Journal of Speech and Hearing Research, Volume 27, 306-310, June 1984
tleseareh Note
RELATION
BETWEEN
REACTION
LARRY E. HUMES
TIME
AND
LOUDNESS
J A Y N EB. AHLSTROM
Vanderbilt University School of Medicine, Nashville, Tennessee
The loudness of one-third octave bands of noise eentered at either 1, 2, or 4 kHz was measured in 10 normal-hearing young
adults for sound levels of 50-90 dB SPL. Reaction times (RT) in response to these same stimuli were also measured in the same
subjects. A moderate-to-strong correspondence was observed between the slopes for functions depicting the growth of loudness
with sound level and eomparable slopes for the reaction-time data. The correlation between slopes for the RT-intensity function
and the loudness-growth function was comparable in magnitude to the test,retest eorrelcttion for the loudness-growth function
except at 1 kHz.
exist and are reliable, although the factors responsible for
these individual differences remain obscure. Statistically
significant test-retest correlations of .60-.85 have been
established, for example, for the slopes of loudnessgrowth functions at 1000 Hz repeated six times with a 24hr intersession interval (Wanschura & Dawson, 1974) and
two times with an l l - w e e k intersession interval (Logue,
1976). There is not much evidence, however, supporting
directly the supposition that those subjects demonstrating
the greatest increase in loudness with increase in stimulus level also exhibit the greatest decrease in reaction
time for identical stimulus conditions. Reason (1972), for
example, obtained only a moderate, although statistically
significant, correlation (r = .45) between the slopes of the
loudness-growth function and the slopes of the RT-intensity (RT-I) function. Data were obtained only at 1000 Hz
by this investigator.
It was the purpose of this study, therefore, to obtain RTI and loudness-growth functions from n0rmal-hearing
young adults and to establish correlations between the
slopes of these two functions at frequencies of 1, 2 and 4
kHz. In this way the strength of the relation between
loudness and RT might be evaluated in more detail. The
strength of this relation will determine whether RT might
be substituted for direct loudness measurements in laboratory animals or with clinical patients.
More than 40 years ago. Chocholle (19401 demonstrated
that the time it takes to react to a sound varies inversely
with its loudness. That is. the louder the sound, the
shorter the time needed to react to its presentation.
Choeholle further demonstrated that equal-loudness contours derived from the reaction-time (RT) paradigm resembled closely those contours obtained by loudnessbalance methods.
Since this pioneering investigation, several other investigators have observed close correspondence between
changes in loudness measured with the RT paradigm and
loudness-balancing techniques. Loudness changes under
partial masking conditions, for example, have been found
to be comparable when measured with either technique
(Chocholle & Greenbaum, 1966). Recruitment effects
have also been equally apparent utilizing either technique (Pfingst, Hienz, Kimm, & Miller, 1975).
Because of the close correspondence observed between loudness and RT, several researchers doing behavioral work with animals have utilized the RT paradigm to
examine loudness in animals. The use of animals as
subjects obviously necessitates the use of an indirect
measure of loudness. Stebbins 11966) was apparently the
first to suggest this use of the RT paradigm. Since this
investigation, the RT paradigm has been utilized to relate
reaction times to evoked-potential latencies ~Miller,
Moody, & Stebbins. 1969), to study the effects of permanent and temporary hearing loss on loudness perception
(Moody, 1973; pfingst et al., 1975). and to examine the
effects of partial masking on loudness perception (Moody,
1979). Although all the above animal studies utilized
monkeys as experimental subjects, the RT paradigm has
proven effective in other species, including cat (Sandlin
& Gerken, 1977), chinchilla (Luz & Nugen, 1972), and the
house finch (Dooling, Zoloth, & Baylis, 1978).
One approach to better establish the relationship between loudness and RT that has not been pursued previously by many investigators is to correlate loudness
growth functions and RT-growth functions within the
same set of subjects. Individual differences in loudnessgrowth functions obtained with magnitude estimation
© 1984, American Speech-Language-Hearing Association
METHODS
Subjects
Ten normal-hearing young adults aged 22--31 years (~?=
25.2 yrs.) comprised the subject sample of this study. All
subjects had pure-tone air-conduction hearing thresholds
~<10 dB HL at octave intervals of 250-8000 Hz in both
ears. In addition, middle ear function appeared normal
bilaterally in that all subjects had tympanograms of normal shape, amplitude, and peak-pressure point and present acoustic reflexes for a signal level of 100 dB HL
(500-4000 Hz). None of the subjects had participated
306
0022-4685/84/2702-0306501.00/0
HUMES & AHLSTROM: RT and Loudness
previously in the procedures to be described here. All
subjects were paid for their participation. Each subject
participated in one 2-hr session.
Apparatus
All stimuli utilized in this study were one-third octave
bands of noise that were created by routing the output of a
random-noise generator (GenRad Model 1390B) through
low-pass and high-pass sections of a solid-state filter
(Krohn-Hite Model 3848). Noise-band rejection rates
were 48 dB/octave. Attenuation of stimuli was accomplished through a combination of decade attenuators
(Hewlett-Packard Model 350D) and programmable attenuators (Coulbourn Instruments Model $85-08). Stimuli
were gated by a programmable electronic rise/fall gate
(Coulbourn Instruments Model $84-04). All stimuli were
amplified (Crown Model D-75) prior to transduction.
Sound-field stimuli were delivered via six JBL E-110
loudspeakers wired in parallel. The loudspeakers were
positioned to produce a diffuse field around the listener's
head. Reverberation times for octave-band signals centered at 250-8000 Hz were all approximately 0.95 s (-+0.1
s). A D E C PDP-11/03 laboratory minicomputer controlled all experimental paradigms including gating of
stimuli, attentuation of signals, activation of lights on the
subject's response box, and the collection of responses.
The computer was also involved in the measurement of
reaction times.
Procedures
The loudness magnitude-estimation procedure utilized
in this study was developed from the guidelines for such
procedures provided by Stevens (1975). Subjects simply
assigned numbers to one-third octave bands of noise
presented in the diffuse sound field with the numbers
corresponding to the perceived loudness of the noise
bursts. When the geometric mean of these estimates is
plotted on a log scale against noise level in decibels, the
loudness-growth function is defined. Loudness-growth
functions were obtained for noise levels spanning a 40-dB
range from the lowest to the highest sound level at onethird octave band center frequencies of 1, 2, and 4 kHz.
The maximum output levels for each of the center frequencies of the one-third octave bands of noise used in
the magnitude-estimation procedure were 94, 92, and 88
dB SPL, respectively. The loudness-growth functions
were obtained over the 40-dB range of noise levels by
attenuating these noise levels by 0, 4, 10, 16, 20, 24, 30,
36, and 40 dB. Five random sequences of these attenuation values were generated and stored in the computer.
The nine attenuation settings described above appeared
twice in each sequence for a total of 18 attenuation
settings per sequence. A given loudness-growth function
307
was derived by randomly selecting two of the five sequences. Thus, each loudness-growth function was derived using a total of 36 trials with each of the nine
attenuation settings being repeated a total of four times.
The availability of five random sequences and the use of
any two of these for a given loudness-growth function
yielded 9.5 possible quasi-random sequences of attenuation settings for the 36 trials.
For the magnitude-estimation procedure, the subject
was seated in the diffuse sound field with a response box
positioned in front of him or her. Each trial began with a
500-ms warning signal followed 500 ms later by a 1-s
stimulus presentation (25-ms linear rise/fall time). Activation of a response light followed stimulus termination by
500 ms. The subject then wrote down a number that
corresponded to the perceived loudness of the noise
burst. When finished, the subject pressed a button which
turned off the response light and initiated another trial
500 ms later. This process was repeated until all 36 trials
had been completed.
As was the case for the magnitude-estimation procedure, RT-I functions were obtained from each subject for
one-third octave noise band center frequencies of 1, 2,
and 4 kHz. RT-I functions were obtained for noise levels
spanning a range from the maximum output levels described previously to levels corresponding to the maximum output level minus 40 dB. An RT-I function was
derived for noise levels representing successive 10-dB
decrements in noise level. Fifteen trials were employed
for each of the five attenuation settings, yielding a total of
75 trials per RT-I function. The attenuation setting on a
given trial was selected randomly by the computer.
Each trial in the reaetion-time paradigm began with
two lights on the response box flashing alternately. This
served as a "ready" signal, informing the subject that the
computer was ready for another trial. The subject, when
ready, initiated a trial by pressing a button on the response box. A randomly varying delay followed the subject's button press. The delay had a minimum value of
500 ms and a maximum value of 4,500 ms. Following this
delay, the noise band was presented at one of the five
intensities described above. The stimulus had a 25-ms
linear rise time. If the subject responded before 1,000 ms
had elapsed since signal onset, the reaction time was
measured using the internal clock of the computer and
was stored for later retrieval. The signal was then terminated (25-ms fall time). If the subject failed to respond by
pressing the button during the 1,000-ms interval following stimulus onset, the trial was terminated and a reaction
time of 1,000 ms was retained. Any responses on a given
trial prior to stimulus onset or after the 1,000-ms response
interval were ignored by the computer. Thus, possible
reaction times ranged 0-1,000 ms. The use of a lengthy
1,000-ms response interval enabled this same program to
be used for a different experiment using threshold-level
signals. For the more intense stimuli used in this experiment, the reaction times never exceeded a value of 567
ms. A new trial was initiated 500 ms following the
308 Journal o f Speech and Hearing Research
27
subject's response. Following completion of 75 trials, the
computer determined the mean reaction times for each of
the five noise levels. RT-I functions were repeated twice
in immediate succession.
Finally, the loudness-growth functions were obtained a
second time in a manner identical to that described
previously. Hence, the 2-hr session was structured as
follows: First, three loudness-growth functions were obtained from the listener, one at each center frequency,
with each making use of 36 stimulus presentations. The
order of center frequency was randomized across subjects. Next, six RT-I functions were obtained, each based
on 75 trials, with two functions obtained in immediate
succession at each of the three center frequencies. Finally, the loudness-growth function was repeated at each
center frequency at the conclusion of the 2-hr session.
RESULTS
The magnitude estimates for the first loudness-growth
function and the reaction times for two replications of the
RT-I functions were averaged. Geometric means plotted
as a function of stimulus level are displayed in Figures 1
and 2 for 1, 2, and 4 kHz. Because the RT-I and loudnessgrowth functions appeared on visual inspection to be
linear on log-log coordinates for most subjects, as shown
in Figures 1 and 2, the slopes of these functions were
calculated for each subject using a least-squares solution
to the following equation: log Y = log b + a log P, where
Y was either the magnitude estimate or the reaction time
at a particular stimulus level expressed in units of sound
pressure (P). Pearson r correlation coefficients b e t w e e n Y
2KHz
1KHz
4KHz
306-310
June 1984
soo
1KHz
2KHz
4KHz
._c
4
SO
0
10
0
0
N O I S E L E V E L in d B A T T E N U A T I O N
40
,
,0
3
~0
0
re: M A X I M U M O U T P U T
FIGURE2. Geometric means of reaction times (RT) for each of the
10 subjects at the three frequencies examined in this study. The
abscissa is labeled identically to that of Figure 1.
and P were also calculated to evaluate the appropriateness of a straight-line fit to these data. The correlation
coefficients and the calculated slopes for each subject are
provided in Tables 1, 2 and 3 for the first set of loudnessgrowth functions, the second set of loudness-growth functions (obtained after the RT-I functions), and the RT-I
functions, respectively. The high correlations (85% > .9)
indicate that it is appropriate to use linear regression
techniques to derive the slopes of both the loudnessgrowth and RT-I functions. In general, the average slopes
of the RT-I functions are much shallower than those for
the loudness-growth functions. The average slope value
for the RT-I functions approximates .06, whereas that for
the loudness-growth functions approaches .30.
Pearson r correlation coefficients were calculated between the slopes for the first loudness-growth function
(Table 1) and the second loudness-growth function (Table 2) at each of the three center frequencies. The
resulting correlation coefficients were .76, .63, and .69 at
1000, 2000 and 4000 Hz, respectively. All are statistically
significant (p < .05). These test-retest correlations enable
one to evaluate the correlations b e t w e e n the slopes of the
first loudness-growth function and the RT-I function. It
TABLE 1. Correlations and slopes for first loudness-growth function.
~'o
b-
i
40
i
SO
NOISE
ZO
10
LEVEL
?
4i
4O
SO
20
i n d(3 A T T E N U A T I O N
TO
O
4=0
re: MAXIMUM
30
,
,o
, I0
;
OUTPUT
FIGURE 1. Geometric means of magnitude estimates of loudness
for each of the I0 subjects at the three frequencies examined in
this study. The abscissa expresses the noise level relative to
maximum output at that frequency, and the numbers positioned
above the zero point indicate the maximum sound pressure level
for that frequency. Thus, the "'0""point on the abscissa of the left
panel coincides with an overall sound pressure level of 94 dB
SPL for the 1-kHz test signal.
I kHz
slope
Subject
r
1
2
3
4
5
6
7
8
9
10
~(SD)
.881
.986
.978
.886
.939
.942
.981
.948
.873
.97s
0,335
0,350
0.332
0.362
0.435
0.402
0.319
0.184
0.352
0.163
0.323
(0.086)
r
.945
.985
.965
.942
.982
.955
.985
.976
.959
.990
2 kHz
slope
0.246
0,549
0.183
0.386
0.331
0.251
0.388
0.202
0.307
0.219
0.306
(0.112)
r
.964
.969
.943
.855
.970
.870
.908
.950
.934
.965
4 kHz
slope
0.241
0.309
0.177
0.175
0.407
0.356
0.357
0.134
0.257
o.204
0.262
(0.092)
HUMES & AHLSTROM: R T and Loudness
TABLE 2. Correlations and slopes for second loudness-growth
function.
1 kHz
slope
Subject
r
1
2
3
4
5
6
7
8
9
10
.975
.966
.886
.982
.962
.964
.976
.977
.962
.908
0.247
0.378
0.362
0.360
0.388
0.239
0.349
0.149
0.289
0.151
0.291
~(SD)
r
2 kHz
slope
.958
.970
.966
.870
.887
.858
.977
.969
.924
.970
0.226
0.413
0.216
0.309
0.316
0.298
0.349
0.187
0.267
0.395
0.302
(0.091)
r
4 kHz
slope
.904
.965
.928
.944
.898
.966
.971
.982
.969
.933
(0.071)
0.256
0.284
0.153
0.168
0.341
0.170
0.286
0.148
0.292
0.238
0.234
(0.069)
would not be reasonable to expect the RT-I function to be
correlated more strongly to loudness-growth functions
than loudness-growth functions are correlated to themselves. The correlations observed between the slopes of
the first loudness-growth function (Table 1) and the RT-I
function (Table 3) were .38, .69 and .69 at 1000, 2000, and
4000 Hz, respectively. The correlations observed for the
data at 2000 and 4000 Hz are statistically significant (p <
.05), whereas that at 1000 Hz is not.
DISCUSSION
The relationship between reaction time and loudness
appears to be somewhat frequency dependent according
to the present results. That is, a much stronger and
statistically significant correlation between these two
measures was observed for test frequencies of 2 and 4
kHz, as opposed to 1 kHz. All correlations observed,
TABLE3. Correlations and slopes for RT data. All values shown
are actually negative.
1 kHz
slope
Subject
r
1
2
3
4
5
6
7
8
9
10
.959
.902
.980
.976
.981
.979
.952
.918
.836
.951
~(SD)
0.056
0.053
0.064
0.056
0.052
0.061
0.078
0.038
0.044
0.048
0.055
(0.011)
r
2 kHz
slope
.955
.988
.975
.932
.891
.980
.964
.962
.988
.999
0.054
0.083
0.044
0.071
0.073
0.066
0.089
0.077
0.063
0.059
0.068
(0.013)
r
4 kHz
slope
.971
.833
.974
.933
.958
.971
.935
.845
.883
.961
0.053
0.042
0.051
0.038
0.075
0.044
0.065
0.023
0.042
0.054
0.049
(0.15)
309
however, were at least of moderate strength and all were
positive. This suggests, therefore, that those individuals
exhibiting the steepest RT-I functions also tend to exhibit
the steepest loudness-growth functions. Again, this is a
much less accurate conclusion for the data obtained at
1000 Hz.
The only comparable data available for comparison are
those obtained by Reason (1972). As noted previously,
Reason observed a correlation of .45 between slopes of
RT-I and loudness-growth functions obtained at 1000 Hz.
Stimuli in his investigation were pure tortes and were
apparently delivered to the subject via earphones. Despite these procedural differences, a correlation of.38 was
observed in the present study at this same test frequency.
Thus, the present data are in excellent agreement with
previous results.
The strength of the correlations between the slopes of
the RT-I functions and the first loudness-growth functions is more impressive, however, when viewed in
relation to the test-retest correlations for the loudnessgrowth functions. At 2000 and 4000 Hz, the relation
between the reaction time and loudness-growth data is at
least as strong as the test-retest correlation of the loudness-growth data. At 1000 Hz, however, the relation
between the slopes of the RT-I and loudness-growth
functions is considerably weaker than the test-retest reliability of the loudness data. The test-retest correlations
observed here for the loudness-growth functions obtained with magnitude estimation procedures are consistent with those reported previously by other investigators
(Logue, 1976; Wanschura & Dawson, 1974).
The mean slopes for the RT-I and loudness-growth
functions, however, are considerably smaller than expected. Loudness-growth functions referenced to sound pressure should have a slope of approximately .6 for the
procedures used in this study according to previous
results (see Marks, 1974, for review), whereas RT-I functions typically yield slopes of.2 (Scharf, 1980). The slopes
observed here were one half to approximately one third
their expected value. The procedures and stimuli used
were not novel. Perhaps the use of a diffuse sound field as
opposed to an anechoic environment or the use of headphones had an influence on the observed slopes. It is not
immediately apparent to the authors, however, why a
diffuse sound field would yield lower exponents. It is
interesting, however, that both the slope for the RT-I and
the loudness-growth functions were lower than estimates
published previously. This can be taken as further evidence for the covariation of" these two functions.
The results from the present investigation suggest that
the RT procedure may be used to provide an indirect
estimate of loudness, especially for signal frequencies >1
kHz. This finding is of considerable importance to animal
researchers who use RT data to estimate loudness. Its
incorporation as a clinical procedure, however, must be
tempered by careful consideration of variables other than
loudness that must apparently contribute to the observed
RT.
310 Journal of Speech and Hearing Research
ACKNOWLEDGMENTS
This research was supported in part by grant 1-R01-OHO0895
from the National Institute of Occupational Safety and Health.
The authors express their gratitude to Anne Marie Tharpe for her
assistance in this project, and to Barb Coulson and Kathy Lambert for typing the manuscript.
REFERENCES
CHOCHOLLE,R. (1940). Variation des temps de reaction auditifs
en fonction de l'intensite a diverses frequences. Annee Psyehologique, 41, 65-124.
CIaOCHOLLE, R., & GREENBAUM,H. B. (1966). La sonic de sons
purs partinetlement masques. Journal de Psychologie, 4, 385414.
DOOLING, R. J., ZOLOTH, S. R., • BAYLIS,J. R. (1978). Auditory
sensitivity, equal loudness, temporal resolving power, and
vocalizations in the House Finch (Carpodacus Mexicanus).
Journal of Comparative and Physiological Psychology, 92,
867-876.
LocuE, A. W. (1976). Individual differences in magnitude estimation of loudness. Perception & Psychophysics, 19, 279-280.
Luz, G. A., & NUGEN, R. D. (1972). Auditory discrimination in
the chinchilla: I. Reaction time. U,S. Army Medical Research
Laboratory Report No. 967. Fort Knox, KY, Department of the
Army.
MARKS,L. E. (1974). On scales of sensation: Prolegomena to any
future psychophysies that will be able to come forth as science.
Perception & Psyehophysics, 16, 358-376.
MILLER, J. M., MOODY, D. B., & STEBBINS,W. C. (1969). Evoked
27 306-310
J u n e 1984
potentials and auditory reaction time in monkeys. Sciene~
163, 592-594.
MOODY, D. B. (1973). Behavioral studies of noise-induced hearing loss in primates: Loudness recruitment. Advances in
Otorhinolaryngology, 20, 82-101.
MOODY, D. B. (1979). Partial masking in the monkey: Effect of
masker bandwidth and masker band spacing. Journal of the
Acoustical Society of America, 66, 107-114.
PFINGST, B. F., HIENZ, R., KIMM,J., & MILLER, J. (1975). Reaction-time procedure for measurement of hearing. I. Suprathreshold functions.Journal of the Acoustical Society of America, 57, 421-430.
REASON',J. T. (1972). Some correlates of the loudness function.
Journal of Sound and Vibration, 20, 305-309.
SANDLIN, D., & GERKEN, G. M. (1977). Auditory reaction time
and absolute threshold in eat.Journal of the Acoustical Society
of America, 61,602-607.
SCHAtlF, B. (1980). Loudness. In E. C. Carterette & M. P. Friedman (Eds.), Handbook of perception (Vol. 4), Hearing. New
York: Academic Press.
STEBmNS, W. C. (1966). Auditory reaction time and the derivation of equal loudness contours for the monkey.Journal of the
Experimental Analysis of Behavior, 9, 135-142.
STEVENS, S. S. (1975). Psyehophysics. New York: Wiley.
WANSCHURA,R. G., & DAWSON,W. E. (1974). Regression effect
and individual power functions over sessions. Journal of
Experimental Psychology, 102, 806-812.
Received October 12, 1982
Accepted September 14, 1983
Requests for reprints should be sent to Larry E. Humes,
Division of Hearing and Speech Sciences, Vanderbilt University
School of Medicine, Nashville, TN 37232.