PHONOLOGY

PERCEPTUAL STUDIES ON CO-ARTICULATION

PRESENTERS: Bhavana.P.V and Rashmi Singh

FACULTY: Dr. Sreedevi.N

Coarticulation is a phenomenon that helps in maintaining the continuity of speech at the rate of speech of 12
speech sounds per second. Perception too will suffer due to the listener’s inability or delay in stringing together
all the uttered speech sounds and to make sense out of them.

Perceptions of coarticulation in adults have been studied in fricatives and stop the more, as it’s easy to apply the
cut and splice technique for these sounds. Also, the nasals have been studied. In the fricatives, the friction
portions of /f/ and /o/ carry more consonantal information. Transition is also a major cue for the perception of
the fricatives (Lariviero Otal, 1970 Repp And Mann 1980).

Generally, adults were able to make use of the coarticulatory information, even when tokens were cut and
presented in segments which did not physically contain cues. They make use of both place and manner cues but
more of place cues has been found (Kuehn And Moll 1972) syllabic stress and length of the syllable duration
also played important roles in the perception of coarticulatory information from syllables (fowler 1981).

The adult listeners sort the phones into categories prior to the point in time at which the closest approach to a
target (for that phone) is reached (Kuehn & Moll, 1972). They make use of such strategies in this process,
whereby there is modification of acoustic pattern of preceding and following sounds. For such sounds to be
perceived or identified, there requires a degree of invariance in production. This is brought about by
normalization where the listeners system compensates for the discrepancy between the encoding system of the
speaker and the decoding system of the listener. It has been found in all of adult studies that adults do make use
of this principle, whereby they make use of the coarticulatory information, even when the stimuli are presented
to them with as minimum acoustic cues, as possible.

In particular, the fact that the influence of a segment often extends well beyond its own boundaries means that
information about that segment is available to perception longer than would be the case if all cues were
confined inside its boundaries. As pointed out by early perceptual theories, the possibility of ‘parallel
processing’ of information for more than one phoneme probably allows speech to be perceived more rapidly
than would otherwise be feasible. The possibility that the origin of co-articulation lies not only in the
requirements of the articulatory mechanism, but in those of our perceptual system, cannot be discounted.
On the side of perception of co-articulated speech, there is experimental evidence that co-articulatory influences
are perceived and that listeners can correctly identify the intended segment by normalizing the percept as a
function of context, i.e. by factoring out the contextual influences. What exactly normalization means is an
object of debate: do listeners use the acoustic information for recovering the invariant co-articulated gestures as
propounded by the theory of direct perception (Fowler 1986; Fowler and Smith 1986) or is the sound itself and
its acoustic structure the object of perception, as propounded by a number of other speech scientists (Diehl
1986; Ohala 1986)

Theories of speech perception and of listener speaker interactions differ in their accounts of the perceptual
efficacy of co-articulated speech signals. Lindblom’s hyper and hypo-speech theory postulates that, in listener-
directed speech, speakers will minimize co-articulation (by increasing articulatory effort) so as to maintain
sufficient distinction between contrastive differences (Lindblom, 1990).

Tatham and Moreton (2006) have similarly argued that speakers reduce co-articulation in order to increase the
likelihood of being understood by listeners. There is abundant evidence, some of which is reviewed below, that
the acoustic consequences of Co-articulatory overlap can be perceptually advantageous. In general, the findings
show that, when provided with appropriate Co-articulatory information, listeners respond more quickly and
more accurately (e.g., Martin and Bunnell,1981; Whalen, 1991; Connine and Darnieder, 2009) and can predict
an upcoming or deleted segment with better than chance accuracy (Ostreicher and Sharf, 1976; Alfonso and
Baer, 1982; Jenkins et al., 1999) If listeners use co-articulatory information in their moment-by-moment
processing then, as the acoustic signal unfolds over time, listeners’ perceptual assessments should evolve in
ways that precisely use the time-varying information present in the signal.

Perceptual evidence
Most of the research on the perception of nasal coarticulatory effects has focused on the influence of the vocalic
environment on nasal consonants. Concerning /m, n, ŋ/, it was generally agreed that two cues, i.e. formant
transitions and nasal murmurs, were mainly responsible for the identification of these sounds. In seeking to
determine the vowel contextual influence, several investigators have used either synthetic speech (Liberman et
al. 1954) or natural speech (Kuehn and Moll 1972; Repp 1986a) to demonstrate that formant transitions are
perceptually more salient in a low-vowel context than in a high-vowel context. According to Sharf and Ohde
(1981) this trend can be ascribed to the fact that low vowels have a greater intrinsic intensity. The importance of
the nature of the adjacent vowel on the consonant’s perception was confirmed by Zee (1981).

In an experiment on the effect of vowel quality, he used natural speech signals with different signal-to-noise
ratios to demonstrate that, whereas /m/ and /n/ are often misidentified after the front vowels /i/ and /e/, the same
consonants are correctly perceived after the vowel /a/ even in the noisiest conditions. However, it should be
pointed out that, in addition to the intensity factor, the spacing of formant transitions, may play an important
part as well. Sussman, McCaffrey and Matthews (1991) and Sussman, Hoemeke and McCaffrey (1992) have
shown a high correlation between the locus equation coefficients and consonant place of articulation. Among
other things, it was demonstrated that, whereas the place-dependent transitions for any manner of consonant
(nasal or stop) are very much alike near /i/ (this being also true for labials and velars near /u/), formant
transitions are clearly separated close to /a/. Carlson, Grandstrom and Pauli (1972) reported an inverse
relationship in terms of vowel contextual influence.

In a perceptual experiment based on synthetic stimuli, they found that the perceptual relevance of the nasal
murmur Fn2 was greater in a high-vowel context for the nasal consonants /m, n/ in Swedish. Further evidence
of this interactive influence was provided by Abramson et al. (1981) who demonstrated through articulatory
synthesis that, in American English, the perceptual boundary along an acoustic continuum /d–n/ shifted as a
function of the adjacent vowel height.

1. All, et al (1971)
Material: CVC & CWC in which final consonance were (M or N) are non-nasal consonants (b, v, f, z,
d) the entire final consonant along with v-c transition was spliced away.
Task: To determine the perception of coarticulated nasality.
Conclusion: Results indicated that the presence of 1 nasal consonant should be predicted and that
listeners utilize this information to lengthen the phoneme- processing load
Consonants which allowed low back vowel/a/perceived as nasal with significantly greater frequency
than vowels/u/,/ei/,/i/ (high vowels). Perceptually significant coarticulation of velar are opening across
the vowel in CWN type sequences is supposed to have counterparts in the coarticulation of other
articulatory gestures and this way perception is related to production.
2. La Riviere, et al (1970)
Materials: CV syllables (C: s, sh, f, th and V: a, i, u )
Task: To study the presence of useful perceptible information in the fricative that precedes the vowels.
Conclusion: It was found that: f, s, use vocalic information for perception and s,sh use information
based on frications.
Transition is a cue for f and th. Friction portion of sh and th carry more consonantal information.
3. Lehiste, et al (1972)
Material: VCV syllabol (C;p.t.k and V;I,ae,a,u)
Task: To determine whether the change in formal transition due to the anticipation of the following
vowels where perceptually significant
 Conclusion: Results showed that listeners performance is supplying the (missing) initial or final viowels
was at chance. A final vowel may have an implosive transition or a n initial vowel may have on
explosive transition or not perceptible to listeners.

I. CROSS-LANGUAGE STUDY OFTHE PERCEPTION OF COARTICULATED SPEECH

Beddor.P.S., et.al Project: University of Michigan, Ann Arbor, USA
National science foundation (1997)

The aim of the study was to determine whether language specific patterns of co-articulatory organization give
rise to language-specific patterns of perceptual compensation. Study includes series of experiments that explore
two types of co-articulation: co-articulatory vowel nasalization and vowel-to-vowel co-articulation.

a. COARTICULATORY VOWEL NASALIZATION

In Thai, but not English, longer vowels were proportionately less nasalized than shorter vowels. Thus
anticipatory vowel nasalization in the two languages differs in both temporal extent and its constant (English)
vs. variable (Thai) proportional nature.
Rena Krakow investigated whether these language differences in patterns of co-articulatory vowel nasalization
have consequences for the perception of vowel nasalization by native speakers of English and Thai. To assess
perception of oral and nasal vowels in different co-articulatory contexts, in this study they have tested 16 native
English-speaking listeners and 15 native Thai-speaking listeners in two perceptual paradigms, a nasality rating
test and a nasality (4IAX) discrimination test. Both tests involved manipulating original [CVC] and [Nv)N]
words ( b e d , b ode , m en , and m oa n ) produced by a male native speaker of American English. English and
Thai listenersÕ responses to tests of nasality rating and nasality discrimination show that listeners perceptually
compensate Ñ albeit partially rather than fully Ñ for vowel nasalization under co-articulatorily appropriate
conditions. The response patterns of the two language groups were similar in many respects, but the cross-
language perceptual differences that were found appear to be linked to the differences in co-articulatory
structure in the two languages.

b. VOWEL-TO-VOWEL COARTICULATION
Languages also differ in their patterns of vowel-to-vowel co-articulation. This study investigates vowel-to-
vowel co-articulation in the productions of several speakers each of Shona and American English. Stimuli are
disyllables with initial stress, ’CV1CV2, and tri-syllables with medial stress, CV1’CV2CV3, where V= /i e a o
u/ and C = /b/ for English and /p/ for Shona (Shona /p/ is relatively un-aspirated; the voiced bilabial stop in
Shona is implosive) Patterns of co-articulatory nasalization and vowel-to-vowel co-articulation are influenced
by a range of segmental and prosodic factors.In addition, not all speakers of a language show particular patterns
to the same extent. Listeners must be able to deal with both types of variability in co-articulation. At the same
time, systematic language differences emergedin these perceptual studies, and these differences support a
theoryof perceptual compensation that is sensitive to language-specific co-articulatory patterns.

II. A RATIONAL ACCOUNT OF PERCEPTUAL COMPENSATION FOR CO-

ARTICULATION
Sonderegger.M., Alan Yu (2009)

The term perceptual compensation (PC) has often been used to characterize this type of context-induced
adjustment in speech perception. For example, the identification of an ambiguous target syllable as /da/ or /ga/
is shifted by preceding /ar/ or /al/ contexts (Mann, 1980): the same /Ca/ token is less likely to be heard as /ga/ in
/arCa/ context than in /alCa/ context. This effect has been argued to result from perceptual reduction of the co-
articulatory fronting effects of /l/ on a following velar consonant: listeners are compensating for the effect of /l/
on /g/. PC effects have been observed in many phonetic settings. The fricative /S/ has lower noise frequencies
than /s/, and lip rounding lowers the resonant frequencies of nearby segments. Synthetic fricative noises ranging
from /S/ to /s/ are more often identified by English listeners as /s/ when followed by /u/ than by /a/ (Mann
&Repp 1980; see also Mitterer 2006), presumably because listeners take into account the lowering effect of lip
rounding from /u/ on the noise frequencies of /s/ in natural co-articulated speech. As another example, the
perception of a fundamental frequency (f0) contour can change as a function of vowel height (Hombert, 1978;
Silverman, 1987) or consonant voicing (Pardo& Fowler, 1997): /i/ is perceived as lower in pitch relative to an
/a/ with the same f0, presumably because high vowels typically have higher f0 than low vowels. The generality
of PC effects is accentuated by evidence for contextual compensation with speech and non-speech sounds in
human and non-humans (Holt, Lotto, &Kluender, 2000; Lotto, 2004). For example, when /da/–/ga/ syllables are
preceded by tone glides matching in frequency to the third formant (F3) transition of /al/ or /ar/, listeners’
syllable identification responses shifted in the same direction as when targets were preceded by real speech (/al/
or /ar/). The same effect was observed even when steady-state tones at the offset frequency of /al/ or /ar/ F3
were used (Lotto &Kluender, 1998; cf. Viswanathan, Fowler, & Magnuson, 2009)

III. SEGMENTATION OF CO-ARTICULATED SPEECH IN PERCEPTION

Fowler.C.A. etal Perception and psychophysics 1984,34 (4), 359-368

Aim: to investigate how listeners segment the acoustic speech signal into phonetic segments and explores
implications that the segmentation strategy may have for their perception of the context-sensitive allophones of
a phoneme.
Method: 2 manner of segmentation are contrasted. (stimulus used are /gi/ and /gu/)
 1. Listeners segment the signal into temporally discrete, context-sensitive segments.
2. Talker’s production of the segments, they partition the signal into separate, but overlapping, segments
freed to their contextual influences.

Results: In this study, the listeners being using the acoustic information for a segment as such, whenever it
occurs in the speech signal. This leads to a reaction time advantage for original and spliced over cross-spliced
stimuli in the choice reaction time study if the information is co-articulatory, the listeners do not integrate it
with information for a segment with which it co-occurs in time.

IV. FRICATIVE–STOP CO ARTICULATION: ACOUSTIC AND PERCEPTUAL EVIDENCE:

Bruno H. Repp Virginia A. Mann (1982).
Aim:
FCV and VFCV utterances from which the fricative noise and release burst had been removed might provide
another means of assessing any co articulation between fricative and stop a procedure used successfully by
Repp and Mann (1981).
Method:
Eight native speakers of American English each produced 10 tokens of all possible CV, FCV, and VFCV
utterances with V = [a] or in], F = is] or [J'],and C = it] or [k]. Acoustic analysis showed that the formant
transition onset following the stop consonant release was systematically influenced by the preceding fricative,
although here were large individual differences.
Conclusions: They found additional evidences for fricative induced shifts in stop production in listeners'
perception of the vocalic formant transitions, although the correlation between the acoustic and perceptual
findings were weak.

V. PRODUCTION AND PERCEPTION OF CO ARTICULATED TONES:

Journal of Acoustic Society of America(1994)
Aim: To examine the perception of coarticulated tones, naturally produced words and phrases were used as
stimuli for tone identification. However, their semantic information was removed through waveform editing
while all the tonal information contained in the signal remained intact.
Method:
Subjects:
Chinese students, 3 males and 2 females served as Subjects. There are 16 possible tonal combinations in
Mandarin.
Stimuli: Trisyllabic words or phrases whose second syllables have either the rising tone or the falling tone were
used in this experiment.
Results: However, the amount of deviation of a tone from its canonical form due to co articulation varied
depending on the nature of the tonal context. In a context where, adjacent tonal values agree (a "compatible"
context), the deviation was relatively small. In a
context where adjacent tonal values disagree (a "conflicting" context), the deviation was much greater,
sometimes even to the extent of changing the direction of a dynamic tone.

Study:
VI. ACOUSTIC ANALYSES AND PERCEPTUAL DATA ON ANTICIPATORY LABIAL
COARTICULATION IN ADULTS AND CHILDREN
Joan A. Sereno, Shari R. Baum, G. Cameron Marean, and Philip Lieberman(1986)
Aim: The present study investigated anticipatory labial coarticulation in the speech of adults and children.
Method:
Stimuli and subjects: C V syllables composed of [ s] , [t ], and [d ] before[i] and [u ], were produced by four
adult speakers and eight child speakers aged 3 -7 years.
Each stimulus was computer edited to include only the aperiodic portion of fricative-vowel and stop-vowel
syllables.
Perceptual data were obtained by presenting the aperiodic consonantal segments to subjects who were instructed
to identify the following vowels as [i] or [u].
Results:
Both the acoustic and the perceptual data show strong coarticulatory effects for the adults and comparable,
although less consistent, coarticulation in the speech stimuli of the children.