Content uploaded by Aude Noiray
Author content
All content in this area was uploaded by Aude Noiray on Apr 21, 2018
Content may be subject to copyright.
1
The Development of Motor Synergies in Children:
Ultrasound and Acoustic Measurements
Aude Noiray
1
Haskins Laboratories, 300 George Street, Suit 900, New Haven CT 06510, USA
Lucie Ménard
Center for Research on Language, Mind, and Brain,
Département de linguistique, UQAM,
Case postale 8888, Montréal (Québec) H3C 3P8, Canada
Khalil Iskarous
Haskins Laboratories, 300 George Street, Suit 900, New Haven CT 06510, USA
---
Copyright (2013) Acoustical Society of America. This article may be downloaded for personal
use only. Any other use requires prior permission of the author and the Acoustical Society of
America.
The following article appeared in The Journal of the Acoustical Society of America 133, 444
(2013); https://doi.org/10.1121/1.4763983
---
Author to whom correspondence should be addressed. Electronic mail:
noiray@haskins.yale.edu
1
Also at: Linguistic Department, Center for Excellence Cognitive Science, Potsdam
University, 14459 Potsdam, Germany.
2
ABSTRACT
The present study focuses on differences in lingual coarticulation between French
children and adults. The specific question pursued is whether 4-5 year old children have
already acquired a synergy observed in adults in which the tongue back helps the tip in
the formation of alveolar consonants. Locus Equations, estimated from acoustic and
ultrasound imaging data were used to compare coarticulation degree between adults and
children and further investigate differences in motor synergy between the front and back
parts of the tongue. Results show similar slope and intercept patterns for adults and
children in both the acoustic and articulatory domains, with an effect of place of
articulation in both groups between alveolar and non-alveolar consonants. These results
suggest that 4-5 year old children 1) have learned the motor synergy investigated and, 2)
have developed a pattern of coarticulatory resistance depending on consonant place of
articulation. Also, results show that acoustic locus equations can be used to gauge the
presence of motor synergies in children.
PACS numbers: 43.70.Ep, 43.70.Mn, 43.70.Jt
Key words: coarticulation, speech motor control, language acquisition.
3
I. INTRODUCTION
Coarticulation is generally defined as the articulatory overlapping of a sound on
another one. Beyond this basic definition, describing the nature of this process; i.e.,
whether it is a crucial motor control for speaking a language or the “on-line”
consequence of the interactions among articulators, has remained an object of
longstanding controversy (see Coarticulation, Theory, Data and Techniques, Hardcastle
& Hewlett, 1999).
In this study, we consider coarticulation to be a complex mechanism, involving
multiple articulators (e.g., the tongue, the lips) whose actions are finely coordinated in the
space of the vocal tract as well as over time to produce intelligible and fluent speech. It
involves functional articulatory synergies whose crucial function is to permit efficient
and stable language-specific coordination among muscles and articulators to achieve
speech tasks. From a developmental point of view, a main goal for the child is therefore
to develop these functional synergies and reduce the number of possible articulatory
coordinations to the ones that are the most consistently produced by adults in a given
language (Smith & Zelaznik, 2004). Indeed, although children’s articulations of
phonemes may be perceptually intelligible very early in age, their articulatory strategies
differ from adults’ because of immature speech motor control. In addition, differences in
the anatomy of their vocal tract between children to adults, children may require to adapt
their articulations in order to achieve an acoustic target that is comparable to adults and
intelligible to others.
A main question in this study regarded therefore whether 4-5 year old children
have acquired one particular synergy: the use of the back of the tongue to assist the tip in
4
the formation of alveolar closures in CV syllables. Both the jaw and the tongue back
might assist the tongue tip to make contact with the palate. Adults consistently use the
tongue back to push the tip forward (e.g., Iskarous et al., 2010, Sussman et al., 1999),
whether or not they also use the jaw, but it is unknown whether young children do the
same, since at 4-5 years of age, they may lack fine control over the functional subparts of
the tongue.
To achieve our goal, we transposed measures of Locus Equation (LE), commonly
employed in acoustics to the articulatory domain. LE was used as a metric to investigate
whether 4-5 years old children differentiate the tongue tip and tongue body to achieve
adult like patterns of CV coarticulation according to consonantal contexts, i.e., large
coproduction between the vowel and labial or velar consonants, but lesser coarticulation
in the alveolar context (Sussman et al., 1999) as a result of a motor synergy between the
tongue back and the tongue tip for achieving the main constriction at the alveolar ridge.
To our knowledge, no study investigating coarticulatory patterning in
preschoolers has ever provided any direct account from the tongue (Zharkova et al.,
2012; 2011; 2008 in school-aged children from 6 to 9 years of age). Unlike other muscle
systems in the human body, the tongue is a muscular hydrostat - like octopus tentacles or
elephant trunks - that does not rely on a distinct skeletal system (Kier & Smith 1995;
Stone, 1993) but can produce a large variety of movements and complex shapes. In a
developmental (and clinical) perspective, it is important to study the maturation trajectory
of the tongue as it is central to the production of all vowels and most consonants, and
therefore it is a crucial articulator to be controlled for coarticulation.
5
A. Development of coarticulation in CV syllables
In children, the development of spatial and temporal organization of speech
actions, that is, of articulatory gestures, is poorly understood because its investigation has
mostly been limited to measures of the acoustic output or phonetic transcriptions (e.g.,
Goodell & Studdert-Kennedy, 1993; Lee et al., 1999; Munson, 2004; Nittrouer et al.,
1996; Sussman et al., 1999) that give only incomplete evidence of the underlying
articulations. Although acoustic measurements of the speech signal provide important
insight into the ontogeny of coarticulation, the lack of corresponding information from
articulation itself limits theoretical conclusions about 1) the maturation of the speech
motor system (Zharkova et al., 2011); and 2) the convergence of coarticulatory strategies
on adults’ patterns (Noiray et al., 2009).
There is general agreement that in the first years of life, children’s vowel
productions exhibit high variability in acoustics, which suggests that they also vary in
their articulatory strategies to match the perceived targets of adults (Lee et al., 1999;
Ménard et al., 2007). The development of articulatory skills required for fluency varies a
lot both across children and within a child, with spurts and plateaus, (cf. Kent, 1976;
2004). Regarding consonants, children’s production accuracy differs depending on
whether they are produced in isolated forms or within words (with differences in
accuracy depending on word position; Canadian French: McLeod et al., 2010; English:
McLeod et al., 2001; Stemberger & Bernhardt, 2002). Results from previous studies (de
Boer 2000; Fowler & Goldstein, 2003; Goldstein, 2003; McLeod et al., 2009 for
consonant clusters) suggest coarticulation of phonemes distinguished by motion from two
distinct articulators (e.g., lip motion for /b/ and tongue motion for /i/ in “beep”) would be
6
mastered earlier in typically developing children than those requiring contrastive actions
from a single articulator (e.g., successive tongue motions for /t/ and /æ/ in “tack”). In the
latter case, young children may be expected to show more spatiotemporal overlap
between Cs and Vs than adults because of an immature control over the functional
subparts of their tongue. However, such a hypothesis has not yet been empirically
demonstrated in young children with articulatory data partly because of methodological
constraints associated with child studies.
B. Locus Equation as a measure of lingual synergy
1. LE measures in adults
Locus equations (LE) have been identified as relational invariants for consonants
(Sussman et al., 1991), as a measure of the degree of coarticulation between consonants
and vowels (Krull, 1987), and as a measure of the coarticulation resistance of consonants
(Fowler, 1994). LE are linear regressions calculated between F2 at the beginning of a
CV transition and F2 at the acoustic midpoint of the vowel for a given consonant
produced in the context of a variety of vowels, (Lindblom, 1963; Nearey & Shammass,
1987). The regression equation parameters (slope, intercept) provide insight into the
magnitude of coarticulation depending on the consonant’s place of articulation. A steep
slope of 1.0 is evidence for a high degree of coarticulation between C and V, because it
means that for every 1 Hz change in the vowel midpoint, there is a corresponding 1Hz
change in the CV transition onset. On the other hand, a low slope indicates that the
consonant’s F2 shows a smaller change for each 1 Hz change in the vowel, an indicator
of a smaller degree of coarticulation. The intercept value indicates the value for F2 at the
7
consonant release for a zero F2 value at the midpoint of the subsequent vowel. Empirical
work has shown that the magnitude of the slope characterizing CV coarticulation differs
according to consonant place of articulation (e.g., Krull, 1987; Nearey & Shammass,
1987; Sussman et al., 1998); it is larger in labial context than in the velar and especially
the alveolar contexts. The difference between alveolars (e.g., /t, d/ and non-alveolars
(e.g., /k, p/) is consistent across all studies, whereas labials and velars often have similar
slopes. Intercept magnitudes are negatively related to slopes, with labial intercepts
smallest and alveolars largest. Using articulatory and acoustic data, Iskarous et al. (2010)
showed that the reason that alveolars have a lower slope and higher intercept than non-
alveolars is the particular way in which the tongue back interacts with the main
constrictor for the consonant. Specifically, alveolars have a low slope, because the tongue
back is pushed forward to assist the tip and that prevents it from assisting in the
constriction for the following vowel (Manuel & Stevens, 1995). This means that
coarticulatory overlap is limited, and, correspondingly, coarticulation resistance is high.
The high intercept is a direct indication that the tongue back is more advanced in the
vocal tract for the alveolars than for other consonants. Iskarous et al. argued that the
synergy involving the tongue back for the achievement of alveolars is the basis for the
difference in slopes, and is the basic reason why alveolars have a lower degree of
coarticulation and a higher coarticulation resistance. Note that depending on the nature of
the vowel, the synergy between the tongue body and tongue tip may be facilitated. For
instance, in /ti/, the front position of tongue body is required both for the alveolar and the
high front vowel. However, as regression slopes are computed across a range of vowels,
they reflect general patterns of coarticulation for places of articulation.
8
Contrary to adults, children may first mainly use the jaw synergy to assist the
tongue tip in making constrictions before starting to use the tongue back to move the tip
as the jaw requires fewer muscles for its motion and may therefore be easier to control.
This would be measureable using locus equations, because, if the tongue back is not
occupied by helping the tip for the alveolar constriction, it could start to coarticulate
earlier with a following vowel, raising the slope for the alveolar to the level where there
may not be a difference between alveolars and non-alveolars (e.g., bilabials and velars) in
coarticulation degree (or in locus equation slope). The children we examine are 4-5 year
olds. This age group does still show many differences with adults in measures of their
speech, and our interest is in establishing whether these differences extend to this motor
synergy.
2. LE measures in infants and children
So far, several studies have reported LE measures of infants/children’s speech
productions (Gibson & Ohde, 2007 from 17 to 22 months; Goodell & Studdert-Kennedy,
1993 from 22 to 32 months; Sussman et al., 1996 at 12 and 21 months, Sussman et al.,
1999 from 7 to 40 months; Sussman et al., 1992 at 3,4 and 5 year olds) to characterize
children’s modifications of articulatory controls with age. The main results of these
studies are: 1) high variability in infants’ coarticulatory patterns, 2) gradual distinctions
in stop place of articulation with lexical development, 3) a general trend toward greatest
coarticulation magnitude in the labial context (illustrated by having the steepest slopes),
intermediate in the velar contexts, and lowest in the alveolar context.
9
However, results are contradictory across studies, showing either more
coarticulation in children than adults (Nittrouer et al., 1989; Nittrouer & Studdert-
Kennedy, 1996; Studdert-Kennedy, 1987) or less (Green & Moore, 2002; Kent, 1983;
MacNeilage & Davis, 1998) or finally no substantial difference (Sereno & Lieberman,
1987; Katz et al., 1991) between the groups. These differences are compounded by the
fact that it is difficult to measure formants from children’s speech due to their high F0
and consequent wide separation between harmonics. Developmental studies using
standard LE measures provide only a partial explanation for the maturation of articulatory
coordination, because analyses are conducted on the acoustic outputs of the articulatory
mechanisms responsible for coarticulation rather than on the articulatory actions
themselves. In one of the most important studies, Sussman (1999) observes a decrease in
the slope for alveolar consonants, as children develop, and attributes that decrease in
slope to the development of separate control for the tip and dorsum. We believe that this
would be quite an important result if confirmed by articulatory measures and it is one of
the motivations of our ultrasound data analysis. This type of articulatory examination has
been lacking in typically developing children. Moreover, normative data would be
valuable for diagnosis and development of treatment strategies of speech and/or language
disorders (e.g., detection and treatment of early stuttering disorders that manifest as
differences in formant transitions in CV syllables, e.g., Cheng et al., 2002; Subramanian
et al., 2003).
In this work we use non-invasive ultrasound imaging, together with acoustic
measures, to establish whether the coarticulatory difference involving alveolar /t/ and
non-alveolars /p, k/ is present in 4-5 year old children. While this relation has been
10
demonstrated recently in adults (Iskarous et al., 2010), it has not yet been evidenced in
preschool children with direct measures of tongue motion. This work therefore aims at
providing new insight on the control of a crucial articulator for language acquisition.
Based on the literature, we predict children‘s coarticulatory patterns as young as 4
and 5 years old to differ from adults. Variability in both F2 and horizontal position of the
tongue body between the consonant and the vowel (indicated by correlation coefficients)
are expected to be higher in children than in adults. We expect the lowest correlation
coefficients to be observed for both measures in alveolar context as a consequence of
immature speech motor control (Noiray et al., 2010; Terband et al., 2009; Walsh et al;
2006). Also, we expect children to exhibit higher slopes in alveolar context than adults,
as they may not have mastered fine control over the lingual subparts to achieve
articulatory synergies as in adults. If this prediction is verified, it will bring articulatory
evidence that children display greater coarticulation degree than adults in alveolar
context.
II. METHOD
A. Subjects and stimulus material
Six Canadian French children aged 4 and 5 years old were recruited in Montreal
among monolingual French families. Their coarticulatory strategies were compared with
those of five adults (mean age: 25) who have achieved a mature speech motor system and
full knowledge of the phonological system of their language. Prior to the recording, a
hearing screening as well as a phonological assessment (Chevrie-Muller and Plaza, 2001)
was administered for each participant. The study was approved by an IRB, and consent
11
was obtained from the parents of the children, and the non-invasive methods used were
explained to the children before the experiment.
The task consisted in the production of /V1CV2/ sequences with the consonant C
corresponding to the bilabial stop /p/, alveolar /t/ and velar /k/ and vowel V to the high
front /i/, low /a/ and high back /u/. The three cardinal vowels allowed for testing
diverging tongue positions. In addition to the alveolar stop /t/ that is the target consonant
under investigation in this study, the bilabial and velar provided examples of stops that
either do not require any active motion from the tongue (e.g., bilabial stops) or in the case
of /k/, implies motion from the tongue but with contextual adaptability in the tongue
positioning as observed in adults (i.e., the amount of movement from the tongue body
varies with the surrounding vowel; e.g. Lofqvist 1999; Mosshammer et al., 1995).
Sequences were embedded in short carrier sentences: “c’est VCV ça”. Ten to twelve
repetitions of each VCV sequence were collected in random order for each participant,
young children included (with an average of 30 to 36 sequences for each consonant type:
labial, alveolar, velar). A total of 90 to 108 utterances recorded per participant.
B. Experimental procedure
Because of the young population targeted in this study, children’s recordings were
conducted at school while adults were recorded in a sound booth (at the Laboratoire de
phonétique, UQAM). Individual recordings consisted in a single 20-minute session
during which preliminary screenings and data were collected. Recordings were preceded
by a familiarization period with the experimenter, the task, target sequences, and
ultrasound setup. Note that the duration of familiarization phase was longer for children
12
to stimulate interest in performing the task and to ensure comfort with the experimenter
and set up.
During the recording, tongue data were collected via ultrasound imaging. This
technique has become an appealing tool to be used in the developmental field, because it
is a non-invasive and uncomplicated method for collecting lingual data of high quality
with very young children. Ultrasound imaging has been used in clinical studies with
school-aged children as a tool for speech therapy and on-line feedback (e.g., Adler-Block
et al., 2007; Modha et al., 2008) and recently for tracking the development of vowel
control in children (Ménard & Noiray, 2011).
Also, ultrasound imaging provides a continuous view of the tongue surface, which
is more convenient to measure the position of the highest point on the tongue during
vowels and consonants production than points tracking (e.g., with electromagnetic mid-
sagitttal articulography, Perkell et al., 1992 or x-ray Microbeam, Westbury, 1994).
Subjects were recorded with an ultrasound system (Sonosite 180 Plus, sr: 30Hz)
with an 84-degree probe and audio system (unidirectional microphone Sure). Participants
were seated comfortably in a chair and the ultrasound probe was held under the chin with
a stand similar to microphone stands. While head stabilizing devices are usually used to
hold the head stable during the experiment (e.g., Zharkova et al., 2011), it is not possible
to use them with very young subjects. Even so, for the children, as for adults, the
ultrasound probe was held below the chin via a stand to prevent horizontal and lateral
motion from the probe. As a consequence, the probe follows jaw movement and the
resulting ultrasound images are in jaw-based coordinates. However, it is important to
13
mention that this method also presents some constraints (i.e., the probe provides an
indirect estimate of the vertical jaw motion as the elasticity of the tongue’s floor is likely
to trigger variable distance between the probe and the jaw, Noiray et al., 2008).
In a recent study designed to test various articulatory parameters to characterize
ultrasound data, Ménard et al. (2012) confirmed that in such a setup, the horizontal
position of the highest point of the tongue is a reliable measurement point, which is
robust independently of vertical probe movement. An experimenter monitored the
experiment to insure participants would remain in the same position throughout the
recording and would not remove their chin from the probe. In the present study, both
ultrasound and speech sound signals were simultaneously recorded on a miniDV
Panasonic AG-DVC 30 camera, in NTSC format.
Two types of LE measures were taken: on the acoustic speech signal and on the
lingual data. For each participant and target V1CV2 sequence, a phonetic transcription of
the acoustic speech signal was conducted (via Praat, Boersma & Weenink, 1996).
Acoustic measures of F2 were automatically extracted using Linear Predictive Coding
formant estimation at the consonant-vowel transition (the first glottal pulse of the vowel
onset (T1)) and 25ms window centered at the midpoint of the vowel (T2). The number of
poles varied from 10 to 14, and a 14-ms Hamming window and preemphasis were
applied before formant extraction. Because obtaining measurements of formants in child
speech is more complex and can lead to more formant detection errors than in adult
speech, the automatic formant extraction was compared for each vowel with spectral
slices from a fast Fourier transform (FFT) with a hamming window. Complete details on
14
the method can be found in Ménard et al. (2008). For each participant, LE regression fits
were generated between F2 calculated at two points in time, i.e., at T1 (C release) and at
T2 (the acoustic midpoint of V2).
The adaptation of LE to the articulatory domain was conducted on the lingual data
simultaneously recorded with the acoustic speech signal via the video camera. Relevant
tongue images were extracted from ultrasound movies with Adobe Premiere Pro. These
corresponded to the consonant closure and T2 as identified in the acoustic analysis.
Tongue surface contours were then extracted using a semi-automatic system (EdgeTrak,
Li et al., 2003) and sampled at 100 points. The horizontal x coordinate corresponding to
the highest point y on the tongue body was used to determine the position of the
constriction on the front/back dimension for both C and V (cf. Figure 1). We refer to this
point as TBx.
[Insert Figure 1 about here]
III. RESULTS
Locus equation analyses have been done in two ways (Sussman, 1991). In the first
method, all the data from a group of subjects for a given consonant is pooled and a
regression line is estimated from the pooled data. In the second method, the regression
line is estimated separately for each subject, and then regression coefficients and statistics
are compared across subjects. The first method has the advantage of providing more
reliable statistics than if a regression is estimated for each subject separately, because
more data are available through the pooling. However it is possible that a regression
across a group shows a relation between two variables that is not characteristic of
individuals or subgroups (Raudenbush & Bryk, 2002). We performed both types of
15
analyses in this work. After presenting the data from both methods, we present a
statistical analysis of the difference between children’s and adults’ regressions using a
mixed effects general linear model.
[Insert Figure 2 about here]
Figure 2 compares the slopes for the adults and children based on the articulatory
and acoustic data. Panels a) and b) show the slopes of regression lines estimated using
the first method, from data pooled amongst adults (solid line) or children (dashed line).
The expected difference between alveolar and non-alveolar consonants can be seen in
both the adults’ and children’s data: the slope for /t/ is lower than the slope for /p/ and /k/
for both groups. This is true of the slopes derived from the articulatory data in Figure 2(a)
and the slopes derived from the acoustic data in Figure 2(b). The lower panels show the
mean and standard errors of the slopes for adults and children, where the regressions
were estimated for each subject separately and the mean and standard errors are
calculated across the 5 adults and 6 children. The statistics for individual subjects are
presented in Tables 1 and 2. The patterns within individuals closely resemble those in the
figures. Specifically, the low slope for alveolars is true for individuals, and not an artifact
of data pooling. The difference between adults and children that can be seen by
comparing the standard errors will be discussed later in this section, when the statistical
analysis is presented.
[Insert Figure 3 about here]
Figure 3 presents the intercepts. The upper panels show the intercepts derived
from pooled data, whereas the lower panels show the intercepts derived from individual
data. Both analyses show that /t/ has a higher intercept than /p/ and /k/, in both the
16
acoustic analyses (as in Gibson & Ohde, 2007) and the articulatory analyses, for both
adults and children. The asymmetry between /p/ and /k/ is present in the acoustic analysis
across subjects, and in those within subjects, but it is less obvious in the articulatory
analysis across subjects. These differences, however, are smaller than the difference
between the alveolar and non-alveolar places of articulation, which is the main focus of
this research.
Figures 2 and 3 show that children at 4-5 years have basically the same patterns as
the adults, but there are some small differences. The significance of the differences and
the differences between groups in patterns of slopes and intercepts was examined by
testing the hypothesis that there is no difference in patterns. Two mixed-effects general
linear model tests were performed, one for the articulatory and the other for the acoustic
data. A single value was entered for each subject and for each sequence produced. We
used a conservative p < .001 level for significance. In the articulatory model, the
dependent variable was TBxC, i.e., TBx at the consonant closure. The independent
variables were: 1) TBxV: TBx at the midpoint of the vowel (Continuous); 2) PofA: Place
of Articulation (Levels: /p/, /t/, /k/); 3) Generation (Levels: Adults and Children). Subject
was the random effect. The main effect of PofA is the intercept of the LE regression line,
since it indicates the value of TBxC when TBxV is controlled for. The interaction effect
between TBxV with Place of Articulation yields the slope of the regression lines, since it
indicates how the effect of TBxV depends on PofA. The interactions between these two
effects and Generation gives the difference in patterns between the adults and children.
The model tests for the acoustic slopes and intercepts had the same structure, except that
the dependent variable was F2 at T1 (at C release) and the vowel measure was F2 at T2
17
(the acoustic midpoint of V2). Significance was tested by using Monte Carlo Markov
Chain simulations (Baayen, 2007; Quene, 2008). The effects will be presented in terms of
the magnitude of the effect, as contrasts between levels, and the confidence interval (CI)
of the effect.
[Insert Table 1 and 2 about here]
Across adults and children, there was a significant effect of the interaction
between PofA and TBxV, which expresses slope, as a difference in slope of .44 (CI
[.37,.54], p < .001) between /p/ and /t/, with /p/ having a higher slope. /k/ also showed
significantly higher slope than /t/ by .39 (CI [.29,.50], p < .001). But there was no
significant difference between /p/ and /k/ in slope. There is also a non-significant
tendency for the slope difference between /p/ and /t/ to be larger in children than in adults
by .15 (CI [-.007,.335]). The p-value for this non-significant effect is .07, a marginal
effect. The effects for /p/ vs. /k/ and /k/ vs. /t/ showed no tendency to be different for
adults vs. children.
In the acoustic model, there was a significant difference in slope between /t/ and
the other places, where the /t/ slope is lower than /p/ by .22 (CI[.18,.28], p < .001), and
lower than /k/ by .22 (CI[.18,.27], p < .001). There was no tendency for these differences
to depend on Generation.
To summarize, based on the statistical analysis, we conclude that the slope patterns
are basically the same for adults and children, therefore the null hypothesis that the adults
and children have the same slope pattern cannot be rejected.
For intercepts, across adults and children, /t/ showed a more advanced tongue than
/p/ by 4.75 mm (CI [3.80,5.69], p < .001), and a more advanced tongue than /k/ by 4.23
18
mm (CI[3.25,5.22], p < .001). But the intercept for /p/ and /k/ were not significantly
different. There was a significant interaction between Generation and the intercept
difference between /p/ and /t/, where the tongue back is more retracted for the /t/ in the
adults than the children by 2.14 mm (CI[.55,3.71], p < .01). But we believe that this result
is an artifact of a larger vocal tract for adults than for children. Such an artifact arises for
the intercepts, but not for the slopes, because the slopes are in normalized units, whereas
the intercepts are in mm. The acoustic study showed, across adults and children, that the
intercept for /t/ was significantly higher than for /p/ by 733 Hz (CI[630,834], p < .001)
and for /k/ by 473 Hz (CI[364,573], p < .001). The /k/ intercept is significantly higher
than /p/ by 260 Hz (CI[149,360], p < .01). There were also significant effects of
Generation on intercepts derived from the acoustics. All of these effects show higher
intercepts in the children than adults, in Hz, but all of them are likely to be artifacts due to
the smaller vocal tract sizes for the children. To summarize, based on the statistical study,
we conclude that the intercept patterns are basically the same for adults and children.
Therefore the null hypothesis that the adults and children have the same slope pattern
cannot be rejected.
IV. DISCUSSION
This study aimed at investigating differences in coarticulation degree between 4-5
year old children and adults in CV syllables, employing ultrasound imaging of the tongue
with acoustic measurements. To our knowledge, it is among the first studies providing a
direct estimate of tongue motion in preschool children (cf. Zharkova et al., in press; 2011;
2008 for children aged 6 to 9 years of age). In light of the empirical work conducted over
the past three decades, understanding the development of coarticulated speech from
19
cross-study comparisons has indeed been quite challenging because of differences in
stimuli material: e.g., fricatives (Nittrouer & Whalen, 1989; Nittrouer, 1993, 1995;
Munson, 2004), velar stops (Kent, 1983; Sereno & Lieberman, 1987); the age span
investigated (e.g., babbling period or first words, Sussman et al., 1996, 1999; from 1 to 6
years of age with a gap between 2 and 6 years of age, Green & Moore, 2002) or in
method (acoustic measures: Goodell & Studdert-Kennedy, 1993; Katz et al., 1991;
Sereno et al., 1987; or articulatory: Optotrak: Smith & Goffman, 1998; video: Green &
Moore, 2002; Green et al., 2000; EMA: Cheng, 2007; Katz & Bharadwaj, 2001;
glossometry: Flege, 1983).
In this study, we transposed acoustic measures of Locus Equation to the
articulatory domain to compare coarticulation degree. We further sought to examine
whether at 4-5 years of age, children show articulatory synergy between the tongue tip
and tongue body to achieve adult like patterns of coarticulation for /tV/ syllables with
slopes in the descending order /p/ > /k/ > /t/. We predicted that children would exhibit 1)
more variability in their coarticulation patterns across consonantal contexts than adults
(lower R2) and 2) greater coarticulation degree than adults in alveolar context (higher
slopes) as a result of a global control of the tongue rather than a fine control over the
functional subparts of the tongue to achieve adult like synergies.
Results show that 1) adults and children exhibited similar patterns of
coarticulation magnitude according to stop place of articulation. This was demonstrated
by the absence of significant differences in slopes and intercept patterns between the two
groups, 2) within each Generation group, participants varied in the order of slope
20
amplitude (p ≥ k > t) which indicate that, contrary to a strictly fixed pattern of
coarticulation amplitude in each age group, slight individual differences in slope order
are possible within an age group (cf. Tables 1 and 2).
This study aimed at providing a new type of data to foster understanding of how
children develop articulatory coordination (or synergy) to achieve mature coarticulation
patterns in their native language. Whether young children coarticulate more or less than
their adult peers in the first years of their life has been a controversial subject. We believe
the combination of LE measures on both the acoustic and articulatory productions of
children provide elements allowing us to start addressing the debate. The main result of
this study is indeed that the tongue back is used to advance the tongue tip in 4-5 year old
children as in adults. This conclusion emerges from the results on slopes and intercepts of
locus equations calculated in the articulatory and acoustic domains. Therefore by the age
investigated, this synergy seems to have been learned.
This supports results from the acoustic LE literature, which have all converged
toward a similar order of slopes (e.g., .62 for labial and .52 for alveolars in Gibson &
Ohde (2007) versus .68 and .40 respectively in Sussman et al., 1999 or about .80 and .40
in older children in Sussman, 1992). Moreover this study has gone further in establishing
the articulatory reason for the alveolar/non-alveolar asymmetry, linking the asymmetry to
a synergy, in which the tongue back helps the tip for alveolars.
However, these data do not show that the velar is intermediate in slope between
the labial and alveolar as found in some previous studies (Gibson & Ohde, 2007).
However, note that this difference is not as frequently found as the asymmetry between
21
alveolar and non-alveolar, and indeed several studies have reported data in which velars
have the same slope value as labials (Lindblom, 1963; Sussman, 1991; Sussman et al.,
1999).
Most studies using LE measures to assess coarticulation between C and V have
exclusively relied on acoustic measures on infants or toddler's CV productions. However,
Sussman et al., (1999) provides interesting suggestions about the underlying tongue
behavior responsible for the slope patterns found in acoustics. In that study he followed
the same child from age 7 months to 40 months. In his measurement of alveolar slope, he
found a steep slope for early measures that lowered toward the adult-like slopes by 12
months. He does estimate that the tongue must be fronted for the alveolars, based on the
F2 values. But he conjectured that the early high slopes are due to a lack of
differentiation of the tip and the dorsum, and that the lowering of the slope is due to the
growing ability of the child to control the two parts of the tongue separately. Regarding
the fall in slope, he says “We are suggesting that the child, perhaps by using heard adult
forms as target sounds, has gained the ability to exercise independent motor control over
the tongue body and tongue tip/blade during [dV] productions”. The articulatory data we
have presented shows, rather, that the lowering of the slope indicates how the tongue
back can be used to assist the tip, as observed in adults. That is, differentiation would
lead to the possibility that the tip and tongue back could move independently, whereas
what we suggest is that slope lowering is due to assistance, where the back moves the
front, as if it were not independent of it. Indeed, we believe that the early steep slopes for
/d/, if they are not due to the experimenters’ difficulty of identifying the formants in
young children’s speech, suggest that the tongue body starts to move for the vowel during
22
the formation of the /d/. Such a behavior observed in children but not in adults may
reveal that they have not learned the synergy between the back and the front of the
tongue.
One important observation of Sussman (1999)’s that our study corroborates is
that, even when children have learned an aspect of coarticulation, the resulting pattern my
not be stable. We believe that the non-significant tendency we observe in this study for
children to have a larger difference in slope between /t/ and /p/ than adults, based on
articulatory measures, is in fact due to the immaturity of the pattern. A child’s pattern
may overshoot the adult’s and oscillate, before it stabilizes. This supports Sussman’s
finding from his longitudinal study. A focus of further study is therefore to track
individual development of the synergy longitudinally, as in Sussman’s study, to
determine the trajectory of maturation or lack of maturation in atypical speech. Such
study should advance our understanding of how the development of speech motor control
(together with those of the vocal tract) affects individual articulatory strategies for the
production of distinct acoustic goals.
V. CONCLUSION
The results of the present study investigating children from 4 to 5 years old
suggest that children aged 4 and 5 years old 1) have developed a pattern of coarticulatory
resistance depending on consonant place of articulation; 2) they have a control of the
different functional subparts of the tongue to achieve a proper /t/, and they use the back
part of their tongue to produce these consonants; 3) the lower correlation coefficients
(Table1 and 2) associated with children’s regressions compared to adults’ indicate more
23
variability in coarticulatory patterns (e.g., for alveolar) that can be due to the immaturity
of the speech motor system and organization of the articulatory gestures to produce
distinct goals (e.g. an oral closure in the alveolar region for the alveolar stops). Other
factors of variability such as anatomical growth of the vocal tract that could not be
investigated in this study should also be carefully considered in future investigations.
Overall, the two LE analyses demonstrated that the relation found between C and
V in acoustics is also observed in articulation both in adults and children, aged 4 and 5
years old. Results reaffirm 1) the significance of F2 as a robust indicator of tongue
motion in the front/back dimension, 2) provide further details on coarticulatory resistance
as a possible origin for variation in coarticulation magnitude.
Further studies testing more consonants, voiced and voiceless in various
environments are needed to better understand how children learn to deal with the
articulatory constraints underlying the coarticulation of various Cs and Vs and attune
their control of the functional lingual subparts to the phonological regularities of their
native languages.
ACKNOWLEDGMENTS
This study was supported by FQRSC and SSHRC grants (Quebec’s Government) and
NIH DC-02717. We are grateful to Corinne Toupin for her help in data analysis and to C.
Fowler for valuable comments as well as for the two anonymous manuscript reviewers
REFERENCES
Adler-Bock, M., Bernhardt, B., Gick, B., & Bacsfalvi, P. (2007). “The use of ultrasound
24
in remediation of /r/ in adolescents,” American Journal of Speech-Language
Pathology. 16, 128-139.
Baayen, H. (2007). “Analyzing Linguistic Data: A Practical Introduction to Statistics,”
Cambridge University Press, Cambridge, England, 241-284.
Boersma, P., & Weenink, D. (1996). “Praat, a system for doing phonetics by computer,
version 3.4,” Institute of Phonetic Sciences of the University of Amsterdam, Report,
132, 1-182.
Cheng, H. Y., Murdoch, B. E., Goozée, J. V., & Scott, D. (2007). “Electropalatographic
assessment of tongue-to-palate contact patterns and variability in children, adolescents,
and adults,” J. Speech Hear. Res. 50, 375-392.
Chevrier-Muller C. and Plaza M. (2001). “N-EEL: Les nouvelles epreuves pour l’examen
du langage,” [New tasks for language evaluation], Paris: Editions du Centre de
Psychologie Appliquée.
De Boer, B. (2000). “Self-organization in vowel systems,” J. Phonetics 28(4), 441-465.
Flege, J. (1983). “The influence of stress, position, and utterance length on the pressure
characteristics of English /p/ and /b/,” J. Speech Hear. Res. 26, 111-118.
Fowler, C. (1994). “Invariants, specifiers, cues: An investigation of locus equations as
information for place of articulation,” Perception and Psychophysics, 55, 597–610.
Gibson, T. & Ohde, R. (2007). “F2 locus equations: Phonetic descriptors of coarticulation
in 17- to 22-month-old children,” J. Speech Hear. Res. 50, 97–108.
Goffman, L., Smith, A., Heisler, L., & Ho., M. (2008). “The breadth of coarticulatory
units in children and adults,” J. Speech Hear. Res. 51, 1424-1437.
25
Goldstein, L. (2003). “Emergence of discrete gestures,” In Proceedings of the 15th
International Congress of Phonetic Sciences, Barcelona, Spain, 85-88.
Goodell, E.W., & Studdert-Kennedy, M. (1993). “Acoustic Evidence for the
Development of Gestural Coordination in the Speech of 2-Year-Olds: A Longitudinal
Study,” J. Speech Hear. Res. 36, 707-727.
Green, J. R., Moore, C. A., & Reilly, K. J. (2002). “The sequential development of jaw
and lip control for speech,” J. Speech Hear. Res. 45, 66-79.
Hardcastle, William J., & Nigel Hewlett. Eds (1999). “Coarticulation: Theory, Data and
Techniques”. Cambridge University Press. 1-383.
Iskarous, K., Fowler, C. A., & Whalen, D. H. (2010). “Locus equations are an acoustic
expression of articulator synergy,” J. Acoust. Soc. Am. 128(4), 2021-2032.
Katz, W., & Bharadwaj, S. (2001). “Coarticulation in fricative-vowel syllables produced
by children and adults: a preliminary report,” Journal of Clinical Linguistics and
Phonetics. 15(1/2), 139-144.
Katz, W.F., Kriple, C., & Tallal, P. (1991). “Anticipatory coarticulation in the speech of
adults and young children: acoustic, perceptual, and video data,” J. Speech Hear. Res.
34, 1222-1232.
Kent, R. D. (1983). “The segmental organization of speech,” In Peter MacNeilage (Ed.),
The production of speech, New York: Springer, 57-89.
Kent, R. D. (1976). “Anatomical and neuromuscular maturation of the speech
mechanism: Evidence from acoustic studies,” J. Speech Hear. Res. 19, 421-447.
Kier, W.M., & Smith, K. K. (1985). “Tongues, tentacles, and trunks: the biomechanics of
movement in muscular-hydrostats,” Zool J. Linnean Soc. 83, 307–324.
26
Koenig, L. L., & Lucero, J. C. (2008). “Stop consonant voicing and intraoral pressure
contours in women and children,” J. Acoust. Soc. Am. 124, 3158-3170.
Krull, D. (1987). “Second formant locus patterns as a measure of consonant-vowel
coarticulation,” In Proceedings of Phonetic Experimental Research at the Institute of
Linguistics University of Stockholm. 5, 43–61.
Lee, S., Potamianos, A., & Naryanan, S. (1999). “Acoustics of children’s speech:
Developmental changes of temporal and spectral parameters,” J. Acoust. Soc. Am.
105, 1455-1468.
Li, M., Kambhamettu, C., & Stone, M. (2005). “Automatic contour tracking in ultrasound
images,” Clinical Linguistics and Phonetics. 19(6-7), 545-554.
Lindblom, B. (1963). “Spectrographic study of vowel reduction,” J. Acoust. Soc. Am. 35,
1773–1781.
Lofqvist, A. (1999). “Interarticulator phasing, locus equations, and degree of
coarticulation,” J. Acoust. Soc. Am. 106, 2022–2030.
Macleod, A. A, Sutton, A., Trudeau, N., & Thordardottir, E. (2010). “The acquisition of
consonants in Québécois French: A cross-sectional study of pre-school aged
children,” Int. J. Speech Lang. Pathol. 1-17.
MacLeod, S., J. van Doorn, & V. A. Reed (2001). “Normal Acquisition of Consonant
Clusters,” American Journal of Speech-Language Pathology. 10(2), 99-111.
Manuel, S. Y., & Stevens, K. N. (1995). “Formant transitions: Teasing apart consonant
and vowel contributions,” In Proceedings of the XIII ICPHS, ed. K. Elenius and P.
Branderud KTH and Stockholm University, Stockholm, 436-439.
Ménard, L., Aubin, J., Thibeault, M., and Richard, G. (2012): "Comparing tongue shapes
27
and positions with ultrasound imaging: a validation experiment using an articulatory
model", Folia Phoniatrica et Logopaedica, 64, 64-72.
Ménard L, & Noiray A. (2011). “The development of lingual gestures in speech:
comparing synthesized vocal tracts with natural vowels,” Faits de Langue, 37, 189-
202
Ménard, L., Perrier, P., Savariaux, C., Aubin, J. & Thibeault, M. (2008). “Compensation
strategies for a lip-tube perturbation of French [u]: an acoustic and perceptual study
of 4-year-old children,” J. Acoust. Soc. Am. 124, 1192-1206.
Ménard, L., Schwartz, J. L, Boë, L. J., & Aubin, J. (2007). “Production-perception
relationships during vocal tract growth for French vowels: analysis of real data and
simulations with an articulatory model,” J. Phon. 35(1), 1-19.
Mooshammer, C., Hoole, P. & Kühnert, B. (1995). “On loops,” J. Phonetics 23, 3–21.
Munson, B. (2004). “Variability in /s/ production in children and adults: Evidence from
dynamic measures of spectral mean,” J. Speech Hear. Res. 47, 58–69.
Nearey, T., & Shammass, S. (1987). “Formant transitions are partly distinctive invariant
properties in the identification of voiced stops,” Can. Acous. 15, 17-24.
Nittrouer, S. (1995). “Children learn separate aspects of speech production at different
rates: Evidence from spectral moments,” J. Acoust. Soc. Am. 97, 520–530.
Nittrouer, S. (1993). “The emergence of mature gestural patterns is not uniform:
Evidence from an acoustic study,” J. Speech Hear. Res. 36, 959-971.
Noiray, A., Cathiard, M. A., Ménard, L., & Abry, C. (2011). “Test of the Movement
Expansion Model: Anticipatory vowel lip protrusion and constriction in French and
English speakers,” J. Acoust. Soc. Am. 129 (1), 340-349.
28
Noiray A., Cathiard M. A., Ménard L., & Abry, C. (2009). “Emergence of a vowel
gesture control. Attunement of the anticipatory rounding temporal pattern in French
children,” In Sophie Kern, Frédérique Gayraud and Egidio Marsico (Eds.),
Emergence of language Abilities, Cambridge Scholars Publishing, 100-116.
Noiray, A., Iskarous K., Bolaños, L., and Whalen, D.H. (2008).
(http://issp2008.loria.fr/Proceedings/PDF/issp2008-14.pdf, date last viewed 7/09/12)
“Tongue-jaw synergy in vowel height production: Evidence from American English”
in Proceedings of 8th International Speech Production Seminar, pp. 81-84.
Perkell, J., Cohen, M., Svirsky, M., Matthies, M., Garabieta, I. & Jackson, M. (1992)
“Electro-magnetic midsagittal articulometer (EMMA) systems for transducing speech
articulatory movements, ” J. Acoust. Soc. Am. 92, 3078-3096.
Raudenbush, S. W. & Bryk, A. S. (2002). “Hierarchical Linear Models: Applications
and Data Analysis Methods (2nd Edition)”. Newbury Park, CA: Sage, 491p.
Sadagopan, N. & Smith, A. (2008). “Effects of utterance length and complexity on
speech motor performance: A large-scale developmental study,” J. Speech Hear. Res.
51, 1138-1151.
Sereno, J. A., Baum, S. R., Marean, G. C., & Lieberman, P. (1987). “Acoustic analyses
and perceptual data on anticipatory labial coarticulation in adults and children,” J.
Acoust. Soc. Am. 81, 512–519.
Smith, A. & Zelaznik, H. (2004). “The development of functional synergies for speech
motor coordination in childhood and adolescence,” Developmental Psychobiology.
45, 22-33.
Smith, A., & Goffman, L. (1998). “Stability and patterning of speech movement
29
sequence in children and adults,” J. Speech Hear. Res. 41, 18-30.
Stemberger, J. & B. Bernhardt (2002). “Editorial: Forum on Intervocalic Consonants in
Phonological Development,” Clinical Linguistics & Phonetics. 16, 149-154.
Stone, M., Faber, A., Raphael, L., and Shawker, T. (1992). “Cross-sectional tongue shape
and lingua-palatal contact patterns in [s], [K], and [l],” J. Phon. 20, 253-270.
Subramanian, A., Yairi, E., & Amir, O. (2003). “Second formant transitions in fluent
speech of persistent and recovered preschool children who stutter,” Journal of
Communications Disorders. 36, 59–75.
Sussman, H. M., Duder, C., Dalston, E., & Cacciatore, A. (1999). “An acoustic analysis
of the development of CV coarticulation: a case study,” J. Speech Hear. Res. 42,
1080-1096.
Sussman, H. M., Fruchter, D., Hilbert, J., and Sirosh, J. (1998). “Linear correlates in the
speech signal: The orderly output constraint,” Behav. Brain Sci. 21, 241-299.
Sussman, H. M. Minifie, F. D. Buder, E. H., Stoel-Gammon, C. Smith, J. (1996).
“Consonant-vowel interdependencies in babbling and early words: preliminary
examination of a locus equation approach,” J. Speech Hear. Res. 39(2), 424-33.
Sussman, H. M., Hoemeke, K., & McCaffrey, H. A. (1992). “Locus equations as an index
of coarticulation for place of articulation distinctions in children,” J. Speech Hear.
Res. 35, 769–781.
Sussman, H. M., McCaffrey, H. A., and Mathews, S. A. (1991). “An investigation of
locus equations as a source of relational invariance for stop place categorization,” J.
Acoust. Soc. Am. 90, 1309-1325.
Terband, H., Brenk, F., Lieshout, P., Nijland, L., & Maassen, B. (2009). "Stability and
30
composition of functional synergies for speech movements in children and adults", In
Proceedings of INTERSPEECH-2009, 788-791.
Walsh, B., Smith, A. & Weber-Fox, C. (2006). “Short-term plasticity in children’s speech
motor systems,” Developmental Psychobiology. 48, 660-674.
Westbury, J.R. (1994). “X-ray microbeam speech production database user's handbook,”
Waisman Center, Madison: University of Wisconsin at Madison; 135p.
Zharkova, N., Hewlett, N. & Hardcastle, W.J. (2012). “An ultrasound study of lingual
coarticulation in /sV/ syllables produced by adults and typically developing children,”
Journal of the International Phonetic Association, 42 (2), pp. 193-208.
Zharkova, N., Hewlett, N. & Hardcastle, W.J. (2011). “Coarticulation as an indicator of
speech motor control development in children: an ultrasound study,” Motor Control,
15, 118-140.
Zharkova N., Hewlett N., Hardcatle W.J. (2008). “An ultrasound study of lingual
coarticulation in children and adults,” In Sock Rudolph, Fuchs Susanne, Laprie Yves
(Eds.), in Proceedings of the ISSP, Strasbourg France, 8-12 December, 161-164.
Figures and Tables
Table 1.
ACOUSTIC DATA
/p/
/t/
/k/
Group
Age
P
S
Int.(Hz)
R2
S
Int.(Hz)
R2
S
Int.(Hz)
R2
31
Adults
25
1
.85
230.77
.98
.74
590.25
.8
.98
231.86
.86
2
.93
27.51
.92
.57
744.48
.7
.77
499.17
.88
3
.95
42.35
.99
.84
265.5
.98
1.00
22.75
.98
6
.95
49.85
.99
.68
689.5
.96
.91
217.2
.93
7
.96
57.36
.96
.68
560.12
.83
1.00
92.26
.96
MEAN
.93
81.56
.97
.7
569.97
.85
.93
212.65
.94
Children
5
1
.90
47.35
.92
.53
1279.5
.87
.89
540.56
.84
4
2
.90
225.57
.96
.64
987.88
.86
.88
540.6
.84
4
3
.91
190.18
.91
.52
1070.4
.88
.85
560
.91
4
4
.94
127.08
.93
.73
761.65
.92
.89
502.23
.88
5
5
.94
218.59
.94
.76
766.16
.94
.87
426.6
.94
5
6
.83
332.7
.96
.51
1344.9
.77
.94
291
.96
MEAN
.90
190.25
.94
.62
1035
.87
.89
476.83
.90
Locus data in the acoustic domain for both adults and children. Participant’s age (Age),
participant number (P), Regression slopes (S), Intercepts (Int.) in Hertz, correlation
coefficients (R2) in labial /p/, alveolar /t/ and velar /k/ coarticulatory context. Correlation
coefficients are significant at p< .001.
Table 2.
ARTICULATORY DATA
/p/
/t/
/k/
Group
Age
P
S
Int.(mm)
R2
S
Int.(mm)
R2
S
Int.(mm)
R2
32
Adults
25
1
.72
2.52
.88
.38
5.8
ns.
.7
2.84
.63
2
.9
0.88
.98
.50
4.65
ns.
.92
0.45
.71
3
.9
0.94
.94
.70
2.7
.78
1.00
-0.17
.71
6
.98
0.19
.97
.53
4.38
.32
1.00
-0.74
.83
7
.91
0.80
.86
.60
3.55
.40
.80
1.65
.6
MEAN
.88
1.07
.93
.54
4.25
.5
.88
0.80
.7
Children
5
1
.9
0.86
.81
.45
5.62
.42**
.85
1.49
.84
4
2
.98
0.16
.82
.32
6.85
ns
.95
0.61
.84
4
3
.92
0.61
.87
.68
3.55
.45**
.80
1.91
.91
4
4
.96
0.61
.91
.42
6.23
.ns
.70
2.90
.88
5
5
.84
1.45
.80
.11
8.75
ns
.67
2.95
.94
5
6
.78
2.11
.60
.40
6.45
.3*
.89
1.12
.96
MEAN
.9
.97
.80
.40
6.24
.39
.81
1.83
.90
Locus data in the articulatory domain for both adults and children. Participant’s age
(Age), participant number (P), Regression slopes (S), Intercepts (Int.) in mm, correlation
coefficients (R2) in labial /p/, alveolar/t/ and velar /k/ coarticulatory context. Correlation
coefficients are significant at p< .001.The symbol ** indicates a value significant at p<
.05; * at p< .2. ns values that are not statistically significant.
Figure 1.
Midsagittal tongue contour collected via ultrasound imaging technique. Y-coordinate
represents the highest point on the tongue surface, x-coordinate shows its horizontal
position. The front part of the oral cavity is on the left side of the ultrasound image.
33
Figure 2.
Slopes for LE measure performed in the articulatory domain (panels a and c) and acoustic
domains (panels b and d) for the adults (solid lines) and children (dashed lines) in /p/, /t/,
/k/ contexts. Panels a) and b) show the slopes lines estimated on the data pooled amongst
adults (solid line) or children (dashed line). Panels c) and d) present the slopes derived
from individual data as well as the sd values.
Figure 3.
Intercepts for LE measure performed in the articulatory domain (panels a and c) and
acoustic domains (panels b and d) for the adults (solid lines) and children (dashed lines)
in /p/, /t/, /k/ contexts. Panels a) and b) show the slopes lines estimated on the data pooled
amongst adults (solid line) or children (dashed line). Lower panels c) and d) present the
intercepts derived from individual data.
