J Am Acad Audiol 20:629–643 (2009)
Auditory Steady-State Responses and Speech Feature
Discrimination in Infants
DOI: 10.3766/jaaa.20.10.5
Barbara Cone*
Angela Garinis{
Abstract
Purpose: The aim of this study was to determine whether there was a correlation between auditory
steady-state responses (ASSRs) for complex toneburst stimuli and speech feature discrimination
(SFD) abilities in young infants.
Study Sample: Seventeen infants (mean age 5 9.4 months) and 21 adults (mean age 5 27 years)
with normal hearing had ASSR and SFD tests.
Data Collection: The ASSR test employed an eight-component complex toneburst stimulus; threshold
and input–output functions were determined as level was systematically varied. The SFD test utilized
an observer-based, visual-reinforcement test procedure to determine the infant’s ability to detect the
speech feature change from /ba/ to /da/.
Results: The correlation of the group mean /ba/–/da/ discrimination performance (percent correct) with
the group mean ASSR score (percent responses present) ranged from r 5 0.64 for the 1500 Hz
amplitude-modulated and frequency-modulated tone burst to 0.99 for ASSRs for all stimulus
components; however, correlations between ASSRs and SFD scores for individual subjects were
modest.
Conclusion: The ASSR and SFD results appear to reflect the audibility of the stimuli.
Key Words: Auditory evoked potentials, infants, speech audiometry
Abbreviations: ABR 5 auditory brain stem response; AM 5 amplitude modulation; ASSR 5 auditory
steady-state response; CF 5 carrier frequency; DPOAE 5 distortion product otoacoustic emission;
FFR 5 frequency following response; FM 5 frequency modulation; IAFM 5 independently amplitude
and frequency modulated; MF 5 modulation frequency; RECD 5 real-ear to coupler difference;
SFD 5 speech feature discrimination; SNR 5 signal-to-noise ratio; VRA 5 visual reinforcement
audiometry
T
he ability to discriminate speech sounds relies,
in part, upon the temporal- and frequencyresolving power of the auditory system for
elements of complex acoustic events. Speech signals
are composed of amplitude and frequency variations
occurring over time in different frequency bands.
Prosodic and speech segmentation information is
conveyed by slow (1–50 Hz) amplitude modulation.
Many aspects of the fine structure of speech, such as
consonant–vowel transitions, voiced–voiceless cues
(voice-onset time), and the fundamental frequency of
the voice are linked to very rapid (50–500 Hz)
amplitude and frequency modulations (AMs and
FMs). In the ongoing speech signal, there are multiple
AM components and FM components that must be
processed concurrently. The discrimination of speech
features, such as manner, place, voicing, and vowel
type, requires that the ear and brain simultaneously
*Speech, Language and Hearing Sciences, University of Arizona, Tucson; {Speech and Hearing Sciences, University of Washington, Seattle
Barbara Cone, PhD, Speech, Language and Hearing Sciences, University of Arizona, P.O. Box 210071, Tucson, AZ 85721; Phone: 520-626-3710;
Fax: 520-621-9901; E-mail: conewess@u.arizona.edu
Portions of this work were presented at the International Evoked Response Audiometry Study Group, June 2005, Havana, and at the American
Auditory Society, March 2006, Scottsdale.
This work was supported by grants from the National Organization for Hearing Research, the National Institutes of Health–National Institute for
Deafness and Other Communication Disorders (R55 DC 05646-01A1) and Intelligent Hearing Systems.
629
Journal of the American Academy of Audiology/Volume 20, Number 10, 2009
process low- and high-frequency spectral components,
AMs, and FMs. Thus, one mechanism of speech feature
discrimination is the ability to encode rapidly changing
temporal (modulation) and spectral (frequency) cues
simultaneously across a range of frequencies. Evidence
from psychophysical studies of adults suggests that
sensitivity for AM and FM is needed for speech
perception (Zeng et al, 2005).
AUDITORY STEADY-STATE RESPONSES
A
uditory steady-state responses (ASSRs) from the
brain stem can be obtained for amplitude modulation or frequency modulation or combined amplitude
and frequency modulation of tones or noise (John et al,
2001). ASSRs obtained for AM tones as carrier frequency is varied provide estimates of pure-tone thresholds
(for review, see Picton et al, 2003). ASSRs can be
obtained for simultaneous independently amplitudeand frequency-modulated (IAFM) tones (Dimitrijevic et
al, 2001; Dimitrijevic et al, 2004), offering insight into
how the auditory system encodes simultaneously the
temporal and spectral complexities of speech.
The interactions between amplitude and frequency
modulation (of tones) have been evaluated in adults,
using ASSRs. John and colleagues (2001) and Dimitrijevic and colleagues (2001) show that when a tone is
both amplitude and frequency modulated, the response
amplitude is smaller than that found for the AM
response or FM response alone. The phase of the
response, however, appears to be unchanged by the
presence of another modulated component. Also, the
response to FM appeared to be affected by either an
additional AM or an additional FM component, while
the AM response was affected only by another FM
component. These effects offer some insight into how
the cochlea and brain stem pathways process AM and
FM. Because the decrement in the response to these
IAFM tones was small, relative to the AM alone and
FM alone condition, it has been suggested that the two
types of modulation are processed independently (John
et al, 2001). This ‘‘independence’’ may have been
largely due to the separation of the cochlear place of
excitation patterns for AM versus FM. It is also
possible to combine AM and FM to obtain a response
that is larger than the response to AM alone. This is
referred to as a mixed-modulation stimulus, and the
‘‘improvement’’ in the response amplitude is dependent
upon the phase of the FM stimulus component relative
to the AM component.
ASSRS AND WORD RECOGNITION
D
imitrijevic and colleagues (2001) demonstrate a
correlation between the ASSRs for IAFM tones
and word-recognition scores in normal-hearing adults.
630
First, they measured word recognition as a function of
stimulus level and developed a ‘‘performance-intensity’’ function: the percent correct word recognition for
each level tested. Then, they determined the number of
ASSR components present for an eight-tone complex
that consisted of four carrier frequencies (CFs), each of
which was amplitude and frequency modulated at
different (independent) rates. The ASSR ‘‘score,’’ that
is, the percentage of components present (out of a
possible eight), for each stimulus level was calculated.
They then compared the word-recognition score as a
function of level to the ASSR score as a function of
level. The percentage of ASSR components detected at
each stimulus level had a significant correlation with
the word-recognition score at a similar level. Dimitrijevic and colleagues (2004) also tested the correlation
between ASSRs and word recognition in adults with
hearing loss. They obtained ASSRs in response to
IAFM stimuli with CFs, levels, and modulation depths
designed to have similar characteristics to speech.
Word-recognition abilities were determined using
standardized monosyllabic word lists. There were
strong correlations between the word-recognition score
and the number of ASSR components present for
young and elderly adults with normal hearing and also
for elderly adults with hearing loss. These investigators suggest that ASSRs for multiple IAFM tones
correlate with word-recognition scores because both
speech and multiple IAFM tones contain information
that varies rapidly in intensity and frequency. The
ASSR ‘‘score’’ (i.e., the number of response components
present for the eight-component IAFM stimulus) was
used as an estimate of how much acoustic information
was available to the listener. The more information
available in the speech frequency range, for which the
auditory system can process rapidly changing intensity
and frequency cues, the better the word-recognition
capabilities should be.
SPEECH FEATURE DISCRIMINATION
IN INFANTS
I
nfants have excellent speech-perception skills
(Kuhl, 1992, 2004; Jusczyk et al, 1998; Werker et
al, 1981; Werker and Tees, 1999). They are able to
discriminate speech features, such as a voiced–voiceless contrast (/ta/ vs /da/), place of articulation (/ba/ vs
/da/), or good versus poor exemplars of vowel sounds.
Experience listening to speech in the native language
during the first year of life appears to exert a strong
effect on speech perception and phoneme-discrimination
abilities (Werker and Tees, 1984; Kuhl, 2004). Eilers et
al (1977) used a visually reinforced operant conditioning
method to evaluate the development of speech feature
discrimination (SFD) in infants under the age of one
year. Yet the methods used for investigating the
ASSR and SFD in Infants/Cone and Garinis
development of infant speech perception have had little
carryover to clinical applications. There are no standardized methods for testing infant speech perception or
discrimination procedures that can be used in infants
who have receptive language less than 2.5 years.
Because decisions about amplification, cochlear implantation, and language-learning methods may be based
upon speech-perception abilities, there is a considerable
need to develop clinical assessment methods suitable for
infants and toddlers.
Some methods used to study speech-perception
abilities in young infants involve recording nonnutritive sucking behavior in response to sound or rely on a
habituation/preferential looking protocol (Werker et al,
1998; Houston et al, 2003). These methods, while
useful in the research lab, may not have easy carryover
into clinical settings. One method, based upon a visual
reinforcement paradigm, has been successfully used in
research establishing speech-discrimination abilities
in normal-hearing infants (Eilers et al, 1977; Trehub,
1979; Nozza, 1987; Nozza et al, 1991a; Nozza et al,
1991b; Werker et al, 1998; Kuhl, 2004). In this
procedure, the infant/toddler is reinforced for detecting
when the speech token changes, after having been
exposed to a train of identical speech tokens. The
technique used—observation of stimulus-contingent
behavior, with subsequent reinforcement of ‘‘correct’’
responses—is similar to visual reinforcement audiometry, a readily available clinical protocol (Widen et al,
2000). For instance, the consonant–vowel token /ba/
can be presented repeatedly, and the change token
interspersed in this /ba/ train may be /da/. The infant is
reinforced for detecting (indicated with a head turn or
other stimulus-contingent behavior) when the token
changes. The response to the stimulus change indicates that the infant has detected a change in the
speech feature, in this example, a place-of-articulation
cue for a voiced plosive. Using this technique, Nozza
(1987) showed that infants with normal hearing
required higher levels than adults to reach their
maximum performance on an SFD task. These differences in level were 27 dB for a /ba/–/da/ contrast, 25 dB
for a /ba/–/ga/ contrast (Nozza et al, 1991a; Nozza et al,
1991b), and as much as 30 dB for a /da/–/ga/ contrast.
These ‘‘performance-intensity’’ functions, that is, the
accuracy of speech sound discrimination plotted as a
function of level, provide evidence of a developmental
process in speech-perception abilities for infants that
cannot be accounted for on the basis of a simple
elevation in hearing threshold.
RATIONALE
W
e aimed to address the issue of whether and how
ASSRs are related to infant speech-discrimination abilities. This knowledge could then be applied to
the problem of documenting the effect of intervention
strategy (i.e., amplification) on speech-perception
abilities for infants with hearing loss. The study was
also motivated by the need to develop (clinical)
methods for assessing speech-discrimination/perception abilities in young infants (Eisenberg et al, 2005).
ASSRs for mixed-modulation tones that model
speech characteristics are correlated with speechdiscrimination abilities in adults with normal or
impaired hearing (Dimitrijevic et al, 2001; Dimitrijevic
et al, 2004). Thus, we aimed to test whether the
relationship between ASSRs and speech-discrimination abilities that Dimitrijevic and colleagues (Dimitrijevic et al, 2001; Dimitrijevic et al, 2004) found
would hold true for infants. The hypothesis was that
infant ASSRs for a tone complex composed of multiple
mixed-modulation stimuli (approximating the acoustic
complexities of speech) would be correlated with infant
speech feature discrimination abilities. The theoretical
foundation of this work is based in the realm of speech
acoustics and feature detection (Pickett, 1999) as well
as in methods to predict speech intelligibility of speech
sounds (French and Steinberg, 1947) such as the
Articulation Index and the Speech Transmission Index
(Humes et al, 1986). As the acoustic features of speech
sounds can be analyzed with respect to their frequency
components, it is possible to estimate speech-recognition or -discrimination abilities on the basis of
audibility for those frequency components. The multiple mixed-modulation stimuli used in this experiment
were acoustically similar to speech sounds in terms of
mean formant frequencies for consonant–vowel transitions and amount of amplitude and frequency
modulation found for individual consonants in the
long-term speech spectrum (Dimitrijevic et al, 2004).
ASSRs were obtained in response to complex
toneburst stimuli and the number of response components present, and their amplitudes were measured as
a function of level. Similarly, speech (feature)-discrimination abilities were measured as a function of level,
using an observer-based psychophysical method in
conjunction with visual reinforcement techniques.
METHOD
Subjects
Seventeen infants aged 4–13 months (mean age 5
9.4 months) and 21 adults aged 22–43 years (mean age
5 27 years) with normal hearing participated in both
an ASSR and speech feature discrimination test. Nine
of the 21 adults participated in a second ASSR test,
described below.
To demonstrate normal hearing, adult subjects had a
pure-tone threshold test prior to participation in the
study, and thresholds were # 20 dB HL at 500–8000 Hz.
631
Journal of the American Academy of Audiology/Volume 20, Number 10, 2009
For infant subjects, distortion product otoacoustic
emission (DPOAE) and click-evoked auditory brain stem
response (ABR) tests were used to establish normal
hearing. To be considered normal, DPOAEs were
present at 3 dB over the noise floor in the 1000–
4000 Hz frequency range, and ABRs were present with
normal latencies for click stimuli presented at 30 dB
nHL.
ASSR Test Stimuli
A complex toneburst stimulus, consisting of four
carrier tones each presented at two different modulation rates, with amplitude modulation alone or combined with frequency modulation, was used for the
ASSR test. Stimuli were generated using the Intelligent Hearing Systems software package for the SmartEP-ASSR system.
The carrier frequencies were 500, 1500, 2500, and
4000 Hz. The modulation frequencies were 78 and
85 Hz for the 500 Hz carrier, 83 and 90 Hz for the
1500 Hz carrier, 88 and 95 Hz for 2500 Hz, and 93 and
100 Hz for 4000 Hz. At 78, 83, 88, and 93 Hz (i.e., the
lower modulation frequency for each CF pair) the tones
were amplitude modulated with a cosine ramp. This
created an AM stimulus. At 85, 90, 95, and 100 Hz (the
higher modulation frequency for each CF pair) the
tones were frequency modulated at 35, 30, 20, and 35
percent, respectively, with a phase of 90 degrees, and
also shaped with a cosine ramp, creating an AM+FM
stimulus. The CFs and modulation rates and depths
were chosen to approximate the stimuli used by
Dimitrijevic and colleagues (2004); however, a cosine
ramp was used for all carriers in the present study,
resulting in the creation of a complex, eight-component
toneburst stimulus. This complex toneburst stimulus
was presented monaurally through insert earphones at
levels of 20 to 80 dB SPL in 10 dB steps. A fourcomponent toneburst stimulus was also created,
composed of the AM+FM components described above,
that is, 500, 1500, 2500, and 4000 Hz tone bursts with
additional frequency modulation presented at rates of
85, 90, 95, and 100 Hz, respectively. The four-component stimulus was used to test a small sample (N 5 9)
of adults for comparison to the larger group tested with
the eight-component stimulus. Stimulus calibration
was carried out using a 2 cc coupler with an H1a
adapter in conjunction with a Larson Davis Model 824
sound pressure level meter.
Speech Feature Discrimination Test Stimuli
Consonant–vowel tokens /ba/ and /da/ were used for
the SFD test. The /ba/ versus /da/ contrast has been
used in past research employing psychophysical or
electrophysiologic measures and so was chosen for this
632
investigation. The /ba/ versus /da/ contrast consists of a
place-of-articulation cue (bilabial vs lingual-alveolar
voiced plosive) signaled by the change in second format
frequency in the transition from the consonant to the
vowel. A /ba/ versus /sa/ contrast was used for training.
This contrast consists of changes in place, manner
(plosive vs sibilant), and voicing (voiced vs unvoiced).
This contrast was used in order to provide multiple
cues for speech feature change.
Digital recordings were made of a female speaker
saying the syllables /ba/, /da/, and /sa/. Using Praat
Speech Analysis Software, exemplar tokens were
truncated at 150 msec following the consonant onset,
using a 20 msec tapering function. Each token was
then copied four times, with a 500 msec interval
between the offset of one token and the onset of the
next. The test stimuli were, therefore, 2100 msec in
duration, with four 150 msec speech samples presented
at a rate of 2 Hz. These digitized samples were then
presented through the Intelligent Hearing Systems
software package for visually reinforced infant speech
discrimination tests. Stimuli were presented through a
loudspeaker and were calibrated using a sound
pressure level meter placed in the position of the test
subject. Stimulus presentation levels during the SFD
test were 30–70 dB SPL. While it has been shown that
infants need higher sensation levels than adults for
optimum performance (Nozza, 1987; Nozza et al,
1991b), the aim of the present study was to establish
speech feature–discrimination scores as a function of
level, using levels that ranged from close to threshold
to those typical of conversational speech. These scores
as a function of level were then compared to the ASSR
test results obtained at the same presentation levels.
ASSR Test Method
ASSRs were obtained using the Intelligent Hearing
Systems Smart ASSR System. Infants were tested
during natural sleep, and adults were tested while
resting in a reclining position with their eyes closed.
Many of the adults slept during the test; however,
sleep stage was not formally monitored. EEG electrodes were placed at the vertex and at the ipsilateral
mastoid as the noninverting and inverting electrodes,
respectively, with the contralateral mastoid serving as
ground. The EEG was sampled over a 1024 msec
epoch, with a gain of 80 dB, filtered at 30–300 Hz, and
averaged for spectral analysis. A response at a
particular modulation frequency was deemed to be
statistically significant using an F-statistic to evaluate
the signal (power at the MF)-to-noise (power in
spectral bands adjacent to the MF) ratio. Responses
were significant when p , .05 was achieved. Averaging
was continued until the averaged background noise
level in five frequency bins above and below the
ASSR and SFD in Infants/Cone and Garinis
modulation frequency reached a level of 0.011 mV or
until all components reached significance, whichever
occurred first. The percentage of significant responses
(out of eight possible) was calculated as a function of
stimulus level.
SFD Test Method
Infants were tested using visually reinforced operant
procedures with observer-based psychophysics. Infants
were brought into the test booth while the ‘‘control’’ or
background stimulus was presented continuously, that
is, the four-token sample, such as /ba/-/ba/-/ba/-/ba/, at
70 dB SPL. A test assistant in the booth manipulated
toys to keep the infant quiet and alert. The parent and
test assistant listened to music presented through
earphones to keep them masked as to the nature of the
stimuli being presented. Infants and observers were
first trained to detect a change in the speech stimulus.
The training consisted of pairing the presentation of
the contrast sound, /sa/ (repeated four times), with the
visual reinforcer: an animated toy obscured in a
smoked Plexiglas box. When activated as a reinforcer,
the box was illuminated and the toy was animated. The
pairing of the reinforcer with the change trial was used
to teach the infant to emit a behavior used during the
testing phase. The behavior could be a head turn, an
eye movement toward the reinforcer, a change in facial
expression, or cessation of or increase in body movement. Training consisted of five pairings of the change
trial with the reinforcer. Background and change trials
were presented at the same level, 70 dB SPL. (Between
change trials, the background control stimulus [/ba//ba/-/ba/-/ba/] was being played continuously.) After
five pairings, a probe trial was used in which the
stimulus change was presented but the reinforcer was
withheld until the infant emitted a response. The
response window was 5 sec in duration. If the
observer detected the infant’s response, the reinforcer
was introduced. The observer had to correctly detect
the presence of stimulus change based upon the
infant’s behavior for two consecutive probe trials
before testing was begun. Infants who did not emit
behaviors that could be correctly detected by the
observer during the probe trials underwent further
training trials. Up to five training trials were
undertaken for each session. If the observer could
not meet criterion for that infant after multiple
training trials, the infant was dismissed for the day.
The infants were allowed to participate in up to three
training sessions. If observer criteria were not met
after the third training session, the infant was
dismissed from further participation. When observer
criteria were met, infants were allowed to participate
in as many as five sessions in order to obtain data for
as many levels as possible. Data from 17 infants are
reported; an additional 12 infants were tested who did
not meet the observer’s training criteria.
After training, a test block was initiated using the
/ba/ versus /da/ contrast. Test blocks were initiated at a
level of 60 dB SPL. The next level tested was 40 dB
SPL. Trials at 50 and 30 dB SPL were undertaken
after data were obtained at 60 and 40 dB SPL. An
attempt was made to test at two levels per stimulus
contrast. The test block consisted of 10 trials, with at
least 40 percent of the trials being controls. The
percentage of correct responses for both change and
control trials was determined.
During testing, the observer controlled when a trial
was initiated but was masked as to the nature of these
trials, that is, whether they contained a speech feature
contrast (change trial) or no stimulus change (control
trial). The order of trials was automatically randomized by the computer software. The observer voted on
each trial by pressing a response button. The observer
had to determine whether the trial contained a speech
feature change based solely on observation of the
infant’s behavior. When the observer made a correct
detection, the infant was given the visual reinforcer.
When the observer made a correct rejection (no
response on a control trial) there was no reinforcer.
Likewise, there was no reinforcer for a miss (failure to
detect a change trial) or a false alarm (voting that a
change occurred when there was a control trial).
Adults were tested using similar procedures. Adult
subjects were told to ‘‘raise your hand when you hear
the sound that makes the toy light up.’’ Only three
pairings of the stimulus change with the reinforcer
were used for adult subjects, and only one probe was
used prior to initiating testing. After training, stimulus
trials were conducted at 50 dB SPL and at 10 dB
decrements until the score for change trials was ,40
percent. Adults were tested with both contrasts.
RESULTS
ASSR Tests
Effects of Level, CF, and Modulation Envelope
The percent of ASSRs present for the eight stimulus
components was calculated as a function of level and
collapsed across CF and MF. ‘‘Percent present’’ was
calculated as ([NS 3 NP] 4 [NS 3 8]) 3 100, where NS
5 number of subjects tested at a given level and NP 5
number of ASSR components present. This was
calculated for data from infant and adult groups. For
the data shown in Figure 1, NS decreased with level
because testing was terminated if there were no ASSRs
present for any stimulus component. There were some
remarkable similarities between the percentages of
significant ASSRs as a function of stimulus level for
633
Journal of the American Academy of Audiology/Volume 20, Number 10, 2009
Figure 1. Percent present auditory steady-state response (ASSR) components plotted as a function of stimulus level for adult and
infant groups. ‘‘Percent present’’ was first calculated as ([NS 3 NP] 4 [NS 3 8]) 3 100, where NS 5 number of subjects tested at a given
level and NP 5 number of ASSR components present. Error bars indicate the standard error of the mean. A second calculation of percent
present (filled and open circles for adults and infants, respectively) uses a fixed NS in the denominator: 17 for the infant group and 21 for
the adult group.
infants and adults. Fewer than five out of eight
response components were present, on average, even
at the highest level tested. The difference in adult
versus infant percent present scores never exceeded 9
percent and was observed only at 80 and 50 dB SPL.
When a fixed NS in the denominator was used (17 for
infants and 21 for adults), in order to take into account
differences in threshold across CF and subject group,
there were larger differences in the percent present
scores for adults and infants. These differences were 14
and 15 percent at 50 and 60 dB SPL, respectively.
Otherwise, the differences in percent present scores
were, on average, 6 percent. There was considerable
variability in both infant and adult groups, and the
adult–infant differences in percent of responses present were not statistically significant.
Effect of CF and Envelope
The percentage of responses present was calculated
for each stimulus component. The formula used was
(NP/NS) 3 100, where NS is the number of subjects
tested with that stimulus and NP is the number of
ASSRs present. The percent of responses present is
shown for each stimulus component as a function of
level for adults (Figure 2A–B) and infants (Figure 2C–
D), with stimulus component as the parameter. For
both adults and infants, the response to the 1500 Hz
AM+FM tone burst at a MF of 90 Hz was most often
present. In general, more responses were present for
the AM+FM tone bursts than for the tone bursts
containing only AM. The only significant differences
634
between the distributions of responses present for
adults and infants were at 500 Hz, for which infants
were less likely to have responses (x2 5 8.091, p 5
.004). A chi-square test detected a trend for infants to
have more responses present for 2500 Hz tone bursts
than did adults (x2 5 3.497, p 5 .06).
Interaction of AM+FM and AM
The finding that there were differences in the
number of responses present as a function of CF and
modulation envelope raised the question of how the
presence of multiple stimulus components affected the
response. The effects of the presence of the AM
stimulus components upon the AM+FM responses
were evaluated in two ways. First, the difference in
the percentage of responses present was calculated for
the AM+FM stimulus minus the AM stimulus. The
differences were largest for the 500 and 1500 CFs and
averaged 35 percent (collapsed across level), while at
2500 and 4000 Hz the differences averaged only 9
percent. Second, nine adults were additionally tested
using a four-CF tone complex consisting of the AM+FM
components at 500, 1500, 2500, and 4000 Hz. The
results for the four-CF/MF toneburst complex were
compared to the previous results obtained with the
eight-component stimulus. The percent of responses
present at each CF for the four- minus eight-component stimulus was calculated. Overall, the presence of
the AM component produced a 23 percent decrement in
the percent present score. The largest effect was seen
at 2500 Hz, for which a 38 percent difference was
ASSR and SFD in Infants/Cone and Garinis
Figure 2. Infant and adult auditory steady-state response percent present for each carrier frequency/modulation frequency (CF/MF)
component. The percentage of responses present was calculated for each stimulus as (NP/NS) 3 100, where NS is the number of subjects
tested with that stimulus and NP is the number of ASSRs present. The stimulus component type (CF/MF) is the parameter: filled
symbols indicate the amplitude modulation and frequency modulation (AM+FM) stimulus components, and open symbols indicate the
AM stimulus components. The components have been grouped by type, with A and C showing adult and infant responses for AM+FM,
respectively, and B and D showing adult and infant responses for AM. The number of subjects tested at each level is indicated in
parenthesis on the x-axis.
observed, and the smallest effect was seen at 500 Hz,
with only a 7 percent difference.
ASSR Threshold and Amplitude
Threshold as a function of CF was estimated at the
50 percent point on the functions for the AM+FM
stimuli (Figure 2A–B), shown in Table 1. In general,
thresholds were somewhat elevated with respect to
previously published data, using either single or
simultaneous presentation of CFs (Cone-Wesson et al,
2002; Luts et al, 2004). Thresholds could not be
estimated at 2500 Hz in adults because the percentage
of significant responses remained under 50 percent at
every level tested. The threshold elevations for individual CF/MF components are easily explained by the
fact that the CFs used in the present study were
selected to represent speech, and so one spectral
component could very well mask another. This effect
is evident in the threshold differences estimated for the
four-component stimulus compared to the thresholds
for the same components presented in the context of an
eight-component stimulus (Table 1). Infant thresholds
were 20 dB higher than those of adults at 500 Hz.
Response amplitudes for each carrier frequency were
measured for the AM+FM tone burst when it was used in
combination with seven other carriers and in combination with three other carriers. The values shown are
those for responses that met the a priori statistical
criterion for ‘‘response present.’’ These amplitudes are
plotted in Figure 3 (adult subjects only). At 500 and
1500 Hz the amplitudes were larger in response to the
four-component stimulus compared to the eight-component stimulus, but this was statistically significant only
635
Journal of the American Academy of Audiology/Volume 20, Number 10, 2009
Table 1. Threshold Estimates for Auditory Steady-State Response Percent Present as a Function of Stimulus Level
Carrier Frequency/Modulation Frequency (Hz)
Subject Group
Infant
Adult
Adult (four-component stimulus)
500/85
1500/90
2500/95
4000/100
65 dB SPL
45 dB SPL
35 dB SPL
28 dB SPL
35 dB SPL
28 dB SPL
68 dB SPL
CNE
28 dB SPL
65 dB SPL
60 dB SPL
45 dB SPL
Note: The percentage of subjects with responses present for each carrier frequency/modulation frequency (CF/MF) component was calculated as
a function of stimulus level (see Figure 2A–B). The 50% point on each CF/MF function was used to calculate threshold. CNE 5 could not evaluate.
for the 500 Hz carrier (F[1, 119] 5 6.24, p 5 .01). The
amplitudes appeared smaller for the four-component
compared to the eight-component stimulus condition
for 2500 and 4000 Hz carriers, but this was not
significant.
Analyses of variance were used to determine if
response amplitudes, side-bin noise amplitudes, and
signal-to-noise ratios (SNRs) varied with respect to
stimulus type, that is, the four-CF AM+FM stimulus
versus the same components embedded in an eightcomponent stimulus. The means and standard deviations for these response parameters are presented in
Table 2. The differences in side-bin noise and SNR
were significant (p , .001), with the responses to the
four-component stimulus having larger SNRs, owing to
significantly lower noise levels in this condition.
Adult versus infant differences for response amplitude, side-bin noise, and SNR (amplitude/sideband
noise ratios) were evaluated for ASSRs in response to
the AM+FM signal components. These are summarized
in Table 2. The values shown are those for responses
that met the a priori statistical criterion for ‘‘response
present,’’ collapsed across stimulus level. Analyses of
variance were completed to determine if there were
differences attributable to age group. Those values
marked with an asterisk were significant at the p , .05
level.
Response amplitudes obtained from adults were
significantly larger than those obtained from infants
at 500 and 1500 Hz (Figure 4). Although adults had a
greater percentage of responses present for 500 Hz,
infants and adults had the same percentage of
Figure 3. Auditory steady-state response amplitude in adults in nanovolts plotted as a function of stimulus level for carrier frequencies
at 500, 1500, 2500, and 4000 Hz, presented at modulation frequencies of 85, 90, 95, and 100 Hz, respectively. Filled symbols indicate
response amplitude when the component was presented as part of the eight-component mixed-modulation stimulus. Open symbols
indicate the response amplitude when the component was present as part of the four-component mixed-modulation stimulus. Bars
indicate standard errors.
636
Note: The amplitude of the auditory steady-state response (ASSR), background noise, and the ASSR/noise amplitude ratios are shown for infants and adults. The asterisks indicate values for
which infant and adult differences are statistically significant (p , .05). The number of carrier frequencies (four or eight) is indicated in parentheses.
4.19 (1.3)
3.24 (1.2)*
4.91 (2.4)*
3.42 (1.49)*
5.24 (3.6)*
3.85 (1.29)
2.64 (0.54)
Amplitude/Noise (SD)
Adult (4)
Adult (8)
Infant (8)
9.43 (6.3)
5.41 (2.8)
4.38 (1.8)
3.12 (1.04)
3.09 (1.0)
7.74 (4.5)
5.18 (3.06)
4.77 (1.9)
2.81 (.80)
3.31 (1.4)
4.15 (1.7)
3.8 (3.4)
4.82 (3.7)
0.008 (0.004)
0.010 (0.003)*
0.007 (0.003)*
0.011 (0.004)*
0.007 (0.002)*
0.011 (0.003)
0.011 (0.005)
Noise
Amplitude, mV (SD)
Adult (4)
Adult (8)
Infant
0.009 (0.004)
0.011 (0.015)*
0.007 (0.007)*
0.012 (0.005)*
0.007 (0.003)*
0.008 (0.004)
0.011 (0.003)*
0.007 (0.003)*
0.010 (0.004)*
0.007 (0.003)*
0.008 (0.004)
0.011 (0.015)
0.007 (0.003)
0.033 (0.033)
0.034 (0.016)
0.031 (0.016)
0.036 (0.020)
0.038 (0.027)
0.039 (0.013)
0.027 (0.011)
0.074 (0.048)
0.056 (0.030)*
0.029 (0.009)*
0.035 (0.014)*
0.021 (0.006)*
0.059 (0.025)
0.053 (0.039)*
0.033 (0.012)*
0.029 (0.012)*
0.022 (0.010)*
0.030 (0.012)
0.034 (0.024)
0.033 (0.027)
4000/100
4000/93
2500/95
2500/88
1500/90
1500/83
500/85
500/78
ASSR Parameters
Response
Amplitude, mV (SD)
Adult (4)
Adult (8)
Infant (8)
Carrier Frequency/Modulation Frequency (Hz)
Table 2. Auditory Steady-State Response Amplitude, Noise, and Amplitude-to-Noise Ratios
ASSR and SFD in Infants/Cone and Garinis
responses present at 1500 Hz. Infant and adult ASSR
thresholds were not significantly different, so the
differences in ASSR amplitude cannot be accounted
for on the basis of threshold differences. It is
interesting to note that response amplitudes decrease
with increasing CF for adults but are unchanging for
infants. These observations were confirmed statistically using analyses of variance as a function of frequency
for the adult (F[3, 282] 5 7.67, p , .001) and infant
(F[3, 156] 5 0.43, p 5 .07) groups. Infants had
significantly lower noise in the averaged EEG than
did adults. Response-to-noise ratios were larger for
infants at 4000 Hz only.
Speech Feature Discrimination Tests
The percentages of contrasts detected by the infant
subjects are plotted as a function of level in Figure 5.
Each gray circle represents the trial score for an
individual infant. The percentage of contrasts correctly
detected decreased with stimulus level. The mean
score was 74 percent at 60 dB SPL, 63 percent at 50 dB
SPL, and 57 percent at 40 dB SPL. Performance at
levels below 40 dB SPL showed less than 40 percent
correct detection of contrasts. These performance
differences as a function of level were significant (F
5 4.18, p , .01). The mean score for correct rejections
during control trials was 87 percent. Total correct
scores (correct contrast + correct control trials) are
shown for individuals in Figure 5.
The data were examined to determine if there were
differences in performance as a function of age of the
infant. Although infant age at the time of testing
ranged from 4.7 months to 13.9 months, the mean age
at time of test was 9.4 months. There was no apparent
effect of age on performance, with the correlation
between percent correct detection of contrasts and age
(in months) at r 5 0.04, when tests at all levels were
considered. When considering the effect of age on
performance at 60 dB SPL only, 30 percent of infants
aged four–six months had 100 percent correct detection, while this increased to 44.4 percent of infants
aged seven–nine months and 56.2 percent of infants
aged 10 months or older. While these appeared to be
steady gains in performance with age, they were not
statistically significant.
SFD performance as a function of level is shown for 12
infants for whom data were obtained for at least two
levels (Figure 5). Testing at two or more levels was
possible for the infants who needed fewer training trials
and who had high performance levels early in the test
session. There was considerable variability in performance, especially at levels below 60 dB SPL. Performance increased with level, in a monotonic fashion. For
these subjects, the mean score (% correct contrast trials
detected) was 90.7 percent at 60 dB SPL.
637
Journal of the American Academy of Audiology/Volume 20, Number 10, 2009
Figure 4. Auditory steady-state response amplitude in infants and adults in nanovolts plotted as a function of stimulus level for carrier
frequencies at 500, 1500, 2500, and 4000 Hz, presented at modulation frequencies of 85, 90, 95, and 100 Hz, respectively. Filled symbols
indicate infant response amplitudes, and open symbols indicate adult response amplitudes. Bars indicate standard error.
Adults were tested using the same methods as
infants, as described above. Mean percent correct
discrimination scores (correct detection of the contrast
stimulus) were 8.3, 63.1, 91.8, and 99.0 percent for
stimulus levels of 0, 10, 20, and 30 dB SPL, respectively. Comparison of ASSR and SFD
Comparisons were made between the infant ASSR
and SFD data. For SFD, the percent correct contrasts
detected as a function of level for all infants and for
the select group tested at two or more levels were
examined. The mean detection scores as a function of
level for ‘‘all’’ infants and for the ‘‘select’’ group tested
at two more levels are presented in Figure 6. Also
shown in Figure 6 are the infant ASSR ‘‘percent
present’’ data for all eight stimulus components,
collapsed across CF and MF. The ASSR percent
present scores (collapsed across CF and MF) were
calculated for the four AM+FM tone bursts at 500,
1500, 2500, and 4000 Hz at modulation rates of 85,
90, 95, and 100 Hz, respectively (Figure 6). Finally,
as the responses to the 1500 Hz AM+FM tone burst
at a MF of 90 Hz were most often present in infant
subjects, these are highlighted in Figure 6. Speech
feature discrimination performance and the percentage of ASSRs present are monotonically increasing
functions.
638
Correlation of the mean scores for SFD and ASSR
are shown in Table 3. The correlations of the /ba/–/da/
discrimination data with ASSR data range from r 5
0.64 for the 1500 Hz component (most often present) to
0.99 for all ASSR components. It is obvious that
increasing level is important for obtaining ASSRs to
one or more stimulus components when they are
presented in the context of seven other tone bursts.
Similarly, the infants’ ability to discriminate the placeof-articulation /ba/–/da/ speech feature change is
related to stimulus level. Owing to the variability in
infants, correlation between ASSR and SFD scores for
individuals was low (r 5 .50) but statistically significant (F 5 9.276, p , .01).
DISCUSSION
T
he hypothesis that ASSRs for complex toneburst
stimuli would be correlated to infant speech
feature discrimination performance was partly supported by the results. In previous work, Dimitrijevic
and colleagues (2004) have shown that the results of
ASSR tests using mixed-modulation stimuli designed
to approximate speech characteristics were strongly
correlated with conventional word-recognition scores
in adults with normal hearing and hearing loss. This
ASSR and SFD in Infants/Cone and Garinis
Figure 5. Speech feature discrimination in infants, with individual scores for the /ba/ vs /da/ contrasts shown for infant subjects as a
function of stimulus level. ‘‘Correct contrast’’ is the percent correct for stimulus change trials. ‘‘Total correct’’ is the percent correct for
both stimulus change trials and for (correct) no responses for the no-change trials. The group means are shown for all subjects (mean
correct contrast) and for a select group of 12 subjects with tests completed for at least two stimulus levels (select correct contrast). The
errors bars indicate standard error of the mean.
study aimed to test this concept in young infants, in an
age group for which there are no standardized methods
for evaluating SFD.
The novel ASSR test stimuli were composed of two
different toneburst (modulation) envelopes used for
each of four CFs, and ASSRs for each modulated
component were measured as all CFs were presented
simultaneously. This was a complex stimulus, with
frequency components that were chosen to approximate speech characteristics, particularly the formant
frequencies, and amplitude and frequency modulation
depths and rates that would be present in consonant–
vowel transitions (Dimitrijevic et al, 2004).The number
of ASSRs present for each stimulus component (as a
function of level) was used as a gross estimate of the
amount (and frequency range) of acoustic information
available to the listener, an approach based on
principles similar to the Articulation Index (Humes,
1991).
We did not replicate all of the stimulus parameters
used by Dimitrijevic and colleagues (2004). Their
stimuli were composed of four carrier frequencies that
were independently amplitude or frequency modulated. The stimuli used in the present study were trains of
tone bursts, using tone bursts at the same frequencies
and modulation rates as Dimitrijevic and colleagues.
The linear and brief onset ramps of the toneburst
stimuli used in this study produced larger amplitude
responses than those evoked by IAFM tones (Dimitrijevic et al, 2001).
The overall percentage of ASSRs present as a
function of level was similar for adults and infants.
The mean age of the infants tested was 8.4 months, an
age when cochlear tuning and eighth nerve response
characteristics are thought to be mature (Werner and
Gray, 1998). This maturation may have contributed to
the similarities. There were, however, some significant differences as a function of carrier frequency.
Infants had fewer responses present at 500 Hz than
did adults. The maturation of perceptual thresholds
for 500 Hz stimuli is known to be delayed until later
childhood (Werner and Marean, 1996), likely due to
differences in middle ear function. It is also the case
that tone burst–evoked ABR thresholds for 500 Hz are
elevated in infants and young children, compared to
adults (Stapells, 2000). There was also a trend for
infants to have a greater number of responses than
adults at 2500 Hz. All ASSR tests were conducted
with insert phones, and perhaps the differences in ear
canal sound pressure level in the infant ear, compared
to the adult ear, contributed to this disparity. In
newborns, there may be as much as a 15–20 dB ‘‘gain’’
639
Journal of the American Academy of Audiology/Volume 20, Number 10, 2009
Figure 6. Infant group mean data for three auditory steady-state response (ASSR) and two speech feature discrimination (SFD) scores
shown as a function of stimulus level. The three ASSR scores are (1) % present for all eight carrier frequency/modulation frequency (CF/
MF) components; (2) % present calculated for the four CF/MF components that had amplitude modulation and frequency modulation
(AM+FM); and (3) % present for the 1500 Hz CF/90 Hz MF component. The two SFD scores are (1) % correct contrast discrimination for
infant subjects and (2) % correct contrast discrimination for 12 infant subjects for whom data were obtained for at least two stimulus
levels. Error bars indicate standard error of the mean.
in SPL; however, ABR thresholds for clicks or a
4000 Hz tone burst are elevated with respect to
adults, when the ear canal SPL differences are
considered (Sininger et al, 1997). Rance and Tomlin
(2006) measured ASSR thresholds for 500 and
4000 Hz AM+FM tones together with ear canal sound
pressure levels. They show that by six weeks of age
real-ear versus coupler differences (RECDs) in SPL
were only 1.5 dB at 500 Hz but were 11 dB at
4000 Hz. Yet, even when accounting for the RECDs,
the neonatal ASSR thresholds were elevated with
respect to adults. In the current data set, it may be
that stimulus levels were effectively higher in the
infants compared to adults, owing to differences in ear
Table 3. Correlation of Mean Values for Auditory Steady-State Response (ASSR) and Speech Feature
Discrimination (SFD)
ASSR
Test Conditions
SFD
All (Eight
Components)
AM+FM (Four
Components)
CF 1500
Hz/MF 90 Hz
1.00
0.99
0.86
1.00
0.89
1.00
0.99
0.99
0.99
0.97
0.64
0.53
/ba/–/da/
All
/ba/–/da/
Select
1.00
0.98
1.00
ASSR
All
AM+FM (Hi-MF)
CF 1500 Hz/MF 90 Hz
SFD
/ba/–/da/ All
/ba/–/da/ Select
Note: The mean values for three ASSR responses and two SFD scores were calculated as a function of stimulus level. The three ASSR scores
are (1) % present for all eight CF/MF components; (2) % present calculated for the four CF/MF components that had AM+FM (Hi-MF); (3) %
present for the CF 1500 Hz/MF 90 Hz component. The two SFD scores were (1) % correct contrast discrimination for all subjects and (2)
percent correct contrast discrimination for 12 select subjects for whom data were obtained for at least two stimulus levels. CF 5 carrier
frequency; MF 5 modulation frequency; AM+FM 5 amplitude modulation and frequency modulation.
640
ASSR and SFD in Infants/Cone and Garinis
canal and middle ear properties, thus leading to
‘‘better’’ responses for infants at 2500 Hz.
Thresholds estimated from the data relating number
of responses present to stimulus level (Figure 2) were
elevated with respect to previously published results
(Cone-Wesson et al, 2002; Rance et al, 2005). This is to
be expected, because the stimulus used was a complex
toneburst stimulus with four CFs and eight MFs.
These may be ‘‘masked’’ threshold estimates, as there
could be considerable interaction of one CF/MF
component upon the other (John et al, 2002). Herdman
and colleagues (2002) show that AM+FM tones produce
a place-specific cochlear response; however, the place
specificity is just over an octave band wide. Thus,
cochlear activation areas produced by the AM+FM
carriers at 1500, 2500, and 4000 Hz would be expected
to overlap and affect the responses to one another,
because they were spaced less than one octave apart.
These CFs were used because they approximate the
formant frequencies for some vowels. In previous
studies of ASSR threshold, single CFs were used or
the CFs were spaced at octave intervals so as to avoid
interaction. In contrast, we were particularly interested in how the ear and brain responded when presented
with a complex stimulus with (spectrally) overlapping
stimuli, as this could give a clue into how speech
sounds are encoded.
Studies employing behavioral methods similar to
those used in this study have established that infants
under the age of one year are able to discriminate
place-of-articulation cues. Studies of particular relevance to the current results are those completed by
Nozza and associates (Nozza and Wilson, 1984; Nozza,
1987, Nozza et al, 1991a; Nozza et al, 1991b). They
show that infants with normal hearing required higher
levels than adults to reach their maximum performance on a speech feature discrimination task. We do
not know if we reached the levels needed for maximum
performance in the current study because we did not
test above 70 dB SPL. Compared to the performance of
adults, who achieved close to 100 percent discrimination scores at 30 dB SPL, it is evident that the infants
need at least 25–30 dB greater levels to perform this
discrimination of the place of articulation speech
feature. The adult SFD performance scores suggest
that the task required only detection of an acoustic
difference between tokens, not discrimination or
recognition. This is evident from the performance/level
functions, which were very steep, with greater than 50
percent performance at 10 dB SPL and 99 percent at
30 dB SPL.
The correspondence between the (infant) group
mean SFD performance intensity function (% correct
3 level) and the group mean ASSR performance
intensity function (% components present 3 level)
was high (r . 0.95), although the correlation for
individuals was modest (r 5 0.50). There was considerable variability in both the ASSR and SFD scores.
The variability in ASSR, particularly among normalhearing infants, has recently been demonstrated by
Rance and Tomlin (2006). They found variability in
ASSR threshold, even when using single tones, rather
than simultaneous presentation of modulated tones.
ASSR is known to be of smaller amplitude than its
ABR cousin and so, more easily obscured by myogenic
and electrophysiologic noise. Using a stimulus paradigm with spectral overlap among the components, as
in the present study, would increase the likelihood of
masking and further variability. Another source of
variability between the ASSR and SFD data was the
difference in stimulus transducer used for each test.
Insert earphones were used for ASSR, with one ear
tested, whereas soundfield testing was used for the
SFD test, and so both ears could contribute to the
response. There were likely calibration differences
between the two test conditions, especially considering
the RECDs for the insert phone condition, as discussed
previously. Admittedly, presentation of stimuli via
insert phone for the SFD test would preferable. Earspecific testing with insert phones is the clinical
standard for visual reinforcement audiometry (Widen
et al, 2000). One application for SFD tests given in the
sound field, as was done in this study, would be to test
the performance of an infant with hearing loss as
amplification or cochlear implant parameters were
adjusted.
The method used for testing infant speech feature
discrimination incorporated familiar elements of visual
reinforcement audiometry (VRA), including a number
of training trials during which the contrast stimulus
(speech feature change) was paired with a visual
reinforcer. Unlike conventional VRA, the observertester did not know whether the trial contained a
contrast or control stimulus. The percentage of correct
detections is therefore based upon observer’s response,
which in turn is contingent upon the infants emitting a
behavior (head turn, eye widening, vocalization) that
indicates that they have detected the stimulus change.
This observer-based technique (Werner, 1995) controls
for observer bias, unlike conventional VRA. When
observer-based techniques are used to study infant
psychophysical abilities, a lengthy training period is
the norm for both observer and infant (Werner, 1995),
so that the observer can learn to detect the infant’s
responses. The training period used in this study may
not have been long enough; or perhaps the training
criterion was not of sufficient stringency to ensure
optimal results. Another possibility may be that
infants need training for each stimulus used in the
discrimination task. The infants were trained to detect
a /ba/ versus /sa/ speech feature change but were tested
on the stimulus contrast /ba/ versus /da/. Although
641
Journal of the American Academy of Audiology/Volume 20, Number 10, 2009
infants were required to meet a performance criterion
for the /ba/–/sa/ contrast, this may not have generalized to /ba/ versus /da/.
Eisenberg and colleagues (Eisenberg et al, 2004,
2005, 2007) have reported their experiences with
infant speech feature discrimination tests, and they
are similar to the results in the present study. They
used consonant voicing, manner and place contrasts,
and a visually reinforced operant procedure to test
normally hearing infants aged 7–11 months. Confidence level scores, their measure of test performance,
ranged between 65 and 77 percent for these consonant contrasts. This is the same level of test
performance achieved by infants in the present study,
when tested using stimuli presented at levels of 50–
70 dB SPL.
Overall, infant SFD performance increased with
level, and variability decreased, an expected stimulus-response relationship. Certainly, there will always
be a strong relationship between audibility and speech
perception performance, as shown in adults with
hearing loss (Yellin et al, 1989; Jerger et al, 1991),
and various Articulation Index methods for estimating
performance (Humes et al, 1986; Humes, 1991). It may
be that the ASSR and SFD results of this study simply
reflect the audibility of the stimuli, as both improve
with increased SPL. And yet, the perceptual task was
not merely to detect the presence of the stimulus,
which would obviously improve with level, but, rather,
to discriminate the difference between two stimuli
based primarily upon a formant transition. There was
a reasonably high correlation between the ASSR for
the 1500 Hz AM+FM tone and SFD results. The
acoustic cue for place of articulation is signaled by
the change in the second formant transition, and for
these sounds, the formant transition is in the range of
1500 Hz. It is not clear why the response to the
1500 Hz AM+FM tone was remarkably robust as a
function of level, even in the presence of all the other
‘‘competing’’ stimulus components.
There are other electrophysiologic methods that can
be used to indicate brain stem mechanisms underlying
speech perception. The frequency following response
(FFR) is an evoked response that reflects the activity of
phase-locking neurons of the auditory brain stem
(Moushegian et al, 1973). When vowel sounds are used
to evoke FFR, phase locking to the fundamental
frequency of the vowel sound and its formants, at least
up to 1500 Hz or so, can be detected (Krishnan, 2002).
Reduced amplitudes or prolonged latencies of FFR in
response to vowel tokens are evident in children with
language-based learning disorders (King et al, 2002;
Banai et al, 2005). The FFR is also influenced by
language experience (Krishnan et al, 2005), which is
further evidence that phase locking to vowel formants
is a speech-encoding mechanism. There are, however,
642
no studies that have directly linked vowel perception or
discrimination to FFR presence, amplitude, or latency.
This is the first report of infant ASSRs evoked by a
complex toneburst stimulus designed to represent the
rapid amplitude and spectral modulations found for
certain consonant–vowel transitions. This is the first
report of infant speech feature–discrimination abilities
as a function of level, using a methodology that could
be developed for routine clinical use. This is the first
report of the correlation between infant ASSR and SFD
abilities. Although the correlations for individual
subjects were modest, the grand mean results are
encouraging. Next steps include refining the training
for the speech feature discrimination test; evaluating
the effects of carrier frequency, modulation type (AM
or FM), and modulation depth on infant ASSR
responses; and using high-pass masking noise during
the ASSR and SFD tests to evaluate how stimulus
degradation (simulating hearing loss) affects infant
performance.
Acknowledgments. We gratefully acknowledge the expert advice of Dr. Lynne Werner with regard to observerbased psychophysical procedures and that of Dr. Andrew
Dimitrijevic with regard to auditory steady-state responses.
Dr. Dimitrijevic also provided a helpful critique on a previous
draft of this article.
REFERENCES
Banai K, Nicol T, Zecker S, Kraus N. (2005) Brainstem timing:
implications for cortical processing and literacy. J Neurosci 25:
9850–9857.
Cone-Wesson B, Rickards FW, Poulis C, et al. (2002) The
auditory steady-state response: clinical observations and applications in infants and children. J Am Acad Audiol 13(5):270–
272.
Dimitrijevic A, John MS, Van Roon P, Picton TW. (2001) Human
auditory steady-state responses to tones independently modulated in both frequency and amplitude. Ear Hear 22:100–111.
Dimitrijevic A, John MS, Picton TW. (2004) Auditory steady-state
responses and word recognition scores in normal-hearing and
hearing-impaired adults. Ear Hear 25(1):68–84.
Eilers RE, Wilson WR, Moore JM. (1977) Developmental changes
in speech discrimination in infants. J Speech Hear Res 20:766–
779.
Eisenberg LS, Martinez AS, Boothroyd A. (2004) Perception of
phonetic contrasts in infants: development of the VRASPAC. In:
Miyamoto RT, ed. Cochlear Implants. International Congress
Series 1273. Amsterdam: Elsevier, 364–367.
Eisenberg LS, Johnson KC, Martinez AS. (2005) Clinical
assessment of speech perception for infants and toddlers. Audiol
Online. www.audiologyonline.com.
Eisenberg LS, Martinez AS, Boothroyd A. (2007) Assessing
auditory capabilities in young children. Int J Pediatr Otorhinolaryngol 71:1339–1350.
French NR, Steinberg JC. (1947) Factors governing the intelligibility of speech sounds. J Acoust Soc Am 19:90–119.
ASSR and SFD in Infants/Cone and Garinis
Herdman AT, Picton TW, Stapells DR. (2002) Place specificity of
multiple auditory steady-state responses. J Acoust Soc Am
112(4):1569–1582.
Houston DM, Ying EA, Pisoni DB, Kirk KI. (2003) Development
of pre-word-learning skills in infants with cochlear implants.
Volta Rev 103(4):303–326.
Humes LE. (1991) Understanding the speech-understanding
problems of the hearing impaired. J Am Acad Audiol 2(2):59–69.
Humes LE, Dirks D, Bell TS, et al. (1986) Application of the
Articulation Index and the Speech Transmission Index to the
recognition of speech by normal-hearing and hearing-impaired
listeners. J Speech Hear Res 29:447–462.
Jerger J, Jerger S, Pirozzolo F. (1991) Correlational analysis of
speech audiometric scores, hearing loss, age, and cognitive
abilities in the elderly. Ear Hear 12(2):103–109.
John MS, Dimitrijevic A, Van Roon P, Picton TW. (2001) Multiple
auditory steady-state responses to AM and FM stimuli. Audiol
Neurootol 6:12–27.
John MS, Purcell DW, Dimitrijevic A, Picton TW. (2002)
Advantages and caveats when recording steady-state responses
to multiple simultaneous stimuli. J Am Acad Audiol 13:246–259.
Jusczyk PW, Houston D, Goodman M. (1998) Speech perception
during the first year. In: Slater A, ed. Perceptual Development:
Visual Auditory and Speech Perception in Infancy. East Sussex:
Psychology Press, 357–388.
King C, Warrier CM, Hayes E, Kraus N. (2002) Deficits in
auditory brainstem encoding of speech sounds in children with
learning problems. Neurosci Lett 319:111–115.
Krishnan A. (2002) Human frequency-following responses: representation of steady-state synthetic vowels. Hear Res 166(1–2):192–201.
Nozza RJ, Wilson WR. (1984) Masked and unmasked pure-tone
thresholds of infants and adults: development of auditory
frequency selectivity and sensitivity. J Speech Hear Res 27(4):
613–622.
Pickett JM. (1999) The Acoustics of Speech Communication:
Fundamentals, Speech Perception Theory and Technology. Needham Heights, MA: Allyn and Bacon.
Picton TW, John MS, Dimitrijevic A, Purcell D. (2003) Human
auditory steady-state responses. Int J Audiol 42(4):177–219.
Rance G, Roper R, Symons L, et al. (2005) Hearing threshold
estimation in infants using auditory steady-state responses. J Am
Acad Audiol 16(5):291–300.
Rance G, Tomlin D. (2006) Maturation of auditory steady-state
responses in normal babies. Ear Hear 27(1):20–29.
Sininger YS, Abdala C, Cone-Wesson B. (1997) Auditory
threshold sensitivity of the human neonate as measured by the
auditory brainstem response. Hear Res 104:27–38.
Stapells DR. (2000) Threshold estimation by the tone-evoked
auditory brainstem response: a literature meta-analysis. J
Speech Lang Pathol Audiol 24:74–83.
Trehub SE. (1979) Reflections on the development of speech
perception. Can J Psychol 33(4):368–381.
Werker JF, Gilbert JH, Humphrey K, Tees RC. (1981) Developmental aspects of cross-language speech perception. Child Dev
52(1):349–355.
Werker JF, Pegg JE, Polka L, Patterson M. (1998) Three methods
for testing infant speech perception. In: Slater A, ed. Perceptual
Development: Visual Auditory and Speech Perception in Infancy.
East Sussex: Psychology Press, 389–420.
Krishnan A, Xu Y, Gandour J, Cariani P. (2005) Encoding of pitch
in the human brainstem is sensitive to language experience. Cogn
Brain Res 25(1):161–168.
Werker JF, Tees RC. (1984) Developmental changes across
childhood in the perception of non-native speech sounds. Can J
Psychol 37(2):278–286.
Kuhl PK. (1992) Psychoacoustics and speech perception: internal
standards, perceptual anchors and prototypes. In: Werner LA,
Rubel E, eds. Developmental Psychoacoustics. Washington, DC:
American Psychological Association, 293–332.
Werker JF, Tees RC. (1999) Influences on infant speech processing: toward a new synthesis. Annu Rev Psychol 50:509–535.
Kuhl PK. (2004) Early language acquisition: cracking the speech
code. Nat Rev Neurosci 5:831–843.
Werner LA. (1995) Observer-based approaches to human infant
psychoacoustics. In: Klump GM, Dooling RJ, Fay RR, Stebbins
WC, eds. Methods in Comparative Psychoacoustics. Boston:
Birkhauser, 135–146.
Luts H, Desloovere C, Kumar A, et al. (2004) Objective
assessment of frequency-specific hearing thresholds in babies.
Int J Pediatr Otorhinolaryngol 68:915–926.
Werner LA, Gray L. (1998) Behavioral studies of hearing
development. In: Rubel EW, Popper AN, Fay RR, eds.
Development of the Auditory System. New York: SpringerVerlag, 12–79.
Moushegian G, Rupert AL, Stillman RD. (1973) Scalp recorded
early responses in man to frequencies in the speech range. EEG
Clin Neurophysiol 35:665–667.
Werner LA, Marean GC. (1996) Human Auditory Development.
Boulder: Westview Press.
Nozza RJ. (1987) Infant speech–sound discrimination testing:
effects of stimulus intensity and procedural model on measures of
performance. J Acoust Soc Am 81(6):1928–1939.
Widen JE, Folsom RC, Cone-Wesson B, et al. (2000) Identification
of neonatal hearing impairment: hearing status at 8 to 12 months
corrected age using a visual reinforcement audiometry protocol.
Ear Hear 21(5):471–487.
Nozza RJ, Miller SL, Rossman RN, Bond LC. (1991a) Reliability
and validity of infant speech–sound discrimination-in-noise
thresholds. J Speech Hear Res 34(3):643–650.
Yellin MW, Jerger J, Fifer RC. (1989) Norms for disproportionate
loss in speech intelligibility. Ear Hear 10(4):231–234.
Nozza RJ, Rossman RN, Bond LC. (1991b) Infant–adult differences in unmasked thresholds for the discrimination of consonant–vowel syllable pairs. Audiol 30(2):102–112.
Zeng F-G, Nie K, Stickney GS, et al. (2005) Speech recognition
with amplitude and frequency modulations. Proc Natl Acad Sci
USA 102(7):2293–2298.
643