1 Introduction

The present monograph deals mostly with the results of auditory ex­
periments: those which, under defined conditions, examine auditory
perceptions. Here and throughout this work, we restrict ourselves to
conscious perceptions.
The different branches of science approach perceptual phenomena
from different perspectives. Fundamental distinctions are between:

l. The approach from the perspective of the theory of cognition: "What

are perceptions?"
2. The approach from the perspective of the natural and technical
sciences: "How do (particular) perceptions come about?"

The first question concerns the essence of perception. The answer

to it indicates that, at the moment of perception, the perceived and
the perceiver encounter each other in such a way that the perceiver
becomes conscious of what is perceived. If the common terms "subject"
and "object" replace the terms "perceiver" and "perceived," one arrives
at the statement that perception is the conjunction-in-opposition of
subject and object or, briefly, the "subject-object relationship" (Lungwitz
1923, 1933a, Bense 1961). There is no perception without both a subject
and an object.
The second question concerns the physical, physiological, and psy­
chological details of perception. Since perception is inseparably linked
to the physiological processes of organisms, the question may be asked
in a more detailed way: "What conditions must be fulfilled inside and
outside an organism for a particular object to appear in the sensory
world of that organism?" This is the general question that underlies
the present study. The organisms to be studied are the human subjects
of the auditory experiments; the external conditions are the stimuli (for
the most part sound signals) presented to these·subjects; and what the
subjects hear are the objects of interest.

Downloaded from [Link] by guest on 11 March 2024

 2 Introduction

1.1 Auditory Events and Auditory Space

Human beings are primarily visually oriented. Compared with the visual
world of the human being, the worlds of the other senses (auditory,
tactile, etc.) are much less highly developed. Correspondingly, concepts
and descriptions are based primarily on visual objects (Becker 1971).
For example, we say "the bell sounds" and not "the sound bells";
similar observations may be made about scientific and technical
descriptions.
The German Standard DIN 1320 (1959) defines "sound" from this
perspective as "mechanical vibrations and waves of an elastic medium,
particularly in the frequency range of human hearing (I 6 Hz to 20
kHz)." This is a description of physically measurable changes of position,
primarily perceived visually. What is heard, what is perceived auditorily,
is only implicitly included in the final phrase "particularly in the fre­
quency range of human hearing."
Satisfactory description of auditory experiments requires a more
thorough analysis of the state of affairs that underlies this definition.
The underlying assumption is that a normal human being generally
hears something when in a medium in which vibratory or wave phe­
nomena are occurring whose frequency is between 16 Hz and 20 kHz.
This does not, however, mean that the vibrations and waves of the
medium are what is heard. For example, if the ears of a human ex­
perimental subject were temporarily plugged or were to suffer from a
loss in sensitivity, then this person would hear nothing, although the
vibrations and waves of the medium would still be perceptible (for
example, by means of physical measuring instruments).
For the sake of terminological clarity, and as has been proposed
previously (Blauert 1966), the word "sound," as defined in the DIN
standard, will be used in this monograph to describe the physical aspect
of the phenomena of hearing-particularly in the compound "sound
event." Terms such as "sound source," "sound signal," and "sound
wave" will always be used to describe physical phenomena that are
characteristic of sound events. What is perceived auditorily will be
denoted by the adjective "auditory," as in the term "auditory object"
or, preferably, "auditory event."
It is commonly believed that auditory events are caused, determined,
or elicited by sound events. Auditory events, however, also occur without

Downloaded from [Link] by guest on 11 March 2024

 1.1 Auditory Events and Auditory Space 3

corresponding mechanical vibrations or waves. Examples are the ringing

in the ears that occurs under certain disease conditions (tinnitus) and
the auditory events perceived when the acoustic nerve is artificially
stimulated. More generally, the fact that a sound event does not nec­
essarily produce an auditory event, and that not every auditory event
is connected with a sound event, must exclude the interpretation that
one leads to the other in a causal sense. The human visual system is
more highly differentiated than the human auditory system, and so the
visual world is more highly differentiated than the auditory world. The
common belief that sound events cause auditory events is, consequently,
understandable, but nonetheless incorrect.
A careful description would go no farther than to say that particular
precisely definable sound events and particular precisely definable au­
ditory events occur with one another or one after the other under certain
precisely definable conditions (Blauert l 969a,b). It may also be said
that sound events and auditory events are related to each other, as­
sociated with each other, or classified in relation to each other.
Sound events and auditory events are distinct in terms of time, space,
and other attributes (Lungwitz 1933b); that is, they occur only at par­
ticular times, at particular places, and with particular attributes. The
concept of "spatial hearing" acquires its meaning in this context. Since
this concept implies that auditory events are inherently spatially distinct,
it is in fact tautological; there is no "nonspatial hearing." Defined more
narrowly, the concept of spatial hearing embraces the relationships
between the locations of auditory events and other parameters-par­
ticularly those of sound events, but also others such as those related
to the physiology of the brain.
The "locatedness" of auditory events may be more or less precise.*
Thus, for example, the position and spatial extent of the auditory event
called a "sustained tone" in a reverberant room cannot be determined
with any precision. The locatedness of the tone is diffuse. A click in
an anechoic chamber, on the other hand, is precisely located and sharply
limited in extent. The locatedness of an auditory event is described in
terms of its position and extent, as evaluated in comparison with the

*We introduce the new term "locatedness" for the spatial distinction in order to avoid
certain misunderstandings that might arise from use of "localization" at this point (see
Blauert 1969a). The word "localization" will be newly defined in another context in
section 2.1.

Downloaded from [Link] by guest on 11 March 2024

 4 Introduction

positions and extents of other objects of perception, which might be

other auditory events or the objects of other senses-in particular,
visual objects.
Auditory events may occur at positions where nothing is visible: in
connection with sounds inside one's own body or another opaque object,
behind walls, beyond the horizon, in the dark, and so forth. Auditory
events, in contrast to what is perceived visually, occur not only in front
of the observer, but in all directions from the person who is perceiving
them. Similar comparisons may be made with what is perceived via
the other senses, such as the tactile, olfactory, and gustatory senses.
The totality of all possible positions of auditory events constitutes au­
ditory space. The word "space" used in this expression is to be under­
stood in the mathematical sense, as a set of points between which
distances can be defined.
Especially in the older literature, the opinion is frequently expressed
that locatedness is not an inherent characteristic of auditory events,
but that it develops only during the course of the differentiation of the
organism and, specifically, to the degree that the human being learns
by experience to ascribe the "correct" location to auditory events. The
"correct" location is taken to be the position of the sound source. This
approach implies that there exist auditory events which, so to speak,
wait in a "nonspace" to be ascribed, and only by an experienced in­
dividual, to a particular location. The error of such a point of view has
long been a matter of record (see, e.g., Hombostel 1926). The actual
situation is the following: During the development of the individual,
the auditory world differentiates itsel£ Auditory events, at first relatively
diffuse in their locatedness, become more precisely defined spatially;
the correspondence to the visual world and to the other senses also
becomes more precise.
It is certainly true that the position of the auditory event and the
position of the vibrating body that radiates the sound waves (the sound
source) frequently coincide. Nonetheless, the conclusion that the position
of the sound source is also the intrinsically correct position of the
auditory event is, at the very least, problematic. The sound source and
the auditory event are both sensory objects, after all. If their positions
differ, it is an idle question to ask which is false.
The telecommunications engineer, of course, is especially interested
in just those cases in which the positions of the sound source and the

Downloaded from [Link] by guest on 11 March 2024

 1.2 Systems Analysis of the Auditory Experiment 5

auditory event do not coincide. The telecommunications engineer seeks

to reproduce the auditory events that occur at the point where a recording
or transmission originates, using the smallest possible number of sound
sources (e.g., loudspeakers). Sound events must be generated at the
receiving end of the electroacoustic telecommunications chain in such
a way that auditory events occur in the same directions and at the
same distances as at the point of origin. Auditory events must, therefore,
also occur at other positions besides those of the loudspeakers.

1.2 Systems Analysis of the Auditory Experiment

Generally, in an auditory experiment, a sound event with a known

spatial and temporal structure is presented in a precisely defined way
to an experimental subject. The subject then describes, with regard to
the particular attributes that are of interest, the auditory events that
occur under these conditions. Description may be in the form of verbal
statements, but may also consist of actions such as hand gestures,
operation of electrical contact switches, or note-taking. It is, however,
essential that the description allow a quantitative evaluation of the
auditory attributes of interest.
Two fundamentally different types of auditory experiments may be
distinguished: those in which the experimenter is also the subject and
those in which one or more other persons are the subjects. Experiments
performed on oneself are useful only for preliminary investigations.
But in an experiment in which another person is the subject, the auditory
events perceived by the subject are accessible to the experimenter only
indirectly, by way of the subject's own description. From this description,
the conclusion may be drawn (by analogy with the experimenter's own
experience) that, with a certain degree of probability, the subject has
perceived a given auditory event. The experimenter can never perceive
the subject's auditory event directly!
Except in physiology and medicine, the physiological processes within
the subject are also, as a rule, not observable for the purposes of the
experiments. The experimental arrangement may, however, be rep­
resented symbolically by a "black box" in the sense of systems theory,
which in the simplest case treated here has three ports: one input port
and two output ports (figure 1.1). Note, however, that the black box

Downloaded from [Link] by guest on 11 March 2024

 6 Introduction

output I:
auditory event

observable only in an r--------7

introspective experimentt-----.__ I
i �� I
I I
I I I
so
I ,------,
I I I
I r------7
i : I bo
o--++� �i_J �++----o
I I I
input: 1 1 1 1I I output 2:
sound event
I L _____ .J L ____ _J
I description

L_: _______ ::J

Figure 1.1
A simple schematic representation of the subject of an auditory experiment.

serves here only as a way of portraying functional relationships and is

not to be taken as a physiological model of the subject.
Now let us suppose that a sound event with a certain attribute of
interest, s0 , is present at the input port. An auditory event with the
(spatial) attribute h0 appears at the output port. This auditory event
appears in its location in auditory space, at a position that is only in
unusual cases inside the subject's body (see section 2.3.2). The auditory
event is directly accessible only when the experimenter himself is also
the subject. The two system elements and their connections inside the
black box that are drawn with dashed lines at this stage of the discussion
are intended only to indicate that the outputs h0 and h0 are not defined
as identical. If more than one auditory experiment is to be considered,
whether with one or more subjects, the following sets of inputs and
outputs may be distinguished:
1. The set of the attributes of sound events S0 , whose elements are s0 •
2. The set of attributes of auditory events H0 , whose elements are h0 •
3. The set of descriptions of the attributes of the auditory events B0,
whose elements are h0 •
These "fundamental" sets are related: the sets H0 and B0 are both
functions of the set S0 • The goal of auditory experiments is to arrive
at quantitative statements about these sets, or about subsets of them.
Before this goal can be achieved, certain intermediate steps are necessary.

Downloaded from [Link] by guest on 11 March 2024

 1.2 Systems Analysis of the Auditory Experiment 7

Clearly, if there are to be quantitative statements, measurements must

be taken.
Measurement, defined generally, is the assignment of numbers to
objects according to consistent rules (Campbell 1938, Stevens 1951).
"Consistent rules" means, in this context, that the assignment must
be undertaken in such a way that isomorphic relationships exist between
the attributes under consideration that form the fundamental sets and
between the numbers. In each specific case, a set of numbers is first
chosen which can be correlated with the elements of a fundamental
set, or a subset of it, in such a way that each element under consideration
can be assigned to a number. Such a set of numbers is called a "scale."
The elements of the scale are used to describe the elements of the
fundamental set in a quantitative manner.
In the theory of measurement, distinctions are made between scales
of different complexity, namely, nominal, ordinal, interval, and ratio
scales. These scales differ according to which of the following properties
of numbers are applied: identity (each number is identical only with
itself), rank order (the numerals are arranged in a specified order), and
additivity (rules for addition are defined).
Nominal scales are based only on the property of identity. A number
is used only as a label for an element or a group of elements of a
fundamental set that are identical with respect to the attribute under
consideration. The number has no other meaning. For example, auditory
events can be ordered into groups numbered 1 and 2, depending on
whether they are impulsive or sustained.
Ordinal scales use identity and also rank order. A group of n auditory
events, for example, can be denoted by the numbers 1 through n cor­
responding to increasing distance from the experimental subject. A
higher number corresponds to a greater distance, but the differences
in distance between auditory events need not be equal; that is, the steps
of the scale are not necessarily equidistant.
Interval scales use the properties of identity, rank order, and additivity
of intervals. What is not required is that the attribute of the particular
element corresponding to the number zero be equal to zero; that is, no
"absolute" zero of the scale is required. A well-known example is the
Celsius scale of temperature.
In ratio scales, finally, all three of the listed properties of numbers
are true for the elements of the fundamental sets assigned to the numbers

Downloaded from [Link] by guest on 11 March 2024

 8 Introduction

fundamental sets scales

� so= f(s)
�
f so=f(ho) s =f(h)t

So = f(bo) � s=f(b)
� --h-o=f
- (- h_)_

tho =flbo) h=f(b)t

� bo = f(b) Ca {bl)

Figure 1.2
Relationships between fundamental sets and scales in an auditory experiment.

in the scales; for example, the distance of two auditory events from
one another is denoted by zero when it disappears. The distance denoted
by the number two is twice as great as that denoted by the number
one, and so forth. The great majority of physical measurements are
based on ratio scales.
The type of scale used in a measurement procedure determines which
mathematical operations may be used in interpreting the measured
results. Guilford (1954), Siegel (1956), and Sixtl (1967), among others,
examine this topic in more detail.
In connection with the schematic representation of the experimental
subject in figure 1.1, three scales may be established with which to
measure the three fundamental sets:
1. A scale of sound events S with the numbers s.
2. A scale of auditory events H with the numbers h.
3. A scale of descriptions B with the numbers b.
In figure 1.2, the three fundamental sets and the scales pertaining to
them are represented by enclosed areas. Also shown are the relationships
between elements of the different sets:
1. Scaling functions-essentially the rules of measurement, indicating
how the assignment of the elements of the fundamental sets to the
numbers that form the scales is to be carried out.
2. Psychophysical functions-represented vertically in the diagram, these
indicate the relationships between the elements of the fundamental sets

Downloaded from [Link] by guest on 11 March 2024

 1.2 Systems Analysis of the Auditory Experiment 9

b0=h

perceiving describing system

system conceived of as a
psychophysical
measuring instrument

Figure 1.3
Measurement of the psychophysical function h = J(s).

and, in a narrower sense, the corresponding relationships between the

scales.
In practice, every measurement is oflimited accuracy; the number that,
according to a rule of measurement, corresponds to an element of a
fundamental set may be determined only within certain limits. When
measurements are repeated, it also becomes clear that reproducibility
is limited; the results of repeated measurements differ more or less
from one another. Problems relating to the accuracy of measurement
and to reproducibility occur in several ways in auditory experiments;
figure 1.3 indicates where they occur. Let us take as an example the
determination of the psychophysical function h = f(s), representing
the relationships between measured values ofspecific attributes ofsound
events and auditory events.
The measurement of a specific attribute s0 of a sound event at the
input is undertaken by means of a physical measuring instrument and
corresponds to the measurement procedure effected through this in­
strument: s = f(s0). The attribute h0 of the auditory event that appears
at output port 1 is not accessible to direct measurement. Its measurement
is undertaken indirectly, in this way: The subject is considered sche­
matically as being divided into two parts, which are interposed into
the system. The first of these is the "perceiving part" and the second
the "describing part." The subject in the experiment is instructed to
give a numerical description of the .attribute of interest of the auditory
event according to consistent rules, or is at least instructed to give a

Downloaded from [Link] by guest on 11 March 2024

 10 Introduction

description from which a numerical result can be derived. The methods

that make this possible are described in section 1.3.1. In our schematic
representation, the output h0 represents a numerical description of h0,
that is, a measured value. We can then say that h0 = h. The subject
serves at the same time as an object being measured in the experiment
(the perceiving part) and as a psychophysical measuring instrument
(the describing part).
If this experimental situation is examined with regard to accuracy
and reproducibility of measurement, it can be seen that inaccuracies
occur twice: once in the physical measuring instrument, and inde­
pendently in the psychophysical measuring instrument. Limited re­
producibility of measurement also occurs twice, as variations in the
measured values of the attributes of the sound events and of those of
the auditory events. These variations are not independent of one another,
since the sound events and auditory events are related as the inputs
and outputs of the system. If s0 varies, so does h0 • The perceiving part
of the system also usually changes from one experimental run to the
next, whether several measurements are performed using a single subject,
or measurements are performed on each of several subjects; h0 would
vary, then, even if s0 could be held constant.
In practice, it is possible to simplify the treatment of these relatively
difficult relationships, because in most cases the following assumptions
may be made:
1. The reproducibility of sound events and the accuracy of measuring
them is so great that s0 may be taken as being constant during an
auditory experiment.
2. The psychophysical measurement procedures are chosen and the
subject is instructed in such a way that the psychophysical measuring
instrument may be regarded as nearly invariant from one experimental
run to another or between subjects.
If these assumptions are made, variations in the attributes of auditory
events, and of their measured values, are to be explained only as resulting
from variations in the perceiving part of the system.
Changes in the perceiving part with regard to measurement, however,
are neither predictable in the case of each individual experimental run
nor controllable by choice of experimental procedure; therefore, the
measured results must be regarded as being influenced by a random

Downloaded from [Link] by guest on 11 March 2024

 1.2 Systems Analysis of the Auditory Experiment 11

variable. The collection of these results, then, must use the methods
of descriptive statistics, and their interpretation must use statistical
methods of parameter estimation and tests of significance (see, e.g.,
Guilford 1950, Siegel 1956, Graf et al. 1966, Kreyszig 1967).
These statistical procedures rest on the assumption that the available
measured values are samples taken randomly and independently of
one another from a specified ensemble. The subjects in auditory ex­
periments should fulfill these conditions; for example, they might be
chosen at random from a given population. However, these conditions
are as a rule not fulfilled in an ideal sense, since the experimenter is
usually limited to volunteer subjects. From these volunteers, the suitable
ones-e.g., those with normal hearing-will be chosen. In any particular
experiment, failure to fulfill the conditions for a statistical sample may
reduce the validity ofa generalization ofthe results ofthe measurements.
Another point that deserves brief mention is the question of whether
any general validity, or "objectivity," can be attributed to measured
results obtained by means of statements made by individual experi­
mental subjects. With physical measuring techniques it is assumed as
a rule that measured results are independent of the experimenter and
thus have general validity. This assumption rests on the presumed
ability of most persons to arrive at the same results when bringing an
indicating needle of a meter into line with a mark on its dial, or when
reading a number off the dial. A closer look at this situation, however,
reveals that each person is, in fact, describing his own indicating needle
and his own number-his own objects of perception. Consequently,
there is no difference between physical and psychophysical measure­
ments such that the one may be called, in principle, "objective" and
the other "subjective." Differences exist only in the degree ofagreement
among descriptions given by different persons.
Indices of this degree ofagreement, known as "indices of objectivity,"
have consequently become a common feature of the technique of psy­
chophysical measurement. According to these indices, objectivity is
complete when the measured result is always the same, either over
several measurements using one subject or over one measurement using
each of several subjects. Another way of expressing this is to say that
objectivity is complete when the hypothesis holds true that the system
under test is ergodic. Measurements of objectivity are seldom provided
along with the conclusions about spatial hearing in the published lit-

Downloaded from [Link] by guest on 11 March 2024

 12 Introduction

erature; the general validity of many hypothesis must therefore be held

in doubt.
In the schematic representation of the experimental subject (figures
1.1-1.3), it was assumed that the inputs are attributes of sound events.
At this point, however, we generalize the sets of inputs to include all
attributes of physical phenomena and processes that might correlate
in any way with the location of an auditory event. The relevant phe­
nomena and processes are sought out by asking which of the experi­
mental subject's senses take part in the creation of the physiological
situation in the brain that is related to the formation of an auditory
event. Attributes may be correlated if they are suitable for particular
senses, that is, if they are received by particular sensory organs and
result in transmission of nerve impulses to the central nervous system.
These are called "adequate stimuli."
Table 1.1 lists the physical phenomena and processes for which a
correlation to the position of auditory events has been proven or hy­
pothesized. The first column lists the physical processes and phenomena.
The second lists the sensory organs involved in receiving and processing
these. The third gives the usual name for the psychophysical theory of
spatial hearing that describes the relationship between the particular
physical attribute and the location of the auditory event. The fourth
shows several ways of categorizing these theories. Thus, for example,
theories of sound transmitted through air are categorized as the basic
theories; every theory of spatial hearing is based on sound transmitted
through air to the eardrums. Only in cases in which the location of the
auditory event is not unequivocally correlated with the sound signals
at the eardrums are further physical characteristics drawn into consid­
eration, through supplementary theories. According to whether one or
more senses are involved, theories of spatial hearing are categorized
as homosensory or heterosensory. Finally, theories are categorized ac­
cording to whether they apply to experimental subjects whose head is
held still (fixed-position theories) or whether they require head move­
ments (motional theories).

1.3 Remarks Concerning Experimental Procedures

In the following discussion of experimental procedures, it should be

noted that references to position in connection with spatial hearing

Downloaded from [Link] by guest on 11 March 2024

 l .3 Experimental Procedures 13

Table 1.1
Psychophysical theories of spatial hearing.
Physical
phenomena and
processes Participating
considered sensory organs Usual designation Categorization
Sound conducted Hearing (one ear Monaural theories B,Ho,F
through the air to suffices) for air-conducted
one or both sound
eardrums
Interaural differ- Hearing (both ears Binaural theories B,Ho,F
ences for air-con- necessary) for air-conducted
ducted sound at sound
both eardrums
Sound conducted Hearing Bone-conduction S,Ho,F
through the air to theories
the eardrums and
sound conducted
through bone in
the skull (gene-
rated by air-con-
ducted sound)
Sound conducted Hearing,vision Visual theories S,He,F
through the air to
the eardrums and
light on the retinas
Sound conducted Hearing,sense of Vestibular theories S,He,F
through the air to balance
the eardrums and
to the cochlea and
vestibular organ
Sound conducted Hearing,sense of Tactile theories S,He,F
through the air to touch
the eardrums and
sound received by
tactile receptors
(such as the hair at
the nape of the
neck)
Head movements Hearing,sense of Motional theories S,He,M
during which air- balance; receptors
conducted sounds of tension,posi-
are modified at the tion,and orienta-
eardrums tion; vision
Categories: Basic (B) vs. Supplemental (S); Homosensory (Ho) vs. Heterosensory (He);
Fixed-position (F) vs. Motional (M).

Downloaded from [Link] by guest on 11 March 2024

 14 Introduction

forward
<P=O"
6:0"
horizontal plane

Figure 1.4
A head-related system of coordinates used in auditory experiments; r is distance, "' is
the azimuth, and o is the elevation.

(position of the auditory event, position of the sound source, point of

measurement, etc.) are usually made in terms of a head-related system
of spherical coordinates. In other words, the system of coordinates
shifts in conjunction with movements of the subject's head. The system
of coordinates is also practically constant relative to the position of the
ears, since we humans are almost incapable of moving our ears with
relation to our heads.
In this book, unless otherwise noted, the system shown in figure 1.4
will be assumed. Angles will be denoted as shown. The origin of the
system of coordinates lies halfway between the upper margins of the
entrances to the two ear canals. These, with the lower margins of the
eye sockets, define the horizontal plane, in conformity with international
standards for measurement of the skull. The frontal plane intersects
the upper margins of the entrances to the ear canals and lies at right
angles to the horizontal plane. The median plane (median sagittal plane)
lies at right angles to the horizontal and frontal planes. The three planes
intersect at the origin. If the head may be assumed symmetrical, it is
then symmetrical about the median plane.
1.3.1 Psychometric methods
Psychophysical and psychological measuring techniques, to which the
name "psychometric methods" has been given, may be categorized in
different ways. For example, they may be categorized according to the
level of judgment the experimental subject is required to make. Cor-

Downloaded from [Link] by guest on 11 March 2024

 1.3 Experimental Procedures 15

responding to the four levels of complexity of the measurement scales,

measuring methods are distinguished as requiring nominal, ordinal,
interval, or ratio judgments. Or they may be categorized according to
how the inputs (e.g., sound signals) are presented to the subject.
With regard to the latter ways of categorizing, there are two principal
possibilities. It is a prerequisite of the first that the input attribute under
consideration be continuously variable. This attribute is then adjusted
during the experiment until the experimental subject perceives a given
judgment as being fulfilled. It is of secondary importance in this respect
whether the adjustment is carried out by the experimenter, the subject,
or automatically. Methods using this procedure are called "methods
of adjustment." They are especially efficient when different subjects
are used successively for a series of auditory experiments.
The second possibility is to present to the subject an input whose
attribute of interest remains constant throughout each interval of pre­
sentation. From a consistently organized list of judgments, the subject
chooses the one that seems most appropriate for each presentation.
The experiment is repeated several times, with variations in the input
attribute each time. The judgments are then evaluated statistically using
suitable procedures. The variations in the input attribute follow a pattern
unknown to the subject; usually they are stochastic. Methods of this
type are called "constant methods" or "interrogative methods." They
are especially efficient when auditory experiments are conducted using
several subjects simultaneously.
Here we shall confine ourselves to discussion of a few methods com­
monly used in conducting auditory experiments. Additional methods
and further details are given by Fechner (1860), Stevens (1951, 1958),
Guilford (1954), Sixtl (1967), and Robinson and Jackson (1972). Because
the results of measurements with experimental subjects depend to some
extent upon the measuring technique, the method used should always
be indicated along with the results.
The most important methods in auditory experiments are based on
nominal and ordinal judgments. These methods are especially well
adapted to determine thresholds of perceptibility, difference thresholds,
and points of perceptual equality. Figure 1.5 clarifies these concepts.
Let us take as an example an auditory experiment seeking to determine
the relationship between the intensity of a sound and the loudness of
the corresponding auditory event. The scales appropriate to such an

Downloaded from [Link] by guest on 11 March 2024

 16 Introduction

1 range of perceptible intensity 14-

I
r
2.4(1 ,,..::.4(I1lmin
�
Imin r;. � I,
Im., intensity of sound I--
\ \ \ I J ! ; --

',,\'\\\:: :: /
I

\
\ ,: , loudness A___.

/// i \\\ ! \\
i \
50 % present 5b% Jrea,er more alike
50% not present 50% lesser than not
! i
75 % lesser 75% greater 50% equal
25 % greater 25 % lesser 50% unequal

Figure 1.5
Definitions of thresholds of perceptibility, difference thresholds, and points of percep­
tual equality.

experiment are one-dimensional: a scale of intensity for the sound e:vents

and a scale of loudness for the auditory events. Thresholds of percep­
tibility are, in this example, the two points /min and /max on the scale
of intensity, which set the limits for the range of readily perceivable
intensities. We shall not discuss /max here; it may be understood as a
threshold of discomfort or pain. For the measurement of the lower
threshold of perceptibility /min, the subject is required to make the
nominal judgment whether an auditory event is "present" or "not
present." The threshold of perceptibility is defined as the point on the
intensity scale where both judgments occur with equal probability (0.5).
Difference thresholds are segments of the intensity scale corresponding
to a just noticeable difference in loudness; for example, �(/1 )min and
�(/2)min• The subject is required to make either nominal or ordinal
judgments. For example, if the judgments "equal" or "unequal" loud­
ness are required, then the difference threshold corresponds to the
segment of the intensity scale from a point /1 to another point at which
the probability of the answer "unequal" becomes 0.5. If the ordinal
judgments "greater" or "lesser" are required, then the two points at
which the probability of one or the other judgment is 0.75 define the
limits of a segment that corresponds to twice the difference threshold.
(These two procedures can produce different measured values for the
difference threshold; there is no agreement in the literature as to which
procedure is preferable.)
Points of perceptual equality are, in the example, the points /eq and
/'eq that correspond to auditory events of equal loudness. Let us take

Downloaded from [Link] by guest on 11 March 2024

 1.3 Experimental Procedures 17

overage of I and 2
1.0 t--------�------�
direction of
adjustme nt 2

scale of the -
threshold or attribute of the input
point of
perceptual equality

Figure 1.6
Evaluation of results obtained by using the method of adjustment.

as an example two sound sources whose outputs are pure sinusoids,

one of 1,000 Hz and the other of 200 Hz. Let us say that the auditory
events corresponding to the two sources are equally loud when the
intensity of the 1,000 Hz tone is 10 dB and that of the 200 Hz tone
is 15 dB. Ordinal judgments are required of the subject. If only the
answers "greater" and "lesser" are permitted, then the point of per­
ceptual equality is where each of these answers occurs with equal
probability.
Two ofthe better-known methods ofadjustment suited to determining
thresholds of perceptibility, difference thresholds, and points of per­
ceptual equality are the method of average error and the method of
minimal changes. When the method ofaverage error is used, the subject
adjusts the variable attribute of the input until a given judgment is
fulfilled. In the method of minimal changes, the input variation, in
small increments, is effected by the experimenter, until the subject
indicates that a given judgment is fulfilled. The experimenter repeats
this measurement procedure several times, until the relative frequency
of each ofthe judgments, as given by the subject, can be plotted against
a variable that characterizes the input attribute ofinterest. This usually
leads to a function resembling a normal (Gaussian) distribution. Ac­
cording to whether the input attribute was varied toward higher or
lower values, two different curves may result, from which an average
curve may be calculated (figure 1.6). For example, in measuring a
threshold of perceptibility, the input attribute of interest may be varied

Downloaded from [Link] by guest on 11 March 2024

 18 Introduction

,-,
t (a)
nominal judgments
(bl
nominal judgments
(cl
ordinal judgments
� 1,0 t---a=-------,----,:;;,-------.;;.---------+----,-------,;.--..;:---,------,--,-�-

� :?
; � 0.5 +----------ll-------lF--+--lif-------+-----ll-----+---

·-
.� en
.,,
- "0
:,
:;::
�0 0
threshold point of point of
of percept Ion perceptual equality perceptual equality
-I l- -1 1-
2x difference 2 x difference
threshold threshold

scale of the attribute of the input -

Figure 1.7
Evaluation of results obtained by using the constant method.

first from lower to higher values until the subject gives the response
that an auditory event is present; then the input attribute may be varied
from higher to lower values until the subject indicates that the perception
of an auditory event has disappeared. The final, calculated value for
the threshold is the peak of the averaged curve. To measure difference
thresholds and points of perceptual equality, a similiar procedure is
used, with the subject giving the answers "equal" or "unequal," "greater"
or "lesser." A useful variation of the methods of adjustment is von
Bekesy's (194 7) method of oscillating adjustment (see also section 2.2.3).
When the constant method is used to determine thresholds and points
of perceptual equality, a series of inputs is presented successively to
the subject, each of which includes the attribute of interest to a different
degree. The order of the series is unknown to the subject. For example,
if a threshold of perceptibility is to be measured, the subject is required
to answer whether the auditory event is present or not. It is evident
that the series of different inputs must be chosen by means of preliminary
experimentation so that the range of the attribute of interest in which
the threshold is expected will be as evenly covered as possible. After
several repetitions of the series of inputs, it becomes possible to derive
a curve of the relative frequencies of the two answers as a function of
the input attribute, as shown in more detail in the curves of figure 1.7a.
The SO-percent point of both curves is the calculated value for the
threshold. In order to measure difference thresholds and points of per-

Downloaded from [Link] by guest on 11 March 2024

 1.3 Experimental Procedures 19

ceptual equality, on the other hand, at least two inputs that contain
the attribute of interest to a different degree must be presented either
simultaneously or successively. One input (the standard) remains the
same for all test runs, and the other input is varied from one run to
the next. According to whether the judgments "equal" or "unequal,"
"lesser" or "greater" are required, the curve of the relative frequencies
of judgments of figure 1.7b or 1.7c may be calculated. The figures also
show how the calculated values for difference thresholds and points of
perceptual equality are derived. A sometimes-used variation on the
constant method is the method of triads, in which three inputs-two
equal and one different-are presented to subjects who are then required
to identify the auditory event corresponding to the input that is different.
The foregoing methods do not without additional assumptions allow
assignment of one particular point on a scale of attributes of auditory
events (a perceptual scale) to a scale of input attributes (e.g., a physical
scale) and vice versa. In other words, these methods do not by them­
selves permit a point-by-point assignment between the two scales. This
task may, however, be undertaken directly by means of techniques
based on interval and ratio judgments. These techniques differ from
those already described in the following important way: the subjects,
when making judgments, ascribe numerical values "directly" to attri­
butes of auditory events, or else, when using a constant procedure,
numerical values among which they can choose are prescribed by the
experimenter. These methods are therefore called "direct methods of
measurement."
In the methods based on interval judgments, subjects are instructed
to pay attention to differences between auditory events. An auditory
event to whose attribute of interest a subject assigns the number 2
should, with regard to this attribute, differ equally from other events
to which the subject ascribes the numbers 1 and 3. Two procedures
lead to particularly good reproducibility of results. In the method of
equal perceptual intervals, a number of inputs are presented to the
subject, either successively or simultaneously. The subject must place
each corresponding auditory event into one of a group of numbered
categories, which are usually organized so as to appear equidistant along
a scale. In the halving method, the subject's task is to adjust an input
attribute so that, with respect to the auditory attribute of interest, the

Downloaded from [Link] by guest on 11 March 2024

 20 Introduction

corresponding auditory event lies exactly in the middle on a scale be­

tween two other, fixed auditory events.
The basic assumption underlying methods based on ratio judgments
is that the subject is capable of distinguishing ratios between attributes
of interest. Here, for example, is one outline of a measurement pro­
cedure: An input is presented to the subject, and it is explained that
the attribute of interest of the corresponding auditory event has the
numerical value 1. Then another input is presented, and the subject is
asked for the new value of the attribute of interest. It is also possible
to ask by what factor, ratio, or percentage the new attribute is different
from the one presented first. The judging is made easier if the instructions
are for the subject to represent the relationship between the two char­
acteristics so that their sum is 100-as 50/50 for 1/1 and 80/20 for
4/1. This is called the method of constant sums. The corresponding
method of adjustment requires the subject, by varying an input attribute,
to adjust an auditory event so that the attribute of interest is changed
according to a given ratio, or until the attribute exhibits a given ratio
with respect to a standard, which may be presented simultaneously or
successively.
This has been a general review of psychometric methods, but we
must now consider how these techniques may be applied to spatial
attributes, that is, to the locations of auditory events. These are the
positions of the auditory events and, with spatially extended auditory
events, all positions within and on the surfaces of the spaces occupied
by these events. It has already been noted that positions and their
descriptions cannot be absolute, but must always be in terms of angle
and distance relative to another position. What is measured is, in the
final analysis, quantities such as distance and angle. A complete de­
scription of these requires a three-dimensional coordinate system, such
as the one in figure 1.4.
In such a coordinate system, the scales of distance and angle are
ratio scales. Consequently, all four classes of judgments-nominal, or­
dinal, interval, and ratio-are possible, and all of the previously de­
scribed measuring methods are usable.
When measuring thresholds of angles and distances, it is pointless
to deal with thresholds of perceptibility separately from difference
thresholds, since these two types of thresholds can be converted into
each other by means of a transformation of coordinates. What is meant,

Downloaded from [Link] by guest on 11 March 2024

 1.3 Experimental Procedures 21

in each case, is the smallest change in the input attribute of interest

that leads to a change in the position (direction and/or distance) of the
auditory event. In section 2.1 we shall designate this threshold by the
term "localization blur." One way to measure it, for example, is to use
two sound sources that emit the same sound signal in alternation and
are capable ofbeing displaced from each other spatially. It is important
for the subject to pass judgment only about spatial changes in the
auditory event and to pay no attention to changes in loudness, tone
color, or other variable attributes.
So-called pointer methods are the most widely used in determining
points of perceptual equality for spatial attributes of perceptual events.
The task of the subject may be to identify a pointer that indicates
exactly the direction of an auditory event and/or whose length corre­
sponds to the distance ofan auditory event. Visual and tactile pointers,
as well as auditory ones, may be used. An example of a method using
an auditory pointer is for the subject to displace a movable sound source
until the auditory event that corresponds to it is in the same position
as another auditory event to which it is to be compared. As one example
of a visual pointer, consider a movable source of light; the position of
the visual event corresponding to it may be compared with that of the
auditory event. The pointer is also primarily a visual one if the subject
points toward the auditory event with a finger or a stick. Von Bekesy
(1930a) gives as an example ofa tactile pointer the use ofa small nozzle
to direct a stream of air onto the forehead ofthe subject. Using pointer
methods, it is, for example, possible to measure differences between
the positions of different sound sources, or between sound sources and
light sources whose corresponding perceptual objects are equivalent
with respect to a certain spatial attribute. The pointer method is, how­
ever, insufficient to assign the position of a sound source to that of an
auditory event. The direction ofan auditory event cannot be determined
from what the subject indicates by pointing unless the relationship
between the physically measurable direction of the pointer and the
direction ofthe perceptual event corresponding to the pointer is known.
This important consideration is often not taken clearly into account.
For both direction and distance, direct measurement of the location
of auditory events as related to attributes of sound events is in all cases
possible using interval and ratio judgments. One common method is
to have the subject displace a sound source so that the auditory event

Downloaded from [Link] by guest on 11 March 2024

 22 Introduction

appears in an agreed direction or at an agreed distance. Another method

is to give the subject the task of making an interval or ratio judgment
for the direction or distance of a given auditory event. A third procedure
belonging to this same class of methods is for the subject to organize
auditory events into defined, adjacent categories according to their di­
rection and distance. All of these methods make it possible to determine
the position of the auditory event in a system of spatial coordinates,
though some of them are more accurate than others.
1.3.2 Signals and sound fields
In an auditory experiment, let a sound event be presented to a subject.
The sound event consists of one or more sound signals, which may be
the same or different, radiated into space by one sound source or by
multiple sources at different positions in space. The signals propagate
as sound waves in the medium surrounding the sources, usually air,
finally reaching the eardrums of the subject. The signals at the eardrums
can be described as functions of sound pressure against time. The sound
pressure at the eardrum, p(t), depends, for a given subject, upon such'
parameters as determine the spatial and temporal structure of sound
fields; namely, upon the type, number, and position of the sound sources
and upon the sound signals radiated by these sources.
In theory, sound fields of any desired complexity may be used in
auditory experiments. Because complicated fields present great diffi­
culties in analysis, however, most experiments employ sound fields of
the simplest possible temporal and spatial structure, though these are
chosen so as to allow the furthest possible extrapolation of measured
results to more complex sound fields. Some factors that play a role in
the choice of signals and sound sources in this context will be examined
below. Many textbooks give detailed explanations of signal theory, for
example: Lee (1960), Ktipfmtiller (1968), Fischer (1969), and Unbe­
hauen (1969); with regard to the theory of sound sources, see Skudrzyk
(1954), Meyer and Neumann (1967), Reichardt (1968), and Cremer
(1971), among others.
Almost any function x(t) whose independent variable is time, whether
x represents sound pressure, velocity, voltage, or any other variable,
can be decomposed into series of elementary signals, which are com­
monly used as test signals in auditory experiments. For example, a

Downloaded from [Link] by guest on 11 March 2024

 1.3 Experimental Procedures 23

impulse (click) sinusoid Gaussian signal

(from t=-� to t=+�I

t
B"'
1
.,.
very narrow
t
:;:;
-;;
t
...."'

rectangular pulse
,_ , _ f -

t===
t t t
bandwidth f·
,;
o f the
� �> �
measuring
·;;; ·;;;
.,
<:
,::,
t3
t filter
.,
<:
,::,
0
>, C. >, 0,21£'
.,e'
,Jf
.,e'
f(lin)- f(lin) -
f(lin) -
., fa
<:
<: 0 0 Q) 0

t t
.; t OdB
>
.;> � >
� � � -7dB
� ·;;; � ,Jf
.,
to
<:
·;;; ·;;;
f(log) - f(lag) -
.,
<:
,::,
,::,
t3 + fa .,
<:
,::,
>, >,
e' 0 e'
.,
., C.

.,
<:
.,
<:

lal lb) lei

Figure 1.8
Temporal functions and energy or power density spectra of three elementary signals.
Note that the lowest row of drawings are log-log representations.

given time function can be analyzed into a series of very short impulses.
This analysis is based on the superposition integral
+oo

x(t) = f x(T) o(t - T) dT,

-00
(1.1)

where o(t - T) represents the so-called unit impulse, or Dirac impulse,

at the time t. The unit impulse is a fictional concept that can be ap­
proximated by decreasing to zero the duration of an actual impulse
while the area under it, 1, remains constant. A short impulse has the
property as a test signal that its energy is concentrated at a definite
point in time but is distributed evenly over all frequencies. The time
function and energy density spectrum (energy per bandwidth) of a short
impulse are represented in figure 1.8a. In practice, a sufficently even

Downloaded from [Link] by guest on 11 March 2024

 24 Introduction

distribution of energy density in the frequency range of human hearing

(16 Hz to 16 kHz) is attained if the duration of an impulse is less than
approximately 25 µs.
A function x(t) can also be analyzed into a series of oscillations at
single frequencies (sinusoidal time functions, pure tones). This analysis
is carried out by means of the Fourier integral

J %(/) ei rJ1
+oo

x(t) = 2
df, (1.2)
-00

where %(/) is the complex Fourier spectrum of the signal under con­
sideration, and the real part of exp(j21rft) represents a sinusoidal signal
of frequency f and amplitude 1. The useful property of a sinusoidal
test signal is that the position of its energy on the frequency axis is
known exactly; however, there is no specifying its position in time, as
a sinusoidal signal by definition is infinitely long. Figure 1.8b shows
the time function and spectrum of a sinusoidal signal. Because of the
infinite duration of a sinusoidal signal, its energy is infinite and its
energy density cannot be defined meaningfully. Instead we show the
spectrum of power density, which is defined as the power measured
by means of a bandpass filter divided by the bandwidth of the filter
(power per bandwidth).
To sum up this discussion, sinusoidal test signals are used when it
is desired to concentrate energy or power at one point on the frequency
axis, and impulses are used when it is desired to concentrate energy at
one point on the time axis.
A function of time may be analyzed into many kinds of more ele­
mentary signals. For all of these, the more concentrated the energy is
in time, the greater its bandwidth in the frequency domain; conversely,
the more concentrated the energy is in the frequency domain, the more
indefinite the signal in the time domain. An optimum compromise of
these conditions is afforded by the so-called Gaussian tone burst (figure
1.8c), which is therefore often used as a test signal in auditory exper­
iments. The time function and complex Fourier spectrum of a Gaussian
tone burst are described in the following expressions:

Downloaded from [Link] by guest on 11 March 2024

 1.3 Experimental Procedures 25

2
(t) = Ae-,r[(I - lol/At] Re(ej2.-Jot), (1.3)
X

%(/) = !::.t � A [e-.-[(f - fol!AfF e-j2.-(f - folio + e-,r[(f + fol!AJJ' e-j2.-(f + folio}

(1.4)
Here A is the maximum amplitude of the burst signal, and !::.t is the
width of a rectangle of the same amplitude whose area is equal to that
under the envelope of the signal. The equation !::.t !::.J = 1 applies to
these expressions.
When a broadband excitation of a system is desired (e.g., with an
input signal covering the entire range of human hearing), it is possible,
as previously described, to apply a short impulse. However, the energy
such an impulse can introduce into the system is limited. The limitation
exists because a definite maximum sound pressure cannot be exceeded
without risking damage to the sound transmitter and the human auditory
system. This difficulty can be avoided by substituting a series of short
impulses for the single impulse. The series of impulses may be of any
convenient duration, with the temporal spacing between each pair of
impulses and the mathematical sign (positive or negative) of each im­
pulse being random; that is, any spacing, and either sign, must be
equally probable. The result is what is called "white noise." There is
no limit to the energy in white noise, since it may have any desired
duration. Its power density is the same at all frequencies. The probability
of any particular instantaneous value of a white noise signal follows a
Gaussian distribution (figure 1.9).
By means of linear filtering, random functions of any desired band­
width and power density spectrum may be derived from white noise.
Such signals are used to ensure that the system into which they are
introduced is not set into a steady-state oscillation at its resonant fre­
quencies-for example, that standing waves do not build up in a room.
Filtered noise is therefore often used to simulate speech and music, at
least with regard to their power density spectra. Typical power density
spectra of speech and music are shown in figure 1.10.
Power density spectra, and under certain conditions energy density
spectra as well, may be measured by means of tunable bandpass filters.
In acoustics, two types of filters are customarily used for this type of
measurement. In one, the bandwidth remains constant regardless of

Downloaded from [Link] by guest on 11 March 2024

 26 Introduction

..t t
:;::;
-;:;

t-

Figure 1.9
Probability density w'(x) of the instantaneous value x(t) of white noise.

t
0
r
dB
-1 0 I music
1>

j /_ I, ... ',l \

\'
.!! -20
I'
I/
:,_,A
''
I
-30 speech

;
8.
-40
\\
0,02 0,050,1 0,2 0,5 1 2 5 10kHz
frequency-

Figure 1.10
Typical power density spectra of speech and music (after Skudrzyk 1954).

changes in the filter's center frequency (tl.f = constant, so that the filter
has a constant absolute bandwidth, as in a heterodyne-tuned filter). In
the other type, the ratio of the bandwidth to the center frequency is
constant (tl.f/fo = constant, so that the filter has a constant relative
bandwidth, as in a 1/3-octave or octave filter). If the power level at
the output of the filter is plotted against the filter's center frequency,
a different curve results depending upon the type of filter used. For
example, in figure 1.11, the measured power level at the output of the
filter is shown for both white noise and so-called pink noise, for both
types of filter. For white noise, the power is constant with constant
absolute bandwidth; for pink noise, the power is constant if the ratio
of bandwidth to the filter's center frequency remains constant. Since,
roughly speaking, the human auditory system analyzes sound signals

Downloaded from [Link] by guest on 11 March 2024

 1.3 Experimental Procedures 27

,E- t +20

- ...
S dB
� +10 white noise
-.l!! a
o;;: 0
"ii C,
�,,, b
-
8/octa____ -
3d

---
> C: -
.!!! ·;: -10 pink noise
Ve fat/ _ - --
---✓ a
>,:::,
- (I>

·;;;i; g -20
�--�-�-�--�-�-
E
�� ;: 0,125 0.25 0.5 2 4 8 16 kHz
;
frequency -
g_ 'E
Figure 1.11
Power density spectra of white and pink noise represented two different ways. a:
Power for constant absolute bandwidth (here 60 Hz). b: Power for constant relative
bandwidth (here 1/3 octave).

in terms of constant relative bandwidth, pink noise has considerable

importance as a test signal in auditory experiments.
In certain phenomena related to spatial hearing, the envelopes of
signals play a role. Band-limited time functions (at least those without
a constant term-a condition that always holds for common sound
signals) may in most cases be transformed according to the equation
x (t) = A(t) Re(ei1*>), (1.5)

cp(t) being a monotonically increasing time function. For information

about the algorithm by which this transformation is carried out, see,
for example, Voelcker (1966). In the most general case, x (t) represents
an oscillation whose amplitude and frequency change constantly with
respect to time. Figure 1.12 shows such a signal. Here A(t) is the envelope;
its plot is the dashed line.
Sound fields may be analyzed into elementary fields according to
iaws that are to some extent analogous to those by which time functions
are analyzed into elementary signals. One possible mode of analysis is
based on a principle of Huygens and Fresnel according to which every
point in any system of waves propagating in space can be regarded as
the source of a spherical, or elementary, wave. By the superposition of
such elementary waves, the wave field at any point within the system
may be determined.
A spherical wave is a point-symmetric wave, that is, a wave whose
parameters depend only on the distance from its origin and not on the

Downloaded from [Link] by guest on 11 March 2024

 28 Introduction

t---

Figure 1.12
The envelope of a signal.

direction of propagation. Sound sources that radiate spherical waves

are �ailed spherical radiators of the zeroth order, elementary radiators,
or pulsating spheres. The term "pulsating sphere" arises from the fol­
lowing considerations: A sound source that radiates only spherical waves
must itself be spherical in shape. Also, all points of its surface must,
as seen from its center, move inwards and outwards radially in phase;
hence "pulsating."
The sound field generated by a pulsating sphere is described by the
two equations
2
p(t, r) = Re {const · j 1Po e i 2.-.11 e-i 2.-,1;'} (1.6)

v(t, r) = Re {const · (j2;1A + :2 )

e i 2.-.11 e-i2.-,11} (1.7)

where Po represents the density of the medium. Recognizable in these

equations is the decrease in sound pressure with increasing distance
from the center of the sphere, becoming half as great with every doubling
of this distance r. Figure 1.13 plots this decrease on a log-log grid. The
amplitude of the sound (particle) velocity at large distances from the
sphere (r � 21rrlA) also falls off as 1/r, but nearer the sphere (r �
21rrlA) it falls off as 1/r 2• At large distances, pressure and velocity
oscillations are in phase. The ratio p(t)lv (t) then represents a purely
resistive impedance and takes on the value of the specific impedance
of the medium, z0 = p0C.
Under certain conditions, the equations for fields generated by pul­
sating spheres are very nearly correct also for pulsating radiators of

Downloaded from [Link] by guest on 11 March 2024

 1.3 Experimental Procedures 29

0 sound velocity falls as l/r 2 in the near fi"'!ld

�I..:. dB ' � 12 dB/doubling of distance

Ol
_g -10 ,____..,__-+-�"'--+ sound pressure falls as fir
� � 6 dB/ doubling of distance

�i �
Ol
-20 ''
a
N
''
',
�I,,_ -30

_g near field far field

a
N
region region

.; -40 2nr/J..­
> 0,25 0,5 2 4 8 16

"+
.! r/J.. -
C:
"t2 "t distance relative to wavelength ____,..
e"
,,"
Figure 1.13
Pressure, velocity, and intensity of the sound field near to and distant from a spheri­
cal radiator of the zeroth order.

t=t, t, t, t.

J)
Figure 1.14
Any pulsating sound source that is small compared with the wavelength generates
spherical waves at a moderate distance. As the distance is increased, the spherical
waves become more and more plane.

Downloaded from [Link] by guest on 11 March 2024

 30 Introduction

other shapes, and even for radiators only parts of whose surfaces oscillate
(figure 1.14). The equations are nearly correct if the point of mea­
surement is far enough away from the radiator and if the radiator is
small in comparison with the wavelengths of the sounds. The first
condition can be fulfilled without any special considerations as to the
design of equipment; the second is more difficult to fulfill. As an example,
a loudspeaker in a sealed box of normal dimensions (the longest edge
approximately 30 cm) can be considered to be a spherical radiator only
for frequencies up to approximately 100 Hz. At higher frequencies, it
no longer radiates equally in all directions. Nonetheless, pulsating spheres
may be approximated at higher frequencies by a device such as a hose
or pipe, one end of which opens into a pressure chamber transducer.
The free end of the hose or pipe then behaves as a pulsating radiator
whose dimensions are approximately determined by the hose diameter.
This method of approximating small pulsating spherical sound sources
was used, for example, by Mills (1958) and by Shaw and Teranishi
(1968).
With increasing distance from any radiator, the sound field becomes
more and more similar to a plane wave. Pressure and velocity are in
phase for all practical purposes, and, with respect to any object of
reasonable size, the curvature of the wavefront becomes less and less
noticeable. At a distance of 3 m from loudspeakers in rooms of the
usual size, a level difference of less than 1 dB is measured if a microphone
is displaced perpendicularly to the axis of the loudspeaker over a distance
approximating the dimensions of the human head (±9 cm). Hence the
microphone is, for all practical purposes, being displaced in a plane of
equal sound pressure. The curvature of the wavefront is no longer
significant, and the sound field is nearly planar.
The sound field of a source located in front of a planar reflecting
wall may be represented simply as a superposition of the original sound
field upon that of its virtual mirror image (figure 1.15). This description
is especially useful in explaining the processes of spatial hearing in
enclosed spaces. In chapter 3 it will be shown that the first wavefronts
that reach the ears in an enclosed space are particularly important. The
propagation of these first wavefronts may easily be traced by constructing
mirror-image sound sources.
Headphones are a special type of sound source often used in auditory
experiments. They are particularly useful when it is desired for exper-

Downloaded from [Link] by guest on 11 March 2024

 1.3 Experimental Procedures 31

,,,,
4' ---- ......... I st wavefront

I
I
I ,-----t--:si..n-----�--
\ mirror image \
sound source\
\
''
', .... ____ _ ,,

Figure 1.15
Representation of a reflected wavefront as if generated by a mirror-image sound
source.

imental purposes to eliminate the dispersion, shadowing, and resonance

effects of the external ears or the acoustical coupling between the two
ears that occurs in free sound fields. Headphones are distinguished as
intraaural, circumaural, and supraaural, according to their construction.
Intraaural headphones are tightly plugged into the ear canals; circu­
maural headphones surround the pinna, often making an airtight seal
to the head; and supraaural headphones rest on the pinna.
It has often been assumed that sound pressure is the same everywhere
inside the volume enclosed by each headphone, as if this volume were
a pressure chamber. The assumption, in this generalized form, is in­
correct. This error has been an important reason why some of the
characteristics of spatial hearing have long gone without explanation.
If the dimensions of the volume enclosed by the headphone are ex­
amined carefully, it becomes clear that, with some variation according
to the type of headphone, the sound inside this volume must be regarded
as a wave phenomenon for frequencies above about 1 kHz (see, e.g.,
Villchur 1969). It is precisely the frequency range above 1 kHz that is
especially important for some effects relevant to spatial hearing.
1.3.3 Probe microphones
The most relevant acoustical inputs for the subject are the sound signals
at the eardrums. Special transducers are needed to measure these signals,
or signals at a nearby location inside the ear canal. These transducers
must fulfill the following conditions, among others:

Downloaded from [Link] by guest on 11 March 2024

 32 Introduction

diameter of the probe tube:

I mm OD, 0.6 mm ID

microphone capsule,
BS. K type 4134

r=I4
diaphragm

Figure 1.16
Dimensions (mm) of a probe microphone used to measure sound pressure in the au­
ditory canal.

1. They must not appreciably disturb the sound field in the ear canal,
pinna, or anywhere near the head.
2. They must allow the measurement of a defined signal parameter
throughout the entire range of human hearing.
3. They must not be hazardous to the subject.
So-called probe microphones fulfill these conditions. These usually
consist of a condenser microphone (dielectric microphone) coupled by
means of the smallest possible chamber to a thin hose or pipe. This
construction fulfills the conditions for a sound pressure-dependent mi­
crophone, that is, one at whose output a voltage appears that is pro­
portional to the sound pressure at the opening of the tube. The
proportionality may, however, vary strongly with frequency due to
resonances and attenuation inside the tube. A probe microphone has
a spherical directionality characteristic; in other words, it is equally
sensitive to sound from all directions. This characteristic is easily
understood according to the principle of reciprocity: If the microphone
were used as a sound generator-as is, in fact, possible with dielectric
microphones-then the opening of the tube, being small in proportion
to the wavelength, would approximate a pulsating sphere or radiator
of the zeroth order.
For measurements inside the ear canal, probe microphones with
tubes 1-2 mm in diameter are used. Often a length of soft plastic hose

Downloaded from [Link] by guest on 11 March 2024

 1.3 Experimental Procedures 33

Figure 1.17
The probe microphone in use.

is added to the end of the tube for measurements near the eardrum.
It can be shown that the sound field is not appreciably distorted by
introducing a second probe into the ear canal while observing the changes
in output of the first microphone. The resonances in the tube can be
damped by packing mineral wool or metal wool lightly into the tube,
or they can be compensated electrically.
Section 2.2 surveys measurements using probes described in the lit­
erature. Here, by way of example, we shall describe one particular
microphone developed by Laws (1972), which the author has used in
a series of measurements. The dimensions of this microphone are shown
in figure 1.16. The essential parts of the microphone come from a kit
manufactured by Brliel and Kjaer of Copenhagen. The probe tube is
bent so that sound pressure measurements can be made 0.5 cm inside
the ear canal (at the entrance to the bony part of the ear canal), even
when the subject is wearing headphones. Figure 1.17 shows the mi­
crophone in use and also shows how the subject's head is immobilized.

Downloaded from [Link] by guest on 11 March 2024

 34 Introduction

101

Figure 1.18
Circuit of the equalizing amplifier. R in ohms, L in henries, and C in farads. Opera­
tional amplifiers are Motorola type MC 1439.

t +10
V \
-
>
dB -
e";; / '\I
N
0
probe equalized A A'\
---
z-
d'.; -10
probe unequalized \-
\11
-�
-., .,ii
>E
0- -20
�
broadband equivalent noise
>-
.: u -30 I I I ,,
... =
E""

-
/'
,,,...
., 0 113 octave graph of

"'w ....
>-
- 0 equivalent noise
>
> -40

'
.!?
- ·o3: /
s_ I
_..,
:, C
C
-50
V
0 0
0,063 0,125 0,25 0,5 2 4 8 16 kHz
frequency --+-

Figure 1.19
Magnitude response of the equalized and unequalized probe microphone, and a 1/3-
octave level graph of equivalent noise.

Downloaded from [Link] by guest on 11 March 2024

 1.3 Experimental Procedures 35

One special feature of this microphone is that, by means of a matched

equalizing amplifier (figure 1.18), its output is corrected so that its
sensitivity varies no more than 1 dB throughout the range of human
hearing. Figure 1.19 shows the frequency response curve with and
without the equalizing amplifier. Also shown are the noise levels at the
end of the system of microphone, cathode follower, preamplifier, and
equalizing amplifier. These noise levels were measured using 1/3-octave
filters. The noise is equivalent to a loudness level of 50-55 phon at the
input to the microphone.* Such a signal-to-noise ratio is clearly too
low for high-quality reproduction of music, but it is nonetheless high
enough for many purposes having to do with measurement in spatial
hearing, especially when sinusoidal test signals are used.
*Three definitions of pressure or intensity level will be used in this book. First there is
sound pressure level Lin dB: L = 20 log(p/p0), where Po = 20 µNm- 2; additional tags
such as dBA or dBB indicate that the measurement has been made through a special
weighting filter. Second is sound pressure level L in dBSL, where SL stands for "sensation
level": L = 20 log( p/pJ, where p, is the threshold of audibility of the sound event to
be measured. Third is the loudness level in phon: A = 20 log(p,."/p0), where Po =
20 µNm- 2 andp,.Hz is the sound pressure level ofa 1 kHz sinusoidal signal at the position
of the subject that gives rise to an auditory event equally as loud as the event induced
by the sound to be measured. For further details see DIN 1318 (1970).

Downloaded from [Link] by guest on 11 March 2024