You are on page 1of 26

THE STABILITY OF PHONOLOGICAL FEATURES WITHIN

AND ACROSS SEGMENTS


THE EFFECT OF NASALIZATION ON FRICATION*

MARIA-JOSEP SOL
Universitat Autnoma de Barcelona


Abstract
This paper argues that the articulatory-acoustic stability of phonological features
may be affected not only by concurrent features, but also by features in adjacent
segments which may coincide in time due to coarticulatory overlap. Specifically,
the paper illustrates how frication may be endangered by concurrent and
coarticulatory nasality. We review aerodynamic and acoustic evidence showing
that fricatives tend to be impaired and become unstable with co-occurring
nasalization. Then we examine the stability of fricatives when they come in
contact with nasality in adjacent segments. An experiment is described where
aerodynamic and acoustic data were obtained for fricative + nasal sequences at
slow and fast rates. The results show that anticipatory velophrayngeal opening
during the acoustic duration of the fricative vents the high oral pressure required
for audible frication, thus providing support for the claim that the same physical
principles disfavoring the combination of frication and nasality within a segment
are at play when these features combine across segments. It is argued that the
instability of frication when combined with nasalization may be at the origin of
a number of phonological patterns.



1. Introduction
It is known that the articulatory-acoustic stability of phonological
features may be endangered by their combination with other features within
segments. In this paper we suggest that the stability of features may be affected
not only by concurrent features, but also by features in adjacent segments
which may coincide in time due to coarticulatory overlap. Specifically, we
examine the stability of fricatives when they combine with nasality within a
segment and when they come in contact with nasality in adjacent segments
with varying degrees of coarticulatory overlap.
In this paper we focus on the phonetic grounding of the combination of
features within a segment and across segments, and the implications for

*
Work supported by grants HUM2005-02746, BFF2003-09453-C02-C01 from the Ministry of
Science and Technology, Spain, and by the research group 2005SGR864 of the Catalan
Government. The insightful suggestions and comments of Daniel Recasens and an anonymous
reviewer are gratefully acknowledged.
MARIA-JOSEP SOL
42
phonological patterns. Much work is available on the physical, physiological
and auditory principles that account for the combination of features within a
segment, and characterize feature co-occurrence restrictions, phonological
universals (i.e. why some feature combinations are universally preferred) and
system gaps (i.e. why certain feature combinations do not occur) (e.g. Ohala
1983; Sol 2002a; Westbury & Keating 1986). How speech features affect
each other across segments is at the origin of restrictions on the sequencing of
sounds, the likelihood that segments follow one another, and feature spreading
or blocking in consonant/vowel harmony. The interaction between features in
contiguous segments, however, does not end there. Sound change and
phonological processes may also result from the way in which features
combine across segments. After all, segments change in certain segmental
contexts, due to interaction with features in neighbouring segments, but not in
others. Work on how features combine across segments has thus far focused on
perceptual and physiological constraints. For example, Kawasaki (1986, 1992)
addressed the perceptual discriminability of sound combinations.
Coarticulation theories have explored the contextual restrictions on the
temporal extension and magnitude of nasality, labiality, laryngeal and lingual
gestures for consonants and vowels (e.g. Hardcastle & Hewlett 1999; Huffman
& Krakow 1993). In contrast, the role of aerodynamic factors in the
combination of features across segments has been little addressed (but see
Ohala 1981, 1997a; Sol 2002b). Furthermore, not much emphasis has been
placed on the relationship between how features combine across segments and
how they influence each other when they co-occur within a segment.
This paper goes beyond previous work in suggesting that aerodynamic
factors may be at the origin of the incompatibility of features and phonological
patterning. In addition, it is suggested that the physical and physiological
principles that account for the paradigmatic arrangement of features can also
explain their syntagmatic arrangement. We will explore how the stability of
features may be endangered by their combination with other features within
segments and by features in contiguous segments due to coarticulatory effects.
We hypothesize that the physical and physiological constraints which are at
work within a segment will also play a role in the combination of features
across segments. If this is the case, it follows that features that do not combine
well within a segment, due to physical and physiological constraints, are not
likely to combine in adjacent segments since the same constraints will apply
when the two features overlap in time.
Before addressing our hypothesis we need to consider the theoretical
framework underlying the stability of features. According to Stevens (1972,
1989) quantal theory, gradual and continuous articulatory movements and
aerodynamic variation may have a categorical acoustic-auditory result; that is,
some variations along the continuum involve abrupt acoustic changes whereas
certain others do not. Features that are used in speech are those that fall in
stable regions in the acoustic/auditory space to allow for articulatory,
contextual, rate and prosodic variation while preserving a robust acoustic
THE STABILITY OF PHONOLOGICAL FEATURES

43
result. The quantal nature of speech has been illustrated by varying features in
isolation along the articulatory (or acoustic) dimension and observing the
acoustic/auditory result of such variations (e.g. laryngeal adduction and
presence of voicing; movement of the velum and percept of nasalization). The
range of allowable articulatory/aerodynamic variation within which the percept
of the feature is not affected will define the stability of the articulatory-acoustic
correlation.
Elaborating on this view, we may consider that this stable range may
vary, i.e. may be expanded, reduced, or shifted, with co-occurring features in
the same segment and, as claimed in this paper, in adjacent segments.
Combinations of features which result in tightly constrained articulatory or
aerodynamic requirements are unstable because they allow a narrow range of
variation, that is, they may easily fall into a different category with small
variations in the articulatory/aerodynamic parameters. Such unstable
combinations may easily change into a different percept and will tend to be
disfavoured (as shown, for example, by gaps in segment inventories, and a
lower lexical frequency of certain segment types). A classic example is the
difficulty in maintaining the co-occurrence of voicing and obstruency. The
partial or full blockage of the air exiting the oral cavity for obstruents leads to a
rapid increase of oropharyngeal air pressurerequired to generate turbulence
for fricatives and to create an audible burst for stopsbut tends to impair the
transglottal flow required for voicing within a few tens of milliseconds. Unless
the obstruent constriction is kept very short or the oral cavity is enlarged to
accommodate more air and thus prolong voicingboth maneuvers to the
detriment of a high pressure build-up for obstruencyvoiced obstruents will
tend to devoice. Thus, voiced obstruents require very finely tuned aerodynamic
conditions in order to maintain voicing and obstruency (Ohala 1983).
Similarly, combinations of features which result in a poor acoustic signal, for
example voiceless nasals (as the low frequency amplitude modulation for
nasals is impaired by voicelessness), are auditorily unstable (as measured from
confusion studies) and will tend not to be used.
In order to describe the interactions between features it is necessary to
vary the parametersphysiological, aerodynamic or acousticthat
characterize such features not only singly but in combination (e.g. changes in
oral pressure and duration of the obstruent constriction in the first example
(Westbury & Keating 1986), or changes in glottal excitation for nasals and
non-nasals, or during different degrees of velopharyngeal opening for the
nasal, in the second example). We can then identify a set of categorial values
along these parameters which remain stable with variations in the other
parameters. These categories or optimal settings across the different
parameters, if wide enough in range, are the more likely combinations of
features into segments.
The claim made in this paper is that the physical and auditory principles
that account for how features interact within segments may also account for the
interaction of features in contiguous segments. It is known that when two
MARIA-JOSEP SOL
44
segments are in contact their articulations necessarily overlap. Variations in the
shape, position and temporal coordination of the articulators due to
coarticulation with neighbouring segments may cause modifications of the
articulatory trajectories or the aerodynamic conditions in the vocal tract that
can, in turn, affect the acoustic and auditory result. Thus, the execution of the
articulatory gestures in implementing a particular feature will vary
considerably with the features in contiguous segments. As described earlier,
voicing is difficult to maintain during an obstruent; however, if an obstruent is
preceded by a nasal, voicing during the obstruent is facilitated. In nasal +
obstruent sequences, voicing continuation into the obstruent is facilitated by
nasal leakage before full velic closure is achieved and, after velic closure, by
the velum continuing to rise toward the high position for obstruents, thus
expanding the volume of the oral cavity. Both mechanisms, nasal leakage and
oral cavity expansion, lower the oropharyngeal pressure which accumulates in
the oral cavity and thus prolong transglottal flow for voicing (Hayes & Stivers
1996). Such phonetic effects have phonological significance in languages with
a phonological post-nasal voicing rule, in phonotactic patterns, and in sound
change. Thus, aerodynamic principles in the maintenance of voicing within
segments may account for the extension (or cessation) of voicing when
segments are combined.
In order to test the hypothesis that the factors governing the interaction
between features in the same sound may also govern the interaction of features
across segments when they overlap due to coarticulation, we designed a series
of experiments to explore the impact of co-occurring and coarticulatory
velopharyngeal opening for nasality on the stability of segments requiring a
high pressure build-up in the oral cavity, such as fricatives. The results may
throw light on why some feature combinations fail to occur, e.g. nasal
fricatives, and why certain segment combinations, e.g. fricatives followed by
nasals, tend to change. We report on research in which the aerodynamic
conditions in these segment types were varied (1) by venting the oral pressure
with a pseudo-velopharyngeal valve, and (2) by increasing speaking rate and
thus articulatory overlap of velopharyngeal opening on the pressure build-up
for the fricative.
The remainder of this paper is structured as follows. In Section 2 we
review the aerodynamic and perceptual effects of combining frication and
nasalization within a segment. In Section 3 we address the conflicting
requirements of frication and nasalization when they occur in contiguous
segments and we review how fricatives tend to lose their friction preceding
nasals historically and synchronically. In Section 4 we provide aerodynamic
and acoustic evidence that in fricative + nasal sequences anticipatory velum
lowering during the acoustic duration of the fricative reduces or extinguishes
the pressure difference required for frication. In Sections 5 and 6 we argue that
the principles that explain why the features [nasal] and [fricative] do not
combine within a segment also explain why they do not combine across
segments.
THE STABILITY OF PHONOLOGICAL FEATURES

45
2. Co-occurring features: Nasality and frication
Languages of the world have nasal stops, nasal taps, nasal approximants,
nasal glides and nasal vowels but no nasal fricatives (Ohala 1975; Ohala &
Ohala 1993). Segments reported as nasalized fricatives (e.g. in Umbundu
(Schadeberg 1982), Coatzospan Mixtec (Gerfen 1999), and Waffa (Stringer &
Hotz 1973)) are more adequately described as frictionless continuants due to
the lack of high frequency aperiodic noise (Ohala & Ohala 1993; Ohala, Sol
& Ying 1998). The reported nasalized fricatives in Bantu languages, Kwa
languages, and Igbo appear to involve sequencing of the nasal and the fricative
configuration and are, in fact, better described as prenasalized fricatives
(Welmers 1973:70-73).
Formal phonology attributes the lack of nasal fricatives to antagonistic
constraints: The rarity of such segments [nasalized liquids, glides and
fricatives] can be attributed to an antagonistic constraint NAS/CONT: A nasal
must not be continuant (Pulleyblank 1997:76); or to a constraint hierarchy
ranking segments according to their incompatibility with nasalization:
*NASOBSTRUENTSTOP >> *NASFRICATIVE >> *NASLIQUID >>
*NASGLIDE >> *NASVOWEL, where the less compatible a segment is with
nasality, the higher-ranked its constraint (Walker 2000).
The phonetic grounding for such antagonistic constraints or constraint
hierarchy is provided by work by Ohala and coworkers. Ohala and Ohala
(1993) provide an explanation based on aerodynamic principles. They suggest
that obstruents require a build-up of oral pressure behind the constriction in
order to create audible turbulence. An open velopharyngeal port for nasality
would vent the airflow through the nasal cavity, thus reducing or eliminating
the required pressure difference across the oral constriction for frication.
Ohala, Sol and Ying (1998) further explored this explanation by examining
the aerodynamic and acoustic effect on fricatives of concurrent nasality and its
perceptual result. They vented oro-pharyngeal pressure (P
o
) with a pseudo-
velopharyngeal valve (i.e. catheters of varying cross-sectional areas: 7.9, 17.8,
31.7, and 49.5 mm
2
; all 25 cm long, inserted into the mouth via the buccal
sulcus and the gap behind the back molars), and quantified how much
velopharyngeal opening for nasality was allowed before frication would be
extinguished.
The results of Ohala et al. (1998) show that in producing a fricative there
can be some opening of the velic valve, but the impedance of this valve has to
be high relative to that in the oral constriction so that the air will escape
through the aperture with lower impedance and create friction at the oral
constriction. Thus venting fricatives with catheters with a higher impedance
(7.9 mm
2
-area catheter) than that at the oral constriction did not affect the
quality of the fricative, it just slightly attenuated the fricative noise. Catheters
with values for impedance similar (17.8 mm
2
area) to those at the oral
constriction had noticeable effects on fricatives: they lost much of their high-
frequency aperiodic energy (e.g. the spectral peak at 6 kHz for [s] disappeared
and the energy level dropped 20 dB). The effect was most dramatic for voiced
MARIA-JOSEP SOL
46
fricatives, which became frictionless continuants. Sibilant fricatives sounded
non-sibilant. Larger area catheters (31.7 mm
2
), with a lower impedance than
that in the vocal tract, extinguished frication, since airflow exited through the
aperture with lower impedance, thus reducing the required pressure drop across
the oral constriction to generate turbulence. Voiceless fricatives only retained
the glottal friction. Voiced fricatives were more seriously affected than
voiceless fricatives, becoming vowel-like.
The results show that if impedance at the velopharyngeal port is lower
than that at the oral constriction the air will escape through the nose (i.e. the
fricative will be nasalized), but supraglottal frication will be impaired. Velic
openings which do not impair frication (<17.8 mm
2
) would be insufficient to
create the percept of nasalization in the fricative or even adjacent vowels. A
greater coupling between the oral and the nasal cavity is required for vowels
and sonorants to be perceived as nasalized. Data by other investigators suggest
that a velo-pharyngeal opening of 36 mm
2
(Whalen & Beddor 1989), 40 mm
2

(Maeda 1993) or more is needed to create a robust percept of nasalization in
vowels
1
. Consequently, if impedance at the velopharyngeal port is high enough
not to affect the fricative quality, the fricative will not sound nasalized.
In summary, to the extent that a fricative is a good fricative perceptually,
it cannot be nasalized (without added biomechanical cost, e.g. increased
subglottal pressure). In other words, along the independent physical parameters
of frication and nasalization there are categorial values which show stable
perceptual properties, i.e. a certain range within the continuum where a reliable
identification of frication (or nasalization), say 80%, can be obtained. The solid
line in Figure 1 illustrates that, for a sufficient rate of flow across the oral
constriction, a stable percept of a fricative can be obtained with some
velopharyngeal opening, but if the velic opening is larger than approximately
18 mm
2
, as shown by Ohala et al. (1998), the resulting sound will not be heard
as having friction. The dashed line shows that the same sound will not be heard
as nasalized unless the velic opening is at least 40 mm
2
, as found by Whalen
and Beddor (1989), and Maeda (1993). The two ranges of reliability for
frication and nasality, however, do not overlap, i.e. there is not a range of
values for both frication and nasalization where we may get 80% identification
for both features.
Shosted (2006) obtained similar results to those of Ohala et al. (1998)
with a mechanical model of the vocal tract with which he generated fricatives
with different degrees of velopharyngeal opening. He found that nasalization
on fricatives reduced the intensity of friction and increased the bandwidth of
spectral peaks, thus changing the percept of fricatives. The difficulty in
achieving a stable percept of concurrent nasalization and frication has been

1
Hajek and Watson (1998) found that a velopharyngeal opening of 16.8 mm
2
was sufficient to
give a strong nasal percept in vowels. Even with such a small magnitude of velum opening our
argument still holds. In Figure 1, right panel, the identification curve would shift towards lower
values (i.e. 80% identification of a nasalized segment would be obtained at approximately 20
mm
2
), but the two curves would still not overlap.
THE STABILITY OF PHONOLOGICAL FEATURES

47
noted by other investigators. For example, research on a variety of languages
has shown that nasalized voiced fricatives produced with perceptible
nasalization tend to lose audible frication and become approximants (e.g. in
Guarani, Gregores & Suarez 1967). In contrast, nasalized voiceless fricatives
with audible frication do not differ much auditorily from non-nasalized
fricatives, that is, the acoustic cues for nasalization are hardly detectable (Cohn
1993; Ladefoged & Maddieson 1996:132).


0
20
40
60
80
100
0 10 20 30 40 50 60 70
velo-pharyngeal opening in mm2
%

i
d
e
n
t
i
f
i
c
a
t
i
o
n

fricative nasalized










Figure 1: Diagrammatic representation of the reported percentage of identification of a
fricative and a nasal for continuous variation in velic opening for a given flow rate. See text.

In the next section we address whether the weakening or loss of fricatives
before nasals can be explained by the same principles.

3. Contiguous frication and nasality
Related to the difficulty of combining velopharyngeal opening with
frication within a segment is the precise timing of gestures in nasal + fricative
and fricative + nasal sequences if both segments are to be preserved. The
antagonistic requirements of turbulence generation (i.e. a tightly closed velum
to allow turbulent airflow in the vocal tract) and nasal coupling (i.e. a lowered
velum) in contiguous fricatives and nasals severely constrain the timing of
velic movements. The relative synchronization of articulatory movements in
nasal + fricative sequences has been investigated by Ali, Daniloff and
Hammarberg (1979), Ohala and Bus (1995), Ohala (1997b) and Bus (in
press) amongst others. Historically, such sequences may result in (i) nasal
consonant loss, with associated nasalization and lengthening of the preceding
vowel (e.g. Latin institutione > Italian istituto; PGmc *fimf > Old English five),
and (ii) an epenthetic stop (e.g. English Hampstead, Hampshire < O.E. ham +
stede, scir). Interestingly, one or the other outcome has been related to specific
coarticulatory patterns (Bus in press). Ohala and Bus (1995) have argued that
the first outcome, nasal consonant loss, is due to anticipatory vowel
nasalization resulting from coarticulatory lowering of the velum for the
upcoming nasal consonant and to associated perceptual factors. They provide
perceptual evidence that in these sequences the listener attributes the acoustic
MARIA-JOSEP SOL
48
effects of velum lowering for the nasal (attenuated amplitude and increased
bandwidth of F1 in adjacent vowels) to the similar effects of a wide glottal
opening required for high airflow segments, such as /s/, on neighboring
vowels, thus discounting the nasal consonant. The second outcome, epenthetic
stops, reflects anticipatory raising of the velum (and anticipatory glottal
abduction) during the oral constriction for the nasal, which ensures sufficient
time and rate of flow to build up pressure for the fricative (Ali et al. 1979).
Reverse fricative + nasal sequences, although not the object of extensive
investigation, require equally precise coordination of velic and oral gestures.
These sequences have resulted in several sound changes, including (i) fricative
weakening and loss
2
, and (ii) stop epenthesis. Importantly, the tendency for
fricatives to weaken, or disappear, before a nasal may be related to the
difficulty involved in combining frication with velopharyngeal opening within
a segment observed in the preceding section. Examples of prenasal fricative
weakening in historical sound change, morphophonological alternations and
dialectal-stylistic variation are shown in Table 1.
Prenasal fricative weakening may result in vocalization or gliding,
rhotacism, nasal assimilation, and elision. Examples 1-3 and the first example
in (c) in Table 1 illustrate vocalization; 4 exemplifies rhotacism; 5-9 illustrate
fricative loss, and various examples of nasal assimilation and fricative loss are
presented in Section (c). The examples in (c) and the examples of /s/
vocalization in (a) illustrate processes affecting only certain frequent words or
combinations of words, thus illustrating the role of frequency in phonological
change.
The weakening of prenasal fricatives in the Romance data in Table 1(a),
examples 1-5, may be argued to be part of a more general historical process of
coda weakeningdue to a drecreased oral gesture syllable finallyin Late
Latin and Gallo-Romance (Straka 1964; Gess 1999) by which not only
fricatives but also other obstruents were lost syllable finally, regardless of the
nature of the following segment. We suggest that the weakening of fricatives
before nasals found in a variety of languages may be attributed not only to
articulatory reduction but more crucially to varying aerodynamic conditions
due to the temporal sequencing of velic gestures.
Although fricative weakening in Romance also occurs before non-nasals
(e.g. Latin insula island > French le; Germanic *bruzdon to embroider >
Old Occitan broidar), a number of scholars (Pope 1934:151; Rohlfs 1966;
Torreblanca 1976; Recasens 2002) have noted that this process is favored by a
following voiced consonant and, in particular, by a following [n], [m], [r] or

2
The term fricative weakening is used here to indicate attenuation of the high frequency
noise which characterizes fricatives, due to gestural reduction or aerodynamic factors. Fricative
loss is considered the endpoint of the weakening continuum, i.e. extreme attenuation leading to
the segment becoming inaudible. In perceptual terms gradient attenuation of the friction noise
may result in identification of a discrete segment (e.g. a frictionless continuant, a vowel, a tap,
an assimilated segment, or /h/) or in the perceptual loss of the segment (i.e. deletion).
THE STABILITY OF PHONOLOGICAL FEATURES

49
[l]
3
. Recasens (2002:342) attributes fricative weakening before voiced
consonants in general to anticipatory glottal gestures for voicing during the
fricative, which result in reduced transglottal flow (due to higher glottal
impedance) and lower intensity of frication vis--vis pre-voiceless fricatives,
thus making frication more likely to be missed. Other factors besides voicing,
however, may be at play in the common weakening of fricatives preceding
nasals, laterals, and trills. Fricative weakening may arise from variation in the
relative timing of antagonistic positional requirements of the tongue for
contiguous lingual fricatives and laterals (raised tongue sides and central
critical constriction for the fricative and lowered tongue sides and central
contact for the lateral; Ohala 1997b), and lingual fricatives and trills (raised
and advanced tongue dorsum and a central groove for /s/ vs predorsum
lowering and postdorsum retraction with a lax tongue-tip for the trill; Sol
2002b); in such sequences anticipatory movements for the lateral or the trill
may bleed the positional and aerodynamic requirements for audible frication
for [z, s]. In a similar way, in fricative + nasal sequences variation in the
relative timing of the velic opening gesture and the preceding oral constriction
for the fricative (requiring a sealed velum) may result in fricative weakening.
In particular, anticipatory velum lowering for the nasal may affect the
aerodynamic requirements for the generation of turbulence for the fricative.
The examples in Table 1 illustrate that fricatives are weakened into
frictionless continuants, or are lost altogether, more often when followed by a
nasal segment, involving the same or different articulators, than when followed
by non-nasal segments, when they retain their fricative quality. Fricatives have
also been found to disappear in connected speech in French before nasal
vowels (D. Duez, p.c.), which involve a lower position of the velum than nasal
consonants. Consonants other than fricatives may assimilate the velic state of
the upcoming nasal in casual speech (e.g. let me ['Icmmij, wouldnt [wonntj,
Gimson, 1962), but they do not result in sound change or phonological
alternations
4
. The question is then why fricatives are weakened more often
than other segment types, and why they are weakened more often before nasal
as opposed to non-nasal sounds.
We hypothesized that the same aerodynamic factors disfavouring the
combination of frication and nasality within a segment are responsible for the
weakening/loss of fricatives before nasals: anticipatory movements to lower
the velum for the nasal may reduce the oropharyngeal pressure necessary for
the generation of turbulence for the fricative. Coarticulatory nasal leakage

3
Thus, for example, in Old French /s/ weakening and loss is found earlier in blmer < blasmer
< Lat. *blastemare blame and mler < mesler, medler [l] < Lat. misculare meddle than in
fte < feste < Lat. festa holiday (Pope 1934:151, 449).
4
This is most likely because stops tend to show up as oral in all other contexts (i.e.
prevocalically, preconsonantally, and prepausally), thus listeners can attribute nasality in stops
in stop + nasal sequences to the conditioning effect of the nasal and reconstruct an oral stop. In
contrast, fricatives tend to show diminished amplitude of frication not only when followed by
nasals but also preconsonantally and prepausally (Sol 2003), thus it is less likely that speakers
attribute the weakened fricative to a variety of contexts.
MARIA-JOSEP SOL
50
would have a larger effect on voiced fricatives as opposed to voiceless
fricatives (Ohala et al. 1998; Sol 2002b), since to generate friction at the oral
constriction while air is flowing out through the nasal passage due to
anticipatory velopharyngeal opening would require a high volume of flow, and
vocal fold vibration reduces airflow through the glottis. Thus, voiced fricatives
with overlapping velic movements for the nasal would have two independent
systems reducing the pressure drop across the oral constriction required for
frication, high impedance to ingoing airflow at the glottis and low impedance
to outgoing flow at the velum. In addition, voiced fricatives are known to be
shorter than voiceless fricatives and, due to the lesser rate of flow through the
vibrating glottis, they take longer to achieve the pressure difference for
frication and result in a lower intensity of friction than their voiceless
counterparts (Sol 2002b). This makes anticipatory velopharyngeal opening for
the nasal more likely to impair audible friction in voiced than in voiceless
fricatives. This prediction is in accord with the fact that the majority of
examples of fricative weakening in Table 1 involve voiced fricatives.
Note that our aerodynamic and perceptual explanation is not at odds with
Ohala and Buss (1995) claim presented above that in N + fricative
sequences, the nasal disappeared due to acoustic-perceptual factors. The N +
fricative sequences they examine (e.g. Latin institutione > Italian istituto), in
fact, illustrate that anticipatory velum lowering can be accommodated by a
preceding vowel, nasalizing it, and that the acoustic effects of nasalization on
the vowel may be misidentified. What we suggest is that in fricative + N
sequences nasalization cannot be accommodated by a preceding fricative
without compromising its spectral identity. The active role of the listener here
would involve reconstructing a frictionless continuant at face value (for
example, [j], [w] or a rhotic) or missing the fricative altogether.
A second outcome of contiguous fricatives and nasals (which require a
tightly closed velum and a lowered velum, respectively) is an epenthetic stop in
the transition between the two articulatory configurations, due to a prolonged
velic occlusion of the fricative during the oral constriction for the nasal (e.g.
Old English glisnian > glisten; Sanskrit krsn > Krishna ~ Krishtna). Along
similar lines, the insertion of an epenthetic schwa in /sm/ >[s
c
mj and /sn/ >
[s
c
nj sequences in Montana Salish (Ladefoged and Maddieson 1996:109-110)
reflects a delayed velic lowering and oral closure for the nasal relative to the
end of the fricative (i.e. an increase in the temporal distance between
articulatory gestures), which avoids an overlapped lowered velum during the
fricative in order to preserve frication. Thus a very precise synchronization of
the velic and oral movements is required in order to sequence segments
involving frication and nasality.





THE STABILITY OF PHONOLOGICAL FEATURES

51
(a) Historical change

1. [znj, [zm] > [jnj, [jm] Latin mesnata kids, elemos(i)na alms > Catalan mainada,
almoina (Badia 1951).
Latin asinu donkey > Gascon aine (cited in Recasens 2002).
Standard Catalan besnt grandchild > Majorcan Catalan beint.
Old French ae(s)mer, Standard Catalan esma > Majorcan Catalan eima
(Alcover & Moll 1978-1979); English aim.
2. [yn] > [jn], [wn] Latin agnu lamb, ligna line > S. Italian dialects ['ajcnc], ['Icwna] (cited
in Recasens 2002).
3. *[d
z
m], *[m] > [wm] Latin decimare, decimu > Catalan deumar lessen, reduce, deume
tribute.
4. [znj, [zm] > [rnj, [rm] Latin asinu donkey > Old Picard arne.
Latin *dis(ju)nare to eat breakfast > Old Occitan dis/rnar (also Cat. dinar).
Latin spasmu spasm > Roussillon esparme (cited in Recasens 2002).
S. Spanish mismo ['mirmo] same (Recasens 2002).
5. [zm] > [m] Old French ble(s)mir > English blemish.
Vulg.Latin blastemare > Old French bla(s)mer > Fr. blmer, English blame.
Latin rosmarinu, -aris > *romarinu, -ari(u)s rosemary > Catalan roman,
Spanish romero.
Standard Catalan quaresma Lent > Majorcan Catalan [ko'rcmc].
6. *sn > [n] IE snus daughter-in-law, OE snoru > Latin nurus, Greek nus, Armenian
nu, Spanish nuera (Watkins 1985).
7. *sn > [n] Burmese *sna > [n a] nose.

(b) Phonological alternations

8. [sn] > [n] IE *dhus-no > Welsh dwn dull, brown colour, OE dun(n) dark brown;
BUT IE *dhus-ko > Latin fuscus, OE dox, English dusk (Watkins 1985).
IE *dhs-no > Latin fanum temple, English fanatic, (pro)fane;
BUT IE *dhes-to > Lat. festus festive, German Fest, Spanish fiesta; IE
*dhes-ya > Lat. fesiae, feriae, English fair, Spanish feria (Watkins 1985).
9. [sm] > [m] IE *gras-men > Latin gramen fodder, English grama, gramineous;
BUT IE *gras-ter > Greek gaster stomach, English gastric, epigastrium
(Watkins 1985).

(c) Stylistic variation

[zn j > [nn j, [jnj isn't [inntj, aint [cint]; doesnt [d\nn tj; wasnt [wonn tj (Gimson 1962).
[mOVn] > [mn j, [m
1
n j something [s\mn j, [s\m
1
nj.
[nsVn] > [nVn] Vincent that [v i nc n c1] Tyneside English (Local 2003:325).
[mIr Vn] > [mr Vn] San Francisco [s mrc nsiskoo] American English (N. Hilty, p.c.).
[vm] > [mm] give me, gimme ['gimmi], have mine [ hm'main] (Gimson 1962).
[VN] > [VN] like them ['Iaikc mj, tell them ['tcIc mj.
BUT like this ['Iaik isj, tell this ['tcI isj.
[OVN] > [VN] thank you, thanks [p kju:j, [c pksj.
[zn] > [nnj business ['bidnisj, ['binnisj.
Table 1: Examples of fricative weakening/loss in fricative + nasal sequences.
MARIA-JOSEP SOL
52
4. Experiment. Variations in articulatory overlap
The following experiment was designed to find out whether the tendency
for prenasal fricatives to weaken or disappear can be attributed to anticipatory
velopharyngeal opening for the nasal overlapping the acoustic duration of the
fricative, thus diminishing the oropharyngeal pressure required for frication, as
argued for concurrent features. In order to test this hypothesis we examined
whether anticipatory velic opening occurred in such sequences and, if so, how
it affected the aerodynamics and acoustics of fricatives.

4.1 Experimental procedure
Simultaneous oropharyngeal pressure (P
o
), oral and nasal flow, and audio
signal were obtained with PCquirer (Scicon) for five American English
speakers producing words containing fricative + nasal sequences at slow and
fast speech. The two speaking rates allowed us to observe the effects of
increased articulatory overlap on the relative timing of velic and oral gestures,
and how such changes in timing affect the pressure build-up for fricatives. The
audio signal was digitized and sampled at 12 kHz, and the dc channels were
sampled at 1 kHz. P
o
was obtained by a catheter introduced at the side of the
mouth and bent behind the rear molars and connected to a pressure transducer.
The volume flow from the mouth and the nose was collected simultaneously
with a Rothenberg mask, using one of the outlet holes in the mouth mask for
the pressure tube. The pressure and airflow signals were low-pass filtered at 50
Hz. P
o
and airflow were calibrated as described in Sol (2002a). Words
containing fricative + nasal sequences with C1 = [s, zj and C2 = [n, mj in
word medial position (e.g. Dessna ['dcsncj, Fresno ['Ircznooj, Missmer
['mismcj, Mesmer ['mczmcj) were read in a carrier phrase. Control sequences
with voiced and voiceless alveolar fricatives followed by laterals, stops,
fricatives and approximants, all dento-alveolar (e.g. Grizzly [zIj, Gizder [zdj, is
the [zj, Ezra [z+j; Esling [sIj, Esda [sdj, less the [sj, Esra [s+j) were also
analyzed for comparison. The test and control tokens were randomized in the
reading list.
The aerodynamic and acoustic data were collected in two different
sessions. In the first session, three speakers (JO, MS and DM) read five
repetitions of each token at self-selected slow and fast rates. Because the three
speakers showed different patterns of velic-oral coordination, two more
speakers were recorded in a second session under the same conditions.
Speakers JE and RS produced six repetitions of each token at slow and fast
rates.

4.2 Measurements and analysis
The measurements made on the data are illustrated in Figure 2, which
shows the aerodynamic and acoustic data for Say Mesmer again in slow and
fast speech for one of the speakers. Measurements were made at the following
points in time for the fricative + nasal sequence: onset and offset of friction on
the spectrographic records; onset of increased nasal flow (channel 5), indicated
THE STABILITY OF PHONOLOGICAL FEATURES

53
by a vertical line in Figure 2, reflecting velo-pharyngeal port opening for the
nasal /m/; onset of oral closure for the nasal, indicated by a drop in oral flow
(dotted line in channel 4), and offset of the nasal, indicated by an increase in
oral flow (dashed line in channel 4) and an increase in amplitude and formant
structure for the following vowel on the spectrogram. Figure 2 illustrates
anticipatory velopharyngeal opening: the vertical lines mark the onset of nasal
flow (channel 5) due to velum lowering for the nasal. At this point in time oral
pressure decreases (channel 2) and the high frequency noise disappears (in
slow speech) or is attenuated (in fast speech) (The sliding center frequency
noise after the vertical line in fast speech reflects the effect of anticipatory lip
movement, which is known to filter out higher frequencies). The increase in
nasal flow leads the drop in oral flow (dotted line in channel 4) for the
complete oral constriction for the nasal. Thus, for a few tens of ms there is
concurrent (increasing) nasal flow and (decreasing) oral flow, characterised
acoustically by some low amplitude aperiodic noise around 3 kHz. In other
words, the velum starts to lower during the acoustic duration of the fricative,
resulting in a sudden drop in amplitude of high frequency noise.

SLOW FAST
[ s c i 'm c z m c + c g c nj [ s c i 'm c z m c + c g c nj
Figure 2: (1) Audio signal, (2) filtered P
o
, (3) unfiltered P
o
, (4) oral airflow, (5) nasal airflow
and 0-5 kHz spectrogram of Say Mesmer again in slow and fast speech. Speaker JO. See text.

The duration of fricatives in test and control sequences was measured on
spectrograms and aerodynamic records. In fast speech a considerable number
of cases involved blending of the gestures for the two contiguous consonants
generally resulting in deocclusivized following stops (as noted by Honorof
2003) or fricativized following sonorants [l, r, n, m]. The aerodynamic records
were of great help in segmenting these blended sequences.


MARIA-JOSEP SOL
54
4.3 Results
4.3.1 Patterns of velic movements. The patterns of velic coordination observed
in the data are described in patterns 1 through 4, below, and schematized in
Figure 3, which shows the oral gestures for the fricative (C1) + nasal (C2)
sequence followed by traces of the observed velic patterns. Time 0 is the onset
of the oral constriction for the nasal, indicated by the drop in oral flow (dotted
line in Figure 2). A delayed oral gesture for the nasal, resulting in an epenthetic
vowel, is represented at the bottom of Figure 3.

(1) If onset of velopharyngeal opening is synchronized with the oral
constriction for the nasal, nasal flow starts at 0. We take this pattern to
reflect a precise synchronization of the supraglottal and velic gestures
for the nasal
5
, which allows sequencing of the fricative and the nasal
consonant.
The oral and velic gestures for the nasal, however, may not be precisely
synchronized.

(2) Onset of velum opening, i.e. nasal flow, may occur prior to the complete
oral constriction for the nasal (time 0). That is, anticipatory
velophrayngeal opening overlaps the acoustic duration of the preceding
fricative. In this case we find concurrent oral flow (from the fricative
slit constriction) and nasal flow before time 0 in the aerodynamic data,
and lack of friction or attenuated friction on the spectrogram during this
time interval.

(3) Velopharyngeal opening may be delayed relative to the oral constriction
for the nasal (time 0) resulting in a transitional epenthetic stop. In this
case the aerodynamic data shows neither nasal flow (velum up) nor oral
flow (nasal constriction formed) immediately after 0, and a silent gap
on the spectrogram. We found the epenthetic stop to be audible if the
oral and velic closure overlapped for over 15 ms. Figure 3 shows that
the transitional stops emerging in these sequences are always nasally
released, that is, they end when the velum lowers.

(4) Finally, the oral constriction for the nasal may be delayed with respect to
the release of the preceding fricative constriction, resulting in an
epenthetic vowel. In this case oral and nasal flow is found immediately
before 0. Only one such case was found in the data.


5
Whereas it is clear that the oral target gesture for the nasal is a complete constriction, the
velic target gesture may not be opening the velum but rather achieving a certain magnitude of
velum opening. If this were the case, onset of the oral closure for the nasal would be
coordinated with a lowered velum, and nasal flow would begin before 0, as in the case in VN
sequences. However, since most of the observed values cluster around 0, we will assume the
suggested target coordination.
THE STABILITY OF PHONOLOGICAL FEATURES

55
It is worth noting that these patterns parallel those found in the historical
and synchronic data reviewed in Section 3, in line with Ohalas (1989, 1993)
claim that sound change emerges from synchronic phonetic variation.




















1. synchronous, onset of nasal flow at 0

2. bleeding Po, oral & nasal flow before 0

3. epenthetic stop, no nasal or oral flow after 0





4. epenthetic vowel, oral & nasal flow before 0

Figure 3: Diagrammatic representation of patterns of velic coordination in fricative-nasal
sequences. Traces for the supraglottal movements for C1 and C2 followed by different traces
of velum position (dashed lines). See text.

4.3.2 Coordination and timing of velic movements. The patterns of
coordination of oral and velic movements in fricative + N sequences for each
token at slow and fast rates are presented in Figure 4. This figure plots the time
interval between onset of velopharyngeal opening and onset of the oral
constriction for the nasal (time 0) for each token for the five speakers. The
production of each individual token has been arranged in decreasing duration
of that interval. Bars to the left of 0 are cases of anticipatory velopharyngeal
opening (onset of nasal flow precedes onset of oral constriction for the nasal;
pattern 2 in Figure 3); tokens at 0 represent synchronous onset of oral and velic
gestures for the nasal (onset of nasal flow at 0; pattern 1 in Figure 3) and,
consequently, a precise sequencing of the fricative and the nasal segment; bars
to the right of 0 represent cases of velopharyngeal opening lagging behind the
oral closure for the nasal (transitional stops; pattern 3 in Figure 3). The single
case of vowel epenthesis (Fresno pronounced ['Ircz
c
nooj), where the oral
constriction for the nasal consonant is delayed relative to fricative release and
velum lowering, is indicated by a lined bar (speaker DM, slow speech). Since
inspection of the patterns of sequences involving voiced (e.g. Fresno) and
voiceless (e.g. Dessna) fricatives showed no differences, both types of
sequences were pooled in this graph. White bars represent homorganic
sequences, [sn, zn], and grey bars represent heterorganic sequences, [sm, zm].
Each bar represents one token and the number of plotted tokens ranges
MARIA-JOSEP SOL
56
between 5-6 for each sequence, [sn, zn, sm, zm], depending on the speaker and
session.
Figure 4 shows that although all observations cluster around time 0, the
five speakers show three distinct patterns of velic-oral coordination. Subject JO
exhibits extensive anticipatory velic opening during the acoustic duration of
the fricative for homorganic and heterorganic sequences. Such anticipatory
velic opening vents the required high oral pressure for turbulence and, thus,
frication is attenuated or extinguished (see Figure 2). Speakers JE and MS
exhibit a majority of cases of anticipatory velic opening in heterorganic
sequences (/sm, zm/), but not in homorganic sequences (/sn, zn/), which exhibit
a greater number of transitional stops (i.e. delayed velum lowering)
6
. However,
both speakers show cases of a precise synchronization (time 0) and of
anticipatory velic opening for homorganic as well as heterorganic sequences.
Finally, speakers DM and RS show mostly epenthetic stops, with a few cases
of anticipatory velopharyngeal opening. The difference between homorganic
and heterorganic sequences for these speakers seems to point in a different
direction from that observed for speakers JE and MS: the epenthetic stops,
resulting from a prolonged velic raising while the oral occlusion for the nasal
has been achieved, appear to be more common and longer for /sm, zm/
sequences than for /sn, zn/ sequences. In spite of speaker-dependent
differences, all speakers show cases of anticipatory velic opening bleeding the
high oral pressure for frication. Overall, anticipation of velar activity during the
fricative was found in 40% of the tokens in slow speech and 26% of the tokens
in fast speech.
We now turn to differences in the timing of gestures in slow and fast
speech. We expected to find greater overlap of anticipatory movements of the
velum with the fricative in fast vis--vis slow speech. Contrary to our
expectations, Figure 4 shows similar patterns of coordination of velic and oral
gestures at slow and fast rates for each speaker, and no major differences in the
absolute values of velic timing across rates. Any differences are in the direction
of more cases (i.e. more bars with negative values) and a longer period (i.e.
larger negative values) of anticipatory velic lowering in slow than in fast
speech. That is, the velum appears to be freer to lower before the oral
constriction for the nasal is achieved at slow rates, most likely due to the time
pressure in fast speech imposing tighter time constraints. However, since
fricatives are slightly shorter at faster rates, the same period of anticipatory
velopharyngeal opening has a slightly greater percentage effect in fast than in

6
Assuming that the motor instructions for the oral and velic gestures for the nasal are
synchronic, and ignoring differences in velocity of articulators, the difference between
homorganic and heterorganic sequences could in part be accounted for in terms of gestures
involving independent articulators overlapping in timeanticipatory overlap of the labial and
velic gesture for /m/ during the alveolar fricative in /sm, zm/ sequences. In homorganic /sn, zn/
sequences involving the same articulatorthe tongue tipthe oral and velic gestures for the
nasal would be delayed till the tongue tip was available for repositioning. However, this
interpretation does not explain why velic opening lags behind the oral closure for the nasal in
homorganic sequences.
THE STABILITY OF PHONOLOGICAL FEATURES

57
7
slow speech. This is illustrated in Figure 5, which plots the modal duration of
anticipatory velopharyngeal opening found for each speaker and rate as a
percentage of the average duration of the fricative for that speaker and rate, for
voiced and voiceless fricatives separately. Figure 5 shows a slightly larger
percentage of overlapping velopharyngeal opening (i.e. longer black bars) in
fast than in slow speech for all speakers and sequences. Figure 5 also allows us
to observe the larger percentage of anticipatory velopharyngeal opening in
voiced as opposed to voiceless fricatives, due to the shorter duration of the
former.

SLOW FAST
12
11
10
9
8
7
6
5
4
3
2
1
T
o
k
e
n

n
u
m
b
e
r
60 40 20 0 -20 -40 -60
Time in ms
JO_sn_slow
JO_sm_slow
12
11
10
9
8
7
6
5
4
3
2
1
T
o
k
e
n

n
u
m
b
e
r
60 40 20 0 -20 -40 -60
Time in ms
JO_sn_fast
JO_sm_fast

12
11
10
9
8
7
6
5
4
3
2
1
T
o
k
e
n

n
u
m
b
e
r
60 40 20 0 -20 -40 -60
Time in ms
JE_sn_slow
JE_sm_slow
12
11
10
9
8
7
6
5
4
3
2
1
T
o
k
e
n

n
u
m
b
e
r
60 40 20 0 -20 -40 -60
Time in ms
JE_sn_fast
JE_sm_fast


7
The mode was plotted rather than the mean because of extreme values in the data.
MARIA-JOSEP SOL
58
12
11
10
9
8
7
6
5
4
3
2
1
T
o
k
e
n

n
u
m
b
e
r
60 40 20 0 -20 -40 -60
Time in ms
MS_sn_slow
MS_sm_slow
12
11
10
9
8
7
6
5
4
3
2
1
T
o
k
e
n

n
u
m
b
e
r
60 40 20 0 -20 -40 -60
Time in ms
MS_sn_fast
MS_sm_fast

12
11
10
9
8
7
6
5
4
3
2
1
T
o
k
e
n

n
u
m
b
e
r
60 40 20 0 -20 -40 -60
Time in ms
DM_sn_fast
DM_sm_fast

12
11
10
9
8
7
6
5
4
3
2
1
T
o
k
e
n

n
u
m
b
e
r
60 40 20 0 -20 -40 -60
Time in ms
RS_sn_fast
RS_sm_fast
12
11
10
9
8
7
6
5
4
3
2
1
T
o
k
e
n

n
u
m
b
e
r
60 40 20 0 -20 -40 -60
Time in ms
RS_sn_slow
RS_sm_slow

Figure 4: Coordination of oral and velic gestures for each production of Fresno and Dessna
(white bars), and Mesmer and Missmer (grey bars) at slow and fast rates for each speaker.
Tokens with voiced and voiceless fricatives have been pooled.

THE STABILITY OF PHONOLOGICAL FEATURES

59
0%
20%
40%
60%
80%
100%

v
l
e
s
s
v
d
v
l
e
s
s
v
d
v
l
e
s
s
v
d
v
l
e
s
s
v
d
v
l
e
s
s
v
d
v
l
e
s
s
v
d
v
l
e
s
s
v
d
v
l
e
s
s
v
d
v
l
e
s
s
v
d
v
l
e
s
s
v
d
SLOW FAST SLOW FAST SLOW FAST SLOW FAST SLOW FAST
JO JE MS DM RS
fricative duration overlapped velopharyngeal opening
Figure 5: Modal duration of anticipatory velopharyngeal opening (heterorganic and
homorganic sequences pooled) as a percentage of the total duration of the voiced or voiceless
fricative in slow and fast speech for the five speakers.

4.3.3 Fricative duration. The hypothesis that nasal leakage due to anticipatory
velopharyngeal opening extinguishes or attenuates frication for a few tens of
ms predicts that fricatives preceding nasal or nasalized segments should be
phonetically shorter than those preceding non-nasal segments. We tested this
prediction by measuring fricative duration for test and control tokens in slow
and fast speech.
The results of the measurements show that, as predicted, fricatives
preceding nasals are generally shorter than those preceding non-nasals.
(Following fricatives and laterals also tend to result in shorter preceding
fricatives for some speakers.) Two-way ANOVAs with fricative voicing
(voiced, voiceless) and following consonant (nasal, non-nasal) as independent
variables, and duration of the fricative as the dependent variable, were
performed for each speaker and speech rate separately. Table 2 shows the
results for the ANOVAs. The durational differences between fricatives
preceding nasal vs non-nasal consonants reached significance for speakers JO,
MS and DM in slow and fast speech. Since the interaction between the two
factors was significant for speaker JE at faster rates, one-way ANOVAS were
carried out for voiced and voiceless fricatives separately for this speaker.
Voiced fricatives were found to be significantly shorter preceding nasal than
oral consonants (F
(1,33)
= 11.629, p<0.01), whereas for voiceless fricatives the
difference did not reach significance. To conclude, in seven out of ten
comparisons fricatives before nasals were significantly shorter than before non-
nasals, as expected. A significant main effect of voicing was found for all
speakers and rates, with voiced fricatives being significantly shorter than
voiceless fricatives.



MARIA-JOSEP SOL
60
JO JE MS DM RS
Slow speech
Prenasal vs
non-prenasal
F
(1,76)
=82.76,
p<0.0001
F
(1,76)
=6.88,
p<0.01
F
(1,76)
=8.92,
p<0.01

Fricative
voicing
F
(1,76)
=50.92,
p<0.0001
F
(1,52)
=147.19,
p<0.0001
F
(1,76)
=7.67,
p<0.01
F
(1,76)
=86.60,
p<0.0001
F
(1,44)
=104.48,
P<0.0001
Interaction
Fast speech
Prenasal vs
non-prenasal
F
(1,76)
=141.0,
p<0.0001
F
(1,76)
=27.02,
p<0.0001
F
(1,76)
=16.78,
p<0.001

Fricative
voicing
F
(1,76)
=31.86,
p<0.0001
F
(1,64)
=183.43,
p<0.0001
F
(1,76)
=30.12,
p<0.0001
F
(1,76)
=9.53,
p<0.01
F
(1,44)
=66.26,
P<0.0001
Interaction F
(1,64)
=14.84,
p<0.001

Table 2: Significant differences for the two factor ANOVAS performed for the independent
variable on the rows and fricative duration as the dependent variable for slow and fast
speaking rates for each speaker.

5. Discussion and conclusions
The results on the timing of velic and oral gestures show that in fricative
+ nasal sequences the velum may lower during the acoustic duration of the
fricative (in 40% of the cases in slow speech and 26% in fast speech in our
data). These results are in agreement with those obtained by other
investigators. Shosted (2006) obtained nasal and oral flow for VCV utterances
where C was a fricative and V was an oral or nasal vowel. He found
coarticulatory nasalization during the fricative, with the same acoustic output
observed for nasalized fricatives: attenuation of high-frequency energy and
increased bandwith of spectral peaks. Recasens (in press) provides
electropalatographic and acoustic evidence that /s/ in Majorcan Catalan may be
lost in /sn/, /sm/ and /sl / clusters whereas /sb/, /sv/, /sd/, and /sg/ clusters show
less extreme cases of fricative weakening. Overall the results indicate that
nasal leakage during the fricative reduces the oral pressure required for the
generation of turbulence and frication is attenuated or extinguished for a few
tens of milliseconds, which may lead to the perceptual loss of the fricative.
This is more likely in voiced fricatives which are shorter (Table 2) and, due to
reduced transglottal flow, have a lower intensity of friction vis--vis voiceless
fricatives (Sol 2002b). Thus, if anticipatory velopharyngeal opening for the
nasal overlaps the latter portion of the voiced fricative, the low amplitude
turbulence may not be heard. Our data show that, in addition to the effect of
voicing in the following segment observed by Pope (1934), Rohlfs (1966),
Torreblanca (1976), and Recasens (2002), anticipatory velopharyngeal opening
for a following nasal may also favor attenuation or loss of friction. Thus, there
seems to be a gradient reduction continuum for coda fricatives: fricative +
nasal > fricative + voiced consonant > fricative + voiceless consonant.
Anticipatory velopharyngeal opening was hypothesized to be larger at
faster speaking rates than at slower rates; however, approximately the same
period of an overlapped lowered velum during the latter portion of the fricative
was found across rates. The same amount of velic overlap had a slightly larger
THE STABILITY OF PHONOLOGICAL FEATURES

61
effect at faster rates due to the slightly shorter duration of the fricative in fast
speech.
The results in Section 4.3.2 also show cases of epenthetic stops in the
transition between fricatives and nasals due to prolonged velum raising for the
fricative when the oral constriction for the nasal has been achieved. Such
transitional stops are always nasally released and the lack of a strong release
burst, which is a perceptual cue for intrusive stops (Ali et al. 1979), is most
likely the reason why these epenthetic stops may not be noticed by speakers
and have not phonologised as opposed to those emerging in contexts where
they are orally released (e.g. nasal + fricative, sense [n
t
s]; nasal + flap, Catalan
cambra < Latin cam(e)ra; nasal + lateral, Spanish temblar < Latin trem(u)lu;
lateral + fricative, else [l
t
s]). Finally, the results in Section 4.3.3 show that the
duration of the fricative tends to be shorter preceding nasal than oral
consonants, suggesting that the effect of nasal leakage during the latter portion
of the fricative is present phonetically.
The results obtained are compatible with the proposed account for the
historical and synchronic defricativization or loss of fricatives before nasal or
nasalized segments: velopharyngeal opening during the latter part of the
fricative vents the intraoral pressure, thus reducing or eliminating the required
pressure difference across the oral constriction for audible frication. These
findings suggest that interarticulatory timing and associated aerodynamic
effects may account for weakening of segments crucially dependent on airflow
conditions. Indeed, data on the perceptual impact of these aerodynamic and
temporal variations is needed to back up the findings of this study.

6. General conclusions
We set out to test the hypothesis that the physical and physiological
principles used to account for paradigmatic aspects of phonology, such as
feature co-occurrence restrictions, can be used to explain syntagmatic aspects,
such as phonotactic patterns, context-dependent phonological processes and
sound change. The results of the experiments reported here show that speech
features requiring high airflow through the oral constriction, such as
fricatives
8
, tend to be impaired and become unstable with co-occurring or
coarticulatory nasalization. The results in Section 2 show that a reduction in
oral pressure during the articulation of a fricative (due to venting oral pressure
with a pseudo-velopharyngeal valve) reduces the pressure difference and the
particle velocity of the air across the oral constriction, and frication is
attenuated or losthence, the constraint against combining the features
[fricative] and [nasal] within a segment. The results of the experiment in
Section 4 show that when these features occur in contiguous segments, as in
fricative + nasal sequences, there can be anticipatory velopharyngeal opening
during the acoustic duration of the fricative, which has the same aerodynamic

8
Tongue-tip trills also require high airflow to set the tongue tip into vibration and,
consequently, cannot be nasalized. For the incompatibility between trilling and nasality see
Sol (2002b).
MARIA-JOSEP SOL
62
and acoustic consequences on the fricative as concurrent nasalization.
Reduction of the oral pressure and subsequent reduction of the intensity of the
high-frequency noise may lead to a non-fricative percept or to missing the
fricative altogether. Thus, the same factors responsible for the difficulty in
combining the two features within a segment may be used to explain why these
features do not combine across segments. Relating constraints on the
combination of features within and across segments illustrates the generality
that can be achieved by a physically based explanation.
The instability of frication when combined with nasalization may be at
the origin of a number of phonological patterns, specifically, feature co-
occurrence restrictions (e.g. lack of nasal fricatives), phonological change,
morphological alternations and stylistic variation (e.g. loss/weakening of
fricatives followed by a nasal), and transitional probabilities in the sequencing
of sounds (lower lexical frequency of fricatives followed by nasals, Sol
(forthcoming)). This is one further example of how phonological structure may
emerge from physical constraints as advocated by Ohala (1974, 1983) and
Lindblom (1986, 1990), and strongly suggests that the same physical principles
may provide an explanation for paradigmatic and syntagmatic aspects of
phonology.


References
Alcover, Antoni M. & Francesc de B. Moll. 1978-1979. Diccionari Catal-
Valenci-Balear. Barcelona: Grfiques Instar.
Ali, Latif, Ray Daniloff & Robert Hammarberg. 1979. Intrusive stops in
nasal-fricative clusters: An aerodynamic and acoustic investigation.
Phonetica 36:2. 85-97.
Badia, Antoni M. 1951. Gramtica histrica catalana. Barcelona: Noguer.
Bus, Maria Grazia. In press. Coarticulatory nasalization and phonological
developments: Data from Italian and English nasal-fricative sequences.
Experimental Approaches to Phonology ed. by Maria-Josep Sol, Patrice
S. Beddor & Manjari Ohala. Oxford: Oxford University Press.
Cohn, Abigail C. 1993. The status of nasalized consonants. Huffman &
Krakow 1993. 329-367.
Gerfen, Chip. 1999. Phonology and Phonetics in Coatzospan Mixtec.
Dordrecht: Kluwer.
Gess, Randall. 1999. Rethinking the dating of Old French syllable-final
consonant loss. Diachronica 16. 261-296.
Gimson, Alfred C. 1962. An Introduction to the Pronunciation of English.
London: Arnold.
Gregores, Emma & Jorge Suarez. 1967. A Description of Colloquial Guaran.
The Hague: Mouton.
Hajek, John & Ian Watson. 1998. More evidence for the perceptual basis of
sound change? Suprasegmental effects in the developent of distinctive
nasalization. Proceedings of the 5
th
International Congress on Spoken
THE STABILITY OF PHONOLOGICAL FEATURES

63
Language Processing ed. by R. H. Mannell and J. Robert-Ribes. 1763-
1765. Sydney: Causal Productions.
Hardcastle, William J. & Nigel Hewlett, eds. 1999. Coarticulation. Theory,
Data and Techniques. Cambridge: Cambridge University Press.
Hayes, Bruce & Tanya Stivers. 1996. A phonetic account of postnasal
voicing. Ms., Department of Linguistics, UCLA, Los Angeles, Calif.
http://www.linguistics.ucla.edu/people/hayes/phonet.htm#postnasal.
Honorof, Douglas N. 2003. Articulatory evidence for nasal de-occlusivization
in Castilian. Proceedings of the 15
th
International Congress of Phonetic
Sciences ed. by Maria-Josep Sol, Daniel Recasens & Joaqun Romero,
vol. 2, 1759-1763. Barcelona: Causal Productions.
Huffman, Marie K. & Rena A. Krakow, eds. 1993. Nasals, Nasalization and
the Velum. San Diego, Calif.: Academic Press.
Kawasaki, Haruko. 1986. Phonetic explanations for phonological universals:
The case of distinctive nasalization. Experimental Phonology ed. by
John J. Ohala & Jeri J. Jaeger. 81-103. San Diego, Calif.: Academic
Press.
Kawasaki-Fukumori, Haruko. 1992. An acoustical basis for universal
phonotactic constraints. Language and Speech 35:1,2. 73-86.
Ladefoged, Peter & Ian Maddieson. 1996. The Sounds of the Worlds
Languages. Oxford: Blackwell.
Lindblom, Bjrn. 1986. Phonetic universals in vowel systems. Experimental
Phonology ed. by John J. Ohala and Jeri J. Jaeger. 13-44. San Diego,
Calif.: Academic Press.
----------. 1990. On the notion of possible speech sounds. Journal of
Phonetics 18. 135-152.
Local, John. 2003. Variable domains and variable relevance: Interpreting
phonetic exponents. Journal of Phonetics 31. 321-339.
Maeda, Shinji. 1993. Acoustics of vowel nasalization and articulatory shifts in
French nasal vowels. Huffman & Krakow 1993. 147-167.
Ohala, John J. 1974. Phonetic explanations in phonology. Papers from the
Parasession on Natural Phonology ed. by Anthony Bruck, Robert Fox &
Michael LaGaly. 251-274. Chicago: Chicago Linguistics Society.
----------. 1975. Phonetic explanations for nasal sound patterns. Naslfest:
Papers from a Symposium on Nasals and Nasalization ed. by Charles A.
Ferguson, Larry M. Hyman & John J. Ohala. 289-316. Stanford:
Language Universals Project.
----------. 1981. Articulatory constraints on the cognitive representation of
speech. The Cognitive Representation of Speech ed. by Terry Myers,
John Laver & John Anderson. 111-122. Amsterdam: North Holland.
----------. 1983. The origin of sound patterns in vocal tract constraints. The
Production of Speech ed. by Peter F. MacNeilage. 189-216. New York:
Springer Verlag.
----------. 1989. Sound change is drawn from a pool of synchronic variation.
Language Change: Contributions to the Study of its Causes ed. by Leiv
E. Breivik & Ernst H. Jahr. 173-198. Berlin: Mouton de Gruyter.
MARIA-JOSEP SOL
64
----------. 1993. The phonetics of sound change. Historical Linguistics:
Problems and Perspectives ed. by Charles Jones. 237-278. London:
Longman.
----------. 1997a. Aerodynamics of phonology. Proceedings of the 4
th
Seoul
International Conference on Linguistics. 92-97. Seoul, Korea.
----------. 1997b. Emergent stops. Proceedings of the 4
th
Seoul International
Conference on Linguistics. 84-91. Seoul, Korea.
---------- & Maria Grazia Bus. 1995. Nasal loss before voiceless fricatives: A
perceptually-based sound change. Special issue on The Phonetic Basis
of Sound Change ed. by Carol A. Fowler. Rivista di Linguistica 7. 125-
144.
---------- & Manjari Ohala. 1993. The phonetics of nasal phonology:
Theorems and data. Huffman & Krakow 1993. 225-249.
----------, Maria-Josep Sol & Goangshiuan Ying. 1998. The controversy of
nasalized fricatives. Proceedings of the 16
th
International Congress on
Acoustics and 135
th
Meeting of the Acoustical Society of America. 2921-
2922. Seattle, Washington.
Pope, Mildred K. 1934. From Latin to Modern French with special
consideration of Anglo-Norman. Manchester: Manchester University
Press.
Pulleyblank, Douglas. 1997. Optimality theory and features. Optimality
Theory. An Overview ed. by Diana Archangeli & D. Terence
Langendoen. 59-101. Oxford: Blackwell.
Recasens, Daniel. 2002. Weakening and strengthening in Romance revisited.
Rivista di Linguistica 14:2. 327-373.
----------. In press. Gradient weakening for syllable-final /s, r/ in Majorcan
Catalan consonant clusters. Proceedings of the 7
th
International Seminar
on Speech Production. Ubatuba, Brazil.
http://cefala.org/issp2006/camera-ready/recasens.pdf
Rohlfs, Gerhard. 1966. Grammatica storica della lingua italiana e dei soui
dialetti. Fonetica. Torino: Einaudi.
Schadeberg, Thilo C. 1982. Nasalization in Umbundu. Journal of African
Languages and Linguistics 4. 109-132.
Shosted, Ryan. 2006. The aeroacoustics of nasalized fricatives. PhD diss.,
University of California, Berkeley.
Sol, Maria-Josep. 2002a. Aerodynamic characteristics of trills and
phonological patterning. Journal of Phonetics 30:4. 655-688.
----------. 2002b. Assimilatory processes and aerodynamic factors.
Laboratory Phonology 7 ed. by Carlos Gussenhoven & Natasha Warner.
351-386. Berlin & New York: Mouton de Gruyter.
----------. 2003. Aerodynamic characteristics of onset and coda fricatives.
Proceedings of the 15
th
International Congress of Phonetic Sciences ed.
by Maria-Josep Sol, Daniel Recasens & Joaqun Romero. 2761-2764.
Barcelona: Causal Productions.
THE STABILITY OF PHONOLOGICAL FEATURES

65
----------. Forthcoming. Compatibility of features and phonetic content: The
case of nasalization. Presented at the 16
th
International Congress of
Phonetics Sciences. Saarbrcken, Germany.
Stevens, Kenneth N. 1972. The quantal nature of speech: Evidence from
articulatory-acoustic data. Human Communication. A Unified View ed.
by Peter B. Denes & Edward E. Jr. David. 51-66. New York: McGraw-
Hill.
----------. 1989. On the quantal nature of speech. Journal of Phonetics 17. 3-
46.
Straka, Georges. 1964. Remarques sur la dsarticulation et lamuissement de
ls implosive. Mlanges de Linguistique Romane et de Philologie
mdievale offerts Maurice Delbouille 1. 607-628.
Stringer, Mary & Joyce Hotz. 1973. Waffa phonemes. The Language of the
Eastern Family of the East New Guinea Highland Stock ed. by Howard
McKaughan. 523-529. Seattle: University of Washington Press.
Torreblanca, Mximo. 1976. Estudio del habla de Villena y su comarca.
Alicante: Instituto de Estudios Alicantinos.
Walker, Rachel. 2000. Nasalization, Neutral Segments, and Opacity Effects.
New York: Routledge.
Watkins, Calvert (ed.). 1985. The American Heritage Dictionary of Indo-
European Roots. Boston: Houghton Mifflin.
Welmers, William. 1973. African Language Structures. Berkeley: University
of California Press.
Westbury, John R. & Patricia A. Keating. 1986. On the naturalness of stop
consonant voicing. Journal of Linguistics 22. 145-166.
Whalen, Douglas H. & Patrice S. Beddor. 1989. Connections between nasality
and vowel duration and height: Elucidation of the Eastern Algonquian
intrusive nasal. Language 65. 457-486.

You might also like