ABSTRACT

AUDITORY LOCALIZATION IN PRIMATES:
THE ROLE OF STIMULUS BANDWIDTH

BY

Charles Hawkins Brown

A fundamental characteristic of an acoustic event is its location
with regard to the observer. The acoustic locus dictates the direction
for visual orientation. The ability to accurately determine the
location of a sound may play a fundamental role in predator avoidance,
in mate location, and in other biologically significant activifies.

The objective of this research was to psychophysically measure the
ability of a nonhuman primate to detect changes in the azimuth, or
horizontal coordinate, of high frequency sounds.

TWO species of Old WOrld Monkeys, Macaca mulatta and M. nemestrina,

 

 

were trained by operant conditioning techniques to report the change

in location of a sound in space. When the monkey made contact with

the response disk an acoustical stimulus was pulsed repetitively from
the standard location (zero degrees azimuth). After the presentation
of a variable number of pulses the stimulus changed position in azimuth
from the standard to one of several comparison locations. If the
monkey reported this change in acoustical location by releasing the
response disk within two sec, it received a food reinforcer. Thresholds

for the minimum discriminable change in locus of a sound in space were

Brown
determined psychophysically by the method of constant stimuli.
Threshold, or the 50% detectability locus, was determined by two
methods, linear interpolation and probit analysis. Testing was
conducted under free- field conditions in an anechoic chamber. The
monkeys were tested with pure tones and with bands of noise 250, 500,
1000, 2000, 4000, 8000 Hz wide, geometrically centered at 8000,
11200, 16000 Hz; and with a pure tone and noise bands 250, 500, 8000
Hz wide centered at 13454 Hz. The minimum detectable change in
location was a function of the bandwidth of the stimulus and ranged
from 3° to greater than 20°. Narrow bandwidth stimuli were harder to
localize than wide bandwidth stimuli, and threshold for the detection
of a change in azimuth was found to be inversely related to the
logarithm of stimulus bandwidth. In eight out of nine cases the
correlation coefficient of the regression of threshold on the
logarithm of stimulus bandwidth exceeded .90. This indicates that
stimulus bandwidth accounts for much of the variability in threshold.
The slope of the regression of threshold on the logarithm of stimulus
bandwidth was approximately -6, -11, and -4 for stimuli centered at
8000, 11200, and 16000 Hz respectively. The results indicate that
the stimuli most difficult to localize were pure tones and narrow band-
width stimuli centered at 11200 or 13454 Hz. These data demonstrate
that the macaque's ability to localize sounds in space is dependent
upon the acoustical composition of the signal; as the bandwidth of the
signal was increased, threshold for the detection of a change in

acoustic space decreased.

AUDITORY LOCALIZATION IN PRIMATES:
THE ROLE OF STIMULUS BANDWIDTH

BY

Charles Hawkins Brown

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Department of Psychology

1976

ACKNOWLEDGMENTS

This dissertaion evolved from what initially appeared to be an
academic digression which arose nearly three years age when I was
encouraged by Drs. Mark E. Rilling and James L. Zacks of the Depart-
ment of Psychology, Michigan State University to take advantage of
the CIC Traveling Scholar program to spend a year in residence in the
study of animal psychoacoustics in the laboratory of Dr. William C.
Stebbins, Kresge Hearing Research Institute, University of Michigan.
The research initiated at that time resulted in these two universities
forming an academic partnership, a development upon which this
dissertation was wholly dependent. Inter-institutional cooperation
of this nature is uncommon, and I wish to identify here and celebrate
those uncommon individuals who were the authors of this partnership.

I wish to acknowledge Dr. Rudy A. Bernard, Department of Physiology;
and Drs. Glenn I. Hatton, Mark E. Rilling, David Wessel, and James L.
Zacks, Department of Psychology, Michigan State University who served
on my guidance and dissertation committees. I deeply appreciate the
support and guidance from my major advisor, Dr. James L. Zacks. More
than any other individual, Jim was responsible for introducing me and
kindling in me a taste for the art of critical appraisal. My work has
been greatly dependent on Dr. Michael D. Beecher, Department of
Psychology, Eastern Michigan University. Mike has most directly invested

in the design and execution of this research and is most directly

ii

responsible for my education in auditory localization.

I am indebted to Dr. William C. Stebbins for cultivating in me
an insistence that my work be conducted in appreciation of its
biological and evolutionary context. I owe most to Bill for my
maturation as a scientist, and without his academic and administrative
sponsorship this dissertation would never have been possible.

I also appreciate the guidance and assistance offered by Dr.
David B. Moody, Kresge Hearing Research Institute, University of
Michigan. Dave is largely responsible for the solution to countless
apparatus and instrumentation problems incurred by this research.

I would also like to thank Carol Magoon, Celest Kok, Debbie Olsen,
Gini Struich, and Judy Nowack for their assistance on a variety of
tasks including serving as human subjects, running monkeys, and in the
preparation of this manuscript.

I am indebted to the staff of Kresge Hearing Research Institute,
University of Michigan, and to support from The National Institute of
Mental Health as a predoctoral trainee, awarded to the Department
of Psychology, Michigan State University, and from the National
Institute of Health NS 05077; NS 05785, and the National Science
Foundation B MS74-20050 which made this research possible.

And finally I would like to celebrate my able simian collaborators,
Sidney, Miko, and Oscar who taught me everything I know about their

ears.

iii

TABLE OF CONTENTS
LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . V
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . Vii
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . l

EXPERIMENT I. The Detection of Changes in Azimuth as a Function
of Stimulus Bandwidth . . . . . . . . . . . . . . . . . . . . 24

Method
Results

EXPERIMENT II. The Detection of Changes in Azimuth as a Function
of Stimulus Bandwidth: Systematic Replication. . . . . . . . . 111

Method
Results

DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . . . 129
APPENDIX A. Acoustical Calibration . . . . . . . . . . . . . 148
APPENDIX B. The Proportion of Trials Detected at Each Azimuth 155

REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . 164

iv

Table

10.

11.

12.

13.

14.

LIST OF TABLES

Summary of Filter and Oscillator Parameters
for the Noise Band Stimuli . . . . . . . . . . .

Speaker Locations With Respect to the Zero
Azimuth. . . . . . . . . . . . . . . . . . . . .

Threshold Angles for Bandwidths Centered at
8000 Hz. . . . . . . . . . . . . . . . . . . . .

Test for Speaker Differences: Estimated
Proportions for Two Bandwidths . . . . . . . . .

The Regression of Threshold on Bandwidth
Centered at 8000 Hz. . . . . . . . . . . . . . .

Threshold Angles for Bandwidths Centered
at 11200 Hz. 0 O O O O O O O O O O O O O O O O O

The Regression of Threshold on Bandwidth
Centered at 11200 Hz . . . . . . . . . . . . . .

Threshold Angles for Bandwidths Centered
at 16000 Hz. . . . . . . . . . . . . . . . . . .

The Regression of Threshold on Bandwidth
Centered at 16000 Hz . . . . . . . . . . . . . .

Threshold Angles for Replications at 11200 Hz. .

Threshold Angles for Bandwidths Centered
at 13454 Hz. . . . . . . . . . . . . . . . . . .

Binaural Difference Spectrum Efficiency Quotient

Ratios of Correct to Total Responses for
Bandwidths Centered at 8000 Hz . . . . . . . . .

Ratios of Correct to Tbtal Responses for
Bandwidths Centered at 11200 Hz. . . . . . . . .

Page

33

44

53

68

71

79

90

107

110

117

121

146

156

158

15.

16.

17.

Ratios of Correct to Total Responses for
Bandwidths Centered at 16000 Hz. . . . . . . . . . . . . . . 160

Ratios of Correct to Total Responses for
Bandwidths Centered at 11200 Hz Replication. . . . . . . . . 162

Ratios of Correct to Total Responses for
Bandwidths Centered at 13454 Hz. . . . . . . . . . . . . . . 163

vi

10.

11.

12.

LIST OF FIGURES

The Binaural Distance Difference for a Distant
Sound Source Located at Azimuth d3. . . . . . . . . . . .

Inverse Relationship Between the Availability of

Large Binaural Time Differences and the Availability

of Large Binaural Spectrum Differences in 15 Varieties
of Mammals. . . . . . . . . . . . . . . . . . . . . . . .

Sound Shadows Thrown at Representative Frequencies by
a Loudspeaker Rotating Around the Artificial Head Without

Pinna and With Either Large or Small Pinna. . . . . . . .

Threshold of Hearing Function from 60 Hz to 45 kHz for
Cercopithecinae Based on Data Obtained from Four Genera .

Subject in Primate Chair Positioned.in the Anechoic
Chambe r O O O O O O O O O O O O O O O C O O O O O O O O O

The Location of the Observer With Respect to the
Speaker Array Mounted on the Arc. . . . . . . . . . . . .

The Audio Equipment Used for the Generation of
Pure Tones. . . . . . . . . . . . . . . . . . . . . . . .

The Audio Equipment Used for the Generation of
Noise Bands . . . . . . . . . . . . . . . . . . . . . . .

State Diagram of the Testing Procedure. . . . . . . . . .

The Probability of Detection as a Function of Azimuth
for the 8000 Hz Pure Tone . . . . . . . . . . . . . . . .

The Probability of Detection as a Function of Azimuth
for the 250 Hz Stimulus Bandwidth Centered at 8000 Hz . .

The Probability of Detection as a Function of Azimuth
for the 500 Hz Stimulus Bandwidth Centered at 8000 Hz . .

vii

Page

14

17

22

27

29

32

37

43

52

56

58

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

The Probability of Detection as a Function of Azimuth
for the 1000 Hz Stimulus Bandwidth Centered at 8000 Hz. .

The Probability of Detection as a Function of Azimuth
for the 2000 Hz Stimulus Bandwidth Centered at 8000 Hz. .

The Probability of Detection as a Function of Azimuth
for the 4000 Hz Stimulus Bandwidth Centered at 8000 Hz. .

The Probability of Detection as a Function of Azimuth
for the 8000 Hz Stimulus Bandwidth Centered at 8000 Hz. .

Threshold as a Function of Bandwidth for Stimuli Centered
at 8000 Hz 0 C I O O O O O O O O O O O O O O O O O O O O O

The Probability of Detection as a Function of Azimuth
for the 11200 Hz Pure Tone. . . . . . . . . . . . . . . .

The Probability of Detection as a Function of Azimuth
for the 250 Hz Stimulus Bandwidth Centered at 11200 Hz. .

The Probability of Detection as a Function of Azimuth
for the 500 Hz Stimulus Bandwidth Centered at 11200 Hz. .

The Probability of Detection as a Function of Azimuth
for the 1000 Hz Stimulus Bandwidth Centered at 11200 Hz .

The Probability of Detection as a Function of Azimuth
for the 2000 Hz Stimulus Bandwidth Centered at 11200 Hz .

The Probability of Detection as a Function of Azimuth
for the 4000 Hz Stimulus Bandwidth Centered at 11200 Hz .

The Probability of Detection as a Function of Azimuth
for the 8000 Hz Stimulus Bandwidth Centered at 11200 Hz .

Threshold as a Function of Bandwidth for Stimuli
Centered at 11200 Hz. . . . . . . . . . . . . . . . . . .

The Probability of Detection as a Function of Azimuth
for the 16000 Hz Pure Tbne. . . . . . . . . . . . . . . .

The Probability of Detection as a Function of Azimuth
for the 250 Hz Stimulus Bandwidth Centered at 16000 Hz. .

The Probability of Detection as a Function of Azimuth
for the 500 Hz Stimulus Bandwidth Centered at 16000 Hz.

viii

6O

62

64

66

7O

73

76

78

81

83

85

87

89

93

95

97

Introduction

Biological Transducers and the Consideration of the Acoustic Event

A complete account of the mechanisms by which an organism
determines the spatial origin of a sound must begin with an appre-
ciation of the characteristics of acoustical phenomena to which the
ear is responsive. An acoustical event generates vibrations in a
medium; this disturbance has two components -- displacement and
pressure. These components differ in their characteristics of propa-
gation in specific media. In water it is likely that the displacement
component of sound is relevant for localization by aquatic vertebrates;
however, terrestrial vertebrates are not sensitive to this component
(van Bergeijk, 1967). Although Pumphrey (1940) has noted that many
insects have acoustic receptors which are sensitive to the displacement
component of sound, all terrestrial vertebrates respond solely to the
pressure component. The greatest sensitivity to the pressure component
of sound is realized by the class mammalia (Manley, 1973; Shaw, 1974;
Henson, 1974).

The component of sound to which a species responds has implications
for its ability to locate the origin of an acoustic event. Displacement
is a vector quantity. A receptor sensitive to this vector would be
ideal for determining the location of a sound because it would respond
to changes in the angle of incidence of the acoustical wave front.
Acoustical pressure, on the other hand, is a scalar quantity, and the

1

2
ear of terrestrial vertebrates is insensitive to those changes in
the wave front. Consequently, terrestrial vertebrates cannot
determine the spatial location of a sound through peripheral
mechanism; rather, they must possess central mechanisms which either
monaurally or binaurally resolve the origin of an acoustic event.
The classical research has focused on the putative binaural mechanisms

of auditory localization.

Binaural Differences

 

Binaural differences are the result of two factors -- 1) the
difference in distance (.Ad) the sound must travel to each of the
observer's ears, and 2) asymmetries in the transmission of the
frequency and amplitude of the acoustical signal. The first factor
results in differences in the time of arrival and in the phase of the
signal at each ear. The velocity of sound in air is 343 meters per
sec; consequently, for each additional cm the sound must travel to
reach the far ear, it will arrive 29 [csec later than it will to the
near ear. The second factor results in binaural intensity differences.
Interaural intensity differences can be explained with reference to the
characteristics of acoustical waves, including reflection, refraction,
and diffraction with bodies of a different density or acoustical
impedance. Acoustical reflection is dependent on both the wavelength
of the sound and on the size of the reflector. Binaural intensity
differences are realized with high frequency sounds when the corres-
ponding wavelength is short with respect to the diameter of the obser-
ver's head. The longer wavelengths of lower frequency sounds act as if

they "bend around" the observer's head producing minimal binaural

intensity differences. In this context it sould be apparent that
species with small heads require higher frequency sounds for binaural

intensity differences to be realized.

Geometrical Considerations of Binaural Phenomena

 

Lord Rayleigh (1876; 1945) was the first to propose that the head
be treated as an acoustically opaque spherical object with ears
diametrically opposed. This geometrical simplification is still con-
sidered reasonably precise for audio frequencies below 1000 Hz.
However, for frequencies greater than 1000 Hz this visualization
becomes unsatisfactory because characteristics of the external ear and
ear canal may cause prominent departures from the predictions which
follow from this model (Shaw, 1974).

In this tradition the classical description of binaural differences
was proposed by WOodworth (1938). This description assumes that the
observer's head is immobile and unable to scan the sound field.
Hewever, the duration of most acoustical transients is so brief that
scanning is impossible, and consequently head immobility is not
physiologically unreasonable. Figure l is a construction of the
horizontal plane of an observer for a distant sound source. At distances
from the sound source greater than one meter the wave front may be
approximated by the surface of a plane. Binaural distance differences
occur for all sound locations other than those which lie in the median
plane. As may be seen in Figure 1, for a sound source to the right of
the observer at azimuth «t , the extra distance that the sound must
travel to reach the left ear is given by the sum of the linear distance

r(sin I: ) and the curvilinear distance r(cr ). That is, the distance

4
difference for the two ears is given by the expression
Ad=r(0<+sinoC)
where; .Ad.is the distance difference in
centimeters, r is the radius of the
observer's head in centimeters, and
angle a: is in radians.

The difference in distance to the two ears is acoustically realized
as a binaural difference in the temporal and/or phase domain. Temporal
differences ( A.t) are calculated by dividing the distance difference
by the velocity of sound. If the velocity of sound in air is 343
meters per sec, then the relationship between .4 t and azimuth is

given by the expression

At=r(0( +sina()
3.43 x 10'2

 

where; .At is the temporal difference in
sec, r is the radius of the obser-
ver's head in centimeters, and the
angle 0C is in radians.

In this context two implications merit emphasis. First, for any
given azimuth 0C , Ant varies directly with r. As a result, observers
with large heads will experience a greater binaural temporal difference
than will observers with small heads. Consequently if the neural
ability to resolve binaural temporal differences is approximately equal
across mammalian species, then those organisms with large heads will
be able to discriminate finer changes in azimuth than will organisms
with small heads. If this is true then one may expect to find that
small mammals have exploited alternative mechanisms for localization.
Second, binaural temporal differences do not define a specific locus

in three dimensional space. Rather it is the case that binaural

temporal differences are a function of the binaural distance difference

Figure l

The Binaural Distance Difference for a Distant Sound Source Located
- at Azimuth tfi'. The construction illustrates the geometry of the
formula Ad = r( oc + sin oC. ) . Adopted from Woodworth (1938) .

Zero azimuth

  
 
  

Wave front from distant source
at azimuth at

 

 

r(sinod ’ '
r(od— ‘
a: a

Left 1 \
Ear Rf—J Ear

 

 

Figure l

7
and the same distance difference may be described by a circle of

identity which is normal to and centered on the interaural axis.

Locational Ambiguity

 

Binaural differences in the time of arrival may alternatively be
expressed as binaural differences in phase. The maximum binaural
temporal difference occurs when the acoustic source is at either the
900 or 2700 azimuth. Assuming r = 8.75 cm, the usual value assigned
to man, then maximum zit = 656 /Ksec. That is, at 900 the sound
arrives at the far ear 656 /usec after it arrives at the near ear.
This time lag in arrival at the far ear would result in phase differences
between ears of approximately 900, 1800, and 3600 for pure tones of
380 Hz, 760 Hz, and 1520 Hz respectively. Note that the phase
difference information for a given azimuth can be determined only if
the frequency of the sound is also known; the binaural time difference
is constant for all frequencies at a given azimuth, but the binaural
phase difference is frequency dependent.

The frequency range over which binaural phase differences may be
utilized for sound localization is restricted. More than one location
may produce the same difference in binaural phase when the period of
the impinging sinusoid is equal to or less than twice the maximum
binaural temporal difference. In the idealized case for man, such
locational ambiguities will occur for frequencies with periods less
than or equal to 1312 /xsec. For example, a 760 Hz stimulus located
at the 900 azimuth (i.e. 900 to the right) will produce a binaural
phase difference of 1800; this same difference will be produced by

the stimulus at the 2700 azimuth (i.e. 900 to the left). Thus

binaural phase information alone will not discriminate between these
two locations. Similarly, for all frequencies greater than 760 Hz, the
binaural phase difference for a source at any given azimuth will be
perfectly matched by at least one other azimuth. This ambiguity in
location suggests that phase information can be utilized for sound
localization only with low frequency tones. It should be emphasized,
however, that the stimulus frequency beyond which acoustical locations
are ambiguous is higher for species with smaller heads. For

instance, a small rodent with a maximum (it of 100 /usec will be
confronted with ambiguous locations only for frequencies above 5000 Hz.
For many of the smaller mammals the geometrically defined frequency
limit for unambiguous localization may not, in fact, be fully realized.
A lower limit may be determined physiologically by the refractory
period of auditory neurons and by the limitations of the auditory system

to code and preserve the binaural phase disparity.

The Duplicity Theory of the Localization of Azimuth

The vertebrate ear is confronted by two separate binaural dis-
parities -- interaural differences in phase, or time, and interaural
differences in intensity. The change in azimuth of a sound source
always results in a change in the former and often results in changes
in both dimensions. The relative contribution of either dimension with
respect to the determination of azimuth becomes an empirical matter.
Psychophysical observations relevant to this question have largely
been confined to man; as a result the summary which follows may not
' generalize to all members of the mammalian community.

Stevens and Newman (1934) were the first, with modern techniques,

9
to empirically explore the ability of human observers to localize pure
tones. They positioned their observers on a platform located on the
roof of Harvard University's Biological Laboratory. The stimuli were
presented from a speaker attached to a 12 foot boom which rotated around
the observer in the horizontal plane. The subjects were required to
judge the location of a tone presented in the right hemisphere. Their
_ data were collapsed across azimuths and across observers, yielding
a mean error of localization of about 110 for frequencies below
1000 Hz and above 7000 Hz. The mean error of localization increased to
about 200 at 3000 Hz. Stevens and Newman interpreted these data as sup-
port for a duplicity theory of auditory localization. That is, they
argued that both binaural phase differences and binaural intensity
differences were utilized by the observers in locating the sound source,
the former for low frequencies and the latter for high frequencies.
The degradation in spatial acuity at 3000 Hz was regarded to mark the
physiological boundary between these two mechanisms. Performance
suffered at the boundary because the binaural differences generated
at this frequency in either domain were marginal.

The Stevens and Newman study remained the definitive work on
auditory localization for twenty years until Mills (1958) reexamined
the problem. Mills, one of Stevens's students, conducted his study in
an anechoic chamber and required his observers to report the just
detectable displacement in location of a sound source for changes in
azimuth across the first quadrant locations from straight ahead to
900 to the right. The procedural and acoustical refinements intro-
duced by Mills generated a view of auditory localization strikingly

different on three counts. First, localization was specified

10

in terms of the minimum audible angle -- that is, in terms of the
angle at which the observer reliably reported that the comparison
tone was right or left of the standard tone for the azimuth in
question. The observer's response indicated his ability to resolve
acoustic differences produced by a change in azimuth, rather than

his ability to identify the spatial location on the stimulus.

Second, different azimuths produced different minimum angles. The
ability to resolve locations in space is, therefore, azimuth as

well as frequency dependent. Third, man's ability to resolve changes
in acoustic location is more variable than Stevens and Newman's (1934)
work had suggested. Mills reported that at the zero azimuth man can
resolve a change in location of l0 or less for low frequency sounds
while at a 600 azimuth the minimal audible angle for some frequencies
is indeterminably large. More recently Harris (1972) has presented
data in which pure tone localization did not deteriorate with changes
in azimuth. The reasons for these discrepant results are not clear.
However, Harris employed transducers which were separated by approx—
imately 100 in elevation as well as by their displacement in azimuth.
The difference in elevation may have generated intensity differences
which by themselves were not detectable but which interacted with
horizontal displacements to facilitate detection.

Mills invoked the duplicity hypothesis to account for his data,
considering binaural phase differences instrumental in low frequency
localization and binaural intensity differences instrumental in high
frequency localization. To test the adequacy of this explanation, he
conducted a study employing dichotically presented stimuli (Mills, 1960).

He required his subjects to report the just detectable differences in

11

binaural phase and intensity for pure tones across the frequency
spectrum. His research strategy assumed that phase should be considered
the relevant cue in localization for those frequencies where the
threshold function for dichotic phase difference was found to coincide
with the threshold function for the localization of real sources.
Conversely, at those frequencies where threshold functions for

dichotic intensity difference and the localization of real sources
agreed, binaural intensity difference should be considered the relevant
localization cue. The results indicated that for frequencies below
1500 Hz binaural phase differences best fit the data, while for
frequencies between 1500 and 6000 Hz, intensity differences agreed

well with the localization of real sources. However, neither dimension
accounted for the localization of real source stimuli for frequencies
greater than 6000 Hz. While Mills's curve fitting approach was not
wholly assumption free, his study remains the firmest test of

duplicity theory.

Additional support for the duplicity hypothesis of the localization
of azimuth in mammals has come recently from studies on the localization
of pure tones in cat (Casseday and Neff, 1973) and in sea lion (Moore
and Au, 1975). In both cases, experimental results could be explained
by the operation of two localization mechanisms -- one for low frequen-
cies and the other for high frequencies. As a result, duplicity
theory is generally accepted as characterizing the mechanisms of

auditory localization in mammals.

12

Auditory Localization in an Evolutionary Perspective

 

Masterton has argued that mammalian hearing reveals the operation
of selection for accurate localization mechanisms (Masterton, Heffner,
and Ravizza, 1969; Masterton and Diamond, 1973; Masterton, 1974). This
evolutionary perspective is founded upon the proposition that selection
has operated to favor the individual endowed with the ability to
acoustically orient to biologically significant sounds. According to
this position the early detection of the locus of other animals through
acoustic mechanisms would contribute to an individual's reproductive
advantage by alerting and preparing him to fight, flee, mate, etc., as
appropriate.

Masterton contends that selection for a superior spatial sense
has resulted in an expanded auditory sensitivity range in mammals. The
primary evidence for his position is the correlation between a species's
interaural distance, or max (it, and its high frequency cutoff (Master—
ton, 1974). As seen in Figure 2, mammals with smaller heads are able
to hear higher frequency sounds. The correlation supports the selection
hypothesis within the context of the duplicity theory of auditory
localization. As noted above, species with small heads suffer two
disadvantages -- 1) the range for binaural temporal differences is
restricted, and 2) they experience binaural intensity differences only
for high frequency sounds. Therefore, if an organism with a small head
is to take advantage of binaural differences to locate sounds, it must
possess an expanded sensitivity to the higher frequencies where binaural
differences are produced. Masterton's correlation show that this

expectation is indeed realized.

13

Figure 2

Inverse Relationship Between the Availability of Large Binaural Time
Differences and the Availability of Large Binaural Spectrum
Differences in 15 Varieties of Mammals. Each number represents the
maximum time difference and the upper limit of hearing at 70 dB
SPL for one species: 1) opossum, 2) hedgehog, 5) slow loris,

6) potto, 10) macaque, ll) chimpanzee, 12) man, 13) bat (Myotis),
14) bat (Eptesicus), 15) rabbit, 16) wild house mouse, 17) cotton
rat, 18) guinea pig, 26) bottle-nosed dolphin. Mammals known

to echo locate are circled. Data from several published and
unpublished sources, Masterton (1974). Reproduced by

permission of author.

HIGH - FREQUENCY CUT —OFF

(IN kHz)

I60

80

4O

20

 

14

 

Figure 2

I- @
(9
l6 3 l
_. l7
2 I5
'8\5 IO
6\
Il\
I2
I l l l l
50 . I00 200 400 800
MAXIMUM At (in microsec)

15

The Localization of Complex Sounds

 

As has been reported here the classical investigation of
auditory localization employed pure tones for stimuli. Pure tones have
the advantage that one may characterize localization acuity as a
function of frequency. However, some problems arise regarding the
localization of pure tones. At high frequencies the binaural intensity
difference of tonal stimuli is not necessarily monotonically related
to position in space. Thus it is possible to have two or more
loci with the same binaural intensity difference (Harris, 1972;
Harrison and Downey, 1970; Sivian and White, 1933). In addition, the
intensity dimension may appear to misbehave to the extent that a
stimulus to one side of the observer may actually be more intense at
the far ear and consequently appear to originate from the far side.
This phenomenon is illustrated in Figure 3. This figure represents
intensity measurements obtained from an artificial head by Harris (1972).
Those points in the polar coordinates at which the observed binaural
intensity difference lie within the radius of the circle denote loci
where the sound intensity for the ear near the source was less than
the intensity recorded for the ear farther from the source. In short,
the tonal stimulus may appear to behave capriciously, and this behavior
may be an expression of slight architectural differences between
different individuals' heads, external ears, etc.

Two issues merit clarification. First, while a single frequency
with reference to several points in space may appear to misbehave,
adjacent frequencies in the audio spectra are unlikely to do so; this
is because the wavelengths of different frequencies would require dif-

ferent spatial coordinates to reproduce a similar set of standing

16

Figure 3

Sound Shadows Thrown at Representative Frequencies by a Loudspeaker
Rotating Around the Artificial Head Without Pinna and With Either
Large or Small Pinna. At loci where the observed measurement

lies within the radius of the polar coordinates the sound

pressure level recorded at the distant ear was greater than the
sound pressure level observed at the near ear. Harris (1972).
Reproduced by permission of the author.

17

(T’ELEVMJKNQ

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3

18
waves, etc., leading to the intensity inversion. Second, pure tones
are not naturally occuring stimuli and the vertebrate auditory system
is probably not adapted to localize them. Rather, it is the case that
both biologically and non-biologically produced sounds which occur
in nature are composed of a family of frequencies of a given
spectral extent; and to a first approximation, natural stimuli
may be described by their dominant frequency and bandwidth (Rowell
and Hinds, 1963; see their description of rhesus vocalizations). This
essentially is the level of analysis offered by a sonograph of an
acoustic event. The pitch of a sound is dependent on the frequency
bandwidth greatest energy (dominant frequency) and its location of the
frequency axis. High pitched sounds have a high dominant frequency
and vice versa. Sounds also lie on a continuum from very tonal to
very noisy signals. Tonality is an expression of bandwidth; signals
composed of a narrow range of frequencies are tonal, while broad hand
signals are noisy. Thus in the context of bioacoustics the emerging
question for auditory localization becomes what is the requisite
bandwidth of a stimulus at different points on the frequency spectrum

for an observer to unambiguously resolve its spatial origin?

The Bioacoustical Context of Auditory Localization

 

Marler (1955; 1957) has suggested that species specific vocal-
izations may be differentially locatable according to the frequency
content and bandwidth of the call. Marler (1965) has observed that
throughout the class mammalia alarm calls are characteristically
shrill narrow band barks while screeching, screaming vocalizations

are universally emitted under severe distress. A similar parallel

19

may be observed in the vocal repertoire of many avian species. In this
context the proposition arises that if the acoustical characteristics
which determine the ability with which a call may be localized are
generally held in common by members of the biocommunity (i.e. their
auditory systems are basically similar) then species specific vocal-
izations should possess the acoustical characteristics which make
a call more or less localizable as may be appropriate in the social
context. For example, one may predict that alarm calls should
yield minimal information regarding the location of the vocalizer,
while calls emitted to solicit contact with conspecifics may exhibit
acoustical characteristics optimal for localizing the caller.
Consequently the more generalized case may be described by exploring
the ability of organisms to localize bands of noise of various
bandwidths at different points within the frequency spectrum.

In this regard limited comparisons are available. Casseday and
Neff (1973) reported lower minimum audible angles in cats with broad
bands of noise than with pure tones. In monkeys, Heffner (1973)
found localization thresholds were lower for clicks than for chords
composed of 0.5 and 4 kHz tones. Konishi (1972) found that barn
owls would strike at an acoustic prey with greater accuracy for broad
bands of noise than for pure tones. Nordlund (1962; 1963) has reproted
that humans exhibit a smaller error of estimation of acoustic locus
for low pass noise than for pure tones. Additionally several studies
of human localization in the median plane, or with monaural observers
indicate that the stimulus must contain certain chunks of the auditory
spectrum for subjects to exhibit some spatial sense (Angell and Fite,

1901; Batteau, 1967; Belendiuk and Butler, 1975; Butler, 1969; Butler

20

and Planert, 1975; Hebrank and Wright, 1974a; 1974b; Roffler and

Butler, 1968a; Wright, Hebrank, and Wilson, 1974).

Research Objectives

 

The objective of this research is to psychophysically characterize

the abilities of two species of nonhuman primates, Macaca mulatta and

 

M. nemestrina, to detect changes in the azimuth of narrow and wide

 

bandwidth high frequency signals. This objective serves two ancillary
issues. One, it helps place the human ability to acoustically localize
in a phylogenetic context. Two, it helps characterize the contribution
of high frequency sensitivity to auditory localization. With respect
to these issues, members of the genus Macaca are appropriate subjects.
Their phyletic heritage is sufficiently similar to man's that for
descriptive purposes their aural abilities serve as a model of the
prehominoid condition. Perhaps more important, however, it is far
easier to recognize the selective pressures presently in operation
on nonhuman primates than on man. Consequently propositions regarding
the ability with which species specific vocalizations may be localized
are testable. In addition Masterton, Heffner, and Ravizza (1969)
have observed that man's upper limit of hearing on the frequency axis
is low with respect to the class mammalia. Thus the expanded sensitivity
range possessed by Macaca may provide a more generalized case for
describing the contribution of high frequency sensitivity to auditory
localization.

Stebbins (1973) has fully described the threshold of hearing
function for the macaques; its basic features are shared by all members

of the subfamily of Old World monkeys, Cercopithecinae. As may be seen

Figure 4

Threshold of Hearing Function from 60 Hz to 45 kHz for Cercopithicinae
Based on Data Obtained from Four Genera. Stebbins (1973).
Reproduced by permission of the author.

5
o

m
0

O)
C)

b
O

SOUND PRESSURE (dB re 820 FPO)
N
o

22

 

 

Cercopithecinae

 

l i l

 

I 2 4
FREQUENCY (kHz)

Figure 4

 

23
in Figure 4, hearing in the macaques ranges from 60 Hz to an upper
limit of 45000 Hz. The upper limb of the audiogram shows a peak in
sensitivity at 8000 Hz followed by a rapidly ascending threshold of
hearing function. The research presented here is directed at
characterizing the ability of the macaque to localize changes in
azimuth or the horizontal plane for sound in the region of its
greatest sensitivity to high frequency stimuli -— 5000 Hz to
20000 Hz. This two octave span contains relatively little energy
from the macaque's repertoire of vocalization, thus it participates
little in communication in this genus (Rowell and Hind, 1962;
Grimm, 1967). The highest octave of the macaque's hearing is not
explored here. The macaque is 40 to 80 dB less sensitive to sounds in
this region; and in addition these high frequency stimuli are more
readily attenuated by the environment. Consequently many sounds in
the upper end of the audio spectrum may have insufficient energy

to participate in localization.

Experiment
The Detection of Changes in Azimuth as a Function of Stimulus Bandwidth

Method
Subjects
Three monkeys of the genus Macaca served as subjects. Two species

in this genus were represented, M. mulatta (rhesus) and M. nemestrina

 

 

(pigtail). Miko, a mature adult female pig-tailed macaque was
maintained at an experimental weight of 4.2 to 5.0 kg throughout the
course of the study. Sidney, a juvenile male pig-tailed macaque,
weighed from 3.4 to 4.1 kg. Oscar, a juvenile male rhesus, weighed
from 3.0 to 4.3 kg during the course of experimentation. All subjects
were wild caught and had been acclimated to laboratory conditions for
at least 9 months prior to the commencement of testing.

At the initiation of training Miko was experimentally naive,
while Sidney and Oscar had previously served as subjects in an
audiometric procedure designed to behaviorally measure the threshold
of hearing for nonhuman observers (Stebbins, 1975). The threshold
of hearing function (audiogram) for these two observers was well
within the normal range as reported for the subfamily of Old WOrld
monkeys Cercopithecinae (Stebbins, 1973).

The subjects were maintained under food-restricted conditions.
The bulk of their diet was provided by 190 mg whole banana flavored
pellets (Noyes) which were earned in 1-2 hour daily experimental

sessions. The remainder of their caloric ration was comprised of
24

25

Purina monkey chow and fresh fruit; the latter of which was usually
available three times weekly.

Subjects were individually housed in stainless steel primate
cages located in a large 500 sq ft colony room. Water was continuously
available in the home cage. The colony was maintained on a 12/12

light/dark cycle.

Apparatus

Subjects were seated in a primate chair which had been designed
to reduce acoustic reflections. The chair was positioned on a
pedestal in a 9 by 12 foot anechoic room. As may be seen in the
close—up of the subject in Figure 5, the subject's head was held
in a fixed position in the sound field by a muzzle restraint. The
muzzle restraint and chair pedestal were employed to reduce the
variability in the position of the observer in the acoustical environ-
ment within and between sessions. The error in positioning subjects
between sessions was estimated not to have exceeded 4 or 5 degrees with
respect to the zero azimuth.

In Figure 6 it may be seen that the primate chair was located
2.75 m from an arc upon which the acoustical transducers were mounted.
The speakers were mounted such that their axis was parallel to the
floor of the chamber and directed toward the center of the observer's
head. The subject's head and centers of the speakers were located
.92 m above the floor of the chamber.

The acoustic stimuli presented were pure tones and bands of
noise 250, 500, 1000, 2000, 4000, 8000 Hz wide geometrically centered

at three frequencies -- 8000, 11200, 16000 Hz. The auditory stimuli

26

Figure 5

Subject in Primate Chair Positioned in the Anechoic Chamber.
Note the position of the response disk and the feeder cup.
The subject's left hand is on the response disk; the feeder
cup is to the right of the response disk.

27

Owns-rll-I I

4‘: .:

 

 

FIGURE 5

s

28

Figure 6

The Location of the Observer With Respect to the Speaker Array
Mounted on the Arc.

29

 

FIGURE 6

30
were presented (turned on and off) by enabling a random crossing
tone switch with a rise and fall time of 25 msec. The stimuli were
presented to the subject through a matched array of 5 loudspeakers.
Two different models of transducers were used. Stimuli centered at
8000 Hz were presented through University sphericon T202 loudspeakers;
while stimuli centered at 11200 and 16000 Hz were presented through
60 mm McIntosh tweeters provided through the courtesy of Roger
Russell, McIntosh Laboratories.

Pure tones were generated by an audio oscillator and were
amplified, gated, and attenuated according to the circuit depicted in
the block diagram in Figure 7. Bands of noise were generated by an
analog audio multiplier described by Palen and Gourevitch (1970).

The noise generation system produced a double sideband suppressed
carrier noise. The bandwidth of the noise was twice the bandpass

of a filtered noise signal which was amplified (MC 1433 operational
amplifier) and sent to one leg of the multiplier (Analog Devices 428 J).
The noise band was arithmetically centered at a carrier frequency which
was provided and amplified (MC 1433 operational amplifier) at the other
end of the multiplier by an audio oscillator (Hewlett Packard 200 CDR).
Thus the bandwidth and the center frequency of the audio stimuli were
an expression of the upper cutoff value of a filtered (Allison
Laboratories AL 2ABR) noise source (General Radio 1381) which extended
from near DC; and the carrier frequency of the oscillator respectively.
The audio signal generated by the multiplier was subsequently

filtered (Krohn Hite 310 ABR acitve bandpass filter) and amplified.
Table 1 presents the filter and oscillator parameters for the stimulus

employed. Figure 8 presents the equipment employed for the generation

Figure 7

The Audio Equipment Used for the Generation of Pure Tones

a. frequency counter Hewlett Packard 523 DR

b. audio oscillator Hewlett Packard 200 CDR

c. programmable attenuator custom

d. tone switch custom

e. attenuator Daven T 690 CR

f. audio amplifier McIntosh 240

g. speaker selector switch custom

h. attenuator Daven PT 324-M

i. speaker University sphericon T 202;

McIntosh tweeter

32

 

 

 

 

 

.i ..

 

 

E V
mmOLCQDZMFbu

 

 

 

 

 

 

 

 

 

 

 

 

 

3.
10:26
«063%
mqumam

, A 3
«mi—.524
O_OD<

. 3
10.53th...

h onsmfim

.3
105.56
wzo...

.3
mmkznoo
>ozw30w¢u

«02$me 3.
massage a mofimmwmo

33

 

 

m.omlm.NH omH x b.H Noooa NmeH thma oom
m.omlm.NH mh x h.H Hoooa @NHOH @hmma 0mm cocoa
oatw oovm x h.H mmmHH momma mmmh ooom
v.malv.m OONH x h.H hhmaa humma hhmm 000v
NHIOH 000 x h.H mVNHH mvNNH mvNOH OOON
calm 00m x b.H HHNHH HHNHH HHBOH OOOH
calm omH x h.H MONHH mmvHH mmmoa oom
walw m5 x h.H HONHH QNMHH QNOHH 0mm OONHH
MHIm oovm x h.H vvmm vwmma vwmv 000m
N.OHIN.© OONH x h.H wvmm mvmoa mvmo 000v
min 00w x P.H meow meow Noon ooom
manm 00m x h.H maom mama mamh OOOH
manm OmH x h.H voom wmmm vmhh oom
malm mh x h.H ooom wmam mhmh 0mm ooom

 

mmcauumm “madam mafluumm nmuaam
mmmmccmm muonso Momma Isms Ammo name name name
wuumncaone mmam um nwwnnmo mmouso roam “mouse sou Buoa3ocmm mucosumum umuamo

«Hafifium ocmm mmwoz map How mwmumemwmm HoumHHwomo can Hmuaflh mo hamsfidm

H mqmfifi

34

.me cw owwsmmmfi whomoumo

M

 

m.omlm.NH

mHIVH

hHImH

m.omlm.NH

oovN x

OONH x

000 x

00m x

mmvma

mNHOH

Hmoma

moomH

mmvom

mNHmH

HmOhH

momma

A.U.UGOOV H mqmde

mmvma

mmawa

Hmoma

momma

ooom

ooov

OOON

OOOH

35
of noise bands. The audio multiplier method of producing noise
bands has the advantage of yielding slopes much steeper than those
exhibited by filters displaying conventional Butterworth characteris-
tics. The narrowest bandwidths, 250 and 500 Hz, were nearly rectangular
with slopes exceeding 80 dB per octave. Appendix A contains acoustical
measurements of the stimuli employed. The experimental contingencies,
the delivery of stimuli, and the subject's responses were automa-
tically controlled and recorded by solid state logic modules

(Digi Bits) located in an adjoining room.

Procedure

One way to study auditory localization is by the employment of
the detection paradigm. In this procedure the observer is instructed
to report when he has detected a change in acoustic locus. It should
be emphasized that the detection paradigm as used here denotes that
category of experimentation which was identified by Brindly (1970)
as Class A. That is, it was assumed that if the sensations evoked
at two stimulus locations are indistinguishable then those locations in
space may not be discriminable different. Conversely, when the
observer reports that two spatial locations are detectably different,
this circumstance is likely an expression of differences in the sensations
evoked by the different location. This simple assumption linking the
psychological and the physical domains is the logical substrate upon

which this research is based.

36

Figure 8

The Audio Equipment Used for the Generation of Noise Bands

a.
b.
c.
d.
e.
f.
g.
h.
i.
j.
k.
1.
m.
n.

frequency counter

audio oscillator

noise generator

filter

audio multiplier

audio amplifier

filter

programmable attenuator
tone switch

attenuator

audio amplifier

speaker selector switch
attenuator

speaker

Hewlett Packard 523 DR
Hewlett Packard 200 CDR
General Radio 1381

Allison Laboratories AL 2ABR
Analog Devices 428 J
Hewlett Packard 450 AR
Krohn—Hite 310 ABR

custom

custom

Daven T 690 CR

McIntosh 240

custom

Daven PT 324-M

University sphericon T 202;
McIntosh tweeter

REQUEN
COUNTER

(0)

AUDIO
CILLATOR
(b)

LOW PASS
FILTER

NOISE
GENERA

(c)

ANALOG
ULTIPLIER

37

AUDIO
AMPLIFIER Egg
If)

PROGRAMMABLE
TTENUATOR

 

ATTENUATOR

 

 

 

 

I ATTENUATOR

 

 

 

 

ATTENUATOR .
(m)

 

 

 

 

ATTENUATOR

 

 

[334.5
D3“—

AT TENUATOR

 

 

 

Figure 8

TONE
SWITCH
In

0)

AUDIO
AMPLIFIER
(k)

SPEAKER
SELECTOR
SWITCH

U)

38

Pretraining

 

With nonverbal observers the burden rests with the experimenter
to invent a language through which one may communicate the response
requirements to the subject. The goal of the training procedure
was to instruct the monkey to maintain contact with the response disk
when the sound was repetitively presented at zero azimuth from the
referent speaker and to release the response disk when the sound was
presented at a different location from one of the four comparison
speakers. The method by which this was accomplished was as fOllows:

The subject's ad libitum weight was estimated by averaging the
body weight maintained under nonrestricted feeding. Subjects were
subsequently food deprived until their body weight dropped to 90% of the
free-feeding level. They were seated in a primate chair and trained
to extract and eat banana flavored pellets delivered to the food
cup. After several daily sessions the monkey habituated to the primate ~
chair and would readily take food when it was presented into the
food cup. At this point the subjects were ready for the initiation
of training.

The subject was seated in the primate chair, and the chair was
positioned in the anechoic chamber. The pellet dispenser was attached
to the feeder cup. The subject was then shaped by the method of
successive approximations (Whaley and Malott, 1971) to manually make
contact with the response disk. Typically by the end of one session the
monkey would make contact with the response disk, a banana pellet
would be delivered and the monkey would consume the reinforcer.

The response requirements for the delivery of the reinforcer were

altered in the following session. Contact with the response disk

39

produced a noise burst which pulsed repetitively from the standard
speaker located at zero degrees azimuth. When the subject released the
response disk, reinforcement was delivered. A small light mounted
adjacent to the standard speaker was continuously lit during this
stage. Under these conditions the latency for releasing the response
disk gradually grew shorter and typically by the end of this session
the subject was releasing the disk by the first or second pulse

of noise.

In the next session a time—out was introduced into the procedure.
The time-out was initiated every time the subject released the response
disk. The duration of the time-out was gradually lengthened from a
minimum value of 200 msec to 6 sec over the course of several sessions.
During the time-out the trial light was unlit. If the subject made
contact with the response disk during the time-out, the time-out was
reset and timing was not reinitiated until the subject broke contact
with the response disk. Following several sessions of training the
subject would pause until the trial light was presented, make contact
with the response disk and release synchronously with the presentation
of the noise burst.

In the next stage of training the subject made contact with the
response disk and then was required to wait for successively longer
durations until the auditory stimulus was presented. Releasing the
disk would then result in the delivery of reinforcement. After several
sessions of training the subject would maintain contact with the response
disk until the acoustic stimulus was presented. At this stage of
training the subject was required to maintain contact with the

response disk for an average duration of 8 sec (range 250 msec to

4O

16 sec) at which point the stimulus was presented from a speaker
located to the right of the observer.

In the last stage of training the monkey was subjected to the
identical schedule and reinforcement contingencies as were presented
in the previous stage except that during the period in which the
subject had to maintain contact with the response disk (prior to the
presentation of the auditory stimulus from the subject's right)

a very low intensity auditory stimulus was repetitively pulsed from
the zero azimuth. Thus the subject had to discriminate between
acoustic stimuli which differed along two dimensions, intensity and
location. The intensity difference between the pretrial stimulus
(S-) and the trial stimulus (the stimulus during which releasing the
response disk produced reinforcement, (S+)) was initially 40 dB.
Over the course of 5 to 10 sessions, this intensity difference

was gradually reduced (faded out) by increasing the intensity of

S-.

Psychophysical Testing Procedure

 

A diagram of the testing procedure in state notation is presented
in Figure 9 (Snapper, Knapp, and Kushner, 1970). The testing
procedure differed from the last stage of pretraining along two
dimensions. One difference was the position on the arc of the
comparison speakers. During pretraining the comparison speakers were
clustered at one end of the arc about 30° from the standard transducer.
However, when the method of constant stimuli is employed it is
desirable to adjust the steps within the physical stimulus continuum

to bracket the subject's threshold. In this case threshold was

41
defined as the change in spatial location which was detectably
different from zero azimuth on 50% of the trials. The 50% detec-
tability locus is an empirical entity which varies for different
acoustical parameters. Thus the speakers were positioned and
repositioned until a satisfactory arrangement was determined.
Table 2 presents the spatial configurations employed for the acoustical
parameters explored in this research. The second difference between
pretraining and the last stage of testing was the incorporation of
catch trials. Perhaps the most significant contribution of signal
detection theory to psychophysics is the measurement of the subject's
probability of reporting the presence of the stimulus when it was not
presented. With reference to classical psychophysical procedures
it is desirable to monitor the subject's flase alarm, or random release
rate so that any changes in the observer's response bias do not distort
the estimation of the observer's threshold (sensitivity). In the
data reported here sessions are excluded from analysis when the observer's
false alarm rate exceeded 20%. Only infrequently were data excluded by
this criterion; and the majority of the unacceptable sessions
occurred either when Miko, the only female monkey, was in estrus, or
when the acoustical parameters were recently changed.

With the incorporation of the modification described above the
maintained testing procedure was as follows: The trial light was
presented and the subject was free to initiate the trial by manually
making contact with the response disk. Disk contact, the subject's
observing response, had the consequence of producing the auditory
stimulus which was reiteratively presented from the standard

transducer located zero degrees azimuth.

42

Figure 9

State Diagram of the Testing Procedure.

43

 

.0

 

ZthZSu

6.

m wuzmwm

33 SI

zo_m_ow
>._. .4530qu

.C

room Q§

         

TABLE 2

44

Speaker Locations With Respect to the Zero Azimuth

 

 

 

 

 

 

 

Speaker
Configuration A B C D E
l 0.0 2.9 12.0 17.2 28.6
2 0.0 2.9 8.6 17.2 28.6
3 0.0 2.9 8.6 17.2 La
4 0.0 1.3 7.8 15.5 29.5
Stimulus Bandwidth (kHz)
Subject .001 .25 .50 1 2 4 8 Center Frequency
Sidney 1 1 l 1 l 3 3 8000 Hz
Miko 2 2 2 2 2 3 3
Oscar 2 2 2 2 2 3 3
Sidney 4 4 4 4 4 4 4 11200 Hz
Miko 4 4 4 4 4 4 4
Oscar 4 4 4 4 4 4 4
Sidney 4 4 4 4 4 4 4 16000 Hz
Miko 4 4 4 4 4 4 4
Oscar 4 4 4 4 4 4 4

 

a .
Speaker mounted at zero a21muth below standard speaker. See control
for speaker differences in text.

45

At some random point in time (number of pulses) later the
auditory simulus was presented from one of the four comparison
speakers from the observer's right - then it was presented from the
standard speaker - then once more from the comparison speaker. If
the monkey detected this transition in acoustic locus and reported this
detection by releasing the response disk (within the 2 sec trial
duration) it received a banana pellet in the food cup. A 6 sec time-
out followed reinforcement, then the trial light was presented and the
monkey could make its observing response - hand on the disk - reinitiating
the sequence. The comparison speaker at which the acoustical transition
occurred was randomly determined. The failure to report a transition
had no consequence and the sequence was recycled. The release of the
disk at all times other than in a transition trial did not produce
reinforcement and entered the subject into the time-out.

Contact with the response disk led to the presentation of a
trial on the average once every 8 sec (range 0-16 sec). The subject
was confronted with two different kinds of trials -- transition
trials which when reported produced reinforcement and catch trials
through which the experimenter monitored the subject's random
release rate. Trials were unmarked other than by the presence or
absence of an acoustic transition. The probability of a trial being
a catch trail was typically .3; however, the proportion of catch trials
was elevated on occasions to .4 or .5 when the suject's catch trial
rate began to exceed the criterion level of 20%. Experimental
sessions were conducted daily and ranged in duration from 60 to 120
minutes. In the data to be reported here each daily session was

partitioned into two halves during each of which a different bandwidth

46
of the same center frequency was presented. The stimulus bandwidths
within a session were arbitrarily selected. Usually a narrow and
a wide band were employed. In each half of a daily session 200 to
300 trials were presented, and subjects typically received 100 to

200 banana pellets within this period.

Stimulus Parameters

 

Each acoustical pulse was presented with a duration of 300 msec
with a 25 msec rise and fall time. The pulsed stimuli were repeated
at a rate of 1.5 per second. The sound pressure level of the
auditory stimulus was randomly varied for each pulse. The random-
ization was in 3 dB steps within a 21 dB range. Thus the subject
was required to detect changes in the location of an acoustic event
under conditions in which the intensity of the sound varied. The
adoption of this requirement served two purposes. One, under natural
conditions biologically significant sounds vary in intensity. Thus
the testing procedure was consonant with the conditions under which
an observer must localize sounds in nature. Two, the randomization of
intensity minimized the likelihood of the subject detecting slight
intensity differences between speakers. The sound pressure level
'between speakers was equated within 1 dB and the significance of this
slight difference in intensity was reduced by this technique. (See
Appendix A).

As presented in Figure 4 the threshold of hearing is a function
of frequency. The threshold of hearing for the macaque is about 1 dB,
9 dB, 23 dB re 201/4Pa at 8000, 11200, 16000 Hz respectively. The

acoustic stimuli were presented at a mean intensity level of 40 dB SL.

47
Thus at 8000, 11200, 16000 Hz the corresponding sound pressure levels
were 41 dB, 49 dB, 63 dB re 20 /(Pa. As the bandwidth of an acoustic
stimulus in increased, the energy level per cycle must be decreased
to maintain a constant sound pressure level. with rectangular bands
of noise this requires that the energy level per cycle be decreased

3 dB per doubling of the bandwidth.

The Control for Speaker Differences

 

Psychophysical investigations of auditory localization Which
employ different transducers at different loci require that the
experimenter demonstrate that the observed level of detection was
dependent on location in space and independent of any differences
which could be expressed by the transducers. In the data presented here
the possibility of this artifact was dismissed in the following manner:
For all acoustical parameters at least one comparison speaker was
positioned in a location at which its associated probability of
detection was below threshold. Consequently a high level of detection
at other comparison locations was not likely due to differences
between speakers. At the stimulus bandwidths of 4000 Hz and 8000 Hz,
centered at 8000 Hz a location in azimuth for a comparison speaker
was mounted at zero azimuth directly under the standard speaker. The
threshold for changes in the median plane is greater than the threshold
for changes in azimuth; thus a low probability of detection associated
with the lower speaker served as the same control.

At the 4000 Hz and 8000 Hz stimulus bandwidth centered at 8000 Hz
an additional manipulation was executed. The comparison speakers at

two loci with markedly different levels of detection were exchanged.

48

Thus this manipulation served to differentiate if the observed level
of detection was dependent on differences between speakers or

dependent on differences in location.

Results
The Method of Constant Stimuli

While the method of constant stimuli is a relatively inefficient
procedure, it is regarded as the most accurate and one of the most
generally applicable psychophysical methods (Guilford, 1954). In
the procedure employed here one standard location in space, zero
degrees azimuth, was specified and four comparison locations were
arbitrarily selected such that they sampled the entire range of the
observer's psychometric function. That is, the observed probabilities
of detection typically ranged from less than .2 to greater than .9.
These conditions were met for those cases in which detection increased
monotonically with angle. As was stated previously, the measurements
made by Harris (1972), Harrison and Downey (1970), and Sivian and
White (1933) demonstrate that with some acoustic stimuli binaural
differences may not be monotonically related to spatial location.
The shape of the psychometric function then serves as an index of the
presence or absence of monotonicity. Thus one of the advantages of
the method of constant stimuli is that monotonicity of the psychometric
function is directly displayed, while in alternative procedures, such
as the method of tracking, it is often only assumed and may or may not

be demonstable by inspection of the raw data.

49

The Theory of Data

The observer was presented via some random order the values, or
change in location, which composed the stimulus set. For each stimulus
presentation the observer was required to judge if the stimulus
‘location was different from the standard locus, zero degrees azimuth.
For each acoustical parameter, the 50% detectability locus was
computed from the proportions of the correct detections associated
with each location. The values which constituted each proportion were
typically collected over 8 or more daily sessions. These data are
available in Appendix B.

With data obtained by the method of constant stimuli, the
anaysis of choice is probit. A probit is the entity produced by
transforming the proportions of correct detections at each comparison
location to Z—scores and adding the constant 5 (Finney, 1952). These
transformed data are then fit to a normal distribution following the
principle of maximum likelihood. In the derivation of the solution,
the probabilities of detection are weighted in a similar manner as
employed in the Muller-Urban least squares determination. That is,
the Z value is more likely to be in error for small or large proportions
than for proportions near 0.5. Thus the 2 values are weighted so
that the solution is not greatly affected by errors in Z (Guilford,
1954). The major advantage of probit analysis is that the 50%
detectability locus is computed using all of the psychophysical data.
A second feature of probit analysis is that a chi—square estimate is
computed for the goodness of fit of the observed data to a gaussian
distribution.

The data presented here were subjected to a probit analysis.

50

Sixty-three psychophysical functions were entered in the analysis. Of
these 63 functions, in only 13 cases was the null hypothesis not
rejected, indicating that these data did not satisfy a normal
distribution. That is, in 75% of the cases 7E2.) 7.8; at p < .05,

3 degrees of freedom. Thus any analysis that assumes that the data
conform to normality is subject to this objection.

An alternative procedure for the estimation of the 50% detec-
tability locus is linear interpolation. As a procedure its only
assumption is that the distribution underlying the data is symmetical.
Given any symmetrical distribution, the values near the midpoint of
the cumulative representation of the distribution lie on a linear
segment of the curve. Linear interpolation has the advantage that it
is more conservative in being free of assumptions regarding the nature
of the distribution, but it is statistically less powerful by not
taking advantage of all the data in the psychometric function.

The data reported here were analyzed by both procedures. The
50% detectability point for each psychophysical function was estimated

by probit analysis and by linear interpolation.

Acoustical Stimuli Geometrically Centered at 8000 Hz

Figure 10 presents the psychometric functions for the three
monkeys for the 8000 Hz pure tone. The ordinate is percent correct
detection, the abscissa is azimuth. The subject's random release rate
is displayed over the zero azimuth point. In all three observers the
proportion of the trials detected increases monotonically with azimuth.
As may be seen in this figure the location of the comparison speakers

bracketed the observer's threshold. The estimations of the 50%

 

 

51

Figure 10

The Probability of Detection as a Function of Azimuth for the
8000 Hz Pure Tone. The ordinate is percent correct detection;
the abscissa is azimuth. The subject's random release rate is
displayed over the zero azimuth point.

o

 

O.

manor...

“mummoue muoz<

 

 

 

on om c. on ON 0. o
p p _ F — _ _
. Iow Iom
I ow I o...
I oo I8
I om Iom
r 00. I oo.
:38 8::
N: oooo ammuhzmo
u: _ :Eiozé

 

1-0

.ION

low

.low

>w29w

low

 

F 00.

NOI10313CI 1.038803 lNEICItIBd

53

mums .8. V a I
mums .8. v a ..

wufiamsuo: Eoum wusvwmmoo :ufl3 UquMUOmmm mumsquflnu

M

 

 

 

 

 

 

 

«Im.eamwomm m.o .m.oomm «m.m v.e Iv.ma «m.mv «mumswmIaao
H.m m.m v.5 m.m m.oH m.HH m.oa “shone
m.m v.m m.s e.m 6.0H m.OH q.oa coflamaomuwnaH Humane

waomo
e.o m.o Ie.om~ .m.mm Im.om «m.ov 4H.umH mmnmswmIano
H.m m.m v.s e.s m.m m.HH m.m unnoam
o.m 0.4 o.e s.s m.m m.oH 8.6 cOnumHomnmuaH amazed
one:
m.H e.o «o.HHm .H.meo «H.Hmm «m.NHH «6.6mH «mumswmIaao
m.o H.» m.oa R.HH m.HH o.vH m.mH uaaoam
H.o o.» q.m o.oa m.mH m.va o.ma conpmaomamucH “amend
smcoam
ooom oooe coon oooH oom omm H

 

Anmv nuofi3©cmm
Nm ooom um powopcmo mnpoflzpamm How moamcd paonmwune

m mammfi

54

detectability locus as provided by probit analysis as well as by linear
interpolation are presented in Table 3. These three psychophysical
functions are very similar, yet the differences between subjects
observed at 8000 Hz are greater than those observed for any of the
bands of noise centered at 8000 Hz.

The psychophysical functions for the narrowest band of noise,
250 Hz, may be seen in Figure 11. The functions for 500, 1000, 2000,
4000, and 8000 Hz noise bands are presented in Figures 12, 13, 14, 15,
and 16 respectively. Careful inspection of these functions reveals two
characteristics. First, the functions are highly similar among obser—
vers. Second, as the bandwidth is increased, the slope of the func-
tion becomes steeper and the locus of threshold decreases. The effect
of increasing the bandwidth of the stimulus is well indexed by
the change in the proportion of the trials detected at the 9° azimuth.
At the 250 Hz bandwidth these proportions were .35, .40, and .42 for
Sidney, Miko, and Oscar respectively. At the 8000 Hz bandwidth the
corresponding values were .76, .95, and .99. Consequently at 9°, the
probability of detection dramatically increased from well below to well
above threshold when the bandwidth of the stimulus was increased from
250 to 8000 Hz.

As may be seen in Figures 15 and 16, when the two widest bands
of noise were employed, threshold had decreased sufficiently that the
detection of the comparison speaker adjacent to the standard (i.e. 3°)
was approaching 50%. This may be observed in the psychometric
functions for two of the three subjects, Miko and Oscar, at the 4000 and
8000 Hz stimulus bandwidths. At these two bandwidths one comparison

speaker was mounted at the zero azimuth location, directly under

Figure 11

The Probability of Detection as a Function of Azimuth for the
250 Hz Stimulus Bandwidth Centered at 8000 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth point.

56

 

on 8. 0. on on o. 0 on ow o. o
p _ _ _ L _ _ _ IF _

I L _ d

3

8

0.

IS I8. I8 3

N

l

I ow I 8 I2. m

8

8

3

100 10$ low 3

I...

m.

I8 . .8 I cm 1

3

0

u

I 8. I 8. I8. 0

«<08 9:: 5.35 N

5. oooo oucupzuo

. . manor.

anemone 3oz...

 

 

 

 

N: ODN 15.990de

 

57

Figure 12

The Probability of Detection as a Function of Azimuth for the
500 Hz Stimulus Bandwidth Centered at 8000 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth point.

58

N.

manor.

Ammumouov ”.3024

 

 

 

 

on ow o. c on on o. o
p — _ _ _ _ b _
Iom Iow
I ow I 8
I om I06
I om Iom
f 00. r oo.
«38 9...).
N: ooom oummpzuo
59324....

N: 000

 

on ow o. o
_ _ _ _ d
3
H
3
Iom 3
N
I.
r2. w
8 .
8
3
Tom 0
l
m
.81
3
O
u
I8.
529m m

 

59

Figure 13

The Probability of Detection as a Function of Azimuth for the
1000 Hz Stimulus Bandwidth Centered at 8000 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth.

60

n.

manor...

GUM—meme m402<

 

 

 

 

on 8 o. 0 on 8 o. o
_ _ _ _ L _ _
IoN ION
I ow. I 9..
I om I8
I06 Iom
r 8. r 8.
:38 8...:
N: oooo 35:23
N: coo. :Biazqm

Om

ON

0. O

>mzo. m

ION

IOV

low

low

 

I00.

N0l103130 1038800 1N3083d

61

Figure 14

The Probability of Detection as a Function of Azimuth for the
2000 Hz Stimulus Bandwidth Centered at 8000 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth.

2000 HZ

8A NDW'DTH
Isl—An *l-‘nl-A

v. Mano—u

Amummouov m...oz<

62

 

 

 

 

 

 

 

 

on 0.0. o. 0 on 8 o. o
L _ b _

d

3

8

0

I om ION I8. 3
N

1

I o... I 8 I o: m
8

8

3

I8 I06 I8 0
I.—

m

I8 Iom I om 1
3

0

u...

I 8. I oo. roo. 0
:33 8:: 529m N

N: ooom ammupzuo
N: 008 :85523

63

Figure 15

The Probability of Detection as a Function of Azimuth for the
4000 Hz Stimulus Bandwidth Centered at 8000 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth. The probability of detection of the lower speaker is
displayed over L. See text for details.

m. maze...

.mu mmomn: muozc.

 

64

 

 

 

 

 

 

o 3 om . o. o ._ ON 0. o

_ (I. _ _ _I\\ e _ _ d
o 3
8
ION o I m

om IoN
N
II-
Iov Io¢ Iocm
8
w.
I om Ice Ice 0
I.—
. 0
Iom I2. I 8B
3
0
c H...
00. I 02 I090
m<omo 8:: 329m N

N: 0000 owmwhzuo
N... ooov 15.3024m

65

Figure 16

The Probability of Detection as a Function of Azimuth for the
8000 Hz Stimulus Bandwidth Centered at 8000 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth. The probability of detection of the lower speaker is
displayed over L. See text for details.

66

3
f\

ON

0.

w.

umao:

.mummomo.muoz<

 

\

 

 

cdomo

 

 

3 ow 2 o
r\\L _ _
o
ION ION
10¢ 10....
i om low
low Tom
I 00. r. 00.
052
N... 0000 oummhzuo
N... 0000 72.930240

ON
{\N.

_

o.

I I I
O O
m c 8

I
0
103130 1038800 1N3083d

 

Ion.
.N

>wzo.m

67
the standard speaker. The probability of detection for the lower
speaker is denoted by L on the abscissa for Figures 15 and 16; and
as may be seen in these figures the proportion of trials detected at
this location approximated the observer's random release rate. Once
a speaker location (L) had been determined with a low associated
probability of detection, the transducers associated with this point
and the 9° azimuth point were arbitrarily exchanged. The consequences
of this manipulation are presented in Table 4. As may be seen in
this table the proportions of the trials detected were determined by the
locations of the transducers and not by which transducers occupied
that location.

Figure 17 provides a summary of the psychophysical functions
presented in Figures 10-16. The panel A shows the 50% detectability
locus according to the linear interpolation; panel B presents the
corresponding analysis given by probit analysis. As the width of
the stimulus band increases the threshold for detectability decreases.
It may also be seen from this figure that the threshold varies more
widely between observers for the 8000 Hz pure tone than for any of
the noise bands and that the threshold value for the pure tone is not
directly predictable from the bandwidth data. As may be seen by the
comparison of panel A and B of Figure 17 or from Table 3, differences
in the estimation of threshold between procedures are essentially
trivial. The inspection of Figure 17 suggests that the log of stimulus
bandwidth is linearly related to the threshold for the detection of
changes in azimuth. That is, threshold is related to bandwidth

according to the expression in Table 5.

Test for Speaker Differences: Estimated Proportions for Two Bandwidths

68

TABLE 4

 

 

 

 

 

 

 

Subject and
Bandwidtha Speaker Speaker Location

L 8.6°
Sidney A 10/177 or .056 121/193 or .627
4000 Hz B 25/295 or .085 84/102 or .824
Miko A 28/198 or .171 65/81 or .802
4000 Hz B 17/74 or .229 190/196 or .969
Oscar A 31/299 or .104 72/72 or 1.00
4090 52 B 3/54 or .056 235/243 or .967
Sidney A 5/222 or .023 152/189 or .804
8000 Hz B 14/264 or .053 93/96 or .969
Miko A 14/186 or .075 137/150 or .913
8900 Hz B 30/155 or .194 195/198 or .985
Oscar A 22/308 or .071 64/64 or 1.00
8000 Hz B 10/58 or .172 216/218 or .991

 

a .
Bandw1dths are centered on 8000 Hz.

Figure 17

Threshold as a Function of Bandwidth for Stimuli Centered at
8000 Hz. Panel A presents the 50% detectability locus

as determined by linear interpolation; panel B presents the
same locus as determined by probit analysis.

THRESHOLD (DEGREES)

THRESROLD (DE GREESI

24

20

24

22

20

 

70
CENTERED 3000 H2

PANEL A
SIDNEY
mm o
OSCAR ' o

 

 

 

 

o
o
D
l_/1 I I . I I I J
OOI'r—JS .5 I z ' 4 a
BANDWIDTH (RHZI
PANEL 3
SIDNEY O
MIKO
OSCAR
O
A
0
1 I l J I L 4]
.DDI r—.25 .

.a I z
BANDVIIDTH (IIHZI
FIGURE I7

71

TABLE 5

The Regression of Threshold on Bandwidth Centered at 8000 Hz.

 

Observer Probit Analysis Linear Interpolation

Sidney t=-5.l log b + 26.4 t=-5.5 log b + 27.1
(r=-.97) (r=-.99)

Miko t=-5.9 log b + 25.9 t =- 5.2 log b + 23.3
(r=-.98) (r=-.99)

Oscar t=-6.0 log b + 26.1 t=-5.9 log b + 25.8
(r=-.97) (r=-.96)

 

Note: t is threshold in degrees, b is bandwidth in Hz, r is the
correlation coefficient.

Figure 18

The Probability of Detection as a Function of Azimuth for the

11200 Hz Pure Tone. The ordinate is percent correct detection;

the abscissa is azimuth. The subject's random release rate is displayed
over the zero azimuth.

73

m. manor...

.mummouo. muoz<

 

 

 

 

On ON O. O Om ON 9 O
p . . _ p . . —
ION ION
I o: I O1.
I O6 IOO
IOO‘ IO:
.. OO. I OO.
:38 or:
N: OON: 3:328

N... . x..b;>oz<m

O. O

>uzo. m

I I I
O O O
(D V N

.4:
NOI10313CI 1338803 iNBOHEd

 

I00.

74

Acoustical Stimuli Geometrically Centered at 11200 Hz

Figure 18 presents the monkeys' three psychophysical functions
for the 11200 Hz pure tone. The proportion of trials detected at the
largest angle, 30°, was less than .70 for Sidney and Miko and only
.83 for Oscar. In all the previous psychometric functions the
probability of detection was approaching 100% for this change in
azimuth. Figures 19 and 20 present the corresponding functions for
the 250 and 500 Hz band stimuli. Again, the figures reveal that the
subjects had difficulty discriminating changes in azimuth of these
stimuli. The irregularities in the shape of the psychophysical
functions are reflected in the differences between the estimations of
threshold given by probit analysis and by linear interpolation
shown in Table 6. This difference is greatest for Miko at the 500 Hz
bandwidth stimulus -- 13.0 by linear interpolation vs. 20.2 by probit
analysis. These discrepant results are likely the result of the
departure from monotonicity in Miko's psychometric function (Figure
20). Figures 21-24 present the psychophysical functions for stimulus
bandwidths of 1000, 2000, 4000, 8000 Hz respectively. These functions

are monotonic and relatively steep for each observer.

Figure 25 presents the summary functions relating stimulus band-
width to threshold. Panel A of the figure presents the estimations of
threshold determined by linear interpolation and panel B the thresholds
determined by probit analysis. While the two methods of analysis
yield different absolute values of threshold, the effect of stimulus
bandwidth on localization is clear: threshold varies inversely with

the bandwidth of the stimulus. The slope of the function relating

75

Figure 19

The Probability of Detection as a Function of Azimuth for the

250 Hz Stimulus Bandwidth Centered at 11200 Hz. The ordinate is
percent correct detection; the abscissa is azimuth. The subject's
random release rate is displayed over the zero azimuth point.

76

m. manor.

.mmmmouo. m302<

 

 

 

 

On ON O. O On ON O. O
. . . . L . P _
ION ION
I ow I o:
I O6 IOO
I OO IOO
r OO. .I OO.
:38 9...: .
N: OON: 3:328
N: OON 59.323

 

 

on ON 0. O
. L . P
.
1 ON
.I 0?
Wow
I 00
.IOO.
>Uzo.m

NOLLOBIBCJ 1038800 .LNEOHBd

Figure 20

The Probability of Detection as a Function of Azimuth for the

500 Hz Stimulus Bandwidth Centered at 11200 Hz. The ordinate is
percent correct detection; the abscissa is azimuth. The subject's
random release rate is displayed over the zero azimuth point.

78

on

mdomo

ON

wane.“—

.wummouo. m302<

IO¢

(I00

IOm

 

e 8.

N

 

 

 

On ON O. O
I . . .
IOO
r OO.
9...: 529m
.. oou: cummkzuo
N: 000 z..b_3oz<m

 

I I T
O O O
(.0 Q’ N

5‘:
NOIiOBiBCI 1038800 .LNEIOHBd

 

IOO.

mumc .8. v a...
Emnmoum panama nufl3 poumasoaoo on on manna oou msam>n
Nuflamsuoc Baum musuummmo cufl3 oouaHUOmmm oumaquflnum

 

 

 

79

 

 

 

 

.O.mOemO O.O 4N.mm. IO.OH «m.OO N.m O.m mmumsamIAao
O.m N.m O.O mae. O.mA N.OH m.mA 3.30“:
N.m N.O O.m N.mH m.NA v.8. m.mA cONumHomawucH “86:83

Hmomo

IO.OHemOHH N.e .m.momm «O.HNH .m.OOH .H.Hm IN.NO mmnmsvaNao
m.e m.v m.O O.NH O.OH N.ON A.ON page»:
m.m m.O m.m 3.: a... 0.0H N.OH cannmaoanmucH Hammad

oxNz

.nxxxx IO.mA .m.mNN .m.NN .:.mN .H.ON Is.mm mmumswmINao
m.O N.N e.AH O.NH O.ON O.mN O.ON 38308:
O.m H.N m.m O.mH m.m. m.ON m.NH OONONHOOAIOOH ummcaq

schNm

OOOO OOOO OOON OOOA OOm OmN H

 

.Nm. cupfi3ccmm
Mm oomaa um pououcou mnupwzocmm now moamcd oaonmouza

m mag...“

80

Figure 21

The Probability of Detection as a Function of Azimuth for the
1000 Hz Stimulus Bandwidth Centered at 11200 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth point.

81

On

 

mdomo

_N

UKDOE

.mummomo. muoz...

 

 

 

 

Om ON O. O
. . . .
ION ION
I OO I o:
I I
3 I OO
I OO. r OO.
9...:
N: OON: 3:328
N: OOO. 392.33

Om

>uzo.m

I
O
N

é
NOliOBIBO 1338800 .LNBOtIBd

TOO

Tom

 

I00.

82

Figure 22

The Probability of Detection as a Function of Azimuth for the
2000 Hz Stimulus Bandwidth Centered at 11200 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth point.

83

NN

manor...

.muumouo. mqoz...

 

 

 

 

 

Om ON O. O O» ON O. O
. . — . _ . .
ION ION
I O.» I O.»
I 3 I3
Io: IOm
r OO....I I r OO.
:38 9......
N: OON: 3:328
N: OOON 392623

 

 

Om ON O. O
E . . .

ION

IO...

IOO

IOO

IOO.
329.:

NOIiOBlBG 1038800 1N308Bd

84

Figure 23

The Probability of Detection as a Function of Azimuth for the
4000 Hz Stimulus Bandwidth Centered at 11200 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth_point.

85

 

¢<omo

MN

mmOOE

.mummouo. m..oz<

 

 

 

 

 

 

Om ON O. O Om ON O. O
L _ _ . r L _
ION ION
I O... I 8
I 3 I3
I 3 I3
r OO. 9 I OO. 9i
8...... >329...
N: OON: 3:328
39.323

N... 000....

ION

IO.»

low

Iom

 

rIOO.

NOI103130 1038800 1N30838

86

Figure 24

The Probability of Detection as a Function of Azimuth for the
8000 Hz Stimulus Bandwidth Centered at 11200 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth-point.

8000 H7

BANDW'DT H

87

¢N

MKDOE

.mummmuo. m302<

 

 

 

 

 

 

On ON O. O on ON O. O
_ — — p _ — _ _
ION ION
I 9. I Os.
I OO I3
I 3 I3
.II c OO. V I OO.
:38 9.22
N: OON: 3:328
N... 83 39323

On

ON

>uzo. m

ION

IOV

low

low

 

IOO.

NOI103130 1038800 1N308Bd

88

Figure 25

Threshold as a Function of Bandwidth for Stimuli Centered at
11200 Hz. Panel A presents the 50% detectability locus as
determined by linear interpolation; panel B presents the same
locus as determined by probit analysis.

THRESHOLD (DEGREES)

THRESHOLD (DEGREES)

24

20

I4

24

22

 

89
CENTERED IIZOO HZ

 

 

 

 
 
  
  
   

 

PANEL A
SIDNEY O
O
MIKO D
O
OSCAR
A
J_I 1 I l l L l _l
.00I’ has .5 I 2 4 o
BANDWIDTH (kHZI
PANEL B
8 SIDNEY O
MIKO
OSCAR
o
L l l l 1 l .1
.OOI r—es 4 a

.5 I 2
BANDWIDTH (“12)
FIGURE 25

90

TABLE 7

The Regression of Threshold on Bandwidth Centered at 11200 Hz

Linear Interpolation

 

Observer Probit Analysis
Sidney =-12.6 log b + 54.1
(r=-.99)
Miko t=-12.l log b + 49.2
(r=-.95)
Oscar t=- 8.4 log b + 35.9
(r=-.92)

=-13.8 log b + 54.2
(r=-.98)

=- 7.6 log b + 32.3
(r=-.98)

=- 9.7 log b + 40.9
(r=-.88)

 

Note: t is threshold in degrees, b is bandwidth in Hz, r is the

correlation coefficient.

91
bandwidth to threshold is steeper for stimuli centered at 11200 Hz
than for stimuli centered at 8000 Hz, a half octave lower in frequency.
This is empirically expressed in the linear regressions presented in
Table 7. It should be clear from Figures 17 and 25 that the major
difference between stimuli centered at 8000 and 11200 Hz was the
degradation in detectability of the narrow band stimuli at 11200 Hz.
That is, the minimum change in azimuth required for the detection of the
8000 Hz bandwidth was about equal for either center frequency, while
the minimum change in azimuth required for the detection of the 250
Hz bandwidth was elevated by a factor approaching 2 for the 11200 Hz

center frequency with respect to the 8000 Hz center frequency.

Acoustical Stimuli Geometrically Centered at 16000 Hz

Figure 26 presents the psychometric functions for the three
observers for the 16000 Hz pure tone. As may be seen in this figure
the functions are quite different between observers. The proportion
of the trials detected for the maximum change in azimuth, 30°, was
.66, .91, and 172 respectively, for Sidney, Miko, and Oscar. The
corresponding thresholds were 22.2°,9.6°, and 6.2° given by linear
interpolation; 22.7°, 8.8°, and ll.9° given by probit analysis. These
were the largest inter-subject differences encountered for any
acoustical parameter.

The psychophysical functions for the two narrowest stimulus
bandwidths -- 250 and 500 Hz -- are presented in Figures 27 and 28.
The functions at these two bandwidths are striking in one characteristic;
they are monotonic and more regular in shape than the corresponding

functions centered at 11200 Hz, one half octave lower in frequency.

92

Figure 26

The Probability of Detection as a Function of Azimuth for the
16000 Hz Pure Tone. The ordinate is percent correct detection;
the abscissa is azimuth. The subject's random release rate is
displayed over the zero azimuth point.

93

ON “$50.“.

.mummwwo. m...02<

 

 

 

 

Om ON O. O

 

row

 

Om ON O. 0 On ON O. O
. p . C . _ .
I ON
I O...
I OO.
I ON
I 00.
¢<omo Ov.=2
NI OOOO. memhzuo

N I . 73.05553

I OO.
>wzo.m

ION

T 0.».

Tom

IOm

 

IOO.

NOI103130 1038800 1N3083d

94

Figure 27

The Probability of Detection as a Function of Azimuth for the 250
Hz Stimulus Bandwidth Centered at 16000 Hz. The ordinate is
percent correct detection; the abscissa is azimuth. The subject's
random release rate is displayed over the zero azimuth point.

250 HZ

BANDWIDT H

95

NN

manor...

.muwmouo. m..oz<

 

 

 

 

On ON O. O On ON O. O
P . . . . _ .
I ON I ON
I OV I 0v
I Om I00
I Om Tom
.I OO. I OO.
«demo 02.2
N... 0000. ommwhzuo
N... OON Ibo§oz<m

On

ON

 

O. O

>uzo. m

ION

é
NOI10313CI 1038800 1N3083cl

 

FOO.

96

Figure 28

The Probability of Detection as a Function of Azimuth for the
500 Hz Stimulus Bandwidth Centered at 16000 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth point.

97

ON

manor.

.mummouo. OJOZO.

 

 

 

 

 

 

On ON O. 0 On ON O. O Om ON O. O
_ . — F . . r — b . _ _

Id
3
HO
8. m

I ON I
ON I N
l
I O
I CV I 0.? 0.? O
8
HO
O CO CO A.”
I O I I.-
m
IOO I3 I OO 1
.4.—
3
r .IOO. u

. I . OO.
:38 OO 3.: 329m nNU
N... 000 O. ommwhzmo
IPO.3OZ<O

N: OOn

98

Figure 29

The Probability of Detection as a Function of Azimuth for the
1000 Hz Stimulus Bandwidth Centered at 16000 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth point.

99

 

m<OOO

ON

manor.

.mummomo. m...OZ<

 

 

 

 

Om ON O. O
. . p .
I ON I ON
I O.» I 0v
I OO IOO
I OO IOO
I 8:1 I 8.
Ov...2
N... OOOO. ommuhzwo
N... 000. .......o=soz<m

Om

ON

O. O

>uzo.m

I I I
O O O
(D d" N

S:
NO|103130 1038800 1N3083c|

 

IOO.

100

Figure 30

The Probability of Detection as a Function of Azimuth for the
2000 Hz Stimulus Bandwidth Centered at 16000 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The

subject's random release rate is displayed over the zero azimuth
point.

101

On

umaor.

.mummomo. NJOZO.

 

 

ON

 

 

 

 

 

 

Om ON O. O On . ON O. O Om
_ _ _ .I p _ . . .
ION ION
I 8 I O?
I OO IOO
IOO Ice
:38 9.22

N: ooom. owmmhzuo
N... ooou 1.3.2.8233

O. O

. L... d
3

8

0

ION 3

N

l

IOO m

8

8

3

IOO 0

l.

m...

I om I.

3

0

r u

OO.
>329... m

 

102

The results indicate that the detection of change in azimuth of
narrow bands of noise was less capricious and easier for all three
observers with stimuli centered at 16000 Hz than with stimuli
centered at 11200 Hz.

A similar phenomenon may be observed in the psychophysical
functions for the 1000 and 2000 Hz bands presented in Figures 29
and 30. For these two bandwidths the proportion of the transition
trials on which the change in location was detected ascends monotonically
with azimuth; and in all cases threshold is less than or approximately
equal to the corresponding value obtained from the function
centered at 11200 Hz.

Figures 31 and 32 present the psychophysical functions for
the 4000 and 8000 Hz stimulus bandwidths. In all cases for all
observers the functions are monotonic and the slopes are steep;
however, the 50% detectability locus as derived from these functions
is no longer less than the corresponding estimations from the data
centered at 11200 Hz. Rather, it appears to be the case that the
50% detectabiltiy locus for wide band stimuli is relatively
independent of the center frequency.

Table 8 presents the estimations of threshold as determined by
linear interpolation and probit analysis for stimuli centered at
16000 Hz. In Figure 33 these same data are displayed as a function of
stimulus bandwidth. As may be seen in Figure 33 the slope of the
function relating threshold to bandwidth is less at 16000 Hz
than at the two previous center frequencies. This may be empirically

seen in Table 9.

103

Figure 31

The Probability of Detection as a Function of Azimuth for the

4000 Hz Stimulus Bandwidth Centered at 16000 Hz, The ordinate

is percent correct detection; the abscissa is azimuth. The subject's
random release rate is displayed over the zero azimuth point.

104

.n manor.

.mummcuo. m...oz<

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3 ON O. O On ON O. O Om ON O. O
P P F . _ P . . . .
_ . I... d
8
O
ION ION ION 3
N
l
I O4 I 8 I O.. m
8
8
3
IOO IOO IOO O
I.
m.
IOO IOO I OO 1
.4.
. 0
L L\_ H...
vi III I. I OO.? I OO. «1 IOO. O
:38 . 8...... , 323 N
N: OOOO. 3:328

N... OOOV :...o.§oz<m

105

Figure 32

The Probability of Detection as a Function of Azimuth for the
8000 Hz Stimulus Bandwidth Centered at 16000 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero azimuth
point.

nnnn “7

HA NDWIDT H

106

.mwwmowo. OJOZ...

Nm UNOOE

 

 

 

 

 

 

 

on ON 0. O on ON O. o
L L . . . . . I r
I ON

I ow
I cm
I om
.
anII _ 6L .I OO. O I«\_
:38 8......
N: OOOO. ammuhzuo
N... 0000 1.556240

 

 

 

Om ON O. O
p . . LI.

ION

I9.

IOO

IOm

IOO.? I 329....

1 I I
O O O
(D V’ N

I:
NOI10313CI 1038800 1N3083ci

 

r
o
9.

107

38 .8. V 8..

Emwmoum panama :uH3 pouMHdoamo on on mmme 00» wsaw>

Q

muwHMENoc Eoum wwsuummmp £HH3 OODMHUOmmm mnmnquflnom

 

 

 

 

 

 

 

NO.O 48.888383 m.8 «H.8NNOA .N.Amm IN.AON IO.HON mmumsumINao
8.8 8.m 3.8 8.8 8.NH 3.8 8.OH 338038
3.8 N.8 8.8 8.8 3.8 N.8 8.8 coNuONomamuaH 888883

HMUOO

IO.8>NH «8.388 48.88888. IN.83 «H.883 IO.8O IO.Om mmumswmINOO
8.8 8.8 8.8 3.8 N.8 8.3 O.8 388088
8.8 8.8 N.8 8.8 8.3 O.8 8.8 cowumaomumucH “88:83

0282

.axxxx 88.83 .8.mN58OH 48.383 .N.8NN . «3.88 IN.NN mmumswmIeao
O.N 8.: 8.8 8.8 O.OH 8.83 8.ON 383088
8.8 3.8 3.8 8.8 O.: N.m3 8.ON coflumaomamucH 888883

smceflm
OOO8 OOO8 OOON OOOA OO8 O8N

 

.va nupw3pcmm

Nm OOOOH um pmuoacmu mnupfl3ocmm Mom mmamcm paonmmune

m mqmda

108

Figure 33

Threshold as a Function of Bandwidth for Stimuli Centered at
16000 Hz. Panel A presents the 50% detectability locus as
determined by linear interpolation; panel B presents the
same locus as determined by probit analysis.

THRESHOLD(DEGREESI

THRESHOLD(DEGREESI

24

20

2G

24

20

CENTERED

 

 

109

ISOOO HZ
PANEL A

SIDNEY
NIKD

OSCAR

 

 

 

.5 I 2
BANDWIDTH (kHZI

FIGURE

33

I— D
L A
J 1 L l l l I l _J
ODV’E'JS J I z 4 a
BANDWIDTH (IIIIz)
I.
PANEL 8
I—
O
I' SIDNEY O
” MIKO 0
" OSCAR a
-'A
r-
0
‘ ,1_ L l l I L I
.ooI .25 4 a

110

TABLE 9

The Regression of Threshold on Bandwidth Centered at 16000 Hz

 

Observer Probit Analysis Linear Interpolation
Sidney t=-5.2 log b + 26.1 t=—3.8 109 b + 20.2
(r=-.9l) (r=-.78)
Miko t=-2.6 log b + 14.6 =-l.7 log b + 10.8
(r=-.94) (r=—.75)
Oscar t=-5.1 log b + 22.4 =-1.2 log b + 8.7
(r=-.81) (r=-.70)

 

Note: t is threshold in degrees, b is bandwidth in Hz, r is the
correlation coefficient.

L)!

(I)

5),

'44

'41

L“

I

Ill

[—4

Experiment II
The Detection of Changes in Azimuth as a Function of Stimulus Bandwidth:
Systematic Replication

A very conservative strategy for scientific research advanced by
Sidman (1960) is the systematic replication and extension of the corpus
of empirical results. In this tradition, Experiment II was designed
to (a) replicate the extreme values of the stimulus bandwidths centered
at 11200 Hz and (b) obtain psychophysical functions at selected
stimulus bandwidths centered at the quarter-octave point between
11200 and 16000 Hz. These values were selected to determine if the
differences between the psychophysical functions at 11200 Hz and
16000 Hz were orderly and reliable. Specifically psychophysical
functions were obtained for the 250 and 8000 Hz bandwidths centered
at 11200 Hz, for a 13454 Hz pure tone, and for bandwidths of 250, 500,

and 8000 Hz centered at 13454 Hz.

Method
Subjects, Apparatus, and Procedure

The experimental apparatus, contingencies, and subjects were
the same as employed in Experiment I. However, at the center frequency
of 11200 Hz the locations of the transducers at the 250 Hz bandwidth
stimulus were changed with respect to the same bandwidth in Experiment
I. For the replication the speakers were located at 0.0°, 8.7°, 15.5°,

29.5°, 46° azimuth V.S. 0.0°, l.3°, 7.8°, 29.5°. When these data were

111

112
collected Miko was in estrus and as had been the case before, this
monkey was a poor observer of narrow band stimuli under this condition.
As a consequence, a psychophysical function was also obtained from
Miko with the sitmulus bandwidth of 500 Hz.

As in Experiment I the stimuli presented at 11200 Hz had a mean
sound pressure level of 49 dB re 20 flPa. At 13454 Hz the speakers were
located at 0.0°, l.3°, 7.8°, 29.5°. For the macaques the threshold
of hearing function at 13454 Hz is 16 dB re ZOI/(Pa. The acoustical
stimuli presented at 13454 Hz had a mean sound pressure level of

56 dB re ZO/uPa, or 40 dB SL.

Results
Acoustical Stimuli Geometrically Centered at 11200 Hz

Figure 34 presents the psychometric functions for the three
monkeys with the stimulus bandwidth of 250 Hz and for Miko at
500 Hz. As may be seen in Figure 34 the obtained functions were
monotonic for Sidney and Oscar, but not for Miko. Miko's performance
was better with the 500 Hz stimulus band than it was with the 250
Hz band; however, it was still non-monotonic.

Figure 35 presents the psychophysical function for the 8000 Hz
stimulus bandwidth. As may be seen in the figure the psychometric
functions agree well with the comparable data in Experiment I, pre-
sented in Figure 24. The number of observations constituting this
replication was limited, consequently sampling error should lead
to some discrepancies with the data from Experiment I. However, the
psychophysical functions are qualitatively comparable with the data

presented in Experiment I. Table 10 presents thresholds as provided

113

Figure 34

The Probability of Detection as a Function of Azimuth for Narrow
Band Stimuli Centered at 11200 Hz: A Replication. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero

azimuth point.

PERCENT CORRECT DETECTION

PERCENT CORRECT DETECTION

 

114

   

 

I00
1 SIDNEY
80“
SD-
40..
BANDWIDTH 250 H2
20... CENTERED IIzoo HZ
T I'D Era 3'0 10 5'0
ANCLEIDECREES)

 

MIKO

BANDWIDTH 250 HZ

 

3°... CENTERED ”20on
SD-
40-
ZO-I
Io I'D 210 3'0 4'0 sro
ANGLE (DEGREES)

   

 

 

   

DSCAR
so-
SD—
40-
BANDWIDTH 250 HZ
20_q CENTERED IIZOOHZ
ID I'D 2'0 3'0 4'0 5'0
ANCLEIDECREES)
Ioo-
MIKO
80"
SD-
' 40... SANDWIDTH 500 N2
CENTERED IIZOOHZ
zo—

 

 

I0 I10 2‘0 3'0 4'0 5T0

ANGLE (DEGREES)

FIGURE 34

115

Figure 35

The Probability of Detection as a Function of Azimuth for the
8000 Hz Stimulus Bandwidth Centered at 11200 Hz: A Replication.
The ordinate is percent correct detection; the abscissa is azimuth.

The subject's random release rate is displayed over the zero
azimuth point.

8000 H2

BA NDWIDT H

116

On ON 0. O

 

 

IIIL.

¢<omo

on

umber.

Amummwuov m4¢z<

 

 

 

on ca 0. o
_ _ _
Iou low
I oc I ow
I om row
I om row
r 00. r 00.
9::
N: 03: ommmhzuo
N: oooo 59262.3

on

ON

0. O

 

>wzo.m

I T I
O O O
(D V N

5':
NOI103130 1038800 .LNBOHBd

 

rIOO.

Threshold Angles for Replications

117

TABLE 10

Bandwidth (Hz)

at 11200 Hz

 

 

 

 

 

 

 

250 500 8000
Sidney
Linear Interpolation 25.4 6.6
Probit 25.2 6.3
Chi-squarea 30.1* XXXX *
Miko
Linear Interpolation 27.3 14.5 5.2
Probit 52.6 20.1 5.6
Chi-squarea 48.9* 40.6* 0.6
Oscar
Linear Interpolation 17.2 4.0
Probit 14.5 3.4
Chi—squarea 39.9* 0.002

 

a . . . .
Chi-square assoc1ated With departure from normality

Chi-square value too large to be calculated with Probit program

*p < .05, df=3

118

by linear interpolation and probit analysis. The comparison of

Table 10 with the corresponding values in Table 6 again indicate that
changes in azimuth for the narrow bandwidth stimuli centered at

11200 Hz were difficult for the monkeys to discriminate. This

phenomenon is thus replicable.

Acoustical Stimuli Geometrically Centered at 13454 Hz

Figure 36 presents the psychophysical functions for the three
observers for the 13454 Hz pure tone. As may be seen in this figure
the functions for Sidney and Oscar are non-monotonic and the functions
are individually distinct for different observers. For Sidney only
the 9° azimuth was detected for greater than 50% of the trials. In
contrast, Oscar detected at a greater than .5 level all changes in
azimuth except for the maximum exchange of 30°. The concept of
threshold is inappropriate for these two cases as the locus of 50%
detectability occurs twice in each function. In the treatment of the
data here, the linear interpolation of threshold was arbitrarily assigned
to the smaller of the two 50% detectability loci. For these functions
the two estimations of threshold are presented in Table 11. The
values for Oscar diverge by 25.8°, while Miko's agreed within 0.3°;
and Miko's psychophysical function satisfied a cumulative normal
distribution.

Figures 37 and 38 present the corresponding psychophysical
functions for stimulus bandwidths of 250 and 500 Hz. As may be seen in
these figures the functions for Sidney only approached the 60% level
Of detection at the greatest angle, 30°. The curves for Miko are

Ikeasonably steep monotonic functions, while for both bandwidths

119

Figure 36

The Probability of Detection as a Function of Azimuth for the
13454 Hz Pure Tone. The ordinate is percent correct detection;
the abscissa is azimuth. The subject's random release rate is
displayed over the zero azimuth point.

120

on

umawi

Amummouov mqoz<

 

 

 

 

NI.

 

Ihoioz<m

om om o. o . on ON 0. o

P _ _ _ b _ _ r
A I ON I ON
o/ .w I ow I 8.

// .

l<. I06 I06
Iom row
I oo. I 8.

530 , 9::

N: 53. oummhzuo

Om

ON

>wzo. m

ION

fov

Tom

Tom

 

IOO.

NO|10313CI 1338800 1N30t|3d

121

mumc .mo. v m .

Emuooum uwnoum nufl3 pmumasoamo on on momma oou wsam> mumsvaHnu

n

huHHMEuo: Eouw musphmmmp QUHS UDDDHoommm mumsvmufinom

 

 

 

 

 

 

 

.nxxxx «o.oma .m.om~ .H.mmm mmumswmuflco
m.m m.om H.6H o.nm Cancun
m.¢ 0.0 H.v N.H coaumHomumucH Mmocwq
HMUmO
.o.Hon .H.mo Io.Hm m.v «mumswmIAnu
H.v m.m m.m v.vH uflnoum
m.¢ 6.5 m.o E.vH coflumaomuwuaH ummcflq
oxflz
.nxxxx In.om .m.HRH .m.mam mmumswmnflnu
m.o m.mm m.HN m.6m uflnoum
m.m m.vm o.vm n.o cofiumaomumucH ummcflq
mmcwflm
ooom oom omm H

 

Aumv suufl36cmm
Rm vmvma um cmnmucmo mgucflzucmm

HH mqmfifi

How mmamcd UHonmmHAB

122

Figure 37

The Probability of Detection as a Function of Azimuth for the
250 Hz Stimulus Bandwidth Centered at 13454 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth point.

 

123

O

 

m<omo

hm

mmDoE

“mummome wn_oz<

ION

IOV

IOO

[Om

 

F OO.

N:

 

 

Om ON O. O
L _ _ _
I. ON
I O?
IOO
row
I OO.
9:2
¢n¢n_ omkazuO
NI OnN IkOiOde

 

 

on om o. o
b _ _ _

18

:9.

I8

I8

I8.
32%

NO|13313CI 1338803 1N3083d

124

Figure 38

The Probability of Detection as a Function of Azimuth for the
500 Hz Stimulus Bandwidth Centered at 13454 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth point.

mm manor.

Amummowe m.._oz<

 

 

 

 

 

 

 

on om o. 6 on cm 9 0 on om o. o
_ _ _ P _ _ _ _ _ _ _ LIL .d
3
H
a om ow m
Iou I N
II—
m .. I9. I? . rov m
8
8
oo oo om mu
T 1
.0.
I08 Iom I cm 1
. 3
O
r . u
00. I oo. roo. O
:88 9:: 5.29m N
~: 3.3. oucmpzuo

N: coo . 592623

126

Figure 39

The Probability of Detection as a Function of Azimuth for the
8000 Hz Stimulus Bandwidth Centered at 13454 Hz. The ordinate
is percent correct detection; the abscissa is azimuth. The
subject's random release rate is displayed over the zero
azimuth point.

mm manor... .

Amuumomov m4¢z<

 

 

127

 

 

 

 

 

 

 

 

on om o. 0 on cm 0. c on 8 o. o
p _ _ p r L _ _ _ _ _
Iou Iom Iom
I ow I 3 I oc
I 00 . low low
I oo Iom I om
AL.

01 r 00. r 8. n1! roo.

:38 9:: 3208

~: :21». cumupzuo

N: OOOO Ibo—Boz<m

NOI13313CI 1038800 .LNBOHBd

128

Oscar's functions are non-monotonic. The functions are more similar
between bandwidths for the same observer than they are between obser-
vers for the same bandwidth.

Figure 39 presents the psychophysical functions for the 8000 Hz
bandwidth geometrically centered at 13454 Hz. As may be seen in this
figure the functions for all observers were monotonic and steep as
is characteristic of the curves generated by wide band stimuli.

The data presented here indicate that at the quarter-octave
point between 11200 Hz and 16000 Hz the psychophysical functions for
Sidney and Oscar share some of the attributes of the functions
obtained at 11200 Hz. That is, changes in the azimuth of narrow
band stimuli were difficult to detect. In contrast, the
psychophysical functions obtained from Miko resembled those produced
by stimuli centered at 16000 Hz. Miko's functions were steep and
monotonic and as displayed in the threshold estimations presented
in Table 11 the values were similar to those obtained by the corres-

ponding bandwidths at 16000 Hz.

Discussion

The results of Experiments I and II demonstrate that the ability
to detect changes in the azimuth of an acoustic source is a function
of stimulus bandwidth at any given frequency. This relationship
was summarized by the linear regression of threshold on the log of
stimulus bandwidth. The correlation coefficient was generally higher
for the probit analysis values than for linear interpolation. This
is likely a reflection of greater sampling error in linear interpolation.
That the correlation coefficient was greater than .90 in 8 of the 9
cases with probit analysis and greater than .8 for the remaining
case indicates that bandwidth accounts for a great deal of the var-

iation in threshold.

Interaural Intensity Differences and the Psychophysical Function

 

The graphs of the psychophysical functions show that, in
general, the detectability of a change in azimuth was monotonically
related to the magnitude of the change. Six functions, however,
showed departures form monotonicity (Figures 18, 20, 34, 36, 37,
38). These nonmonotonic functions were restricted to cases of pure
tone, and the 250 or 500 Hz bandwidths centered at 11200 and 13454 Hz.
The departures from monotonicity were also observer dependent;
different subjects displayed nonmonotonic psychophysical functions
at different acoustic parameters.

Both the occurrence of nonmonotonic functions and their observer

129

130
dependence are consistent with psychophysical observations of binaural
intensity differences. Measurements derived psychophysically (Sivian
and White, 1933), by probe microphones (Harrison and Downey, 1970), or
from an artificial head (Harris, 1972) indicate that the interaural
difference in sound pressure level may not be monotonically related to
azimuth, and as a result two or more locations in space may engender
identical binaural intensity differences. The shapes of these functions
vary with frequency, but are in general highly irregular. A study by
Harris (1972; results presented here as Figure 3) demonstrated that
interaural intensities are a function of the size of the pinna. One
might predict, therefore, that individual differences in the architecture
of heads and external ears would produce intersubject differences in

binaural intensities.

The Binaural Spectral Difference Hypothesis

 

In contrast to the functions generated by pure tones and
narrow bands of noise, the psychophysical functions engendered by
wide band stimuli were consistently steep and monotonic. The effect
of stimulus bandwidth on the auditory localization of azimuth is
interpreted according to the following working hypothesis:

An acoustical stimulus is not transmitted without modification to
an observer's ear. The head and external ear of a subject are acous—
tically opaque and differentially attenuate the energy within the audio
spectrum. The shape of this attenuation, or spectral transfer, function
is an expression of two factors: (A) the architecture of an individual's
head and external ears: and (B) of the angle of incidence of the wave

form on the ear. Because an observer's two ears are spatially

131

separated the angle of incidence of the wave form will be different
for each ear and the resulting spectral transfer functions will differ.
The difference between these functions is the binaural difference
spectrum which will be unique for each location in space if the following
four conditions are satisfied: One, the auditory stimulus must be
composed of more than a single frequency. Two, the ear must be capable
of discriminating at least two of the frequencies within the stimulus
and assess their relative energies. Three, the external ear must
differentially attenuate the energy for each frequency within a stimulus.
Four, this differential attenuation must be unique for each location in
space. Thus according to the working hypothesis if the binaural
difference spectrum is unique for each point in space and if the
vertebrate ear is sensitive to the frequencies at which this uniqueness
is realized then an acoustic event with a defined spectral content
could not be changed in locus without the observer detecting that
change.

The effect of stimulus bandwidth on detectability is represented
in Figure 40. The top panel depicts a rectangular band of noise, the
middle panel the spectral transfer function for each ear, and the
bottom panel the resulting binaural difference spectrum. The employ-
ment of the binaural difference spectrum in localization is easily
visualized. For example, in the simplest case if a subject with
perfectly symmetrical head and ears is presented with a pure tone at
the zero azimuth, identical spectral transfer functions would be
produced at both ears and the resulting binaural difference spectrum would
be zero. However, if this pure tone is changed in azimuth, this change

in location may be detected if the binaural difference spectrum at that

 

 

132

Figure 40

Hypothetical Binaural Difference Spectrum. The top panel repre~
sents a rectangular band of noise, the middle panel the spectral
transfer functions for each ear, the bottom panel the resulting

binaural difference spectrum. See text for details.

133

STIMULUS

 

SOUND PRESSURE
LEVE L (d8)

 

 

 

 

 

 

FREQUENCY

HYPOTHETICAL SPECTRAL TRANSFER
FUNCTION

IOO

RELATIVE ENERGY 6°
20

 

 

 

FREQUENCY

BINAURAL DIFFERENCE SPECTRUM

RIGHT m

RELATIVE
SOUND PRESSURE O
LEVEL (dB)

 

 

 

 

 

LEFT FREQUENCY

Figure 4C

134
point is discriminable different from that engendered at zero azimuth.
If the acoustical stimulus is a band of noise rather than a pure tone,
the detectable change in acoustic locus would occur when the binaural
differences summed over the frequencies contained in the noise band
are discriminably different between the two locations.

The binaural difference spectrum hypothesis at present cannot be
rigorously tested because measurements of the spectral transfer function
have not been made. However, measurements of interaural intensity
differences (Harris, 1972; Harrison and Downey, 1970; Sivian and White,
1933) suggest that the following assumptions are probably true. One,
binaural intensity differences are greater for high frequencies than
for low frequencies for any locations not in the median plane. Two,
the change in binaural intensity differences for any given change in
azimuth is greater for high frequency stimuli than for low frequency
stimuli. Three, for any give azimuth outside the median plane,
interaural intensity differences are greater for wide bandwidth

stimuli than for narrow band stimuli or pure tones.

Psychophysical Data in Support of the Binaural Difference Spectrum

 

Hypothesis

 

The Binaural Difference Spectrum Hypothesis predicts that the
threshold for detecting changes in azimuth for wide stimulus band-
widths is less than the corresponding threshold for narrow stimulus
bandwidths or for pure tones. The results reported here agree with
this prediction. Departures from monotonicity in the psychophysical

functions suggest that the binaural difference spectrum may not be

135
monotonically related to azimuth. Theoretically, these departures
will be restrictied to pure tone and narrow band stimuli, a prediction
also consonant with the obtained data. The major findings of the
research presented here are, therefore, in accordance with the
binaural difference spectrum hypothesis.

The functions relating the threshold to stimulus bandwidth may be
considered the product of both the binaural difference spectrum and
the ability of the binaural neural processor to discriminate changes
in the spectrum. Differences in the slope of these functions at any
center frequency must, then, indicate either a difference in the ability
of the head and ears to generate a binaural difference spectrum or a

difference in the ability of the binaural mechanisms to resolve it.

Psychophysical Results Not Predicted by the Binaural Difference Spectrum

 

Hypothesis

The psychophysical functions for the three monkeys include several
cases where localization was poorer for wider bands of noise than for
narrower bands or for pure tones. For example, compare the proportion
of trials detected by Oscar at 7.8° azimuth using a 250 Hz bandwidth
stimulus with the proportion detected at the same angle using a 500 Hz
bandwidth stimulus (Figures 27 and 28). These results were not predicted
from the binaural difference spectrum hypothesis; for as the bandwidth
of the binaural difference spectrum increases, so should detectability.

There are two likely exceptions to the above generalization. The
first exception may occur if detectability is dependent on the energy
level per cycle of the sitmulus. As the bandwidth of the stimulus is

increased from a pure tone, the sound pressure level per cycle decreases.

136
If, as the sound pressure level for the frequency in question approaches
the threshold of hearing, the ability to detect binaural intensity
differences is degraded, then the change in location of a pure tone may
be more easily detected than the same change for a band of noise unless
there is a large difference in the binaural difference spectrum for the
band of noise at the two locations. In this case, the reduction in
sound pressure level per cycle decreases the neural ability to discri—
minate binaural differences and the decrease is not compensated for by
any increase in the difference between the binaural difference spectra.
The second exception falls within the complex realm of masking. In
general, the detectability of a test frequency is maximally degraded
by a masking stimulus composed of frequencies adjacent to the test
stimulus (Fletcher, 1940; Scharf, 1970). Therefore, a wider band
stimulus may be less localizable than a narrower bandwidth stimulus
if the increase in bandwidth effectively masks the region of the
spectrum where binaural differences would have been expressed. As
in the first case, the degradation in detectability of the wider
bandwidth.is not compensated for by any increase in the difference
between the binaural difference spectra generated at the two test
locations; It is interesting to note that low frequency stimuli are
better maskers of high frequency stimuli than the converse (Mayer,
1894; wegel and Lane, 1924); thus any masking that occurs in a band
of noise is most likely due to the low frequency components of the
noise. Furthermore, the possibility of masking is greater with high
frequency bands because the low frequency components of those bands are
louder than the high frequency components due to the fall off in

sensitivity to high frequencies (Figure 4). However, it is likely that

137
bands located higher on the frequency spectrum would generate more
distinct binaural differency spectra, perhaps offsetting the effects
of masking.

The above considerations suggest that the effect of increasing
stimulus bandwidth on localization depends on the local features of
the difference spectra for specific stimuli. While the data support
the general contention that broad band stimuli are more readily localized
than narrow band stimuli, this is only true as long as the salience
of the binaural difference spectrum increases with bandwidth. Whether
this occurs or not is dependent on the frequency of the band and the
structural characteristics of the observer's head and ears, as well as
upon the ability of the binaural neural processor to appreciate these

differences.

Additional Evidence in Support of the Binaural Difference Spectrum

 

Hypothesis

 

One obvious test of the binaural difference spectrum hypothesis is
the comparison of the ability of monaural and binaural observers to
localize. The literature consistently reports that binaural local-
ization is superior to monaural, although the magnitude of the binaural
advantage varies widely among precedures and laboratories. Gatehouse
and Cox (1972) compared localization in normal binaural subjects to
localization in subjects with clinically defined severe to profound
sensorineural monaural deafness. Binaural subjects localized better
than monaural subjects; but the monaural observers did better than
chance would predict. These results agree with early clinical obser-

vations indicating that, while intersubject variability was high, some

138

monaurally deafened subjects were nearly as accurate in localization
as observers with normal hearing (Angell and Fite, 1901; Pierce, 1901).

In the laboratory Perrott and Elfner (1968) and Elfner, Bothe
and Simrall (1970) employed ear plugs to render normal subjects monaural.
The central findings of those studies was that subjects could initially
localize only using loudness cues. However, after training with appro-
priate feedback, subjects could localize according to spectral differences
generated by the acoustic shadow of the head and the differential atten-
uation of high frequencies. That is, subjects could be trained to respond
to differences in the spectral transfer function for the unoccluded ear.

Butler and Nauton (1967) and Belendick and Butler (1975) have also
studied monaural localization in the laboratory. Their data indicate
the error in localization is a function of the spectral content of the
stimulus. The noise band must contain frequencies greater than 5000
Hz for subjects to display an ability to localize the sound. If the
stimulus was a broad band noise, observers were able to localize all
azimuths within about 20°. Furthermore, the error of localization was
not randomly distributed; the perceived source of the sound was
displaced toward the unoccluded ear. The constant error in location
displayed by Butler's monaural observers may be subject to reduction
by a training program. Freedman and Fisher (1968) have reported that
highly trained monaural observers were nearly as accurate at localizing
a sound source as were binaural observers.

Despite the difficulties in making empirical comparisons between
laboratories (due to differences between the angular positions of
speaker arrays, the spectral characteristics of the transducers, and

the room acoustics) the results from the studies on the monaural

139
localization of azimuth support the contention that while the spectral
transfer function for a single ear may be employed to localize an
acoustic source, binaural processing is always more accurate and less
dependent on the spectral content of the stimulus.

Spectral transfer functions may also account for localization in
the median plane. Pratt (1930) and Roffler and Butler (1968b) have
shown that observers are unable to localize pure tones which differ in
elevation in the median plane. Such differences in elevation can,
however, be detected using bands of noise (Thurlow and Runge, 1967;
Roffler and Butler, 1968a; Butler and Planert, 1976; Hebrank and Wright,
1974a; Gardner and Gardner, 1973). The ability to discriminate changes
in location in the median plane can be explained by corresponding
changes in the spectral transfer function. If this hypothesis is true,
then a noise band with a frequency spectrum adjusted to mimic the spec—
tral transfer function associated with a specific location should evoke
that subjective location irrespective of its actual location. This
hypothesis has qualitatively been confirmed by Blauert (1974) and
Hebrank and Wright (1974b). These two studies differed markedly
in the spectral content of the stimulus associated with the apparent
location of a sound. It is unclear if these differences may be
attributable to differences in the spectral transfer functions
generated by the different ears of different observers.

Psychophysical observations of vertical localization demonstrate
that the spectrum of the auditory stimulus must contain high frequen—
cies. Roffler and Butler (1968a) reported that accurate localization
required frequencies between 7000 and 8000 Hz. Butler and Planert (1976)

demonstrated that braod high frequency noise bands were localized more

140

accurately than narrow noise bands. These results are corroborated by
physical measurements of the spectral transfer function. Shaw and
Terishini (1968) and Shaw (1974) have reported that the ear performs
complex spectral filtering primarily for frequencies greater than 5000
Hz. As a result the differential transmission of acoustic energy is
only expressed in that region of the spectrum greater than 5000 Hz.
Hebrank and Wright (1974a) have reported that the spectral transfer
functions change for different median plane locations for frequencies
between 4000 and 16000 Hz.

In the median plane the angle of incidence of the wave front at
the right and left ear is identical, consequently binaural differences
are at a minimum. However, psychophysical data demonstrate a
superiority for binaural over monaural observers in localization within
that plane (Gardner, 1973; Butter, 1969; Butler and Planert, 1976;
Hebrank and Wright, 1974a). Searle (1973) has reported that in the median
plane the spectral transfer functions are asymmetrical. Thus individual
characteristics of each pinna would generate a binaural difference spec-
trum and the localization of sounds in the median plane may in part
be a binaural phenomenon. Hebrank and Wright (1974a) have argued that
median plane localization is a primarily monaural process. In
support of this thesis they showed that trained monaural observers
could localize ripple spectrum noise as well as binaural observers
could. However, this demonstration did not unequivocally prove that
median plane localization is a monaural process because the power
spectra of the ripple-spectrum sounds, adjusted to imitate spectral
transfer functions, may have masked or removed from the spectrum

those frequencies where binaural asymmetries may have been expressed.

141

This interpretation gains support from the observation that neither

monaural nor binaural observers could localize ripple-spectrum

noise as accurately as they could white noise. Thus a binaural advan-

tage in median plane localization may only occur when the power

spectrum of the stimulus allows for the realization of pinna asymmetries.
The phenomena considered here demonstrate that one advantage of the

binaural difference spectrum hypothesis is its ability by extension to

account for the fundamental observations of auditory localization in

vertical as well as horizontal space.

Duplicity Theory and the Relative Contributions of High and Low

 

Frequency Sensitivity to the Determination of Azimuth

 

According to classic duplicity theory, the localization of low
frequency stimuli depends on binaural phase or temporal differences
while localization of high frequency sounds is due to interaural
differences in intensity (Stevens and Newman, 1934). The literature
on the binaural difference spectrum has considered only the latter
case. An important question, then is what are the relative contributions
of binaural temporal and intensity differences to the localization of
acoustic events at frequencies where both mechanisms may be operative.
Yost, Wightman, and Green (1971) attempted to measure the relative
contributions of high and low frequencies to the detection of a
lateralized transient. In the lateralization procedure the stimuli
are presented through headphones and the acoustic image appears to
reside within the observer's head; it is analytically useful because
interaural differences in phase and intensity may be precisely controlled.

The investigators required their subjects to detect the difference

142

between synchronously presented transients and transients where the
presentation to one ear was delayed. The interaural temporal delay
required for 75% detection was measured as a function of the frequency
content of the transient. The results indicated that performance
depended on the low frequency content of the transient; when only
frequencies above 1200—1500 Hz were presented, a greater interaural
temporal difference was required for the subjects to discriminate
simultaneously presented transients from transients in which the signal
to one ear was delayed. These results were interpreted to support the
position that the discrimination between lateral images depends
primarily on the low frequency content of the stimulus. However,

these results do not resolve the question of the relative contributions
of binaural phase and intensity differences to auditory localization.

The stimuli were dichotically presented and only binaural temporal
differences were introduced. Thus the study only measures the upper
frequency limit for the discrimination of interaural temporal dif—
ferences; their value of 1200 to 1500 Hz agrees well with that obtained
by more direct measurements (Yost, 1974).

Banks and Green (1973) repeated the Yost, Wightman, and Green
experiment with some modification -- the stimuli had real spatial
locations and were not dichotically presented. An anechoic chamber
was fitted with three matched and calibrated loudspeakers. The middle
speaker, located at the zero azimuth, emitted a continuous wideband
noise. The subject was required to identify the location of an acous—
tical transient presented randomly from either the right or left side
of the standard speaker. The observation interval was marked and the

subject provided with feedback. The transient signal, a 0.1 msec pulse,

143

was filtered to emphasize different regions in the stimulus spectrum.
Filtering the frequency content of the pulse had little effect; the
change in threshold was no greater than 1°, with a range of about .6°

to l.6°. Within that small range, threshold increased as the low
frequency content of the signal was attenuated. The results tentatively
indicate that in man binaural phase differences are a better clue to
azimuth than are binaural intensity differences. These results,
however, require two qualifications -- l) the experiment used a 0.1 msec
pulse where most of the energy is below 5000 Hz and a bandpass filter
whose upper limit was 6000 Hz. Consequently, the frequency range where
the binaural difference spectrum is best expressed was excluded from
analysis. 2) Masking noise from the referent speaker may have hindered
detection of the higher frequency noise bands. Banks and Green
observed that when the sound pressure level of the masking noise was
increased, threshold for the high frequency bands was elevated while
that for the low pass band was unchanged. Considering these qualifica-
tions, additional research is required to fully demonstrate the

relative contributions of binaural temporal and intensity differences

to the localization of complex stimuli.

Auditory Localization in an Evolutionary Perspective

 

Masterton contends that the mammalian attribute of an expanded
sensitivity range to high frequency stimuli is a consequence of selection
for a superior capability to localize acoustic sources (Masterton,
Heffner, and Ravizza, 1969; Masterton and Diamond, 1975; Masterton, 1974).
If high frequency hearing is an adaptation for auditory localization

then its contribution to a spatial sense should be so expressed. In a

 

144

restricted context this hypothesis may be tested by the research
presented here. The detection of a change in azimuth may be facilitated
by high frequency sensitivity. That is, for a given stimulus bandwidth
the binaural difference spectrum may lead to greater spatial acuity for
high frequency bands than for low frequency bands. Thus selection may
have operated to evolve an external ear and neural binaural processor

for the optimization of spatial acuity. One way to examine this
hypothesis is by the construction of a binaural efficiency quotient.

It is generally recognized for any given stimulus bandwidth the

number of sensory cells encompassed in that band, (or alternatively, the
area on the basilar membrane) decreases as the center frequency of the
band is increased (von Bekesy and Rosenblith, 1951; Scharf, 1970). Thus
the frequency scale is compressed and higher frequencies are subserved
by a smaller proportion of the neural substrate than are lower
frequencies. This compression is approximately expressed by a logarithmic
scale. A behavioral consequence of this comparison is that the just
detectable differences in frequency between two tones increase as the
frequency of the stimuli is increased. The just discriminable difference
in frequency or frequency difference threshold ( AIf) has been determined
for the macaque. Stebbins (1973) reported that the frequency

difference threshold was approximately 50 Hz, 90 Hz, and 200 Hz for
referent frequencies of 8000 Hz, 11200 Hz, and 16000 Hz respectively.
These values may serve as an approximate index of the relative density
of the neural represnetation at each frequency. The relative efficiency
(E) of the binaural difference spectrum may be arbitrarily defined as

the ratio of the threshold for any given stimulus bandwidth to the

frequency difference threshold for the same center frequency. The

145

efficiency quotient for the binaural difference spectrum for the mean
threshold of the three monkeys for the 250 Hz and 500 Hz stimulus
bandwidths for each center frequency are given in Table 12. Though this
quotient is rough and approximate it demonstrates that at 16000 Hz the
detection of a change in azimuth for narrow stimulus bandwidths is
mediated by relatively less cochlear tissue than at lower frequencies.
In this context the results of this research are consistent with the
hypothesis that high frequency sensitivity is an adaptation for sound
localization. The binaural difference spectrum efficiency quotient
does not, however, differentiate between the relative contribution of
the magnitude of the interaural spectral differences or of the acuity
of the neural processor. Thus selection may have operated either
centrally or peripherally. In addition the contribution of high
frequency sensitivity to auditory localization may be more dramatically
expressed in the determination of elevation in the median plane.
Consequently this research only partially demonstrates the role of

high frequency sensitivity in the localization of acoustic events.

Auditory Localization in the Bioacoustical Context

 

Marler (1965) has observed that throughout the class mammalia
alarm calls and distress calls share many acoustical features across
species. That is, alarm calls are characterized as shrill narrow band
barks, while distress calls are typically wide band screeches and
screams. Marler (1955, 1957) has made similar observations regarding
the vocalizations of many birds and has speculated that the acoustical
features of a call may render it more or less readily localized. This

proposition was experimentally explored by Konishi (1972) who found

 

[way -.-I

__II

146

TABLE 12
Binaural Difference Spectrum Efficiency Quotient a

Center Frequency (Hz)

 

 

 

8000 11200 16000 Stimulus Bandwidth
Probit .25 .24 .06
_ 250 Hz
Linear Interpolation .24 .22 .045
Probit .21 .21 .06
500 Hz
Linear Interpolation .21 .19 .04

 

a The smaller the quotient the greater the relative efficiency.

147

that barn owls would strike more accurately at an acoustic prey as the
bandwidth of the stimulus was increased. The research presented here
corroborates Konishi's results and supports Marler's contention. As

a first approximation species specific vocalizations may be described
by their bandwidth and dominant frequency; and high frequency, narrow
band, more tonal stimuli are harder to localize than are wide band Ph?
noisier stimuli. These results are consistent with the proposition .‘J
that the natural variation in the acoustical features of a species's f

vocal repertoire may be tailored to minimize or exploit location cues

 

as appropriate for the social context.

APPENDIX A

 

£43

148

Appendix A

Acoustical Calibration

 

All measurements were made in an anechoic chamber. The acoustical
signal for noise bands and for pure tones was measured in the
following manner: A Bruel and Kjaer 1/2 inch free-field microphone
(4133) was mounted on a microphone stand 10 cm from the transducer on
the speaker's axis. The response of the microphone was measured by a
Hewlett Packard (3590 A) wave analyzer. The wave analyzer was
adjusted to measure the sound pressure in a 100 Hz bandwidth
"window" which was slid along the frequency axis. The acoustic spectrum
of the stimulus measured by the wave analyzer was concurrently recorded
by a Moseley (135C) X-Y plotter. Figure 41 presents the acoustic
measurements for noise bands centered at 16000 Hz. The ordinate is
sound pressure level as recorded by the wave analyzer in the 100 Hz
bandwidth window. The abscissa is audio frequency represented on a
logarithmic scale. The power spectra presented here are for noise bands
which were attenuated to maintain a constant sound pressure level of
93 dB re ZOIfLPa at 10 cm. As a result, as the stimulus bandwidth was
increased the sound pressure level per cycle decreased. As may
be seen in Figure 41 the narrow bands of noise were essentially
rectangular in shape while the wider bandwidths were less regular
in shape and the shoulders of the curves were not as steep. The

degradation of the stimulus for the wider bandwidths was largely due to

 

 

149

Figure 41

Acoustical Calibration at 16000 Hz. The ordinate is relative sound
pressure level as measured by the wave analyzer in a 100 Hz bandwidth
window. The abscissa is audio frequency represented on a

logarithmic scale. The power spectra were recorded from stimulus
bandwidths of 250, 500, 1000, 2000, 4000, 8000 Hz respectively.

The high and low cutoff points for each bandwidth are indicated

above the corresponding power spectrum.

150

 

.v mmaei

.er. >Ozwao um...

#N 4 NW 0..

A
I 11

O. NN O. NN O.

5 is; S... 2.3..

 

 

 

 

 

 

 

 

 

 

v N . n. 0N.

.N... .3 7.50.30240

I

 

(SNOISIAIO 8P0!) EHOSSBHd ONROS 3AI1V138

151

an increase in the noise feedthrough from the analog audio multiplier

as the bandpass of the Allison filter was increased. The high and low
frequency cutoff points for the noise bands as presented in Table l

are represented in Figure 41 above the corresponding power spectrum.

The sound pressure level for the widest bandwidth, 8000 Hz, was 26 dB
down at one octave out from the cutoff points. The slopes for the narrow
bandwidths were much steeper and the sound pressure level fell off as
much as 75 dB per octave to the ambient noise level. Corresponding
acoustical measurements for stimuli centered at 8000, 11200, and

13454 Hz were made. The acoustic spectra for these signals are not
reproduced here. These signals were generated by simply changing the
oscillator carrier frequency to the appropriate value. As a result,

the major difference between the curves generated at 16000 Hz from those
centered at other frequencies was their corresponding location on the
frequency axis. Acoustical measurements of pure tones (not reproduced
here) demonstrated that harmonic distortion was at least 60 dB down

from the fundamental, and electrical measurements indicated that most
of the signal recorded at harmonic intervals was generated by the

oscillator.

Equating Transducers

As previously stated it was important to minimize the contribution
of speaker differences to the detection of a change in acoustic locus.
The matching of speakers is largely opportunistic and the relevance
of an acoustical difference between two speakers must ultimately be
determined by the bahavior of the subject - do they sound different?

As a result, the speakers were initially matched by the ear of the

 

 

 

152

Figure 42

Acoustical Measurements of 5 Transducers for the 8000 Hz Stimulus
Bandwidth Centered at 11200 Hz. The ordinate is relative sound
pressure level as measured by the wave analyzer in a 100 Hz band-
width window. The abscissa is audio frequency represented on a
logarithmic scale. The response characteristic for each transducer
is shifted 20 dB with respect to one another.

 

RELATIVE SOUND PRESSURE (ZOdB DIVISIONS)

 

153

<8kHZ BAND
at

H200 HZ

 

 

ITTIITII I]
3

IO 20 3O
FREQUENCY (kHZ)

FIGURE 42

 

154

experimenter and this match was subsequently confirmed or rejected by
the behavior of the subjects. Figure 42 presents the acoustical
measurements for five speakers for the 8000 Hz bandwidth stimulus
centered at 11200 Hz. The curves are shifted 20 dB with respect

to one another. The intensity of the signal between transducers was
subsequently equated within 1 dB. This was performed by mounting

the microphone in the sound field of the transducer. Measurements
made by this system were easily reproduced and the sound pressure

levels for each speaker were adjusted by Daven (PT 324-M) attenuators.

 

 

APPENDIX B

155

Appendix B

The Proportion of Trials Detected at Each Azimuth

 

 

”23

156

 

 

 

 

 

 

 

mmma\qhmfl omoa\emm Hoaa\maq Hm~H\Hom «RMH\HNH umomo
omm\oam Svn\mao mnn\mnm mmm\oma Rom\oHH ox“:
mvma\vmma HMHH\mvm ammoaxmmv mmva\no mmn\vm smauflm
nucflzocmm a: com
mam\oqm ~mn\mfiw omn\mmm oom\H~H mmoa\mm umomo
mmaaxmoaa H~m\mmo omm\nnm mmHH\GGH Avaa\ova ox“:
omma\mmafl Hmm\oon momm\om~ omma\on vo>\nm smcoflm
a06H36amm a: omm
omn\aan vom\on moo\nqm cmm\mmH N~m\nm Amomo
mam\omn 6H5\m~o R~n\mnv vom\mma moexmm oxflz
ohn\-a m~o\mav mhmo\mmm mom\bv mwa\oa smzeflm
rpoflzcamm R: H
o.mm N.EH o.m m.~ o.o

 

.mwmumwov mamcd

Nm ooom um commucmo mnuow3ocmm How noncommmm kuoe ou pomuuou

MH mqmdfi

mo mowumm

 

 

 

 

157

 

 

 

 

 

 

VIII?

.oo.m um swap umnumu oo.NH um 600mm» smcvflm m
mom\mm o¢v\mmv mmm\nmm oflvxvom mmvxom Amomo
Avmxvv mmm\omm mwm\~mm oom\mma mmm\~m ox“:
omv\mH Nmm\amm mmm\vmm mmm\nn mmm\n smcnflm

spofl36cmm a: ooom
mmm\vm omv\wmq mmm\mnm mmvxmmfi omm\mm umomo
mam\¢m oom\oom mHm\mmm mnm\mea oov\mm oxflz
Rmm\mm omm\mam wpm\mmm mvm\mm oo~\h smcvflm

auoflzocmm n: 0004

a N.RH o.m m.~ o.o

mvoa\avoa ovm\mmm mmw\vmm neoa\oma mRHH\mm “momo
oomxnmm vmh\mmn me>\mov vmm\vam ommxvoa oxuz
mmm\omm «mmxoam amon\aov mmoaxmaa Rmmxoa smcvfim

apoflaocmm N: ooom
mmm\nwm mos\eno vam\vov mflmxvva mmoH\Hn “memo
mmoflxmvoa mmmxsfim onm\mmm mfiaaxamm onaa\ova oxflz
amoa\mnm Hon\mho ammo\omv mooa\mo Hmm\HH smcuflm

 

SHUHSUCMm Nm OOOH

..o.uaoo. maImqmae

158

 

 

 

 

 

 

 

 

 

omm\mmm momxoma mwm\mHH amm\qm mmm\om Amomo
mom\vom m>m\mmm m04xaha nmmxvm vmn\nm oxfiz
mmvxomm hav\mam. v~m\aoa Hmmxma me\mm smanflm
gucfizocmm u: oOOH
Hmm\mvm mmm\moa v~¢\mna mam\mv mmk\maa Amomo
amm\amm wvw\omm Rmvxmma mmmxoq Hoaa\vma oxflz
mmv\mom mmm\aha oomxon mmvxm Hmmxba mwcnflm
gpcwseamm R: oom
mmv\mov mom\onm 4mm\mma mmvxmm evma\mna Amomo
mmv\mmm mom\omm omo\mmH mhm\mm omoa\mma oxflz
mnvxomm mmwxmm mmm\mm ohm\oa Ham\mm smqofim
auwfiseamm R: omm
o~m\nm~ omm\o- vnm\hm Hoa\mm mom\mma umomo
Rmv\mom mmm\mm~ mmm\mma mv~\mm mmoa\ova oxflz
mom\aom mmmxoom mmm\mma Hmm\hv mom\mm smcufim
nueflzeqmm u: A
m.m~ m.mH m.n m.H o.o

 

Nm OONHH “M UOHQUGUU Mfivflufiuqflm .HOM mmmflommmm HMfiOB On. HUQHHOU MO MOflHmm

.mmmumoo. mamcm

wH mflmdfl

159

 

 

 

 

 

Achxaoe onm\mom nmo\vfio Hmn\mma omn\mm Rmomo

onm\mom mam\mom mvm\oam mao\mma mom\mo or“:

oemxmmm Hmm\mvm mmmxoem mmv\qa masxm mmccfim
nucflzocmm R: ooom

mmv\mmv omm\omm Hum\mom mmq\mm «mmxmm “Como

mmm\mmm mom\m0m mmmxavm mmv\m¢ Rmm\am oxfiz

oamxoam wmv\hmv Hmm\mmm mmm\ma GAG\RH swaofim
nuoflzwcmm um ooov

ovm\Hmm HHm\RRm mmmxmam R~m\om mom\mm Dmomo

mmmxamm ammxamm mhm\vmm mmmxmm omo\mm ox“:

Roo\mvm osm\mov omm\mm~ vonxmm mom\na mocvflm

 

 

Sunfisflamm Mm ooom

A.@.UC00V #4.. "NJ—”$9

160

 

 

 

 

 

 

 

 

 

 

F E

Hov\vmm movxqmm hmm\mnv I,vam\am omo\mv Amomo

Hmm\amm mmm\Hmm mmv\mmm mmv\om mmv\mm 0x“:

oom\mmm mov\¢om oom\mmH mmv\mm mnv\ma smccfim
nuofizccmm u: coca

mo4\mnm nmm\mov mmm\omm mmw\n~H mom\vb umomo

mmm\-m mvm\mom man\~mm m¢n\mh «mm\mv oxfiz

omv\vme omv\va oom\mam mmm\vv Ramxma smccflm
aucflzocmm R: oom

4H4\~Em mmm\mam mhv\mvm mmv\oo ammxnm Amomo

mmm\mmm mom\mmm vemxmmm m¢m\mo oam\mm 0tz

Hoqumm mmv\mmm oao\ama chmxmm ovm\vm Smcnam
apoflsocmm u: omm

Emv\nvm mov\qmm mmv\0Hm ommxoma «mm\mm umomo

mmq\mmv H>v\mmv mmo\omm vmo\v~H 4mm\HE ox“:

mom\vmm mmoxmem Han\ooa m~m\m mmo\ma smcoflm
nuoflzocmm R: H

m.mm m.mH m.n m.a o.o

 

.mwwumwov mamCN

mH mqmde

um ooowH um oouwucwo mnuow302mm Mom mmmcommmm Hopes on uomuuoo

mo mowumm

161

 

 

 

 

 

 

mmm\mmm moq\mov RHm\Rom mov\mo Raoxem Amomo
oam\oam wmmxmmm cavxmmm mmv\om HHG\N~ oxflz
mnm\onm mmm\amm omwxmmm mnvxm mmv\v smcnflm
apoflzocmm R: ooom
Nom\H0m omm\mam omm\vnm HHG\HV Rmmxmfl Amomo
cmv\mme anv\onv Hmm\vmm «Ao\mv mom\mm ORA:
Rmm\nmm mmm\mam mmm\mmm cmv\0H emv\a Emcwflm
zucflzccmm u: ooov
mov\mov Hmvxamv mmmxnfim mmmxmv vfioxqm umomo
mmvxvmq mmm\anm omv\omm mom\mm HHG\Rm oxflz
mem\mvm mvm\mmm mmv\mm~ mmv\o~ mmm\m Ewcoflm
nuoflzccmm um ooom
..6.D:00. ma.mqm¢e

162

 

 

 

 

 

 

 

 

 

 

vofi\vofl mma\mma mom\mo~ va\Hm mm~\ea umomo
«na\vaa mwa\mma momxvoa mmmxm Ro~\v oxfiz
mom\Hom mem\bnm Hmm\mam Rhmxm mmm\m smccflm
spoflzocmm a: ooom
m.m~ m.mH m.n m.H o.o
.momumoo. Gamma
«oa\mm om\mn waa\qo mma\mfl HMH\HH oxflz
nuwflzwcmm u: oom
oam\mam omq\omm Hmmxvmm mnm\aom men\mn “Como
moa\hm mm\om mma\am mma\ma owa\ma 0AA:
mam\~aa Rmmxvoa mmmxnva mmmxmm nov\oa mmcwflm
spwfisccmm a: 0mm
ow m.mm m.mH m.h o.o

 

.mmwummo. mamc¢

III:

coflumoflammm "Mm oomHH um owuwuawo mapofl3ocmm How mmmcommmm

0H mdmdB

Hmuoe on uomnuou mo m0flumm

163

 

 

 

 

 

 

 

 

 

vmv\nqv vnv\mov omo\mnm mmm\mn Rmo\mm Amomo

omm\mmm ommxmmm mmm\aom H~m\nn vvv\mm oxflz

om¢\mnv mamxmom vmm\mmv mmm\mv vmm\oa Emccflm
auuflzocmm R: ooom

mmm\vmm movxmma mom\mmm mmo\qmm omm\vm Amomo

mav\oo¢ ma¢\n¢m mae\mqm mmmxmh Rmexnm oxflz

mam\om~ Ram\oma RGG\VRH mmmxnm mvm\ma Emcnflm
nuoflswcmm a: oom

mmv\m~m vmq\anm mam\mmm «mm\mom omm\on Amomo

mvw\omv om«\mom qmm\mam mmo\va owm\ho oxflz

momxamm Hmm\vmm oao\mmm mmmxom ovm\MH Smaoflm
nuoflzoamm a: 0mm

mov\mmH oov\mom mmmxmmm mmm\A0m oom\~m umomo

Hme\mmm oav\oam onm\oma mov\nm mvm\vv 0xflz

oom\mma mmm\mvm mmo\o>m mmmxmm Gamxoa smccflm
nueflzeamm u: H

m.mm m.mH m.h m.H o.o

 

.mmmqupv mamcfl

PH mdmdfi

Mm vmvma um omumusmo mnuoflzocmm How mwmcommwm Hmuoe ou uowuuou mo moHumm

LI ST OF REFERENCES

 

List of References

Angell, J. R. and Fite, W. The monaural localization of sound.
Psychol. Rev., 8, 225-246, 1901.

 

Banks, M. S. and Green, D. M. Localization of high— and low-
frequency transients. J. Acous. Soc Amer., 53, 1432—1433, 1973.

 

Batteau, D. W. The role of the pinna in human localization. Proc.
Royal Soc., 168, No. 1011, Series B, 158-180, 1967.

von Bekesy, G. and Rosenblith, W. A. The mechanical properties of
the ear. in S. S. Stevens (ed.) Handbook of Experimental
Psychology, New York: Wiley, 1951.

 

 

Belendiuk, K. and Butler, R. A. Monaural localization of low-pass
noise bands in the horizontal plane. J. Acous. Soc. Amer.,
58, 701-705, 1975.

 

van Bergeijk, W. A. The evolution of vertebrate hearing. In Neff,
W. D. (ed.) Contributions to Sensory Physiology New York:
Academic Press, 1967, pp. 1-49.

Blauert, J. Raumliches Horen. Leipzig, Germany: 5. Hirzel Verlag,
1974. Cited in Schroeder, M. R. Models of hearing, Proc. IEEE,
63, 1332-1350, 1975.

 

 

Butler, R. A. Monaural and binaural localization of noise bursts
vertically in the median sagital plane. J.Audit. Res.,3,
230-235, 1969.

 

Butler, R. A. and Nauton, R. F. The effects of stimulus sensation
level on the directional hearing of unilaterally deafened persons.
J. Audit. Res., 7, 15-23, 1967.

 

Butler, R. A. and Planert, N. The influence of stimulus bandwidth on

localization of sound in space. Percept. Psychophysics, 19, 103-
108, 1976.

 

Brindley, G. S. Physiology of the Retina and Visual Pathway. Baltimore:
The Williams and Wilkins Co., 1970.

Casseday, J. H. and Neff, W. D. Localization of pure tones. J. Acous.
Soc. Amer., 54, 365-372, 1973.

164

165

Elfner, L. F., Bothe, G. G., and Simrall, D. S. Monaural localization:
Effects of feedback, incentive, and interstimulus interval.
J. Audit. Res., 10, 11-16, 1970.

 

Finney, D. J. Probit Analysis. London: Cambridge University Press,
1952.

 

Fletcher, H. Auditory patterns. Rev. Mod. Physics,12, 47-65, 1940.

 

Freedman, S. J. and Fisher, H. G. The role of the pinna in auditory
localization. In S. J. Freedman (ed.) The Neuropsychology of
Spatially Oriented Behavior. Homewood, Illinois: Dorsey Press,
1968, pp. 135—152.

 

 

Gardner, M. B. Some monaural and binaural facets of median plane
localization. J. Acous. Soc. Amer., 54, 1489-1495, 1973.

 

Gardner, M. B. and Gardner, R. S. Problem of localization in the
median plane: Effect of pinna cavity occlusion. J. Acous.
Soc. Amer., 53, 400-408, 1973.

Gatehouse, R. W. and Cox, W. Localization of sound by completely
monaural deaf subjects. J. Audit. Res., 12, 179-183, 1972.

 

Grimm, R. J. Catalogue of sounds of the pig-tailed macaque.
J. 2001., 152, 361-373, 1967.

Guilford, J. P. Psychometric Methods. New York: McGraw-Hill Book
Co., 1954.

 

Harris, J. D. A florilegium of experiments on directional hearing.
Acta Otolaryngologica Supplement 298, 1972.

 

Harrison, J. M. and Downey, P. Intensity changes at the ear as a
function of the azimuth £0 a tone source: A comparative study.
J. Acous. Soc. Amer., 47, 1509-1518, 1970.

 

Hebrank, J. and Wright, D. Are two ears necessary for localization
of sound sources on the median plane? J. Acous. Soc. Amer.,
56, 935-938, 1974a.

 

Hebrank, J. and Wright, D. Spectral cues used in the localization
of sound sources in the median plane. J5 Acous. Soc. Amer.,
56, 1829-1834, 1974b.

 

Heffner, H. E. The effect of auditory cortex ablation on sound
localization in the monkey (Macaca mulatta). Unpublished
Doctoral Dissertaion, Florida State University, 1973.

 

Henson, 0. W. Comparative anatomy of the middle ear. In Keidel and
Neff (eds.) Handbook of Sensory Physiology. Vol. VII, Berlin:
Springer Verlag, 1974.

 

 

166

Konishi, M. Locatable and nonlocatable acoustic signals for barn owls.
The American Naturalist, 107, 775-785, 1972.

 

Manley, G. A. A review of some current concepts of the functional
evolution of the ear in terrestrial vertebrates. Evolution,
26, 608-621, 1973.

Marler, P. Characteristics of some animal calls. Nature, 176, 6—8,
1955.

Marler, P. Specific distinctiveness in the communication signals of
birds. Behaviour, 11, 13-39, 1957.

Marler, P. Communication in monkeys and apes. In DeVore, I. (ed.)
Primate Behavior: Field Studies of Monkeys and Apes. New York:
Holt, Rinehart, and Winston, 1965, p. 544-584.

 

Masterton, R. B. Adaptation for sound localization in the ear and
brainstem of mammals. Federation Proc., 33, 1904-1910, 1974.

 

Masterton, B, Heffner, H. E., and Ravizza, R. The evolution of
human hearing. J. Acous. Soc. Amer., 45, 966-985, 1969.

 

Masterton, B. and Diamond, I. T. Hearing: Central neural mechanisms.
In Carterette and Friedman (eds.) Handbook of Perception, Vol. 3,
New York: Academic Press, 1973.

 

Mayer, A. M. Researches in acoustics. Lond. Edinb. Dubl. Phil. Mag.,
37, Ser. 5, 259-288, 1894.

 

Mills, A. W. On the minimum audible angle. J. Acous. Soc. Amer.,
30, 237-246, 1958.

 

Mills, A. W. Lateralization of high frequency tones. J. Acous. Soc.
Amer., 32. 132-134, 1960.

 

Moore, P. W. B. and Au, W. W. L. Underwater localization of pulsed
pure tones by the California sea lion (Zalophus californianus).
J. Acous. Soc. Amer., 58, 721-727, 1975.

 

 

Nordlund, B. Physical factors in angular localization. Acta Otolaryng.,
54, 75-93, 1962.

 

Nordlund, B. Studies of stereophonic hearing. Acta Otolaryng., 56, 493-
499, 1963.

 

Palin, J. and Gourevitch, G. An improved narrow-band noise source.
Electroenceph. Clin. Neurophysiol., 29, 523-534, 1970.

 

Perrott, D. R. and Elfner, L. F. Monaural localization. J. Audit.
Res., 8, 185-193, 1968.

167

Pierce, A. H. Studies in Space Perception. New York: Longmans, Green,
and Co., 1901.

 

Pratt, C. C. The spatial character of high and low tones. J. Exptl.
Psychol., 13. 278-285, 1930.

Pumphrey, R. J. Hearing in insects. Biol. Rev., 15, 107, 1940.

Rayleigh, Lord Our perception of the direction of a source of sound.
Nature (Lond.), 14, 32-33, 1876.

Rayleigh, Lord Theory of Sound. 2nd Ed., V01. II, p. 440-443, New York:
Dover Publication, 1945.

 

Roffler, S. K. and Butler, R. A. Factors that influence the localization
of sound in the vertical plane. J. Acous. Soc. Amer., 14, 1255-1259,
1968a.

 

Roffler, S. K. and Butler, R. A. Localization of tonal stimuli in the
vertical plane. J. Acous. Soc. Amer., 43, 1260-1266, 1968b.

 

Rowell, T. E. and Hinde, R. A. Vocal communication by the rhesus monkey
(Macaca mulatta). Proc. Zool. Soc. Lond., 138, 279-294, 1962.

 

 

Scharf, B. Critical bands. In J. V. Tobias (ed.) Foundations of
Modern Auditory Theory, Vol. 1, New York: Academic Press, 1970,
pp. 159-202.

 

 

Searle, C. L. Cues required for externalization and vertical
localization. J. Acous. Soc. Amer., 54, 308(A), 1973.

 

Shaw, E. A. G. The external ear. In Keidel and Neff (eds.)
Handbook of Sensory Physiology, V01. V/I. Berlin: Springer
Verlag, 1974, pp. 455-490.

 

Shaw, E. A. G. and Teranishi, R. Sound pressure generated in an
external ear replica and real human ears by a nearby point source.
J. Acous. Soc. Amer., 44, 240—249, 1968.

 

Sidman, M. Tactics of Scientific Research: Evaluating Experimental Data
in Psychology. New York: Basic Books, Inc., 1960.

 

 

Sivian, L. J. and White, S. D. On minimum audible sound fields.
J. Acous. Soc. Amer., 4, 288-321, 1933.

 

Snapper, A. G., Knapp, J. 2., and Kushner, H. K. Mathematical description
of schedules of reinforcement. In W. N. Schoenfeld (ed.) The Theory

 

 

of Reinforcement_Schedules: New York: Appleton- Century-Crofts,
1970, pp. 247—275.

 

Stebbins, W. C. Hearing of Old World monkeys (Cercopithecinae).
Am. J. Phys. Anthrop., 38, 357-364, 1973.

 

 

168

Stebbins, W. C. Hearing of the anthropoid primates: A behavioral
analysis. In D. B. Tower (ed.) Human Communication and Its
Disorders. New York: Raven Press, 1975, pp. 113-124.

 

Stevens, S. S. and Newman, J. D. The localization of pure tones.
Proc. Nat. Acad. Sci., 20, 593-596, 1934.

 

Thurlow, W. R. and Runge, P. 8. Effect of induced head movement on
localization of direction of sounds. J. Acous. Soc. Amer.,
42, 480-488, 1967.

 

Wegel, R. L. and Lane, C. E. The auditory masking of one pure tone
by another and its probable relation to the dynamics of the inner
ear. Phys. Rev., 23, Ser. 2, 266-285, 1924.

Whaley, D. L. and Malott, R. W. Elementary Principles of Behavior.
New York: Appleton-Century-Crofts, 1971.

 

Woodworth, R. S. Experimental Psychology. New York: Holt, 1938,
p. 521.

 

Wright, D., Hebrank, J. H., and Wilson, B. Pinna reflections as
cues for localization. J. Acous. Soc. Amer., 56, 957-962,
1974.

 

Yost, W. A. Discriminations of interaural phase differences.
J. Acous. Soc. Amer., 55, 1299-1303, 1974.

 

Yost, W. A., Wightman, F. L., and Green, D. M. Lateralization of
filtered clicks. J. Acous. Soc. Amer., 50, 1526-1531, 1971.