Key Words amplitude modulation - central auditory processing disorder - dyslexia - speech envelope
INTRODUCTION
Accurate speech perception relies on efficient processing of temporal cues in the
speech signal ([Specht, 2014 ]). There are two types of temporal cues present in a speech signal: the slowly changing
temporal amplitude envelope of speech, which facilitates accurate perception of syllable
boundaries, and the fine structure formant changes, which are needed for accurate
phoneme discrimination ([Rosen, 1992 ]; [Goswami, 2011 ]; [Goswami et al, 2011 ]). The goal of our research was to develop two tests that assess children’s processing
of each of these temporal cues. The current article describes the development of the
Parsing Syllable Envelopes (ParSE) test that assesses children’s detection of amplitude
modulations. The description of the Phoneme Identification Test (PIT) assessing children’s
categorical perception of formant frequencies is described in [Cameron et al (2018) ].
Identifying smaller units from a continuous acoustic stream is essential in the process
of parsing, or structuring, speech. The incoming acoustic signal contains peaks and
valleys of intensity, which define syllabic boundaries. Although speech recognitions
involve both bottom-up and top-down processes, syllable envelopes provide one of the
most prominent acoustic landmarks in the continuous speech signal. They are therefore
likely to be important for lower level segmental processing ([Stevens, 2002 ]).
The amplitude changes related to the slow modulations of the incoming acoustic signal
are one of the primary cues that young children first attend to when they hear speech
([Nittrouer, 2006 ]). Infants as young as 1 mo old are better able to discriminate syllable-like stimuli
than nonsyllable-like stimuli ([Bertoncini and Mehler, 1981 ]). As infants gain more experience with their native language, they start to discover
the other acoustic properties of the phonetic identity for their language ([Nittrouer, 2006 ]). As long as temporal cues are available within each of several frequency bands,
speech can be perceived with 90% accuracy by adults even when there is no fine spectral
information available within each of these frequency bands ([Shannon et al, 1995 ]). Therefore, having the ability to process temporal cues is vital for a person to
be able to perceive speech accurately ([Poelmans et al, 2011 ]). The amplitude envelope can be analyzed in terms of its constituent temporal modulation
frequencies ([Goswami, 2003 ]). The modulation spectrum of speech in quiet and with low reverberation has a peak
at ∼4–5 Hz, which reflects the rate at which syllables occur ([Houtgast and Steeneken, 1985 ]; [Joris et al, 2004 ]). When the speech envelope is low-pass filtered below 4 Hz, speech becomes unintelligible
due to the loss of syllable boundaries ([Drullman et al, 1994 ]).
In principle, detection of syllable boundaries should be affected by clarity of articulation,
the listening environment, and the psychoacoustic abilities of the listener. Clearly
articulated speech is inherently more understandable ([Picheny et al, 1985 ]). It has longer gaps between words and more intense, fully released obstruent sounds
([Picheny et al, 1986 ]), all of which should make syllable boundaries clearer. Noise and reverberation
each inherently mask lower level sounds more than higher level sounds, which must
make the detection of gaps, and hence syllable boundaries, more difficult. Because
the detection of syllable boundaries is essentially one of gap, modulation, and/or
change detection (albeit with complex, changing signals), boundary detection should
also depend on any relevant individual listener-related psychoacoustic abilities.
These might include rate of decay of forward and backward masking, spectral resolution
(because adjacent syllables will in general have different frequency spectra), the
smallest change in intensity that the listener can detect for sustained signals, and
the time constant with which the listener integrates intensity. For a given talker
and listening environment, differences in any of these psychoacoustic characteristics
could cause individual differences in the ability to segment syllables, and hence
cause individual differences in speech intelligibility ([Gordon-Salant and Fitzgibbons, 1993 ]). Decreased ability to detect amplitude modulations (as assessed with the ParSE)
may therefore indicate that a child is more susceptible to the detrimental effects
of noise, reverberation, or rapidly produced speech.
[Dimitrijevic et al (2016) ] examined envelope-following responses to monaural amplitude-modulated broadband
noise carriers, with varying modulation depths. Participants were younger adults and
older adults with varying pure-tone average hearing levels (19 dB HL versus 35 dB
HL). Significant moderate correlations were found between electrophysiological and
behavioral amplitude modulation detection thresholds across participants. Older adults
had slightly higher amplitude modulation detection thresholds than younger adults,
but the difference was not significant, and thresholds did not correlate with age.
The authors concluded that the behavioral-physiological amplitude modulation depth
threshold relationship was likely too weak to be clinically useful in the population
assessed, who did not suffer from apparent temporal processing deficits.
Unfortunately, however, the ability to process slow-rate temporal auditory cues is
impaired in some populations, such as those who have phonological dyslexia ([Goswami et al, 2002 ]; [Hämäläinen et al, 2009 ]). Children with phonological dyslexia have difficulty reading regular and/or nonwords
(e.g., “cat” or “gop”), which cannot be explained by low intelligence or neurological
damage ([Goswami et al, 2002 ]; [McArthur et al, 2013 ]). An often co-occurring disorder with dyslexia is a central auditory processing
disorder (CAPD) ([King et al, 2003 ]). CAPD refers to a variety of disorders characterized by difficulties in the central
nervous system processing auditory information ([American Speech-Language-Hearing Association [ASHA], 2005 ]). Children with suspected CAPD are highly heterogeneous in nature. Difficulty in
understanding speech when background noise is present is a commonly reported symptom
of CAPD ([Jerger and Musiek, 2000 ]; [Cameron and Dillon, 2014 ]).
Despite 60 yr of research, CAPD is still poorly understood and the definition, diagnosis,
and treatment of the disorder are highly controversial ([Vermiglio, 2014 ]). There is a real need for new tests to be developed that can accurately diagnose
additional specific deficits causing individual children’s listening problems ([Dillon et al, 2012 ]; [Vermiglio, 2016 ]). One potential deficit that [Vermiglio (2016) ] suggests targeting is a test for a temporal resolution disorder. Temporal resolution
can refer to the ability to detect changes in the duration of auditory stimuli, the
presence of brief gaps in a stimulus, the masking of one sound by another that precedes
or follows it, or the detection of rapid variations in intensity. Although there are
a number of paradigms used to characterize temporal resolution, one of the most widely
adopted methods is the gap detection task ([Buss et al, 2014 ]). Gap detection tests in common use clinically include the Gaps-In-Noise test ([Musiek et al, 2005 ]) and the Random Gap Detection Test ([Keith, 2000 ]). Such tasks measure an individual’s ability to detect a silent gap between two
stimuli in milliseconds. The stimuli bounding the silent gap can be spectrally identical
narrow-band markers (within-channel gap detection), identical broadband acoustic markers
(across-channel gap detection), or acoustic markers that differ in frequency, ear
stimulated, or location in free-field space (between-channel gap detection). Decreased
gap detection thresholds are found for across-channel tasks, as the auditory system
can integrate spectral information across very wide frequency ranges to detect the
gap ([Phillips, 1999 ]; [Phillips and Hall, 2000 ]).
The acoustic markers that bound the gap in traditional gap detection tasks are identical
in spectrum and have very short rise and fall times. The temporal task is therefore
simply the detection of discontinuity in the activity aroused in the peripheral auditory
neurons and the perceptual channel supported by that representation ([Phillips and Hall, 2000 ]). [Bellis (2003) ] questions the clinical utility and ecological validity of such traditional gap detection
techniques.
The detection of a temporal gap requires the listener to monitor stimulus intensity
over time ([Buss et al, 2014 ]). The goal of this article was to develop the ParSE test to assess children’s ability
to perceive temporal envelope cues using a controlled, but highly realistic stimulus.
In contrast to traditional gap detection tasks, the ParSE investigates children’s
ability to recognize syllable boundaries using an amplitude modulation detection paradigm.
Perceiving where syllables start and finish based on amplitude modulations at syllable
boundaries is needed for developing accurate speech perception and reading skills
([Stevens, 2002 ]; [Goswami, 2011 ]; [Poelmans et al, 2011 ]). Consequently, we hypothesize that an underlying deficit in temporal resolution
processing may be the cause of some children’s listening and reading difficulties.
This article describes the development of the ParSE test and the analysis of normative
test and retest data. It was hypothesized that children’s ability to identify syllable
boundaries based on changes in amplitude modulation would be poorer than that of adults,
but that it would improve with age.
METHOD
Approval for the study was granted from the Australian Hearing Human Research Ethics
Committee and the New South Wales Department of Education.
Participants
A total of 158 participants were initially assessed. There were 12 adults (aged 23
yr 10 mo to 50 yr 9 mo, mean 32 yr 4 mo) of which nine were females and three were
males. All had normal hearing defined as ≥20 dB HL at all octave frequencies from
250 to 8000 Hz measured bilaterally using an Interacoustics AC40 audiometer (Middelfart,
Denmark) with Telephonics TDH 39P audiometric headphones (Huntington, NY) in H7A Peltor
cups (3M; St. Paul, MN). The child participants were recruited from a Sydney primary
school. Children whose parents reported they had an attention, language, or learning
problem in the study consent form were excluded from participating. Children’s hearing
was tested on the day and only those who passed the pure-tone audiometric screening
test participated in the study. Audiometric testing was as for the adult participants,
with the exception that an Interacoustics Audio Traveler A222 portable audiometer
was used. Data were collected from 146 children. Eight children were excluded posttesting
on the ParSE due to inconsistent performance, as documented in the “Exclusions Based
on Confidence Intervals” section. Further, five outliers (four children and one adult)
were excluded, as documented in the “Calculation of z Scores and Removal of Outliers” section. As such, following exclusions, data were
analyzed from a total of 145 participants. There were 134 children (aged 6 yo 0 mo
to 12 yr 4 mo, mean 9 yr 3 mo), of which 64 were female and 70 were male, as well
as 11 adults (aged 23 yr 10 mo to 50 yr 9 mo, mean 32 yr 10 mo) of which 8 were females
and 3 were males.
Software Development
The ParSE graphical user interface and signal processing application were developed
in the MATLAB programing language ([MathWorks Inc., 2014 ]), and compiled for use on a touchscreen computer. Three screens were developed:
a data capture screen for collection of client information and activation of reference
tone; an operations screen for activation of practice, familiarization, and test materials;
and a test screen for the participant to respond to the ParSE stimuli. An image of
the test screen appears as [Figure 1 ].
Figure 1 ParSE test screen.
Stimuli
A synthesized low central /a/ vowel, 1 sec in duration, was generated with a sampling
rate of 44100 Hz using the source-filter model provided by Praat (version 5.4.04)
([Boersma and Weenink, 2014 ]). The voicing source consisted of a pulse train with fundamental frequency (F0)
of 110 Hz. A filter representing the shape of the vocal tract, consisting of five
formants with steady-state frequencies of 750, 1200, 2350, 3300, and 4000 Hz, was
created based on frequencies used by [Blomert and Mitterer (2004) ]. Formant bandwidths of 50, 60, 110, 160, and 210 Hz were used, based on bandwidths
reported by [Fant (1962) ]. The filter was applied to the voicing source to produce a low, central /a/ vowel.
The resulting vowel has the temporal properties of the source with the spectral properties
of the filter. The middle 750-msec portion was selected as the carrier wave. The carrier
wave was then normalized to a root mean square value of −20 dB (ref. full-scale square
wave, 50-msec window) in Adobe Audition CS6. Using MATLAB ([MathWorks Inc., 2014 ]), the onset and offset of the carrier were then ramped to zero amplitude using a
50-msec linear ramp function (see [Figure 2 ]).
Figure 2 The 750-msec carrier /a/ vowel with 50-msec linear ramped onset and offset. Equivalent
to the 0% modulated token.
A total of 16 different stimuli sets were constructed, including 8 two-syllable sets
and 8 three-syllable sets. Each set consisted of 11 tokens of different modulation
depths. For each token in the set, the modulation depth decreased by 10%, from 100%
modulated to 0% modulated (0% modulation represented a one pseudo-syllable token).
The width of each notch in the token was fixed at 100 msec (50-msec linear down and
up ramps), hence the slope of the linear ramp decreased as the modulation depth decreased.
The 50-msec rise/fall times were selected to be broadly representative of syllable
envelopes in natural speech (e.g., [Howell and Rosen [1983] ] found average rise times of 33 msec for affricates and 76 msec for fricatives).
Given the broadband nature of the carrier, amplitude modulation did not introduce
spectral cues.
The location of each modulation was randomized within a specified range to ensure
that modulation notches did not occur with a high degree of regularity and thus predictability.
For the two-syllable stimuli, the duration of each syllable was always between 350
and 450 msec. For the three-syllable stimuli, the syllables were always between 225
and 300 msec. These durations are within the range of syllable durations observed
in natural language ([Crystal and House, 1990 ]; [Greenberg et al, 2003 ]). Examples of the two and three pseudo-syllable tokens can be seen in [Figure 3 ].
Figure 3 Examples of (A) a two pseudo-syllable token with a 30% modulation depth and (B) three
pseudo-syllable token with a 100% modulation depth.
Task
In an adaptive three-alternative, forced-choice procedure, the listener’s task was
to indicate whether each trial contained one, two, or three pseudo-syllables by pressing
the corresponding numbered button on the touchscreen. The response buttons became
active 300 msec after the offset of the stimulus to ensure that participants did not
respond before hearing the complete stimulus. Participants did not receive any feedback
following responses. A total of 112 tokens were presented in two blocks. The first
block contained 64 randomly ordered presentations consisting of 8 presentations of
the odd continuum steps (0%, 20%, 40%, 60%, 80%, and 100%). At the end of the first
block, the threshold (inflection point) of the psychometric function (PF) was automatically
calculated by the ParSE software. The second block contained 48 randomly ordered presentations
consisting of 4 presentations of each end point token and 8 presentations each of
the 5 tokens clustered around the threshold of the PF. These five tokens included
the tokens nearest to 5% and 95% of the PF and the nearest token above and below the
threshold plus the next nearest token (either above or below the threshold).
Scoring was weighted so that a score of 1 was given if the correct number of syllables
(i.e., 1, 2, or 3) was identified; a score of 0 was given if modulation or the lack
of modulation was not identified (i.e., a response of 1 instead of 2 or 3, or a response
of 2 or 3 instead of 1); and a score of 0.5 was given if modulation was detected but
the incorrect number of syllables was identified (i.e., 2 instead of 3, or 3 instead
of 2).
Practice and familiarization conditions were presented orally by the examiner before
testing, as described in the “Practice and Familiarization Procedure” section. The
practice, familiarization, and test instructions are provided in Appendix. Including
practice and familiarization, the ParSE test took ∼6 minutes to complete for children
aged 8–12 yr, and ∼8 minutes to complete for 6- and 7-yr-olds.
Practice and Familiarization Procedure
Participants completed a brief practice task before the test condition. Twelve practice
stimuli were presented. These comprised three repetitions each of the 100% modulated
two and three pseudo-syllable endpoint tokens and six repetitions of a one pseudo-syllable
(0% modulated) token. Practice stimuli were presented in random order. Following each
response, the participant was provided with visual feedback (the words “correct” or
“incorrect” appearing on the screen).
Following the practice task, participants (children only) performed a brief training
task to ensure familiarity with the ambiguous tokens and to provide an understanding
of how the ambiguous tokens relate to the endpoint tokens. Each continuum step was
presented once in order from 0% modulation to 100% modulation for the two syllable
tokens only. After each token, the participant selected whether they heard one or
two syllables. No feedback was given following responses.
Reporting
PFs were automatically fitted to individual data by the ParSE software using a logistic
curve. A graph of the results was displayed on the computer screen following testing.
[Figure 4 ] provides an image of a typical result for a child on the ParSE (z score 0.0).
The y axis corresponds to the weighted proportion of multisyllabic responses (range 0–1).
Thus, for a specific modulation depth, 0 signifies that no modulation had been detected,
whereas 1 indicates that a participant identified the correct number of modulated
syllables 100% of the time.
The x axis corresponds to the depth of modulation (range 0–100%).
Figure 4 Image of ParSE results screen for a participant 8 yr 2 mo of age with a z score of 0.00. The curved lines represent the PF fitted to the data. The dark-colored
central curved line is the curve fitted to all measured data, and the other two lines
are the curves fitted to odd-numbered responses of the data and even-numbered responses
of the data, respectively. The right-hand dashed line shows the UBUR.
The following performance criteria were calculated by the software and stored in a
spreadsheet.
a) Threshold: The threshold is the point at which a participant detects a two or three
syllable token 50% of the time (weighted for accuracy of number of syllables), and
no modulation 50% of the time.
b) Upper boundary of the uncertainty region (UBUR): The UBUR corresponded to the value
of the PF evaluated at the point at which the asymptotic (midpoint) slope of the function
intersects 100% correct. At this point, the PF crosses the y axis at 0.88. Thus, the UBUR represents the corresponding point on the x axis (modulation depth) at which syllables can be parsed with 88% accuracy (as shown
by the dissection lines in [Figure 4 ].)
c) Confidence interval of the UBUR (CI UBUR): The CI UBUR was obtained using a nonparametric
bootstrapping technique with B = 400 simulations. The observed proportions correct
at each stimulus step (from 0% to 100% modulation) were used directly to generate
the bootstrap simulations. The UBUR is obtained from the threshold and slope of the
best-fitting logistic function for each set of simulated responses and is sorted in
order from highest to lowest. The 95% CI UBUR is determined using the (B*2.5%)th and
(B*97.5%)th values. The width of the CI UBUR is equal to the range of stimulus percentage
values from the corresponding 2.5th percentile UBUR to the 97.5th percentile UBUR.
Note that although the stimuli can have values only between 0% and 100%, both the
UBUR and CI UBUR can exceed 100%.
Procedure
Testing for the adult participants was conducted in a sound-attenuated room at the
National Acoustic Laboratories using a Sony Vaio Duo 11 touchscreen computer (Sony,
Tokyo, Japan). For the child participants, testing was completed in a quiet room at
their primary school. Sound levels in the school testing rooms were measured between
45 and 50 dBA using a Q1362 digital sound level meter (Dick Smith Electronics, Sydney,
Australia). Data were collection using a Microsoft Surface Pro 3 touchscreen computer
(Microsoft, Suzhou, China).
The pregenerated stimuli were presented binaurally, using the ParSE software, through
Sennheiser HD 215 circumaural headphones (Hanover, Germany). All tokens were presented
at a volume control setting calibrated to 77 dB SPL during the steady-state /a/ vowel
using a Brüel & Kjær Head and Torso Simulator Type 4128C (Naerum, Denmark). This level
corresponded to a volume level of 40 on the Sony computer and 18 on the Surface computer.
This level was selected based on the average most comfortable listening level of the
test stimuli chosen by four normal-hearing adult listeners.
Exclusions Based on Confidence Intervals
As noted in the “Participant” section, only children whose parents reported no attention
or learning deficits on the study consent form were assessed with the ParSE. However,
an additional inclusion criterion was implemented posttesting based on the width of
the CI UBUR recorded for the ParSE for each participant. If the CI UBUR was greater
than 200, then the result was rejected as invalid as the reliability of the fitted
PF was very poor. In such cases, the slope of the fitted function was extremely flat
and the measured data were consistent with the responses being essentially random.
Participants were also excluded if their threshold—the modulation depth at which a
participant detects a two or three syllable token 50% of the time—was less than 0.
Based on these criteria, and as noted in the “Participant” section, of the 146 children
assessed, data were excluded for 8 children. The majority of children excluded (63%)
were aged 6 ry 0 mo to 6 yr 12 mo.
RESULTS
Statistical analysis was performed using Statistica version 10 (StatSoft, OK). For
each participant, performance on the ParSE was determined by the UBUR score, which
was converted to a z score, as described in “Calculation of z Scores and Removal of Outliers” section. Effects of age were calculated using raw
scores. All other analyses, including correlations between measures, were calculated
using the z scores. Use of z scores removes the contribution that age makes to correlations between the measures
(because all test scores on average improve with age). Correlation coefficients based
on z score data will therefore be smaller than those based on raw scores. As the data
were skewed toward negative performance, nonparametric analyses were used.
Calculation of z Scores and Removal of Outliers
The UBUR scores, collected from the 150 children and adults remaining following exclusion
based on confidence intervals, were used to create equations that allow the expression
of individual scores in age-corrected population standard deviation (SD) units (z scores).
The raw UR scores were regressed against age using the exponential formula UBUR =
a + b × exp (×age/c ) where a , b , and c are the coefficients that determine the curve. These three coefficients determine
the asymptotic value applicable to adults (a ), the rate of change with age (b ), and the age above which the effect of age starts to diminish (c ). This equation calculated the predicted UBUR score for participants as a function
of age.
Residual scores were calculated as the difference between each participant’s actual
score and the predicted score for participants of that age. The squared residual scores
were regressed against using the formula described earlier, but with new values of
a , b , and c. The square root of this second regression formula was used to predict the
SD of UBUR scores at any age. The coefficients are reported in [Table 1 ].
Table 1
ParSE UBUR (% Modulation Depth), Regression Coefficients Used in the Creation of z Scores
Mean
SD
a1
b1
c1
a2
b2
c2
24.3
72.6
7.76
30.7
1218
5.28
UBUR scores for each participant were then were standardized to a mean of zero and
unity SD and reported as z scores, using the following formula:
The UBUR z scores were examined to determine if they deviated significantly from a normal distribution.
The Shapiro–Wilk W value was 0.91 (p < 0.000001) with z scores ranging from 1.7 to −3.1 (mean −0.0002). Outliers (z scores <2.5 SDs below the mean) were removed from the normative data. At this cutoff
point, it would be expected that approximately one participant would be removed were
the distribution to be normal. However, there were five outliers removed (four children
and one adult), demonstrating a skew toward decreased precision of syllable boundary
perception.
z Scores were recalculated for the remaining 145 participants. The Shapiro–Wilk W value was 0.95 (p = 0.00007), with z scores ranging from 2.3 to −3.0. A histogram is provided as [Figure 5 ].
Figure 5 Histogram of UBUR z scores for the 145 participants on the ParSE.
Gender Effects
The median UBUR was 42% for males (n = 73) and 42% for females (n = 72). The Kruskal–Wallis
H test (one-way analysis of variance on ranks) was used to determine if the threshold
differed significantly between groups. Age was controlled for by comparing z scores. There was no significant difference between males and females [H
(1) = 0.27, p = 0.60].
Age Effects—Thresholds
The mean and median ParSE thresholds, as a function of age, are provided in [Table 2 ]. As there were only a small number of 12-yr-olds with a maximum age of 12 yr 4 mo,
data from 11- and 12-yr-olds were combined. Across age groups, the median threshold
(% modulation depth) was 22%. Because of the skewed distribution of the data, the
Kruskal–Wallis H test was used to determine if threshold differed significantly between age groups.
Overall, there was no significant effect of age on ParSE threshold [H
(6) = 12.465, p = 0.052].
Table 2
Median, Mean, and SDs for ParSE Threshold and UBUR (% Modulation Depth) as a Function
of Age
ParSE
Threshold (%)
UBUR (%)
Age
N
Median
Mean
SD
Median
Mean
SD
Overall
145
22
23
10
42
45
17
6
19
24
24
8
50
55
19
7
18
22
23
11
45
47
17
8
19
25
30
14
54
58
20
9
22
22
22
11
41
42
13
10
25
19
22
8
39
43
15
11–12
31
20
21
9
38
40
12
Adult
11
19
17
5
25
26
8
Age Effects—UBUR
The mean and median ParSE UBUR scores, as a function of age, are provided in [Table 2 ]. Children aged 11 and 12 yr were combined, as noted earlier. Across age groups,
the median UBUR (% modulation depth) was 42%. There was a trend of decreased UBUR
with age. The Kruskal–Wallis H test revealed a significant effect of age on the ParSE UBUR [H
(6) = 35.7, p < 0.00001]. Children aged 6–10 yr had significantly higher uncertainty region boundaries
than adults (see [Figure 6 ]). A scatterplot of individual raw ParSE UBUR data for the 145 children and adults
as a function of age is provided as [Figure 7 ]. The ±2 SD limits, calculated from the regression equations, are delineated.
Figure 6 Box and whisker plot of UBUR as a function of age for the 145 participants on the
ParSE. The filled square represents the median UBUR, the open boxes represent the
25–75% boundaries, and the whiskers represent the minimum and maximum scores.
Figure 7 Scatterplot of the individual raw UBUR scores at for the 145 children and adults
tested on the ParSE. The solid line indicates the mean score as a function of age,
the dashed line shows the ±2 SD limits.
Test–Retest Reliability
Test–retest reliability data were analyzed for 106 children on the ParSE. Results
for an additional 13 children were rejected before statistical analysis due to invalid
confidence intervals. Participants were retested between 16 and 44 days (mean 35 days)
after their initial appointment. The median UBUR z scores were 0.11 SD at test and 0.56 SD at retest (mean −0.01 and 0.41 SD, respectively).
A repeated measures (Wilcoxon matched pairs) test revealed a significant difference
between test and retest (T = 1173, p < 0.0000001). Spearman rank order correlations revealed a strong correlation (r = 0.68, p < 0.0000001) between test and retest UBUR z scores ([Figure 8 ]). The mean difference between retest and test z scores was 0.42 SD.
Figure 8 Scatterplot of the UBUR scores at test and retest for the 106 children retested on
the ParSE. The dashed line represents the least squares regression line.
The Pearson’s product moment UBUR z score test–retest correlation was r = 0.64 (p < 0.00001). Consequently, the proportion of variance accounted for by measurement
error in the test scores is estimated as 36% (equal to 1 − r ).
DISCUSSION
The present article documents the development of a new test of temporal resolution
intended for future clinical use for children with suspected CAPD and/or reading deficits.
The ParSE test uses an adaptive amplitude modulation detection paradigm to evaluate
an individual’s ability to analyze the low-rate peaks and valleys in the envelope
of an incoming acoustic signal that provide cues to syllable boundaries. Stimuli were
randomized one, two, and three pseudo-syllable tokens—the multisyllabic tokens containing
notch modulation depths ranging from 10% to 90%. Performance was determined by the
individual’s UBUR, being the modulation depth at which multisyllabic tokens could
be detected with 88% accuracy.
The threshold for syllable detection—that is, the modulation depth where an individual
indicated that they had heard a two or three syllable token half the time—was 22%.
There was no significant difference between age groups on the threshold for syllable
detection. There was, however, a significant age effect on the UBUR, with younger
children (6–10 yr) needing a greater modulation depth to consistently detect a multisyllabic
syllable. From [Table 2 ] and [Figure 6 ], it can be seen that there is both a gradual variation of performance with age,
and also year-to-year random fluctuations. The method we used to calculate z scores involved regressing the test scores against age as a continuous variable.
This method preserves the gradual variation while smoothing out the random fluctuations.
It is clear that the distribution of normative data test scores deviated from a normal
distribution, with a skew toward below-average performance. That is, the worst performers
were further below the mean performance than the better performers were above the
mean. It is not possible to say whether this reflects an intrinsic skew in the range
of abilities of syllable boundary perception, or is a product of the particular measure
used to describe performance—in this case, the amplitude modulation depth needed for
the children to detect the modulation 88% of the time. Although we could have performed
a mathematical transformation that normalizes the distribution, we have chosen not
to as there is no a priori reason why the ability to perceive syllable boundaries
should be normally distributed. Importantly, any such transforms do not change the
rank order of children’s performance on the task, so the only effect of such a transform
is to change the z score at which one considers performance is sufficiently different from the mean
to represent a problem in real-life perception of speech.
Repeated measures analysis revealed a small but significant average improvement on
retest by 0.4 SDs. Correlational analysis revealed a strong relationship (r = 0.68) between test and retest z scores. Test–retest correlations were used to determine the impact of random measurement
error. The proportion of measurement error in the ParSE test, estimated as 36% of
the variance in test scores, is considerably less than was found for the PIT ([Cameron et al, 2018 ]). Possibly, there are bigger true differences in children’s ability to identify
syllable boundaries than differences in their ability to perceive categorically. Alternatively,
perhaps the task of counting syllables maintains a more constant level of attention,
or attracts a more constant criterion of the boundary between the different sounds,
than the task of identifying which phoneme was heard.
Inspection of individual data before group analysis revealed the greatest variation
in performance in the youngest children. Confidence intervals were calculated on the
slope parameter used to calculate the UBUR and results were deemed invalid for individuals,
whose confidence intervals were substantially wider than those of the remaining children.
Of the children excluded, the majority were in the 6-yr age group. In clinical trials
currently in progress, participation has been restricted to children aged 7 yr 6 mo
to 11 yr 6 mo. If it is found during these studies that the ParSE has clinical validity
in older children, we may reexamine performance on 6- to 7.5-yr-olds using a verbal
or pictorial response method whereby the child indicates to the audiologist the number
of syllables perceived and the audiologist inputs the data. Younger children may be
more motivated to attend to the test stimuli, and less distracted by outside influences,
if an authority figure is more actively involved in the test administration.
Finally, even having accounted for any invalid results by excluding children based
on UBUR confidence interval parameters, before analysis of the normative data, five
outliers were removed from the ParSE results. Based on the sample size, the number
of outliers would be predicted to be around one. Thus the number of children who exhibited
decreased precision of syllable boundary detection outside the normal range was unexpected,
and may be an indication of a distinct clinical population. Alternatively, they may
be children who for some reason were not able to properly follow the test directions,
despite provision of practice and familiarization procedures, or who otherwise did
not respond in a way that reflected their true temporal resolution. It will require
future research to resolve whether poor performance on the test always has adverse
consequences for communication in challenging situations or whether poor performance
can sometimes have no real-life consequences. The same statement can be made about
many, if not all, clinical tests used to assess auditory processing disorders.
The present study forms the first steps toward developing a new and innovative test
of temporal resolution ability suitable for clinical use. Studies are currently being
undertaken in children who present with phonological dyslexia. Patterns of performance
across a range of standardized assessment tasks and measures of cortical auditory
evoked potentials will be documented. These results will be correlated with ParSE
results, as well as results on the PIT ([Cameron et al, 2018 ]), which was developed for the study to investigate the ability of children to process
rapid formant transitions. It would be desirable for future research to also examine
the relationship between performance on this modulation-based measure of temporal
resolution and performance on traditional rapid-onset/offset gap-detection tasks.
It is anticipated that the results from the research presented here, as well as the
clinical studies currently in progress, will contribute to the expansion of diagnostic
targets for auditory processing assessment noted by [Vermiglio (2016 ]), and shed light on any auditory-specific contributions to other diagnostic entities
such as phonological dyslexia.
APPENDIX
Instructions given to participants prior to undertaking the ParSE practice, familiarization,
and test conditions.
Practice
“When you are ready to start press the play button. You will hear a voice saying either
‘ah,’ ‘ah ah,’ or ‘ah ah ah.’ Press the button on the screen that matches how many
‘ah’s you hear each time. If you are not sure how many ‘ah’s you hear please guess.”
Familiarization
“Now you will hear the voice again but the sound will start as one syllable and change
to two. So at first it will sound like ‘ah’ and then it will start to sound more like
‘ah ah.’ Press the button that matches what you hear. If you are not sure, just guess.
There are no ‘ah ah ahs’ this time.”
Test
“Now you will hear the same type of sounds but some sounds may be less clear. If you
are not sure whether you heard ‘ah,’ ‘ah ah,’ or ‘ah ah ah’ please choose the sound
you think it was more likely to be. Halfway through you can have a break. Try to press
the button as soon as you hear the sound but it is more important to be accurate than
fast.”
Abbreviations
CAPD:
central auditory processing disorder
CI UBUR:
confidence interval of the upper boundary of the uncertainty region
ParSE:
Parsing Syllable Envelopes Test
PF:
psychometric function
PIT:
Phoneme Identification Test
TMTF:
temporal modulation transfer function
UBUR:
upper boundary of the uncertainty region
SD:
standard deviation
