Advantages and Caveats When Recording Steady

advertisement
J Am Acad Audiol 13 : 246-259 (2002)
Advantages and Caveats When
Recording Steady-State Responses
to Multiple Simultaneous Stimuli
M. Sasha John*
David W. Purcell*
Andrew Dimitrijevic*
Terence W. Picton*
Abstract
This article considers the efficiency of evoked potential audiometry using steady-state responses
evoked by multiple simultaneous stimuli with carrier frequencies at 500, 1000, 2000, and 4000 Hz .
The general principles of signal-to-noise enhancement through averaging provide a basis for determining the time required to estimate thresholds . The advantage of the multiple-stimulus technique
over a single-stimulus approach is less than the ratio of the number of stimuli presented . When
testing two ears simultaneously, the advantage is typically that the multiple-stimulus technique is
two to three times faster. One factor that increases the time of the multiple-response recording is
the relatively small size of responses at 500 and 4000 Hz . Increasing the intensities of the 500and 4000-Hz stimuli by 10 or 20 dB can enhance their responses without significantly changing
the other responses . Using multiple simultaneous stimuli causes small changes in the responses
compared with when the responses are evoked by single stimuli . The clearest of these interactions is the attenuation of the responses to low-frequency stimuli in the presence of higher-frequency
stimuli . Although these interactions are interesting physiologically, their small size means that they
do not lessen the advantages of the multiple-stimulus approach .
Key Words: Auditory evoked potentials, masking, MASTER, objective audiometry, steady-state
responses
Abbreviations : MASTER = multiple auditory steady-state response ; MINT = multiple-intensity
(technique)
Sumario
Este art(culo evalua la eficiencia de la audiometria de potenciales evocados utilizando respuestas de estado-estable evocadas por estimulos multiples simultaneos, con frecuencias portadoras
de 500, 1000, 2000, y 4000 Hz . Los principios generales de reforzamiento de la relacion sepal/ruido
por medio de la promediacion aportan la base para determinar el tiempo requerido para estimar
umbrales . La ventaja de la tecnica de estimulos multiples sobre el enfoque de estimulo 6nico es
menor que la tasa del numero de estimulos presentados . Cuando se evaluan dos oidos simultaneamente, la ventaja es tfpicamente que la tecnica de estrmulos multiples es dos o tres veces mas
rapida . Un factor que incrementa el tiempo de registro de las respuestas multiples es el relativamente pequeno tamano de las respuestas a 500 y 4000 Hz . Incrementando en 10 6 20 dB las
intensidades de los estimulos a 500 y 4000 Hz puede aumentar esas respuestas, sin modificar
significativamente las otras . El use de estimulos multiples simultaneos causa pequenos cambios
en las respuestas, comparado con las respuestas evocadas por estimulo 6nico . La mas clara de
estas interacciones es la atenuacion de las respuestas ante estrmulos de baja frecuencia, en presencia de estrmulos de alta frecuencia . Aunque estas interacciones son fisiologicamente interesantes,
su pequeno tamano implica que ellas no reducen las ventajas del enfoque de estrmulos multiples .
Palabras Clave : Potenciales evocados auditivos, enmascaramiento, MASTER, audiometria objetiva,
respuestas de estado-estable
Abreviaturas : MASTER = respuesta auditiva multiple de estado-estable ; MINT = tecnica de
intensidad multiple
*Rotman Research Institute, Baycrest Centre for Geriatric Care, University of Toronto, Toronto, Ontario
Reprint requests : M . Sasha John, Rotman Research Institute, Baycrest Centre for Geriatric Care, 3560 Bathurst Street,
Toronto, ON M6A 2E1 ; e-mail : sasha@psych .utoronto .ca
246
Advantages and Caveats of MASTER/John et al
W
hen multiple auditory stimuli are
presented simultaneously at
suprathreshold intensities, separate
steady-state evoked potentials can be recognized for each stimulus, provided that each stimulus is modulated at a frequency different from
the others . If the recordings are evaluated in the
frequency domain, each response shows up at its
"signature" modulation frequency (Lins and Picton, 1995) . Recording auditory steady-state
responses to multiple simultaneous stimuli is an
attractive approach to objective audiometry
because multiple responses can be detected in
the time normally required to identify a single
response .
Provided that the modulation frequencies
are more than 70 Hz, that the intensity is 60 dB
SPL or less, and that stimuli are presented sep-
arately to the two ears or separated by at least
an octave in frequency within the same ear, the
responses recorded with the multiple-stimulus
technique are not significantly smaller than
those recorded when the stimuli are presented
singly (John et al, 1998) . Presenting four stimuli with carrier frequencies of 500, 1000, 2000,
and 4000 Hz to each ear (for a total of eight
stimuli) should therefore bring about an eightfold increase in the speed of a hearing test . The
goal of this article is to discuss why this may not
occur in actual practice and to consider several
methods of decreasing the time required for
objective audiometry.
Several factors must be considered when
determining whether a testing session using
multiple simultaneous stimuli is more efficient
than multiple single-stimulus sessions . This
article will discuss these various issues in relation to data reported previously in the literature
and will describe the results of some simple
experiments conducted to address some of these
issues and increase the efficiency of the multiplestimulus technique.
BACKGROUND ELECTRICAL
NOISE IN THE RECORDING
he electrical noise in the recording derives
T from the brain; from the muscles of the
scalp, face, and neck ; and from the amplifier. The
background electrical noise can be quantified in
many ways . One can measure either its power
or its amplitude, and one can make these measurements in the time domain or the frequency
domain . Our approach is to transform averaged
time domain data into the frequency domain
and then measure the root mean square ampli-
tude of the activity in the spectrum at frequencies adjacent to the frequencies at which the
steady-state responses are measured . Since our
typical data windows or "sweeps" last 16.384 seconds, the resulting amplitude spectrum has a frequency resolution of 0 .061 Hz . We compare the
amplitude of each steady-state response to the
amplitudes in 120 adjacent frequencies (60 above
and 60 below the modulation frequency of the
stimulus, equivalent to ± 3 .7 Hz) .
In relation to the steady-state responses,
this background activity (quantified in the
120-bin noise estimate) can be considered as
random noise. If several sweeps are averaged
together, the amplitude of the background noise
attenuates with averaging according to the
square root of the number of sweeps averaged .
The time taken to reach a required signal-tonoise ratio is
T = S(BR/A)2
(Equation 1)
where S is the sweep duration, A is the amplitude of the response, B is the amplitude of the
background activity in a single sweep, and R is
the signal-to-noise ratio (in terms of amplitude)
required to conclude that a response is significantly different from the background noise . This
equation assumes that, during the averaging
procedure, the amplitude of the response and the
amplitude distribution of the noise both remain
constant . We have provided empirical evidence
that this is generally true (John and Picton,
2000a) . The equation demonstrates that doubling
the amplitude of the background noise (B) will
increase the required averaging time by 4 ; doubling the amplitude of the response (A) will
decrease the time by 4 .
The amplitude of the background electrical
activity decreases with increasing frequency.
Since most multiple-stimulus paradigms use
modulation frequencies that are close together,
the background noise levels for these paradigms
can be considered as relatively uniform across
the range of modulation frequencies . Figure 1
illustrates some noise measurements from the
frequency domain . These data are taken from
experiments reported in John and Picton (2000a) .
The left graph of the figure shows a small
decrease in noise amplitude with increasing frequency. The right graph shows that the noise
amplitudes decrease over time (as sweeps are
averaged together) according to a square root
rule .
The noise should be the same regardless if
one stimulus or more than one stimulus is being
247
Journal of the American Academy of Audiology/Volume 13, Number 5, May 2002
50
40
30
20
10 0
75
80
85
90
95
100
0
N=1
A N=4
v
4
6
8
10
12
Sweep Number (N)
Modulation Frequency (Hz)
m
2
O
N=12
actual data
- " 1/-~_N_
Model
Figure 1 Residual background noise. This figure presents a re-evaluation of some data already published (John and
Picton, 2000a) . On the left, the root mean square noise levels are plotted as a function of frequency over the range at
which the steady-state responses were recorded . Each data point combines noise data over ± 3.7 Hz from the plotted
frequency (the modulation frequency of the stimulus). There is a small decrease in the amplitude of the noise with increasing frequency. In addition, the noise decreases as the number (N) of sweeps averaged increases. This is further evaluated in the right graph, which combines data across all of the frequencies measured in the left graph and then follows
the averaged noise level as N increases. The large circles represent the actual data (averaged over 10 subjects) and the
small dots represent the theoretical decrease according to a VVYrule.
tested during an experiment. In a recent experiment, we evaluated the evoked potentials to
multiple tone pairs. The amplitude spectra of
these stimuli contain only two frequencies
rather than having a carrier with two sidebands, as in the case of a sinusoidally amplitude-modulated carrier. However, the evoked
responses to tone pairs are similar to those produced by amplitude-modulated tones (Dolphin
et al, 1994), and, for the issues addressed here,
these two types of stimuli can be considered
identical. Each tone pair produced a beating
stimulus with a repetition rate that was equal
to the difference between the tones. We examined responses to these beating stimuli at different beat frequencies (82, 84, and 88 Hz or 178,
180, and 183 Hz), at different response ranges
(near 85 Hz or near 180 Hz), and under singlestimulus or multiple-stimulus conditions . We
shall discuss the responses data from this experiment in a subsequent section of this article and
only consider the noise levels here . The noise levels were significantly greater in the 85-Hz range
compared with the 180-Hz range (F = 410.3,
df = 1, 9, p < .001). Within these ranges, there
248
was a small effect of the beat frequency with
smaller noise levels at higher beat frequencies
(F = 6.1, df = 2, 18, p < .05) . There were no significant differences in the noise levels between the
single and multiple conditions (F = .2, df = 1, 9,
p = .63) .
The results reviewed in this section indicate
several characteristics of the background electrical noise in which the auditory steady-state
responses are recorded . The noise decreases regularly according to the square root of the number of trials averaged, decreases slightly with
increasing frequency, and does not change significantly when the number of simultaneous
stimuli is increased.
AMPLITUDE OF THE RESPONSE
A
nother factor that affects the testing time
for multiple-stimulus technique concerns
any changes in the amplitude of the responses
when using multiple rather than single stimuli.
When there is no change in the background
noise, the multiple-stimulus technique remains
more efficient if the decrease in the size of the
Advantages and Caveats of MASTER/John et al
response in the multiple-stimulus condition is
less than
where M is the number of stimuli (John et al, 1998). As a specific example,
when presenting four stimuli together in a single ear, the multiple-stimulus technique is more
efficient as long as the amplitudes of the
responses do not decrease to less than half of
their amplitude when the stimuli are presented
singly. John and colleagues (1998) used this
principle to evaluate the use of multiple simultaneous stimuli at different modulation frequencies, at different intensities, and at different
separations between the carrier frequencies of
the stimuli. In no case was it less efficient to
record responses to multiple stimuli simultaneously. The amplitudes could be sufficiently
reduced so that there was no significant difference between the multiple- and single-stimulus
techniques at modulation rates of 30 to 50 Hz,
at an intensity of 75 dB SPL or when the carrier frequencies were separated by less than an
octave . For example (John et al, 1998, Table 3),
at 75 dB SPL, the amplitudes recorded when four
stimuli were presented were reduced to 56 percent
and 49 percent of what they were in the singlestimulus condition. This approximately equals
the 50 percent obtained from 1/when M = 4.
These findings indicate that, under certain
circumstances, the amplitude of the responses
may decrease when the number of simultaneous
stimuli is increased. However, even when this
occurs, this decrease does not overcome the
increased efficiency of the multiple-stimulus
technique. As long as certain caveats are
respected, in many conditions there is little
change in the response amplitude in the multiplestimulus condition compared with the singlestimulus condition.
EFFECTS OF CARRIER FREQUENCY
ur discussion so far has assumed that all
0 of the multiple simultaneous stimuli evoke
responses of similar size . However, in practice,
this is not true . Responses to carrier frequencies
between 1000 and 3000 Hz are generally larger
than those outside this range (John et al, 2001,
2002). Figure 2 illustrates results from several
studies.
This difference in amplitude across carrier
frequency causes some responses to become significant before others . This increases the time
needed when using the multiple-stimulus technique to assess thresholds . Although it is always
certain that more data will be collected using the
multiple- rather than the single-stimulus technique, the testing time will be prolonged if the
recording period has to be extended so that the
" MM, 30 dB HL, BC (Dimitrijevic et at, 2002)
O AM, 60 dB SPL, D (Herdman and Stapells, 2001)
m MM, 20 dB SL (Dimitrijevic et al, 2002)
13 MM, 40 dB SPL (John et al, 2001)
A AM, 40 dB SPL (John et a1, 2001)
2,
A AM 30 dB pSPL (John et al, 2002)
500
1000
2000
4000
Carrier Frequency (Hz)
Figure 2 Carrier frequency. This figure plots the amplitudes of responses recorded using the multiple-stimulus technique at each of the audiometric frequencies 500, 1000, 2000, and 4000 Hz in several studies. All responses
are for airconducted sounds except the Dimitrijevic and colleagues bone-conduction (BC) responses . The B responses are relatively
large because both cochleae are activated by each stimulus . The Herdman and Stapells data are from their
dichotic
(D) condition (four stimuli in each ear) . The mixed-modulation (MM) stimuli from the Dimitrijevic and
colleagues' and
John and colleagues' studies combined 100 percent amplitude modulation (AM) with 25 percent frequency modulation .
The AM2 stimuli used in the John and colleagues (2002) study were modulated using an exponential envelope .
249
Journal of the American Academy of Audiology/Volume 13, Number 5, May 2002
smallest of the multiple responses becomes significantly larger than noise. For example, when
stimuli with carrier frequencies of 500, 1000,
2000, and 4000 Hz are presented singly at 20
dB SL, the time required to detect each response
will vary. To obtain typical times, we can substitute values in equation 1 from our studies of
subjects with sensorineural hearing loss (Dimitrijevic et al, 2002). The amplitudes (A) of the
responses at 20 dB SL are, on average, 29, 46,
45, and 37 nV (plotted in Fig. 2) . The background noise level of a single unaveraged sweep
(B) in these studies is approximately 80 nV
(data not shown) . This is almost twice as high
as that obtained in the sleeping normal subjects studied in the experiments that provide the
data for Figure 1. Many of the hearing-impaired
subjects were older volunteers who tended to be
less able to sleep for the entire recording period,
and, accordingly, their muscle activity was
higher. The minimum signal-to-noise ratio (S)
that allows us to state that the response is statistically different from noise is 1.7, which is the
square root of F at p = .05 with degrees of freedom 2 and 240. For the purposes of this illustration, we shall make S equal to 2. Given a
sweep of 0.27 minutes (16.384 seconds), the
times required to detect the responses are 8.4,
3 .4, 3 .5, and 5.1 minutes. The testing time for
the single-stimulus technique is the sum of these
times - 20 .4 minutes. The testing time for the
multiple-stimulus technique is the maximum
of these times - 8.4 minutes . Rather than being
4 times as fast as the single-stimulus technique,
the multiple-stimulus technique is only 2 .43
times as fast . If one stimulus evokes a response
that is much smaller than all of the others, the
recording time for the multiple-stimulus technique will approach (but never reach) that for
the single-stimulus technique
These results demonstrate that the efficiency of the multiple-stimulus technique can be
decreased if the amplitudes of the different
responses are not equal. This attenuation cannot render the multiple-stimulus technique less
efficient than the single-stimulus technique.
SLOPING AUDIOGRAMS
n issue that arises when performing threshA old evaluations is that it takes longer to
determine that a response is absent than it does
to recognize that a response is present (significantly different than the background noise levels) . The decision that a response is not present
usually requires that the response is not recognizably different from the background noise
after this noise has been reduced to a criterion
level . The time (T) required to reach this criterion noise level is
T = S(B/N)z
(Equation 2)
where S is the sweep time, B is the single-sweep
noise level, and N is the noise level at criterion.
Using the values from the Dimitrijevic and colleagues' (2002) study (B = 80 nV and S = 0.27
minutes) and a criterion (N) of 10-nV noise
would predict a total test time of 17 .5 minutes.
Using a testing period of 17 .5 minutes would
allow the detection of a 17-nV response using the
F test at the p = .05 level of significance .
If the subject has a sloping audiogram,
there are several intensities at which an investigator must decide that at least one response
is absent . Recordings at these intensities need
to be continued until the criterion noise level is
reached (rather than until the other responses
are recognized). Let us work through an example in which a subject has hearing thresholds
of 30, 40, 50, and 60 dB HL at frequencies of 500,
1000, 2000, and 4000 Hz, respectively. We shall
begin the recording at 70 dB HL and work down
in 10-dB steps for both multiple- and singlestimulus recordings . We shall arbitrarily set
the threshold levels for the steady-state
responses at 10 dB above the hearing thresholds . Table 1 shows the times that would be
required to measure the thresholds for the
steady-state responses. These were predicted on
the basis of typical response amplitudes and
background noise levels from the study of Dimitrijevic and colleagues (2002) . The total time
to measure thresholds using the single-stimulus
technique was 137.6 minutes compared with
83 .5 minutes for the multiple-stimulus technique. The multiple-stimulus technique is only
1.65 times as fast . If both ears were evaluated
simultaneously and if the thresholds in the two
ears were similar, then the multiple-stimulus
technique would be 3.3 times as fast . If there
were an asymmetry in the ranges of the thresholds between the two ears, the improvement in
speed would be less than 3.3 .
Deciding that a response is absent requires
a longer time than deciding that a response is
present. This may attenuate but not completely
remove the advantage of the multiple-stimulus
technique, particularly when the subject has a
sloping audiogram.
Advantages and Caveats of MASTER/John et al
Table l
Threshold Testi ng in a Subject with a Sloping Audiogram
Single-Stimulus Tech nique
500 Hz
1000 Hz
2000 Hz
Multiple-Stimuli Technique
4000 Hz
Simultaneous Stimuli
Intensity
(dB HL)
R
T
R
T
R
T
R
T
R
R
R
R
T
70
60
50
+
+
+
4 .8
6 .3
8 .44
+
+
+
3 .2
3 .4
6 .0
+
+
-
3 .5
5 .7
17 .5
+
0
13 .6
17 .5
+
+
+
+
+
+
+
+
-
+
-
13 .6
17 .5
17 .5
30
-
17 .5
0
-
-
-
-
17 .5
40
+
12 .9
-
17 .5
0
0
0
+
0
-
-
-
17 .5
R = whether a response was recognized (+), not recognized (-), or not recorded (0); T = time taken in minutes to detect response or
decide that it was absent
The time taken for the multiple-stimuli recording (rightmost column) is equal to the longest time for any individual stimulus .
MULTIPLE-INTENSITY
(MINT) TECHNIQUE
ne way to compensate for problems incurred
0 when multiple stimuli evoke responses of
different sizes is to present stimuli with different intensities at different carrier frequencies.
The intensities of the stimuli evoking the smaller
responses can be increased so that the magnitudes of all of the individual responses being
recorded are more or less the same . For simplicity, we shall refer to this approach as the
MINT or multiple-intensity technique .
An obvious concern about this technique is
that the tones presented at higher intensities
might mask those presented at lower intensities .
We therefore examined the advantages of adjusting individual stimuli, within the multiple stimuli, to different intensities and investigated if any
masking effects occurred for the stimuli presented at lower intensities . We used four stimuli in one ear (at carrier frequencies of 500,
1000, 2000, and 4000 Hz) in conditions wherein
all stimuli were of equal intensity or wherein the
500-Hz and the 4000-Hz stimuli were either 10
or 20 dB more intense than the other stimuli
Table 2
ondition
1
2
3
4
5
6
7
8
9
(Table 2) . The stimuli were sinusoidally
amplitude-modulated tones with the depth of
modulation at 100 percent and modulation frequencies as described in Table 2. The conditions
of this study contained two experiments: one
examined a 10-dB enhancement at several intensities (conditions 1-6) and another investigated
a 20-dB enhancement at the highest intensity
(conditions 1 and 7-9) .
Steady-state responses were recorded in 10
subjects with normal hearing using the multiple auditory steady-state response (MASTER)
data collection system (John and Picton, 2000a ;
www.hearing .cjb .net) . The system performed
digital-to-analog conversion of the stimulus waveforms at 32 kHz and routed them to a Grason
Stadler Model 16 audiometer, where they were
amplified to a calibration intensity, attenuated
to achieve the desired intensity levels, and presented using Etymotic-2A insert earphones .
Mixing stimuli of different intensities is
more easily performed using analog circuity
than within the digital-to-analog converter since
the resolution of the converter can cause the
representation of the stimuli of lower intensity
to be less accurate . Stimuli of different intensi-
Experimental Conditions to Evaluate MINT Protocols
Stimulus Intensity (dB SPL)
Recording
Duration
(min)
f = 500
f = 80 Hz
f = 1000
f = 85 Hz
f = 2000
f = 90 Hz
f = 4000
f = 95 Hz
5 .4
5 .4
12 .6
12 .6
50
60
40
50
50
50
40
50
50
40
50
60
40
19 .6
5 .4
40
70
30
30
30
30
30
40
19 .6
5.4
5 .4
30
70
50
40
50
50
50
40
50
50
50
50
70
50
70
Journal of the American Academy of Audiology/Volume 13, Number 5, May 2002
ties were therefore created in separate digitalto-analog channels. For example, in condition 2,
the 500- and 4000-Hz stimuli were created and
routed to the tape A input of the audiometer,
where they were amplified to 60 dB SPL, and the
1000- and 2000-Hz stimuli were sent to the tape
B input, where they were amplified to 50 dB SPL.
Both tape A and tape B were then routed to the
left ear insert, which caused an analog addition of the signals prior to acoustic transduction .
The electroencephalographic (EEG) data
were obtained from an electrode placed at Cz,
using a posterior midline neck electrode as reference and an electrode on the clavicle as ground .
The EEG was obtained using a Grass P55 preamplifier with a gain of 10,000, a low-pass setting of 300 Hz, and a high-pass filter setting of
0.3 Hz . The recorded EEG data were again amplified with a gain of 5 by the analog/digital board
and then analog-digital converted at 1000 Hz .
During data collection, the data were submitted
to an online weighted averaging procedure to
reduce the effects of spurious bursts of unwanted
noise, which occurred in our data (John et al,
2001).
Because the signal-to-noise ratio is better at
higher stimulus intensities, less time is needed
for responses to reach significance . At 50 dB
SPL or higher, several of the eight stimuli will
reach significance within the first minute of
testing, whereas, in most subjects, about 6 minutes may be required for all eight stimuli to
reach significance . At near-threshold intensities, up to 20 minutes may be required to obtain
significant responses for all eight stimuli . Accordingly, we used increased durations for conditions for which stimuli were presented at lower
intensities (see Table 2) . The order in which the
conditions were recorded was randomized across
subjects . For the responses collected when the
stimuli had equal intensity at 50, 40, or 30 dB
SPL (conditions 1, 3, and 5), the number of significant responses was 38, 37, and 33 out of 40,
indicating that the test durations were adequate .
Figure 3 shows representative data from a
single subject and Figure 4 shows the average
amplitudes of the responses across all 10 subjects .
We decided to evaluate the 500- and 4000-Hz
responses separately from the 1000- and 2000Hz responses since the experimental manipulation affected them differently (a change in
intensity of the stimulus or the effect of that on
other stimuli) . For the 500- and 4000-Hz
responses, a repeated measures analysis of variance (ANOVA) (3 intensities x 2 stimulus types x
2 carrier frequencies) indicated significant main
252
effects for intensity (F = 44 .5, df = 2, 18, p < .001)
since amplitude increased as intensity increased
from 30 to 50 dB SPL, for stimulus type (F = 34 .4,
df = 1, 9, p < .001) since the MINT stimuli produced larger responses than equal intensity stimuli, and for carrier frequency (F = 10 .6, df = 1, 9,
p < .01) since the 500-Hz response was significantly bigger than the 4000-Hz response . There
was a significant interaction between intensity
level and carrier frequency (F = 10 .4, df = 2, 18,
p < .01) because the 500-Hz responses grew more
rapidly with increasing intensity than the
responses at other frequencies.
A second ANOVA performed using the amplitudes of the 1000- and 2000-Hz stimuli showed
a main effect of intensity (F = 49 .5, df = 2, 18,
p < .001) but no effect of stimulus type (F = 0.6,
df = 1, 9, p = .45) . There was also a significant
interaction between stimulus type and carrier
frequency, to which we shall return later.
The main findings clearly indicated that
responses
to the 500- and 4000-Hz stimuli can
the
increased
in
amplitude by raising the intenbe
sity of these stimuli. This increase in amplitude
occurs without any general effect on the responses
to the other stimuli. The second experiment
using stimuli that were 20 dB rather than 10 dB
more intense showed similar results, with the
more intense stimuli evoking larger responses,
also without significantly affecting the responses
to the other stimuli (Fig . 5).
In general, the use of the MASTER technique provides a rapid method for collecting
auditory steady-state response data because
multiple stimuli are tested in the time normally
required to test a single stimulus . The technique becomes less efficient if one of the
responses being evaluated is considerably
smaller than the other stimuli and is characterized by an amplitude that is near noise levels. In the results presented here, the 4000-Hz
response shows a smaller response at all three
intensity levels in the equal-intensity conditions. In actual practice, this unequal amplitude would cause the MASTER technique to
become considerably less efficient. The MINT
condition more than compensated for this possibility by using +10-dB relative intensity, suggesting that as little as +5 dB could be used to
make the size of the 4000-Hz response equivalent to the others .
The 500-Hz responses in the equal-intensity 50 dB SPL and 40 dB SPL conditions were
generally larger, rather than smaller, than the
responses to the 1000- and 2000-Hz carriers (see
Fig. 4) . This is unlike previously reported data
Advantages and Caveats of MASTER/John et al
f~
fm
500
80
1000
85
2000
90
4000
95
MINT with 500 and
4000 Hz 10 dB more intense
Equal Intensity
V
V
20 nV
40 dB
SPL
0
30 dB
SPL
70
80
90
100
110
70
80
90
100
110
Modulation Frequency (Hz)
Figure 3
Multiple-intensity stimuli: single subject. This figure plots in the frequency domain responses recorded from
a single subject for the multiple-intensity technique. The carrier frequencies (f) and modulation frequencies (f) for
the stimuli are shown in the upper left of the figure . The responses show up at the stimulus modulation frequencies as
lines that are higher than the background noise levels at adjacent frequencies . Filled triangles indicate responses that
are significantly different from the noise at p < .05 and open triangles indicate responses that cannot be distinguished
from the background noise. The graphs on the left show response amplitudes in the equal-intensity condition. The responses
are smaller for the 500- and 4000-Hz responses than for the 2000-Hz responses . The graphs on the right show how the
response amplitudes increase when the 500- and 4000-Hz stimuli are increased in amplitude by 10 dB (MINT condition). The noise levels at 30 dB SPL are lower than at 40 dB SPL since more trials were averaged (see Table 2) .
from the literature (Fig . 2) and might be related
to interindividual variability in the response
amplitudes . The single-subject data shown in
Figure 3 contain amplitudes that more closely
follow the expected distribution of response
amplitudes for the four frequencies examined .
In general, the 500-Hz response is often difficult to detect at lower intensity levels (Aoyagi
et al, 1994 ; Rance et al, 1995 ; Lins et al, 1996 ;
Perez-Abalo et al, 2001 ; Dimitrijevic et al, 2002).
Several factors might contribute to the small size
of this response . The activation pattern of the
500-Hz stimulus on the basilar membrane spans
a broader area than the activation patterns of
stimuli with higher frequency, and this area is
in a region where the traveling wave is slowing
down . Latency jitter between responses generated through different parts of the activated
basilar membrane may therefore decrease the
size of the compound response . Lins and colleagues (1996) also suggested that recording
the 500-Hz response could be difficult owing to the
masking effect of ambient background noise at the
lower frequencies . Alternatively, higher frequencies within the set of simultaneous stimulus might
mask the response to the 500-Hz carrier. However,
as pointed out by Perez-Abalo and coleagues
(2001), similar difficulties in the estimation of
low-frequency thresholds have been reported for
single-stimulus techniques (e .g ., Aoyagi et al,
1994), suggesting that masking is unlikely.
The 4000-Hz response was the smallest at
all three intensity levels in the equal-intensity
condition. The relatively small size of these
responses is likely not owing to masking since
both this study and previous studies (Dolphin,
1997) have shown that lower frequencies tend
to enhance, rather than suppress, responses to
higher-frequency stimuli. Part of the decreased
amplitude is owing to the higher hearing level
threshold at this frequency compared with
1000 Hz, but the small size persists when the
253
Journal of the American Academy of Audiology/ Volume 13, Number 5, May 2002
80
40 dB SPL
50 dB SPL
,
30 dB SPL
60
40
20
0
500
1000
2000
4000
500
2000
1000
4000
500
1000
2000
4000
Carrier Frequency (Hz)
O
0
500- and 4000-Hz stimuli 10 dB more intense
All carriers at same intensity
Multiple-intensity stimuli: mean data . The figure shows averaged mean response amplitudes computed across
10 subjects for both the MINT (+10 dB) and equal-intensity stimuli presented at 30, 40, and 50 dB SPL. In the MINT
conditions, there is a clear enhancement of the responses to the more intense stimuli. There is also a small enhancement of the 1000-Hz response and a small attenuation of the 2000-Hz response . Although this interaction was significant, only the 1000-Hz enhancement reached significance with post hoc testing. The response to the 500-Hz stimulus
at 40 dB SPL is larger when the stimulus is presented with other stimuli at 40 dB (filled circle in the middle graph)
than when it is presented with other stimuli at 30 dB SPL (open circle in the right graph) . This is likely due to the greater
Figure 4
noise levels in the recordings at 40 dB (owing to less averaging) .
stimuli are calibrated in hearing level or sound
level (e .g., the Dimitrijevic et al, 2002, data
plotted in Fig. 2) . The range of activation on
the basilar membrane is smaller, and fewer neurons respond at 4000 Hz than at lower carrier
frequencies.
The use of MINT stimuli with the 500- and
4000-Hz stimuli being presented at levels 5 to
10 dB higher than the other stimuli may help
to compensate for the decreased response amplitudes and lead to increased recording efficiency
when using multiple simultaneous stimuli. The
small changes in the responses to the other
stimuli following these manipulations are too
small to affect the recording efficiency.
INTERACTIONS BETWEEN STIMULI
n evaluating the multiple-stimulus effect, we
have so far assumed that if the responses
are decreased in the multiple-stimulus condition,
this reduction occurs uniformly across the different responses. The evidence presented in the
literature on the human steady-state responses
has generally found no significant effects of
using multiple stimuli compared with using sin254
gle stimuli (Lins and Picton, 1995 ; John et al,
1998 ; Herdman and Stapells, 2001) provided
that the stimuli are 60 dB SPL or less, are separated by ear or by an octave in carrier frequency, and are modulated at rates above 70 Hz .
However, the literature also reports evidence
for small interactions, and these are shown in
some of our recent data .
If we look closely at the amplitudes of the
responses to the 1000- and 2000-Hz stimuli in
the MINT experiment (see Fig. 4), we note a significant interaction between stimulus type
(MINT versus equal intensity) and carrier frequency (F = 13 .9, df = 1, 9, p < .01) . This interaction was owing to the 1000-Hz responses
having enhanced amplitudes in the MINT condition (F = 6.7, df =1, 9, p < .05) and the 2000-Hz
responses showing a small but not significantly
decreased amplitude (F = 0.7, df = 1, 9, p > .42) .
These data clearly indicate that the presence of
a low-frequency stimulus at a higher intensity
than the other stimuli can enhance the amplitude
of the response to a stimulus with a carrier frequency one octave higher. A similar effect was
noted in an experiment reported by John and colleagues (1998) . The response to a 1000-Hz stimulus (the "probe") was increased from 59 to 80 nV
Advantages and Caveats of MASTER/John et al
80 ,
60
0
L
0
40 u
20 0
500
1000
2000
500
4000
1000
2000
4000
Carrier Frequency (Hz)
O
500 and 4000-Hz stimuli 20 dB more intense
0
500 stimulus 20 dB more intense
0
All carriers at same intensity
A
4000 stimulus 20 dB more intense
Figure 5 Multiple-intensity stimuli: +20 dB . The left graph shows averaged mean response amplitudes computed
across 10 subjects for both the MINT (+20 dB) and equal-intensity stimuli presented at 50 dB SPL. The right graph
shows the amplitudes when the MINT stimuli have either a 500- or a 4000-Hz carrier at the greater intensity level .
when the stimulus was presented in the presence of a 500-Hz stimulus (the "masker") .
Our experiment using tone pairs indicated
that high-frequency sounds may attenuate
responses to low-frequency sounds . The multiple tone-pair recordings employed three tone
pairs with the higher stimulus frequencies of
each pair (f2) separated from one another by an
octave (approximately 900, 1800, and 3600 Hz),
as shown in Table 3 . We employed two groups
of beat frequencies (near 85 Hz and near 180 Hz)
to determine the frequency of the lower tone in
each pair (f1) . Tone pairs were presented at 50
dB SPL, one pair at a time (single condition), and
three pairs together (multiple condition) using
either the 85-Hz or 180-Hz separation between
each f2 and fl .
Figure 6 presents the mean response amplitudes (for 10 subjects) measured under singleand multiple-stimulus conditions . The data were
analyzed using a three-factor (single versus
multiple, f2 frequency, beat-frequency range)
ANOVA. The responses to beats in the 180-Hz
range were smaller than the responses to beats
in the 80-Hz range (F = 10 .3, df = 1, 9, p < .01) .
There was no significant effect of the f 2 fre-
quency or single versus multiple for the threefactor ANOVA . However, there was an interaction between single versus multiple and the f 2
frequency (F = 7.1, df = 2, 18, p < .01) . Post hoc
testing showed a small but statistically significant decrease in the magnitude of the response
for the multiple tone-pair condition only when f 2
was 900 Hz and the beats were in the 180-Hz
range (F = 8 .4, df = 1, 9, p < .05) . There was also
a small decrease when f 2 was 900 Hz and the
beats were in the 80-Hz range; however, it was
not statistically significant (F = 2 .0, df = 1, 9,
Table 3
Response
Range
(Hz)
Tone-Pair Parameters
Stimulus
Pair
f2 (Hz)
f, (Hz)
Beat
Frequency
(Hz)
85
1
2
3
904
1800
3593
822
1716
3505
82
84
88
180
4
5
6
904
1798
3589
727
1618
3406
178
180
183
255
Journal of the American Academy of Audiology/Volume 13, Number 5, May 2002
85-Hz Range
180-Hz Range
50
40
30
20
10
0
" Single
O Multiple
900
1800
3600
900
1800
3600
Frequency of f2 (Hz)
Figure 6 Multiple stimuli versus single stimulus . This graph compares the responses to tone pairs (beats) presented
singly (one tone pair) or multiply (three tone pairs) . The amplitudes were vector averaged across 10 subjects . The responses
were recorded using beating frequencies that were near 85 Hz (left) or near 180 Hz (right) . The stimulus frequency is
plotted as the second frequency of the pair (f2) . For both ranges of beat frequency, the response amplitude for the lowest frequency is slightly decreased in the multiple-stimulus condition (open circles) .
p =19) . The amount of attenuation was -15 percent at 85 Hz and -35 percent at 180 Hz .
These results indicate that when using multiple stimuli, higher-frequency stimuli can attenuate the responses to the lower-frequency
stimuli. Other studies have found similar
decreases in the low-frequency responses,
although these have not been significant on statistical testing. Lins and Picton (1995, Table 4)
found a multiple-stimulus effect of -18 percent
for the 500-Hz stimulus for four stimuli in one
ear and -9 percent for four stimuli in each ear.
Herdman and Stapells (2001, Fig. 2A) found
effects of -9 percent for four stimuli in one ear
and -24 percent for the dichotic condition (four
stimuli in each ear) . Dolphin (1997) found that
the gerbil responses to tone pairs were attenuated when tone pairs of higher frequency were
simultaneously presented. The attenuation was
larger for the tone pairs of lower frequency (both
in terms of the tones and in terms of the beat frequencies). The greatest attenuation occurred
for the 299- to 337-Hz pair, which had a 38-Hz
beat frequency. This attenuation was sufficient
to make the efficiency of the multiple-stimulus
technique equal to that of the single-stimulus
technique. A modulation rate of 38 Hz is lower
256
than normally used for audiometry with the
auditory steady-state responses, and the
response may be generated as much in the cortex as in the brain stem . All other interactions
were smaller; thus, the multiple-stimulus technique was generally more efficient. Indeed, the
presence of the lower-frequency tone pair could
enhance the amplitude of the responses to
higher-frequency tone pairs.
If the tone pairs are very close together,
such that tones are all within the same critical
band, Dolphin (1996) found that the responses
in dolphins were attenuated to a level (-3 dB)
where it was not advantageous in terms of efficiency to record responses to simultaneous stimuli. This is similar to data reported by John and
colleagues (1998, Fig. 4), which found that presenting two sinusoidally amplitude-modulated
tones with carrier frequencies of 1000 and 1050 Hz
reduced the responses to one half of what they were
when the stimuli were presented alone.
Dolphin and his colleagues (Dolphin and
Mountain, 1993 ; Dolphin et al, 1994) looked at
the interaction between an "interfering" pure
tone on the gerbil response to a probe stimulus
that was either a sinusoidally amplitudemodulated tone or a tone pair. The probe
Advantages and Caveats of MASTER/John et al
response was attenuated when the frequency of
the interfering tone was higher than the probe
stimulus . This is the opposite of what is found
for tone-on-tone masking, for which low-frequency tones attenuate the response to (or raise
the threshold for) higher-frequency tones. Very
similar findings were found for the human
steady-state response (John et al, 1998) . The
mechanism of the effect is not known . Dolphin
and Mountain (1993) suggested either some
desynchronization of envelope-following neurons or some effects of suppression . Bernstein
(1994) suggested that there may be some acoustic
disruption of the envelopes when the interfering tone is too close to the probe tone .
Other possible interactions between stimuli
may occur, but these are smaller than those
already described. Lins and Picton (1995) found
that the responses to dichotic stimuli (two stimuli in each ear - 500 and 2000 Hz) were significantly larger than those to single stimuli, but
this effect was not replicated when four stimuli
were presented to each ear.
Interactions occur when stimuli are presented simultaneously. The most reliable of
these are an attenuation of the responses to
lower-frequency stimuli when presented with
stimuli of higher frequency and an enhancement of the responses to higher-frequency stimuli when presented with lower-frequency stimuli.
These interactions are worth studying in their
own right since they indicate physiologic interactions in the cochlea or auditory nervous system . However, these interactions are generally
small and do not substantially affect the efficiency of the multiple-stimulus technique .
AUTOMATICALLY TRACKING
THRESHOLDS
here are two main indications for using
T stimuli with different intensities at differ-
ent carrier frequencies . The first reason is to
enhance small responses by making the stimuli
that evoke these responses more intense . By
making responses roughly the same size, the
recording is not delayed by the requirement
that smaller responses have to have more time
to reach significance . A second reason why different intensities might be used is to evaluate
patients with hearing losses that differ significantly across frequency . Individual stimuli could
be independently increased or decreased in
intensity to determine thresholds independently
at the different frequencies as long as using
stimuli of different intensity did not substantially
attenuate the responses to other stimuli. Even
with differences of 20 dB between the simultaneously presented stimuli, this appears true .
At present, the MASTER system records
responses to a set of modulated tones that are all
presented at the same intensity. The duration of
the recording period at that intensity is determined by how long it takes to recognize the
response with the smallest amplitude or how
long it takes to determine that one or more of the
responses is not present. A more dynamic version
of the MASTER technique could adjust the intensity of the stimulus on the basis of whether a
response has been recognized . For example, the
system could begin with all of the stimuli at
50 dB HL. As the larger responses to some stimuli
become significant, the system could present
these stimuli at 40 dB HL, while maintaining at
50 dB HL the stimuli that have not yet evoked
significant responses . If one of the responses
does not become significant after some criterion
averaging time has passed or some criterion
noise level has been reached, the intensity of
that stimulus can be increased . The program
will assess the responses in the frequency domain
after each sweep . Rather than using one averaged
sweep, the responses (and the appropriate noise
values) will be allocated to separate averages
on the basis of the modulation frequency and
intensity of the stimulus . The computer will thus
store a set of responses at different intensities,
with each response averaged sufficiently either
to recognize it as different from noise or to determine that it is not significantly larger than some
criterion. The software will have to include algorithms to assess responses, decide on the next
intensity, allocate responses to appropriate memory locations, and keep track of these different
responses . The hardware will also have to be
more sophisticated than that used presently.
Producing stimuli at different intensities will
require a high-precision digital-to-analog converter. Alternatively, a digital-to-analog system
with eight channels of audio output would allow
the intensity of the individual carrier frequencies
to be independently adjusted prior to being mixed
and sent to the left or right ear.
A dynamic MASTER system could thus continually adjust the intensities of the stimuli on
the basis of the significance of the recorded
response until the intensities begin to hover
around the thresholds for the response . The
present system automatically determines when
a response is present. In the envisioned system, even the decision about which intensity or
intensities to use would be automatic .
257
Journal of the American Academy of Audiology/Volume 13, Number 5, May 2002
CONCLUSION
his article has considered the efficiency of
T evoked potential audiometry using steadystate responses evoked by multiple simultaneous stimuli. The general principles of signal-tonoise enhancement through averaging have
allowed us to evaluate empirical results and
estimate times using simple mathematical equations. Having obtained normative values for
both signal and noise levels, we can estimate how
long recordings must last for stimuli of different
intensities . These calculations can also be used
to assess the required testing time for an individual subject using the signal and noise values
from the data collected early in the recording
procedure.
In the clinical setting, the advantage of the
multiple-stimulus technique over a singlestimulus approach is generally less than the
ratio of the number of stimuli presented. This
occurs when the responses are of different sizes
or when the thresholds are not equal for the
different stimuli. Response amplitudes evoked
by multiple stimuli can be made similar by
increasing the intensities of those stimuli that
are producing smaller or no responses. Within
the limits that we have evaluated so far (20 dB),
this can be done without producing significant
changes in the responses to the other stimuli.
The issues raised here are important when
using the MASTER technique in clinical audiometry, for which one needs to determine as quickly
as possible if a response is present or absent .
Within the laboratory setting, the concern may
be more related to response parameters (amplitude and phase) rather than to response detectability. When investigating how an experimental
manipulation affects responses evoked by stimuli of different frequencies, the MASTER technique generally decreases the time required by
a factor of the number of stimuli. In addition,
nonexperimental factors such as the arousal
level of the subject or the placement of the stimulus transducer will be the same for the
responses to all frequencies. In John and Picton
(2000b), we examined the phase of responses
evoked by carriers modulated at several different modulation rates and then converted the
phase data into latency. The standard deviations of the latency data were very small because
long recording durations were used to decrease
background noise to very low levels and obtain
stable estimates of phase. Rather than 2 hours,
it would have taken closer to 16 hours to obtain
these data with equivalent noise levels using the
single-stimulus technique.
Our general experience is that using multiple stimuli rather a than single stimulus
increases the speed of measuring response parameters such as amplitude or latency by the
number of stimuli presented simultaneously.
Certain caveats concerning the stimulus intensity and the differences in carrier frequency
between stimuli need to be followed for this to
be true . In the context of evoked potential
audiometry, for which decisions are based on
the presence or absence of a response, the advantage of using multiple stimuli rather than a single stimulus is reduced. When using eight
stimuli, we would estimate the multiple-stimulus
technique as two to three times faster than the
single-stimulus technique.
Acknowledgment. This research was funded by the
Canadian Institutes of Health Research . The authors
would also like to thank James Knowles, the Catherall
Foundation, and the Baycrest Foundation for their support . Patricia van Roon assisted with the data collection
and preparation of this manuscript .
REFERENCES
Aoyagi M, Kiren T, Furuse H, et al . (1994) . Pure-tone
threshold prediction by 80 Hz amplitude modulation following response . Acta Otolaryngol Suppl (Stockh) 511:
7-14 .
Bernstein LR . (1994) . Comments on "The envelope following response (EFR) in the Mongolian gerbil to
sinusoidally amplitude-modulated signals in the presence of simultaneously gated pure tones" [JAcoust Soc
Am 94 :3215-3226, 19931. JAcoust Soc Am 96 :1189-1190 .
Dimitrijevic A, John MS, van Roon P, et al . (2002) .
Estimating the audiogram using multiple auditory steadystate responses. J Am Acad Audiol 13 :205-224 .
Dolphin WY (1996). Auditory evoked responses to amplitude modulated stimuli consisting of multiple envelope
components . J Comp Physiol A 179:113-121 .
Dolphin WY (1997) . The envelope following response to
multiple tone pair stimuli. Hear Res 110:1-14.
Dolphin WF, Mountain DC . (1993) . The envelope following response (EFR) in the Mongolian gerbil to sinusoidally
amplitude-modulated signals in the presence of simultaneously gated pure tones. JAcoust Soc Am 94 :3215-3226 .
Dolphin WF, Chertoff ME, Burkard R . (1994) . Comparison
of the envelope following response in the Mongolian gerbil
using two-tone and sinusoidally amplitude-modulated
tones. JAcoust Soc Am 96 :2225-2234 .
HerdmanAT, Stapells DR. (2001) . Thresholds determined
using the monotic and dichotic multiple auditory steadystate response technique in normal-hearing subjects .
Scand Audiol 30 :41-49 .
Advantages and Caveats of MASTER/John et al
John MS, Lins OG, Boucher BL, Picton TW. (1998) .
Multiple auditory steady-state responses (MASTER) :
stimulus and recording parameters . Audiology 37 :59-82 .
Lins OG, Picton TW. (1995) . Auditory steady-state
responses to multiple simultaneous stimuli . Electroencephalogr Clin Neurophysiol 96 :420-432 .
John MS, Picton TW. (2000a) MASTER: a Windows program for recording multiple auditory steady-state
responses . Comp Methods Prog Biomed 61 :125-150 .
Lins OG, Picton TW, Boucher BL, et al . (1996). Frequencyspecific audiometry using steady-state responses. Ear
Hear 17 :81-96 .
John MS, Picton TW (2000b). Human auditory steadystate responses to amplitude-modulated tones: phase and
latency measurements . Hear Res 141 :57-79 .
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, Dimitrijevic A, Picton TW. (2002) . Auditory
steady-state responses to exponential modulation
envelopes . Ear Hear 23 :106-117 .
Perez-Abalo MC, Savio G, Torres A, et al . (2001) . Steady
state responses to multiple amplitude-modulated tones:
an optimized method to test frequency-specific thresholds in hearing-impaired children and normal-hearing
subjects . Ear Hear 22 :200-211 .
Rance G, Rickards FW, Cohen LT, et al . (1995) . The automated prediction of hearing thresholds in sleeping subjects
using auditory steady-state evoked potentials . Ear Hear
16 :499-507 .
Download