settings
predict an OVE approach (and the one that does not make this prediction is actually inconclusive). Here we have
a breakdown of how Mamaindê fares under each of the previous parameters:
A. ) OralV~NasalV: Mamaindê DOES have contrastive oral/nasal vowels.
Therefore, according to the model, OVE cannot be excluded
B.) P/B~M: Mamaindê DOES contrast nasal stops with non-nasal stops.
Therefore, VSE is excluded.
)))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))
8Q)!)GILL)<DDNC:)?9<?)?9ID)X<=<C:?:=)GILL)<LD>)<XXLF)?>)aJCc)D:[N:HB:D)DNB9)<D)?9>D:)E>NHA)IH)7<C<IHA@K)2BB>=AIH;)
?>)X<=<C:?:=)PAWY)>H:)G>NLA)<B?N<LLF):`X:B?)aJCc)B>A<)D:[N:HB:D)?>)J:)C>=:)E=:[N:H?)?9<H)aCJc)>HD:?)D:[N:HB:D)
IH)B<D:D)G9:=:)%-,)ID)>X:=<?I>H<LK)

C.) g>d>b: Mamaindê DOES prefer both coronal and dorsal biphasic segments to their labial
counterpart, which is found much less frequently. This fact agrees in part with parameter (c), seemingly
predicting VSE. However, in Mamaindê this parameter is skewed by allophonic variation, as we have
already seen that all coda nasals, regardless of place, can be derived from the coronal nasal (or an
underspecified nasal). The labial and dorsal place features of nasal codas are a direct result of vocalic
place spreading from the nucleus to the coda. Thus, the scarcity of the labial biphasic segments is due
not to the distribution of the different contrastive places of articulation of the nasal consonant, but to
the distribution of the previous vowels which derive them. This parameter, then, is inconclusive.
D.) [Vbm] vs [mbV]: Mamaindê biphasic segments are ONLY found in the coda.
Therefore, OVE is predicted
E.) [mb]/[bm] ~ [b]: In Mamaindê, biphasic segments do NOT alternate with voiced stops.
Therefore, VSE is not predicted
F.) [mb]/[bm] ~ [m]: In Mamaindê, biphasic segments DO alternate with nasal stops.
Therefore, OVE is predicted
G.) NasalV/NasalC: In Mamaindê, nasal consonants are not limited to syllables with nasal vowels.
Therefore, OVE is not excluded.
Note that in 6 of the 7 parameters (one of them which was inconclusive), the cumulative settings for the
Mamaindê data clearly point to the Oral Vowel Enhancement strategy. This general consistency in the
parameters shows that they function quite well together in predicting the motivation for the biphasic segments
in this language.
In Mamaindê, then, it seems safe to conclude that the OVE approach treats the oral phase of the
oral/nasal segments as a means of enhancing the contrast between oral and nasal vowels when these are followed
by a nasal coda. The reasoning behind this is simple: it is much more difficult to maintain an oral articulation of

a vowel preceding a nasal coda than when it is followed by an oral (or partially oralized) coda. Therefore, in
languages which contrast oral and nasal vowels, nasal codas pose environments where such a contrast may
potentially be lost. The delayed opening of the velum, then, would facilitate the production of oral vowels in
these contexts, as well as enhancing their perceptual saliency. As we have already seen, Mamaindê does have
contrastive nasality on vowels. Thus the oral portion of the biphasic nasal coda would not need to be accounted
for by phonological rule, but can instead be explained by principles of phonetic enhancement which are in turn
based on the contrastive nature of the oral/nasal vowels themselves.
A final detail that supports this conclusion is the behavior of lexical items with lengthened oral vowels.
When these lengthened oral vowels precede a nasal coda, the biphasic nasal is not realized. Therefore we have
/waan/ ‘return’  [waan], while [wadn]* is ungrammatical. The OVE analysis actually predicts this exception
to the rule, since the extra lengthening of the oral vowel is sufficient enhancement of its orality, thus not
necessitating the oral phase on the nasal.
The overall benefit of the OVE, then, is that it allows us to account for these segments in a way that is
phonetically grounded, requires very little theoretical power, and does not resort to binary oral/nasal contrastive
features.
B?2!L8A!/==1*/:?!-(!12+/>-*(!>*!*>?21!F/$#-N,/1/!+/(&0/&2O!
I will now make a few brief comments regarding the preceding parameters in light of the Nambikwara family as
a whole (excluding Sabanê for the moment since it does not exhibit any contour oral/nasal segments). In the
majority of the Nambikwara languages studied thus far, such as Southern Nambikwara seen here, the responses
to the 7 parameters are identical to the responses we have seen for Mamaindê. In Latundê and Lakondê,
however, the result for parameter C changes, becoming conclusive, whereas it was inconclusive in Mamaindê.
Parameter C.) g>d>b: Latundê and Lakondê violate the hierarchy predicted in this parameter, having labial
and coronal biphasic stops but no dorsal contour segments whatsoever. Although the labial varieties are
allophonic variations of coronals just as they were in Mamaindê, the complete lack of the dorsal biphasic

segment shows that this parameter is clearly violated in these languages. Therefore, for these 2 languages, this
parameter is actually conclusive and VSE is not predicted.
The outcome of this exercise is to determine that for the family as a whole, OVE is overwhelmingly the
predicted strategy.
These last two Nambikwara languages, Latundê and Lakondê, also add an additional twist to the
analysis of biphasic nasals. In these languages, when pre-oralization occurs, the nasal coda can be partially
oralized following oral vowels, or optionally, it can be completely oralized, losing its nasal phase entirely
(apparently these options are in free variation in these languages).
Latundê
/aun-tan-ti/ [aubmdãni] ~ [aubdãni] ‘it escaped’
/sin-tãn/ [sidndãn] ~ [siddãn] ‘it was smoking’
Lakondê
/on-wi-ta-ta/ [odn ʋiɾaɾa] ~ [odʋiɾaɾa] ‘I dont want to eat meat’
For these languages then, Latundê and Lakondê, the above parameters are simply not sufficiently
detailed enough to predict the extent of the oralization. While not included in Wetzel’s parameters, this graded
or continuous nature between partial and total oralization can actually be seen as a direct result of a general
characteristic of phonetic enhancement, namely, variability, something which is not easy to capture in a
phonological approach, but would actually be expected in a phonetic one (Keyser and Stevens, p40).
!

;GLFABJ9!AFG<F9AHAFB!–!LFA!PAEBQIA!LI!B@LR!
)
Finally, I would like to briefly address a question that I have had in regards to phonetic enhancement, one that
does not seem to be altogether clear in the literature at this point, yet it does have a bearing on the Mamaindê
data. This has to do with the number of distinct gestures expected in cases of phonetic implementation.
In their work on Quantal Theory, Stevens and Keyser (2009:9) mention two types of enhancement.
The first type involves a gesture which defines the contrastive feature, and a non-contrastive gesture which is
used for the purpose of enhancement. The enhancing gesture is superimposed over the primary gesture, thus
requiring these two gestures to be distinct (such as lip rounding being used to enhance the [-anterior] feature of
the alveopalatal sibilant in English).
This explains their analysis of biphasic nasals in Southern Central Australian languages (Keyser and
Stevens, 2006:50), which, as we have already observed, accounted for the oral phase of these segments as a
secondary gesture to enhance the different place features which are contrastive in these languages. In specific
terms, a gesture of the velum is appealed to in order to enhance tongue gestures used for place. But when we look
at the Mamaindê data, it becomes obvious that this requirement for two distinct gestures, a primary one and an
enhancing one, simply does not fit, for in this language a single gesture (the closed velum) is employed for both
functions
While the first type of enhancement appears to be more common in the literature, Stevens and Keyser
(2009:10) mention a second type as well. This type occurs in regions adjacent to the contrastive feature being
targeted, and appears to allow for a single gesture to be used for both feature defining and feature enhancing
purposes. This is permissible apparently because, according to the authors, certain gestures can have a particular
acoustic property on consonants, and a somewhat different acoustic property on vowels. If this reading is
correct, then it is this second type of enhancement that Mamaindê employs, where the closed state of the velum
during the oral phase of the contour coda enhances the acoustic properties of that same gesture in the preceding
vowel.

(3%.!"6)
)
In closing, my first task was to show that the biphasic segments in Mamaindê are in fact underlyingly nasal.
Various approaches to these Mamaindê contour segments were then compared, the autosegmental approach of
Oral Spreading versus several phonetic implementation approaches, the last of which was Wetzels proposal of
Oral Vowel Enhancement. We have seen how this later view is clearly an improvement in that it adds
physiological motivation to the process, requires very little theoretical power, and avoids the use of binary nasal
features. I have also pointed out how the Mamaindê data adds additional support to the set of parameters
proposed by Wetzels, which have already been helpful in accounting for similar biphasic segments in other
Amazonian languages.
A final observation must be made regarding an assumption which was not made explicit at the
outset. This paper has touched on the interface between phonetics and phonology, in the sense that
enhancement gestures are phonetic, while the contrasts that motivate them are phonological. Enhancement
strategies, then, require the phonetic component to access crucial phonological information. If this is true, then it
begs the question whether the opposite may also be true – whether the phonological component may access
phonetic information. While not covered in this paper, this second possibility could conceivably be realized by
the use of Optimality constraints which encode phonetic detail. This has been addressed by Hayes and Steriade
(2004:1), among others, who have discussed ways in which “phonological constraints can be rooted in phonetic
knowledge”. As an example, here we have a hypothetical OT analysis of the Mamaindê biphasic segments,
employing a markedness constraint that is clearly related to articulatory tendencies, and which is already an
attested constraint in the literature (*VoralN states that a nasal may not follow a tautosyllabic oral vowel – see
Kager, 1999:28) It also makes use of faithfulness constraints which attempt to preserve a certain amount of
phonetic detail within the nasal coda, such as Nasal closure, Nasal murmur and Nasal release. Although these are
clearly not as variable as the acoustic signal, they are a way in which OT can incorporate some phonetic detail
into its constraints, thus linking phonology and phonetics.

OT ACCOUNT OF MAMAINDÊ BIPHASIC NASALS (RANKING: MAXV > VORALN. > MAX N)
Here is another hypothetical tableau, this time for a word with a nasal vowel.
/mũn/
MAX-V
*VoralN
MAX-Nrelease
MAX-Nmurmur
MAX-Nclosure

a.mũn
b.mũdn
*!
c.mũ
*!
*
*
d.mun
*!
*
e.mudn
*!
*
f.mũdnd
*!
*
g.mũd
*!
*
*
While it has not been my intent or place to compare various interface models between these two
components of grammar, the basic assumption which the Mamaindê data seems to require is that there is an
inter-relatedness between phonetics and phonology to such a degree that crucial information must be allowed to
cross-over between the two.
)
4DNH4)
MAX-V
*VoralN
MAX-Nrelease
MAX-Nmurmur
MAX-Nclosure
)
a.sun
!
ST!
!
!
!
)
b.sudn
!
!
!
!
S!
)
c.su
!
!
ST!
S!
S!
)
d.sũn
ST!
!
!
!
!
)
e.sund
!
ST!
S!
!
!
)
f.sudnd
!
!
ST!
!
S!
)
g.sud
!
!
ST!
S!
S!

]BIBLIO:
Araujo, Gabriel Antunes de. 2004. A Grammar of Sabane, a Nambikwaran language.
Amsterdam: Vrije Universiteit. Doctoral Thesis.
Blust, Robert. 1997. Nasals and Nasalization in Borneo. Oceanic Linguistics 36/1. 149-179.
Boutin, Michael E. 1993. Bonggi phonemics. In Phonological descriptions of Sabah languages: Studies from ten languages, ed. by Michael E.
Boutin and Inka Pekkanen. 107-30. Sabah Museum Monograph, 4. Kota Kinabalu: Sabah Museum.
Chen, Matthen and Harold Clumeck, 1975. Denasalization in Korean: a search for universals. In Nasalfest – papers from a symposium on
nasals and nasalization, ed. by Charles Ferguson, Larry Hyman, and John Ohala, 125-131. Standford, CA: Standford University.
Court, Christopher. 1972. The phonological structure of the word in Mntu Land Dayak. University of Syndey, Oceania Linguistic
Monographs 15. 86-94.
Eberhard, David. 2009. Mamaindê Gammar: a Northern Nambikwara language and its cultural context. Amsterdam: LOT Publications.
Doctoral Thesis.
Hayes, Bruce and Donca Steriade. 2004. Introduction and the phonetic basis of phonological markedness. In ‘Phonetically Based Phonology’.
Eds. Bruce Hayes, Robert Kirchner, and Donca Steriade. Cambridge University Press:Cambridge, UK.
Kager, René. 1999. Optimality Theory. Cambridge: Cambridge University Press.
Keyser, Samuel J. and Kenneth N. Stevens. 2006. Enhancement and Overlap in the Speech Chain. Language 82/1. 33-63.
Kroeker, Menno. 2001. A descriptive Grammar of Nambiquara. International Journal of American Linguistics 76(1). 1-87.
Ladefoged, Peter. 1993. A Course in Phonetics. Fort Worth, Texas: Harcourt Brace Jovanovich.
Mills, Elizabeth. 1984. Senoufo Phonology, Discourse to Syllable. Dallas: Summer Institute of Linguistics.
Piggot, Glyne. 1992. Variability in feature dependency: the case of nasality. Natural Language and Linguistic Theory 10. 33-78.
Rosendall, Heidi James. 1992. A Phonological Study of the Gwari Lects. Dallas: Summer Institute of Linguistics.
Scott, N.C. 1964. Nasal consonants in Land Dayak (Bukar-Sadong). In In Honour of Daniel Jones: Papers contributed on his eightieth birthday,
ed. by David Abercrombie 432-436. London: Longmans.
Silva, Cacio. 2007. Análise Fonológica da Língua Yuhup. ms.
Steriade, Donca. 1993. Closure, release, and nasal contours. Phonetics and
Phonology 5 (1993), pp. 401–470.
Stevens, K. N., & Keyser, S. J. Quantal theory, enhancement and overlap. Journal of Phonetics (2009), doi:10.1016.
Telles, Stella. 2002. Fonologia e gramatica Latunde/Lakonde. Vrije Universiteit: Amsterdam.
Wetzels, Leo. 1995. Oclusivas intrusivas em Maxakali. In Estudos fonologicos das linguas indigenas brasileiras, ed. Leo Wetzels, p93-100. Rio
de Janeiro, Brazil: Eidtora UFRJ.
Wetzels, W. Leo. 2008. Thoughts on the Phonological Interpretation of {Nasal,Oral} Contour Consonants in some Indigenous Languages of
South-America. In Abordagens em Fonética e Fonologia. Estudos Auditivos, Acústicos e Perceptivos. Modelos de Análise Fonológica de Ontem e de
Hoje. Número Especial da Revista Alfa, 52/2. Universidade Estadual Paulista (UNESP).
)
)