Phonetically Governed Voicing

1
This study was supported by R01 DC-05210,
from NIDCD
Phonetically Governed Voicing Onset and Offset
in Children Who Stutter
Rick Arenas M.A.
Patricia Zebrowski Ph.D.
Dept. of Speech Pathology and Audiology
University of Iowa
Speech production is achieved through
the dynamic coupling of the respiratory,
laryngeal and articulatory systems. Several
theories of stuttering hypothesize that the
behavior emerges from a lack of temporal
coordination between these systems. There has
been a long history of looking specifically at
the laryngeal system and the coordination between
the laryngeal and respiratory system with
relation to their role in stuttering. In adults
who stutter (AWS) there is physiologic data from
connected speech that provides evidence of
abnormal laryngeal activity during dysfluent
utterances (Conture, McCall Brewer, 1977
Freeman Ushijima, 1978 Shapiro, 1980).
However, much of this research has looked at AWS,
making it difficult to determine if differences
are related to underlying deficits or whether
they are due to learned compensatory behaviors.
One strategy for avoiding this problem is to look
at the behavior of children close to the time of
stuttering onset. Recently, Watson
(1998), and Robb and Smith (2002) described
developmental changes in F0 variability at the
level of the phonetic segment, specifically as it
occurs across the vowels in a VCV syllable, where
the consonant is a voiceless obstruent. According
to Robb Smith, the change and variability of
change in vowel fundamental frequency in such
phonetically governed devoicing and voicing
gestures are thought to reflect the coordination
of laryngeal and aerodynamic adjustments
necessary for the production of voiceless
obstruents. Close examination has shown
consistent patterns of vowel F0 change and
variability immediately before and after the
production of voiceless obstruents.
Specifically, there appears to be a reduction in
F0 across vocal cycles in the vowel preceding the
obstruent, and a relatively high F0 at voicing
onset in the subsequent vowel, followed by a
decline. These patterns of voicing onset and
offset have been seen in children as young as
four years old but not in children two years of
age (Robb Saxman, 1985). Thus, the development
of these mature patterns occurs between age two
and four, possibly making this measurement a good
index of laryngeal coordination within the same
window of time that stuttering generally begins.
The goal of the present study was to
compare the developmental patterns of
phonetically governed voicing onset and offset in
children who do not stutter (CWNS) and children
who stutter (CWS) that are close to stuttering
onset. The CWS and CWNS groups were divided in
to three year olds and four year olds to identify
developmental trends. It was hypothesized that
the CWS and CWNS would show similar patterns
phonetically governed voicing but the CWS would
show greater within subject variability
suggesting instability in the laryngeal system
that may contribute to their stuttering.

2
Participants and Stimuli
Table 1. provides a breakdown of the number of
participants in each group. The participants
were prompted to say the following three
utterances six times each papa, see two and
I see two papas. The /apa/ and /itu/ were the
target VCV sequences for the analysis. Some of
the participants were not willing or able to say
all the utterances six times, and some of the
utterances we were not able to analyze due to the
acoustic signal. So not all participants
contributed an equal number of utterances for
analysis.
Table 1.
Data Analysis
Within CSL, each utterance was displayed as an
amplitude-by-time waveform. V1-C-V2 segments
were selected and magnified using vertical
cursors. An example of a typical waveform is
presented in Figure 1.
F0 Onset The first ten vocal cycles of V2
following the consonant were identified and
marked by creating voiced period marks on a
peak-to-peak basis. The onset of voicing was
identified by the first visibly periodic vocal
cycle following the voiceless stop consonant.
The time values for each voiced period mark were
then extracted from CSL and the period for each
vocal cycle was calculated. The F0 for each
vocal cycle was determined as the reciprocal of
the period. F0 Offset The last ten vocal
cycles of V1 preceding the consonant were
identified and marked by creating voiced period
marks on a peak-to-peak basis. The offset of
voicing was identified by the last visible
periodic vocal cycle preceding the voiceless stop
consonant. The F0 for each vocal cycle was
determined as the reciprocal of the period. The
F0 values of the ten vocal cycles for V1 and V2
were normalized by converting them to semitone
values relative to a reference cycle. The
reference cycle for both the F0 onset and F0
offset was the vocal cycle furthest from (prior
to and following) the consonant (Watson, 1998).
Figure 1. Waveform for determining F0 offset and
onset. The cycles are demarcated by the voiced
period marks placed at the peaks of each cycle.
3
Results
As mentioned previously, F0 values were
converted to semitone values using the tenth
vocal cycle furthest from the consonant as the
referent. The semitone values for each subject
were averaged and pooled to create group
averages. The semitone values for the V1 and V2
data sets were separately submitted to a series
of 2 x 10 repeated measures analysis of variance
(ANOVA) tests (SPSS, 2005) to evaluate the
interaction of change in average semitone values
across the 10 vocal cycles. Either phenotype
(CWNS vs CWS) or age group (3-year-olds vs
4-year-olds) were used as the between-group
factor in the series of 2 x 10 ANOVAs, with
vocal cycle always being the within-group factor.
An a priori alpha level of .05 was used to
determine statistical significance.
The semitone values for onset and offset
were initially analyzed using the utterances
papa and see two separately. The results of
these analyses were very similar across groups
therefore subsequent analysis were performed by
collapsing the individual data across the papa
and see two utterance.
The average semitone values for voicing
offset and onset from the CWS are displayed in
Figure 2 and 3, respectively. The averaged
standard deviations for the CWS and CWNS are
plotted in figure 4 and figure 5.
Figure 2. Average semitone values for CWS. ? 3
year olds, ? 4 year olds.
Figure 3. Average semitone values for CWNS. ?
3 year olds, ? 4 year olds.
Figure 4. Standard deviation of semitone values
for CWS. ? 3 year olds, ? 4 year olds.
Figure 5. Standard deviation of semitone values
for CWNS. ? 3 year olds, ? 4 year olds.
Voice Onset Results

• 2 x 10 ANOVA test with cycle as the within-group
  factor and phenotype as the between-group factor
  revealed no significant differences between the
  CWS and CWNS in both the 3 and 4 year old groups
• When comparing the 3 to 4 year old CWS, a
  significant age-by-vocal-cycle interaction was
  found F(8,168) 7.384, p lt .001.
• When comparing the 3 to 4 year old CWNS, no
  significant differences were found

4
Results
Voice Offset Results

• 2 x 10 ANOVA test with cycle as the within-group
  factor and phenotype as the between-group factor
  revealed no significant differences between the
  CWS and CWNS in both the 3 and 4 year old groups
• When comparing the 3 to 4 year old CWS, a
  significant main effect was found for age group
  F(1,21) 5.004, p lt .05.
• When comparing the 3 to 4 year old CWNS, a
  significant main effect was found for age group
  F(1,10) 7.067, p lt .05.

Discussion
The present report found no significant
differences when comparing the CWS to CWNS at
either age group. However, our results did
further illuminate the developmental course of
phonetically governed voicing onset and offset.
Our results from the four year old groups closely
resemble the results that Robb Smith (2002)
found for the same age group. That is, by age
four children present with voicing onset and
offset patterns that are similar to adults. Our
data suggest that the mature voicing offset
pattern is not present at age three in both CWS
and CWNS. However, at age three mature voicing
onset patterns are present but are not as
pronounced as they are at age four. Although
there was no significant difference in voice
onset between the three year old CWS and CWNS, it
was interesting that there was a significant
difference between the three and four year old
CWS but not between the three and four year old
CWNS. This may mean that children who start
stuttering at age three may have slightly
delayed, but not aberrant, voice onset behavior.
The hypothesis that the CWS would show
greater variability due to less stable vocal
behavior was not found. In fact, variability, as
measured by standard deviations of the semitones,
were very similar across the CWS and CWNS in both
age groups. It appears that the time
period between ages three and four is very
important in the development of phonetically
governed voicing onset and offset. The data from
the present paper suggest that by age four there
is no difference between CWS and CWNS. However,
there is indirect evidence that CWS may begin
developing a mature pattern of voicing onset
later than CWNS. Due to the fact that both the
CWS and CWNS did not show a mature pattern of
voicing offset at age three it is difficult to
know whether one group may develop a mature
pattern sooner than the other. Future research
should investigate CWS and CWNS longitudinally to
look at the pattern of phonetically governed
voicing by collecting speech samples every month
from age two to age four. Such a study would
help better understand the time course of the
development of phonetically governed voicing.
Assuming that these mature patterns of voicing
onset and offset reflect the coordination of the
respiratory and laryngeal system, having a better
understanding of their development and role in
different speech disorders could provide
clinically relevant tests that noninvasively
evaluate a childs speech motor development.
References
Robb, M., Saxman, J. (1985). Developmental
trends in vocal fundamental frequency in young
children. Journal of Speech and Hearing Research,
28, 421-427. Robb, M. Smith, A. (2002).
Fundamental frequency onset and offset behavior
A comparative study of children and adults.
Journal of Speech, Language and Hearing Research,
45, 446-456. Watson, B. (1998). Fundamental
frequency during phonetically governed devoicing
in normal and aged speakers. Journal of the
Acoustical Society of America, 103, 3642-3647.
A PowerPoint of this poster can be downloaded
from this website
http//www.shc.uiowa.edu/wjshc/research/stuttering
/research.html