Acoustic Specification of Object Properties

Ecological acoustics
Acoustic Specification of Object Properties
Claudia Carello,1,2 Jeffrey B. Wagman,1 and M. T. Turvey1,2

1
Center for the Ecological Study of Perception and Action
University of Connecticut
2
Haskins Laboratories
New Haven, CT
Abstract
A general statement of the ecological approach to perception is that a particular animal-
environment relation or event lawfully structures energy distributions that specify the event that
generated them. A given event can, of course, structure varied energy media simultaneously, but
how richly? Precise information about geometric propertiesthe sizes and shapes of
objectshas long been assumed to be the province of visual perception. But the foregoing
ecological law statement suggests that this limitation is more tradition than necessity. Since an
objects physical parameters and its spatial dimensions determine the acoustic structure it
generates when mechanically disturbed, such structure is potentially informative about object
properties. We describe research showing that the perception of object properties by sound alone
is reliable, surprisingly accurate, and constrained by dynamical law.

Acoustic Specification of Object Properties
A bottle sits unseen on the kitchen counter. Your elbow clips it inadvertently and sends it
hurtling towards the floor. As you cringe, waiting for the crash, what you hear is not a shattering
mess but a harmless bounce. The impact of glass on linoleum has set the materials into vibration,
generating compression waves in the air. Somehow, from this sound structure, you know that the
bottle did not break. Moreover, the people listening in the dining room heard something bounce
as well. Lets focus on those listeners whose only contact with the event is from sound. What
else do they know about what happened? Can they hear, for example, that the fallen object was
made of glass, that it was a bottle, whether it was large or small, full or empty?
The preceding questions concern what listeners perceive about events happening around
them. On the basis of its inattention to such questions, the science of perception can be
considered skeptical that audition makes us aware of our surroundings with anything
approaching the level of precision that vision allows (Jenkins, 1985). Hearings specialties are
thought to lie in perceiving speech and music. Beyond orienting the listener to the direction of a
crash, hearing isnt considered to be of much use in obtaining information about geometric
properties, such as letting us know the sizes and shapes of objects. At least this has been the bias
of orthodox approaches to perception. Shape and space perception are the traditional province of
vision; pitch and loudness perception are the traditional domain of audition. Whereas vision is
about awareness of environmental properties, audition seems to be largely about the awareness
of sound as such.
Orthodoxy, of course, is likely to inspire heterodoxy. For perception, a contrary treatment
can be found in the ecological approach of James Gibson. A major innovation is its focus on
perceiving object properties rather than sound propertieshearing a small, hollow, glass object
falling onto a hard surface rather than a loud, low-pitched, brief sound. In what follows, we
describe this approach in general terms. From this description, it will become apparent that
limiting the perception of geometric properties to the domain of vision is more tradition than
necessity. Subsequent sections will address the lawfulness of acoustic structure that ought to
permit auditory perception of geometric properties. Finally, we summarize research that
documents listeners success.
The ecological approach to perception
The ecological approach to perception is a metatheory. As such, it describes a particular
conceptualization of how we can be aware of our surroundings. In general, metatheories endorse
a particular style of framing questions, promote certain strategies for addressing them, and seek
particular kinds of answers.1 The prevailing metatheory is that perception begins with inadequate
input, which becomes meaningful by virtue of internal computations. These computations permit
awareness of the world indirectly in the form of a mental representation. In contrast, the
ecological approach asserts that input, once properly construed, is rich and lawful and specific to
its source directly, without elaboration by internal mechanisms.
The properness of the construal is an important issue in revealing lawfulness and this
construal is what motivates the epithet ecological. Consider the form that a so-called ecological
law takes:
generates
event structured energy distribution.

specifies
1
Theories are consistent with these frameworks but aim at a more specific level, such as a theory of space
perception or a theory of movement perception.
The inclusion of something as coarse-grained as an event highlights the ecological commitment
to discovering the appropriate level of description of an animal-environment setting. The object
undergoing some style of change is what structures the energy patterns that reach our perceptual
systems. Significantly, the object undergoing some style of change is also what is perceived. In
the place of the kinds of isolated sound properties that are the focus of traditional psychophysical
approaches, the source event and its properties are the focus of the ecological approach.
In the opening example, the bouncing bottle generates a particular acoustic pattern that is
specific to the event that gave rise to that pattern: a bouncing bottle. A breaking bottle or a
bouncing ball would generate different acoustic structure specific to those different events just as
those same events would generate distinct optical structure. In the domain of ecological
acoustics, our task is twofold: (1) to document the capabilities of perception on the basis of
sound, and (2) to identify what in the acoustic structure supports those capabilities. We can
expect to exploit successes in other perceptual systemsmost notably, vision and touchin
guiding our search for the relevant description. We are seeking invariants of energy distributions
that are generated lawfully by a particular event. These reliable patternings may well be
indifferent to the medium.2
Protocol studies and the possibilities of perceiving on the basis of sound
It is not uncommon for experimenters in a traditional psychoacoustics experiment to ask
2
We are endorsing the notion of amodal invariants: The same abstract invariant ought to characterize all of the
energy distributions structured by a given event. Some ecological psychologists consider invariants to be mode
specific, with different invariants in each type of energy distribution specific to the same event. Others argue for
intermodal invariants, a position that an event is specified by one emergent invariant that is defined over all energy
distributions. At the present time, the last of these is very difficult to assess experimentally.
their listeners to match two tones on some low-level dimension such as loudness or pitch. Of
course, loudness is affected by the frequency of the tone as well as its amplitude; pitch is affected
by the amplitude as well as the frequency. From these carefully controlled studies, one generates
the classic equal loudness contours that show how perception is faulty in its ability to faithfully
register physical properties. Listeners in these experiments adopt what Gaver (1993a) has dubbed
a musical attitude. They are paying attention to the sounds as such with no concern for the
source event that produced them. Given that the source event is typically a tone generator, and
the sounds tend to be harmonic with a relatively simple development over time, they really have
no choice. But, of course, this bears little resemblance to how we are guided by hearing under
ordinary circumstance. Everyday sounds tend to be inharmonic, vary complexly over time and,
more than likely, vary along dimensions of pragmatic utility. Everyday listeners care about the
source of the sound, not the sound itself (Figure 1). They want to know what happened and what
it means for them (Gaver, 1993a; Schubert, 1975).
More in keeping with the style of everyday listening are so-called protocol studies.
Listeners are presented with a variety of recorded sounds and asked to identify what they hear.
long, wooden,
E dropped on a hard M
surface
loud,
low-pitched,
muddy
Figure 1. A pool cue drops to the floor and bounces repeatedly, generating a complex sound pattern. An everyday
listener (E) focuses on the properties of the source of the soundits size, material, and type of interaction. Most
psychoacoustics research asks participants to be musical listeners (M), focusing on the properties of the sound.
They are generally quite successful at identifying footsteps, clapping, hammering, filing, tearing
paper, jangling keys, and so on (Gaver, 1988; Vanderveer, 1979). Protocol studies are, of course,
only a preliminary step in ecological acoustics because they permit no systematic control of
individual events and they make no demands of metrical precision on the part of the listener
(Gaver, 1993a). Nonetheless, two nice theoretical points are apparent. First, it is only when
listeners are unable to identify the source of a sound that they resort to reporting its sensory
aspects, in effect converting from everyday to musical listening. Second, listeners confusions
tend to be of events that have similar temporal profiles (e.g., hammering and stepping; filing and
scratching), pointing at the relevance of higher-order structure.3 Moreover, even without an
avenue for demonstrating metrical precision, some of the distinctions are, in fact, quite subtle
(e.g., listeners can distinguish ascending vs. descending stairs; Gaver, 1988). All of these suggest
that systematic physical analyses are possible and are needed.
Consider mechanical events. The vibration pattern in the air is influenced by the materials
involved in the event and the types of interactions those materials undergo (Gaver, 1993b). The
type of interaction, whether it is scraping or splashing or slapping, affects the time-varying
amplitude and the spectrum of vibration. The tension and elasticity of the materials are restoring
forces. The size, shape, and density of material determine its inertia. Both sets of factors
determine how quickly an object returns from the deformed state brought about by its interaction
with other objects or surfaces. This, in turn, determines the frequency of its vibrations. In brief,
3
Gaver (1988, 1993a) identified three broad categories of interactionsvibrating solids, aerodynamic sounds, and
liquid sounds. In his protocol studies, confusion remained within category. As he noted, they need not if the
temporal patterns of the events are similar (e.g., as illustrated in the use of tubes, pegs, and beads to simulate the
sound of rain in the so-called rain sticks of Mexico).
sound is structured reliably by interacting materials. That is the rightward sequence of the
ecological law statement. The leftward sequencewhether this sound structure informs about
those materials and interactionsis the focus of experiments in ecological acoustics. What is the
object? What is it doing? Where is it doing it?
Categorical distinctions
Our choice of the example that begins this chapter was not accidental. The first experiment
to be labeled ecological acoustics was an examination of listeners ability to distinguish breaking
and bouncing events (Warren & Verbrugge, 1984). The first part of the study involved
categorizing natural sounds. Listeners were presented with tape recordings of various instances
of different sized bottles and jars falling to the ground and either breaking or bouncing. They
simply had to identify which event they heard (or whether they could not tell). The success rate
was 99%. The next step was to try to identify what in the sound structure might allow the
distinction to be made. As Figure 2 illustrates, the breaking event has an initial burst of noise.
Even when this burst was eliminated from the recording, however, listeners still achieved 96%
accuracy in the identification of breaking. What remained, of course, was the timing of the
Figure 2. (left) A bounce produces a single, damped, quasi-periodic pulse train

whereas (right) a break is characterized by an initial rupture or burst followed by
multiple, damped, quasi-periodic pulse trains.
subsequent pulses. Artificial sound tokens were constructed, therefore, by means of manipulating
the timing of these pulses while leaving their spectral characteristics constant.
The construction of synthetic tokens began with pieces of a broken bottle. Four of the
largest pieces were dropped individually and recorded as they bounced. In one type of synthetic
token, the onsets of every impact of each of the four pieces were synchronized by inserting
appropriate amounts of silence between subsequent bounces so as to preserve the single,
damped, quasi-periodic pulse train that characterized bouncing. (Despite the presence of four
pieces, the synchrony supported the perception of a single object.) Sound events synthesized on
this basis were successfully identified as bouncing with 92% accuracy. Synthetic tokens
characterized only by a decline in the amplitude of the impacts without concomitant damping of
the temporal pattern (i.e., with the intervals equated between each of the bounces) were not heard
as bouncing. In a second type of synthetic token, the onsets of the four pieces were synchronized,
but then each followed a different damped, quasi-periodic pulse train. These were identified as
breaking with 87% accuracy (whether augmented by a noise burst or not).
The foregoing is an exemplary study in ecological acoustics. Although such a pedigree
makes it more likely that a researcher will be interested in source properties, a handful of more
traditionally oriented psychoacousticians have contributed to this literature. Two studies of
interest concern protocol events that listeners identified successfully: footsteps and clapping.
We already noted one subtlety of footsteps that listeners can pick up on: whether those
footsteps are going up the stairs or down. Given the different contributions of heel and toe strikes
to those two directions, this success is understandable. A possibly more subtle distinction, the
sex of the walker, can also be discerned simply on the basis of the sound of the walkers
footsteps (Li, Logan, & Pastore, 1991). Sixteen different walkers, eight of each sex wearing the
same style shoe (low, solid synthetic heel), walked at their normal pace on a hardwood stage,
taking eight steps directly towards a microphone. Each of these strolls was recorded and a four-
step sequence from the middle was presented to listeners over headphones. The task was a
simple categorization as male or female. Males were identified correctly on 69% of the trials;
females were identified correctly on 75% of the trials. But there were substantial differences
within each gender category. Half of the males and half of the females were identified correctly
at least 85% of the time. Two more of the females were identified correctly at least 70% of the
time. But the remaining walkers were identified at no better than a chance level. A statistical
evaluation of various anthropometric measures indicated that walker height accounted for 70%
of the variance in the judgments of gender. The investigators conjectured that height was really
standing proxy for height of the center of mass, CM, which is known to differ for males and
females. Although they did not determine individuals CMs, we estimated them from
anthropometric standards and, indeed, these estimated CMs account for 76% of the variance in
judgments of gender. Spectral analyses suggested that male judgments were more likely when
the acoustics were characterized by more energy in the low frequency range with rapid spectral
rising and falling. Conversely, female judgments were more likely when the acoustics were
characterized by more energy in a higher frequency range with slow spectral rising and falling.
Clapping is another sound event that could be identified easily in a protocol study, but does
it allow listeners to discern anything about the clappers themselves? Twenty different clappers,
ten of each sex, were instructed to clap for 10 seconds at their normal rate, as they would
normally clap after an average concert or theater performance (Repp, 1987). Recordings were
presented to listeners over headphones, with the entire sequence being presented once for
familiarization before it was presented a second time for response collection. Since the
10
individuals were all known to one another, the task was to label each bout of clapping as
belonging to one of the 20 people whose names appeared on an alphabetic list. Performance
(11% correct) was better than chance (5% correct) but it was not impressive. Self-recognition
was considerably better (46% correct) but, of course, that could have benefited from factors over
and above the acoustic specification of source properties.4 Even when re-scored simply for
whether a clapping bout was classified as having been produced by a male or female (even if the
specific identity was wrong), performance was only 54% (compared to a chance level of 50%;
Repp, 1987; see also Tousman & Pastore, 1989). Nonetheless, listeners performance was
systematic. A search of acoustic characteristics that encouraged the choice of a male or female
label revealed four factors that together accounted for 85% of the variance in judgments. Most
important was the contribution from the inter-clap interval: the slower the applause, the more
likely it was to be labeled male. Next in importance was the amplitude of the clapping: the louder
the clapping the more likely it was to be labeled male. The final factors are both spectral, with
the most intuitive interpretation being that low-frequency resonances were considered male.
None of these factors actually distinguished male and female clappers, which accounts for the
poor performance. But in a different sequence of claps, produced by a single person with
different hand configurations (palm-to-palm, finger-to-palm, or something in between), a new
group of listeners was successful in judging the configuration of the hands that produced the
clapping sequences.
Quantifying the mechanical properties of human effectorsthe legs of a walker, the hands
4
When the record of a movement is a visual trace (e.g., a trajectory of a figure-8 on a computer screen), people are
also better at identifying their own movements than identifying the movements of another (Prinz, 1997). Some
would like to understand that superiority with respect to a common code for perception and action.
11
of a clapperthat give rise to particular acoustic properties is not straightforward. Moreover, the
sample of sound-producers may be idiosyncratic, either in physical dimensions (height and
weight combinations, hand sizes and shapes) or personal style of walking or clapping, that may
introduce variation that masks the dimension of interest. Finally, judgments of characteristics
such as the sex of the sound source may simply reflect general (and often erroneous) sex
stereotypes. Mechanical events, in contrast to biological events, allow more control over the
source characteristics with a consequent increase in the resolution that can be asked of
perceivers. But, in the modern idiom, if you dont ask, they dont tell. We turn next to
experiments that ask for modest performance on the part of listeners, rank-ordering sounds
without indicating their appreciation of relative differences.
Perceiving mechanical events on the basis of sound
We know that listeners can appreciate that a collision event did not entail breaking. Is
anything else known about the objects in the event? A variety of questions could be asked: What
were the materials, how large were the objects, were they solid or hollow? One such
investigation focused on percussive sound eventsa single impact of a mallet on a hollow
receptacle (Freed, 1990). Six mallets that varied in hardness were used to strike four receptacles
(cast aluminum cooking pans) that varied in size. The 24 percussive events were recorded in an
anechoic chamber and presented over a speaker. The guiding force was explicitly ecological, to
focus listeners on physical properties of the objectsthe mallets hardnessrather than abstract
properties of the soundwhether it is bright or dull, thin or full (cf. Lichte, 1941). Even with this
ecological goal, psychophysical timidity was nonetheless in evidence in the study. The sound
pressure level was roughly equalized across tokens, thereby restricting differences to spectral
12
parameters. Each session began with a demonstration of the hardest mallet hitting first the
smallest pan and then the largest pan, followed by the softest mallet hitting the same two pans.
Listeners were told that these were instances of a hard and a soft mallet. The entire sequence of
sounds, which included four repetitions each of six mallets striking four different receptacles,
was then played once to familiarize listeners with the range of sounds. (The untutored
informativeness of the sounds was not strictly assessed.) Perceived mallet hardness was indicated
by continuously adjusting the length of a visible line segment on a scale marked from 1 to 9 for
softest to hardest. Harder mallets generally elicited higher ratings of hardness and seemed
unaffected by the pan that was struck. More germane for the interests of the investigator was the
identification of acoustic parameters that would be useful as predictors of timbre. A combination
of four spectral parameters characterizing the initial 325 ms of each event accounted for 75% of
the variance in hardness ratings. One is essentially a loudness measure (the log of the area under
the spectrum). Its slope represents the softening over time. A third reflects the brightness (the
mean of the centroid of the spectrum) and the fourth (the time-weighted average of the centroid
of the spectrum) represents the darkening over time.5
An inevitable implication of source-oriented auditory perception is that information about
an event is infused throughout the acoustic signal. Mechanical events involve the movements of
masses that are mutually constraining. How an event began influences how it can unfold; how it
is unfolding is informative about how it may have started. In principle, therefore, later-coming
acoustic structure can influence perception of an earlier event. This was the premise of an
5
Since these parameters are not invariant over resonators, they are unlikely candidates as the definitive constraints
on perceived mallet hardness, given that listeners mallet ratings were indifferent to what was struck (Lakatos,
McAdams, & Causs, 1997).
13
investigation of the perception of steepness based on the sound of a ball rolling down a ramp
(Fowler, 1990). A steel ball was recorded rolling down five different ramps (10, 20, 30, 40 and
50 deg) onto either a flat track or an upward sloping track. During such events, the time that the
ball spends on the track is determined by the slant of the initial ramp, but in opposite ways for
the two tracks. On a flat track, a longer duration means that the early part of the event was
shallower; on an up-slope, a longer duration means that the early part of the event was steeper
(Figure 3). The experimental sequences were hybrid stimuli constructed by splicing acoustic
signals from the extremes of the track portions onto acoustic signals from each ramp slope. That
is, for a given ramp the track portion was replaced by four options: the flat and the upward slope
that had accompanied the 10 and the 50 ramps. If the later-coming acoustic structure influences
perception of an earlier event, then listeners judging the slant of the ramp ought to be influenced
by the track portions they hear and in specific ways. For the up-sloped tracks, the shorter
duration track sounds ought to yield flatter perceived ramp slopes than longer duration track
sounds. For the flat tracks, shorter track portions ought to yield steeper perceived ramp slopes
than longer track portions. In other words, duration per se is not the critical acoustic structure.
a c
b d
Figure 3. (a) Initial ramps of different slopes will influence the length of time a ball bearing
spends on a flat track before falling off. (b) The time on the track is longer the flatter the initial
ramp. (c) When the track slopes upward, the ball slows before rolling back down. (d) The time
on the track is shorter the flatter the initial slope.
14
Prior to the experimental test, listeners were presented with recordings of just the ramp
portions from the 10 and 50 ramps five times in alternation and told which was which. They
were instructed that in the test, they would hear the entire rolling event and they were to judge
whether the ramp portion was steep or flat. Ten listeners heard the ramps spliced onto the flat
tracks and another ten listeners heard the ramps spliced onto the up-slopes. Ramps were
generally ordered appropriately in all conditions, even though the track portions did not help this
specific discrimination. More importantly, the track portions did produce the expected
interaction. Listeners in the flat-track condition heard all ramps as steeper if they had been
spliced to a short duration track portion; listeners in the sloped-track condition heard all ramps as
steeper if they had been spliced to a long duration track portion. A subsequent experiment,
limited to the flat-track condition, addressed the duration issue directly. A track portion from the
10 ramp was cut at regular intervals until it was equal in duration to the 50 portion (it was
essentially halved). Ramps appended to this shortened track portion were indeed heard as steeper
than those with the original long, 10 flat track but not as steep as those with the original 50
track. In other words, duration is not the only dimension that is informative about ramp slope.
Still audible, for example, was the balls revolution speed.
The upshot of this research is the emphasis on perceiving the source event. Had conditions
been limited to the flat tracks, one might infer that listeners are simply influenced by a general
auditory cue, durational contrast. The opposite durational pattern from the upward sloping track
instead supports the notion that listeners recover physical event properties from the information
available in acoustic structure.
Both the mallet hardness and the ramp slant studies show that listeners perceive properties
of environmental objects and surfaces with some degree of gradation: mallets are more or less
15
hard, ramps are more or less slanted. However, neither the mallet nor the slope investigations
exploited a specific advantage of mechanical events. Metrical precision of perception, not just
rank-ordering, can be assessed. That is to say, perceived increments in a dimension of interest
can be evaluated relative to physical increments in that dimension. This has been the focus of
recent investigations of shape and size perception on the basis of sound.
A first step toward metrical precision in perceiving mechanical events
The in-principle argument for specification is straightforward. To the extent that an event
structures sound reliably, that structure ought to specify the source event. But what provides the
metric? Characteristic modes of vibration are one possibility. Objects with a high degree of
symmetry, for example, have three orthogonal modes (Figure 4) that are determined by the
physical dimensions (size, shape) and material properties (mass, density, elasticity) of the source.
An object set into vibration will conform to one of its characteristic modes despite variation in
pitch and timbre (Lakatos et al., 1997). The relative contribution of each mode, however,
depends on how an object is struck. This so-called exciter-resonator relationship is often
nonlinear (Fletcher & Rossing, 1991). Consequently, one methodological strategy is to keep the
strike position constant while varying spatial dimensions of the resonating object.
In one such investigation (Lakatos et al., 1997), the objects were long bars made of steel or
a b c
Figure 4. A metal bar tends to vibrate (a) transversely, (b) longitudinally,

and (c) torsionally. The modes of vibration of are orthogonal.
16
wood, suspended lengthwise and struck at the center with a mallet (steel for the steel bars, resin
for the wooden bars). The bars were of a fixed length but varied in thickness and width; bar
girth is what listeners had to discern (the specific dimensions differed for the two materials).
Sound recordings were made in an anechoic chamber, and matched for loudness. Multiple
recordings were made so that only clear samples were selected. During the experiment, sounds
were presented over a speaker in a sound-isolation booth. A trial consisted of a pair of sounds, in
succession, accompanied by two response alternatives presented visually. These alternatives
represented the actual cross-sectional proportions of the bars (scaled to the computer screen), but
in both the correct and the incorrect order. Listeners had to choose which order corresponded to
the order of the sounds they heard (20 practice trials without feedback familiarized them with
this so-called two-alternative forced-choice, or 2AFC, procedure). Prior to the practice trials,
listeners were allowed to strike five sample bars (not from the experiment set) with the
appropriate mallet to familiarize themselves with the type of interaction they would be hearing.
Finally, they were told to use any available timbral cues to make their decision.
Listeners who failed to attain a 75% performance criterion were excluded from the analysis
(out of the original 60 listeners, 5 did not reach criterion for the steel bars and 10 did not reach
criterion for the wooden bars). With these exclusions, listeners performance did indeed vary
with differences in the width to height ratios and improved the greater those differences were.
For metal bars, the more block-like cross-sections tended to cluster together as did the more
plate-like cross-sections. This pattern was not evident for the wooden bars, which listeners found
harder to discriminate (perhaps because the signals were so much shorter). Acoustic analyses
indicated that matching performance correlated strongly with the frequencies of the vibration
modes. In particular, for metal bars, the frequencies of the torsional modes, FT, accounted for
17
86% of the variance in listener responses. As an alternative, the ratio of the transverse bending
modes dependent on width to those dependent on height, FW/ FH, accounted for 88% of the
variance. These relationships were less secure for the wooden bars, not only because the variance
accounted for was less (58% by with FT; 67% by FW/ FH) but also because those components are
sometimes quite weak or absent in those bars. No attempt was made to provide a comprehensive
account of performance with both sets of bars (Lakatos et al., 1997). One problem is that wood is
an orthotropic material (Rossing & Fletcher, 1995); its mechanical properties, unlike metals,
tend to vary along three perpendicular axes.
At some level, the preceding illustrates a finer degree of resolution on the part of listeners.
They were asked to discriminate the cross-sectional shapes of struck bars whose width-height
ratios ranged from .13 to 1.00. Given the 2AFC procedure, however, we only know that listeners
could discriminate different shaped rods, and that discrimination was easier the more different
the rods were. But we have no appreciation for listeners accuracy in perceiving the rods. In the
absence of the visual matches, could they have indicated that one rod was block-like and another
was plate-like? They reached a performance criterion of 75%, but that was after poor listeners
were eliminated. Given the precision in the sound recordingsanechoic chambers eliminate
reverberation, loudness is equalized, only clear recordings are usedshould we be impressed
with listener achievement? How would listeners fare if the sounds came from real objects in an
ordinary room? What if they were provided with some leeway in responding? Do they actually
need all of the practice and restrictions that characterize the research considered so far?
Metrical precision in perceiving struck plates
We have been building a case, at least implicitly, that by the very nature of the research
18
questions asked in auditory experiments, the science of perception doubts the fine-grained spatial
capabilities of hearing. Consider this opinion from Sir Arthur Schuster in 1882, who was
illustrating the challenge to be faced by the then-new science of spectroscopy:
To find out the different tunes sent out by a vibrating system is a problem which may or may not
be solvable in certain special cases, but it would baffle the most skillful mathematician to solve the
inverse problem and to find out the shape of a bell by means of the sounds which it is capable of
sending out (cf. Gordon & Webb, 1996).
Mathematicians generally endorse this skepticism. Kac (1966) posed the question explicitly:
Can one hear the shape of a drum? and it has been answered explicitly, You cant hear the
shape of a drum (Gordon & Webb, 1996). Isospectral companionsidentical spectra produced
by two manifolds that differ geometricallyare to blame. But, of course, mathematicians
operate in idealized space. Just because isospectral companions are possible need not mean that
they are representative or problematic. First, they seem to be the exception (Gordon & Webb,
1996). Second, trying to implement such manifolds with real physical objects would likely
introduce differences, however small, in their physical parameters (Kunkler-Peck & Turvey,
2000). What happens if we ask the shape question of perceivers rather than mathematicians?
The drums were, in fact, flat steel plates (circle, square, and triangle) struck by a steel
pendulum bob released from a fixed location so that it would always impart the same amount of
energy (Kunkler-Peck & Turvey, 2000; Figure 5a). The plates, which had the same mass and
surface area, were simply suspended by fishing line to provide stability with minimal damping
and without eliciting vibration in the support structure. A listener sat on one side of a screen that
hid the shape that was being struck (about 1 m on the other side). On each trial, the bob was
released from the starting location and caught right after it bounced off the plate. This was
repeated three times so that a trial was defined by three strikes. Each of the three objects, was
19
a b c
H W
Dimension
Figure 5. (a) A simply suspended plate is struck by a pendulum bob. (b) A listener whose view is occluded,
adjusts dowels to indicate the height or width. (c) Proportional shape is distinguished.
presented three times in random order. Listeners had to indicate verbally which plate had been
struck, with no prior demonstrations of the sounds and no practice trials. Nonetheless, the correct
shape was chosen 58% of the time (where chance performance would have been 33%).
To make the task a little harder, a further experiment included the same three shapes in
wood and Plexiglas as well as steel, again with dimensions chosen to provide the same surface
area (mass differed across material but was the same for the three shapes of a given material).
Because listeners were asked about the material as well, the verbal response was replaced by
having them point at replicas of the nine objects mounted on the occlusion screen. Once again,
there were no demonstrations or practice trials. Perceiving material was absolutely
straightforward; one of the seven listeners made one misidentification. Shape performance was
comparable to the previous experiment. The correct shape was chosen 56% of the time (where
chance performance on shape again would have been 33%). Moreover, there were no systematic
confusions in the errors. This suggests that circles, squares, and triangles really sounded like
circles, squares, and triangles rather than simply being labeled by a strategy that could be
consistently wrong.
20
The shape experiment demonstrates a degree of geometric sensitivity. With real sounds in
an ordinary room, listeners identified shape reliably. Their responses were constrained, of course,
by the shape categories provided, thereby limiting how wrong they could be. A truly metrical
response would provide a stronger test. In order to remain in the realm of perceiving shape, this
time the steel plates were all rectangles but of different proportions: a square (48.2 48.2 cm), a
medium rectangle (38.1 61.0 cm), and a long rectangle (25.4 91.4 cm). Again, dimensions
were chosen so that the plates were equal in mass and surface area. The occlusion screen was
augmented with a response apparatus that allowed a listener to provide a visual match for a
plates height independently of its width (Figure 5b). The width indicator ranged from 0 to
2.5 m; the height indicator ranged from 0 to 1.5 m. As before, there were no demonstrations or
practice and this time there was no information about the number of objects. Additionally, there
was no indication of the sizes of the objects (other than the 2.5 1.5 m maximum allowed by the
apparatus). On a given trial, the listener was told which dimension to report before the plate was
struck. He or she adjusted the appropriate indicator to provide a visual match for the heard height
or width. Each of the three rectangles was presented six times (three for width, three for height).
The actual linear dimensions accounted for 98% of the variance in listeners responses.
Although perceived dimensions were underestimates of actual dimensions (ranging from 25.2 cm
to 44.5 cm for an actual range from 25.4 cm to 91.4 cm), they were in the approximate range. We
refer to this as definite scaling rather than as relative scaling (Bingham, 1993; Turvey & Carello,
1995) because responses are more than simply ordered, arbitrary magnitudes. They do not use
the entire range, nor do they use either extreme of the range. Listeners appear to have a definite
impression of size (indeed, the average reliability of their responses was 6%, which compares
favorably to the visual impression of size; cf. Norman, Todd, Perotti, & Tittle, 1996). Most
21
impressive is their sensitivity to shape (Figure 5c). When they attended to the height of the
rectangles, their responses were larger than when they attended to the widths of those same
rectangles. For the squares, those reports were the same.
Before addressing the acoustic support for this performance, lets consider one more
demand placed on the listeners. The preceding experiment was replicated with the three
rectangular shapes cut from wood and Plexiglas as well as steel. As before, all plates had the
same surface area; plates of the same material had the same mass. As before, listeners positioned
the report apparatus to indicate the dimension requested on a given trial. As before, perceived
dimensions (23.5 cm to 51.4 cm) were in the approximate range of actual dimensions (25.4 cm to
91.6 cm) with a mean reliability of 5.5%.
The simple support of the plates means that their vibrational dynamics are captured by the
two-dimensional wave equation (Rossing & Fletcher, 1995). The frequencies associated with the
solutions to that equation are given by
mn = 0.453 h E m+12 + n+12 (1)

(1 2) Lx Ly
where h is the thickness of the plate, E is Youngs modulus, is the mass density, is Poissons
ratio, m and n are integers indexing the vibratory modes, Lx is width, and Ly is height. Obviously,
the modal frequencies are influenced by more than the plates linear dimensions. The physical
parameters constrain how the plate bends (Figure 4). This is the hypothesized informational
support for perceived dimension. It accounted for 87% of the variance in perceived dimension in
the preceding experiment. Note that asking a listener to report height or width is asking that
listener to perceive selectively. Out of the flux of stimulation, they are to extract just that
structure relevant to the requested dimension. Analytically, this selectivity was captured by
22
calculating mn with m = 0 for one dimension and n = 0 for the other. While this works quite well
as a way of summarizing data, we really dont understand what it would mean for the listener to
do the same thing functionally.
Metrical precision in perceiving the lengths of dropped rods
The clean strike of a supported object (or a sequence of such strikes) allows sound structure
that is simpler than that normally encountered in everyday listening. The experimental rationale
is that the scientist needs to know whats in the sound in order to determine what listeners can
respond to. A different strategyinspired by the distinction between everyday listening and
musical listeningis to give listeners as much structure as they ordinarily encounter in everyday
events with acoustic consequences. Apart from the dinner gong, not many sounds are simple
strikes. Objects fall to the floor, they clatter and bounce and roll to a stop, with the sound
reverberating in the room. Listeners have access to all of that in ordinary experience so lets give
them access to all of that in the experiment. Once we assess how metrically precise listeners can
be, then we can worry about how to quantify the available structure.
This was the premise of experiments in which listeners were asked to indicate the lengths
of cylindrical rods that fell to the floor (Carello, Anderson, & Kunkler-Peck, 1998). In one
experiment, wooden dowels 1.25 cm in diameter were cut in lengths from 30 cm to 120 cm (in
15 cm increments). In a second experiment (with different listeners), the diameters were .32 cm
and the lengths ranged from 10 to 40 cm (in 5 cm increments). On a trial, a rod was dropped five
times in succession. To standardize the drop as much as possible, a rod was balanced at its center
of mass on a support 72 cm above a hard linoleum surface. The turn of a handle allowed the rod
to fall from a fixed height. Listeners sat at a student desk on the other side of an occlusion screen
23
Perceived Length (cm)

120
r2 = 0.97
90
60
30
0
0 30 60 90 120
Actual Lengh (cm)
Figure 6. (a) A listener adjusted the position of a report board to indicate the length of a rod that had dropped
on the other side of an occlusion screen. The rod was released from a support stand by the turn of a small
lever. (b) Length was perceived with surprising accuracy (perfect performance is indicated by the dashed line)
even though listeners were unaware of the number of rods or their possible lengths.
in front of a response apparatus that allowed them to position a marker anywhere from 0 to 2 m
to coincide with how far they could reach with the rod (Figure 6a). They were provided no
practice and no information about the number or sizes of the rods; they simply listened.
For the large rods, perceived lengths ranged from 24 to 95 cm. Actual length accounted for
95% of the variance in perceived length. For the small rods, perceived lengths ranged from 14 to
27 cm. Actual length accounted for 95% of the variance in perceived length. The slopes of the
perceived-actual functions were different for the two rod sizes, with the discrimination being
sharper for large rods (slope = .78) than for small rods (slope = .44). Nonetheless, the two data
sets in combination provided a strong dependence of perceived length on actual length (an
overall slope of .77 with 97% of the variance accounted for; Figure 6b).6
For the acoustic analyses, the sounds were recorded from the listeners position under the
6
With somewhat less success, Gaver (1988) asked listeners to indicate the lengths of bars, made of iron or oak, that
were struck once at their centers sitting on a carpeted floor when struck. The sounds were recorded and played to
listeners over headphones. Not all listeners discriminated length and they were differentially affected by material.
The fidelity of the recordings may be at fault, along with the less informative impact events.
24
same conditions as described in the foregoing with the exception that a given rod was dropped
only three times. One might think that a simple acoustic variable, something that relates to length
straightforwardly, ought to account for performance. But neither the duration of the signal, its
average amplitude, nor its frequency centroid approached the success of actual length in
constraining performance. Average amplitude fared best, accounting for 70% of the overall
variance but it fared less well on the set of large rods (21%). Frequency centroid accounted for
66% of the overall variance but less for the individual sets (59 % and 37% for the large and small
rods, respectively). Signal duration accommodated the large rods better than the small (65% vs.
12% of the variance) but overall, it accounted for only 9% of the variance in listener responses.
Although actual length predicts perceived length successfully, it cannot be the constraining
variable. Length is a geometric property, not the kind of mechanical property that can affect
acoustic structure. Its success in the preceding experiments was because the rods of different
diameters were also in non-overlapping length ranges. When this is no longer so, actual length is
not a good predictor of perceived length (Anderson, Carello, & Kunkler-Peck, 1996). Using the
same experimental procedure as in the foregoing, listeners judged five rod lengths cut from
dowels of three different diameters. Perceived length increased with increases in actual length
but the latter accounted for only 39% of the variance in overall responses. Similarly, when rod
lengths are fashioned from different materials (steel, Plexiglas, and wood), actual length
accounts for only 18% of the variance in perceived length.
Manipulations of length, diameter, and material density are simply different ways of
manipulating a higher-order property, an objects mass distribution. This property is quantified
through the inertia tensor, essentially, the resistances of an object to being rotated in different
directions. As noted earlier, inertia influences the vibratory pattern in an impact event and, as
25
such, provides a sensible candidate for the mechanical constraint on perceived length by hearing.
Not surprisingly, rods inertia tensors account for nearly all of the variance in perceived length in
the initial experiments with large and small rods and 91% of the variance in the diameter
experiments. Manipulations of material provide an interesting case, however. Material density
has consequences for the inertia tensor, to be sure, but it also is related to elasticity and stiffness
which, in turn, have consequences for how a body returns to an equilibrium state after being
displaced by an external forcing function (such as accompanies dropping to a surface). Indeed,
the rods inertia tensors alone do not predict perceived length of rods of different densities. But
when augmented by Youngs modulus of elasticity, the variance accounted for is 96%.7
As a matter of pedagogical convenience, we have been characterizing the achievements of
listeners with respect to perceiving particular properties of objects (e.g., length, shape, slant) or
types of interactions (e.g., jangling, bouncing, dropping). In its focus on awareness of the source
rather than awareness of the sensations, this emphasis has illustrated a central concern of the
ecological approach. But it is incomplete. Perceptions raison dtre is guiding activity.
Perceivers need to be aware of objects and events because of the consequences for what
perceivers-as-actors can do. From the perspective of ecological psychology, perception is not
awareness of objects and events per se but awareness of their behavioral relevance. Such
opportunities for behavior are what Gibson (1979, 1983) termed affordances. Behavioral
possibilities are central to a good deal of acoustic research to which we now turn.
7
The relevance of the inertia tensor to the perception of length has long been appreciated for dynamic touch (cf.
Turvey, 1996), work which inspired the dropped rods experiments. And the inspiration goes both ways. The inertia
tensor completely constrains perceived length when wielded objects vary in diameter and density (Fitzpatrick,
Carello, & Turvey, 1994), but elasticity is a constraint when the objects are nonrigid (GrandPre & Carello, 2001).
26
Affordances: Perceiving behavioral possibilities
We have already introduced the notion of affordance indirectly in the rod dropping
experiments. Listeners were not asked to provide a report of length in units of inches or
centimeters. Instead, they were asked to position the visible report surface to coincide with how
far they could reach with the rod they had heard. Reconceptualizing absolute geometric
properties such as length in terms of activity-relevant properties such as reach-with-able is a
goal of ecological theory. An affordance is a legitimate occupant of the left slot in the ecological
law statement. Such a reconceptualization is ultimately a goal of ecological research as well. But
fulfilling that goal requires that we formally fill the right slot in the ecological law statement as it
pertains to a particular affordance. What pattern of structure in an ambient energy array is
specific to an affordance such as reachable? Reachable is a very different kind of thing from
length. We know how to measure length; we have to discover how to measure reachable.
Ecological psychologists are faced with the very real possibility that the currently available
scientific toolkit is not up to the job of quantifying information specific to affordances. In the
meantime, a good deal of the discovery process involves identifying the variety of affordances
that are perceived and the kinds of manipulations that affect them. Oftentimes, the targeted
affordances have already been investigated in the visual domain and the hope is that finding
commonalities will give us a foothold on identifying the information. And heres why. We are
seeking invariants of structured energy distributions, patterns that are always produced by a
given affordance regardless of incidental details that can give rise to dramatically different
sensations. An objects shape and material surely matter to the intensities and wavelengths of
reflected light that reach the eye but they do not matter to whether you can reach that object.
Instead, we seek some invariant of structured light specific to whether something is reachable or
27
not. And, on extension, whether an object structures light or compression waves is also irrelevant
to whether you can reach it. We might, therefore, seek a higher-order pattern common to both
ambient energy arrays that is lawfully related to what is reachable (see FN 1).
A key issue is one of prospective control (e.g., Turvey, 1992; E. J. Gibson, 1994). Your
actions are organized in such a way that you can effect certain outcomes. For example, picking
up a pencil from the desk does not begin with your throwing an arm out in its general direction
and hoping for the best. You can see that the pencil is within reach with an outstretched arm or
that it will require an additional bend at the hip. Indeed, examinations of this visual ability have
shown that sensitivity to the boundary of reach is body-scaled, that is, it is the same for tall and
short reachers (with correspondingly long and short arms), once the boundary is scaled to the
appropriate effector (Carello, Grosofsky, Reichel, Solomon, & Turvey, 1989). But if that same
pencil rolled off the desk out of view, could we hear whether it dropped within reach?
Developmental data clearly support the salience of hearing what is reachable in that infants
reach more often for a sound source in the dark when that sound source is within reach than
when it is out of reach (Clifton, Perris, & Bullinger, 1991; Perris & Clifton, 1988). Formal
auditory reaching experiments were modeled after those in vision. Listeners were selected to be
tall or short. The sound-emitting target, a kind of rattle, was placed at different distances from
the listeners (Figure 7a). When the reach was to be with an arm outstretched from the shoulder,
distances ranged from 38 to 110 cm, in 8 cm increments. When the reach was to be with the
outstretched arm augmented by a bend at the hip, distances ranged from 75 to 150 cm, also in 8
cm increments. There were three repetitions of each distance within each type of reach, and each
block began with 10 practice trials. During a trial, listeners simply judged whether the rattle was
within reach. No actual reaches were conducted until after all judgments were collected, when
28
a b 100 c 100
% Reachable
% Reachable
75 75
50 50
25 25
0 0
30 50 70 90 110 0.5 1.0 1.5
Distance (cm) Distance/Arm
Figure 7. (a) A blind folded listener judges whether the rattle is within reach. (b) Short listeners (open circles)
cannot reach as far as tall listeners (filled circles). (c) When target distance is scaled by the appropriate effector,
judgments by tall and short listeners do not differ.
the actual maximum reaches in the two conditions were obtained. In all respects, the results
rivaled or exceeded those from vision. Whereas visual judgments tended to be overestimates
(Carello et al., 1989), auditory judgments of maximum reach did not differ from actual
maximum reach (Rosenblum, Wuestefeld, & Anderson, 1996). As would be expected, tall
reachers had a farther reaching boundary than short reachers (Figure 7b), but these boundaries
were the same when scaled intrinsically (Figure 7c).
Attempts to characterize the informational support for auditory perception of reachability
implicate a higher-order combination of sound intensity and the ratio of direct-to-indirect sound
reaching the listener (Wightman & Jenison, 1995). But these simply address distance-relevant
structure. The key to perceiving what is reachable must address the body-scaling. How do I
know that rattle is reachable by me? One intriguing conjecture received modest experimental
support. Judgments of what is reachable were less consistent (though no less accurate) when
acoustic structure was restricted to one ear (with the other ear being substantially muffled).
Given the allometric relationship between inter-aural distance (i.e., cranium size) and arm length
(Snyder et al., 1974), binaural superiority suggests the possibility of an intrinsic metric in
acoustic structure that, in effect, scales what is reachable for the listener (Rosenblum,
29
Wuestefeld, & Anderson, 1996). An analogous intrinsic metric in visual structure, inter-ocular
distance, has been promoted as serving a similar function for optically specified reaches in the
preying mantis (Michaels, Prindle, & Turvey, 1985) and distance perception, in general, for
humans (Michaels, 1986).
Just as questions of perceiving what is reachable ecologize the problem of distance
perception, other issues of space perception can be made similarly functional. Consider the
question of localizing a sound source, one of the few routinely source-oriented classical
questions. In classical hands, it is simply a question of the distance and direction to a sound-
emitting object. But a sound-source does not sit in isolation. It is found, for example, among
surfaces that reflect its sound.8 Its relation to these other surfaces may have consequences for
behavior. For affordance-minded researchers, one consequence that has been addressed
experimentally is whether the gap between the sound source and a vertical surface would allow
the listener-as-walker to pass through. A recording of a complex sound, the assembly call of a
mallard duck, was played through a loudspeaker that faced a side wall in a long room. The
speaker, which was at ear-height, was placed at 12 distances ranging from 20 to 75 cm from the
wall. The listener stood with eyes closed, one shoulder next to the wall 2 m from the aperture
and, on a given trial, judged whether he or she could walk through the gap between the sound
and the wall. For all eight listeners, narrow gaps were heard as not allowing passage, wide gaps
were heard as allowing passage (Figure 8a) and, as is standard in these kinds of experiments,
variability in judgments was greatest around the transition from passable to not passable (Russell
8
Our discussion is limited to situations involving direct sounds. A considerable literature addresses the perception of
surface layout on the basis of echolocation, in which the listeners use reflections of self-produced sounds (e.g.,
footfalls, cane taps, spoken syllables). An ecological treatment can be found in Stoffregen & Pittenger, 1995).
30
& Turvey, 1999; cf. Fitzpatrick, Carello, Schmidt,

100
a b
80 & Corey, 1994). This transition occurred at 1.11
% Yes
60 times the listeners shoulder width which

40
compares favorably with the value of 1.16
20 E1
E2 obtained with similar work with visual apertures
0
0 20 40 60 80 100 120 140
Distance (cm) (Warren & Whang, 1987).
Figure 8. (a) Judgments of whether an aperture
could be walked through are categorical. (b) Since the right shoulder was adjacent to the
When the sound source and listener are shifted so
that all gaps are passable, the percentage of wall, listeners may have been aware of the walls
passable responses increases although the pattern
is still somewhat categorical. location not by reflected sound but by touching
the wall with that shoulder during ordinary
postural sway. Judging what allowed passage, then, would simply be a matter of locating the
sound source to the left of the left shoulder, perhaps constrained by a certain magnitude of
azimuth. To address this possibility, the sound source and listener were positioned 60 cm away
from the wall. The relationship between the speaker and the sound source remained the same, but
all apertures were, in fact, passable. Judgments based on body-scaled information about gap size
should have been 100% passable; judgments based on the location of the speaker relative to the
left shoulder should simply have shifted the category boundary rightward. Figure 8b shows that
performance by a new group of eight listeners was reasonably categorical with the boundary
shifted to the right. However, listeners were much more likely to judge gaps as passable in
Experiment 2 (70% of the time compared to 40% in Experiment 1), suggesting some body-
scaling. The remaining categorical pattern may reflect a methodological quirk: Listeners may
have found it difficult to say yes on every trial. Clearly, unlike Experiment 1, there were no
apertures that were judged not passable all of the time.

31
A more straightforward test of the azimuthal hypothesis was provided by locating the
aperture at three distances from the listener. If a gap is judged passable when the sound source
exceeds a certain magnitude of azimuth, then closer gaps ought to be judged more passable than
farther gaps. However, performance was equivalent with the aperture at 1 m and 2 m; it
deteriorated completely with the aperture at 3 m. Minimally, this result suggests that listeners
were not using the azimuthal direction of the sound source. Interestingly, listeners who had seen
the spatial layout of the room before beginning the experiment produced a cleaner category
boundary than those who were led into the room with their eyes already closed. That is, they saw
the size of the room and where they would be standing (but not where the speaker would be).
This suggests that the scaling of perception on the basis of sound is very abstract indeed.
Although obviously ecological in context, these kinds of affordance experiments may
encourage participants to be too analytic (Heft, 1993). If listeners dwell on what the
experimenter wants them to do, they may try to elect a conscious strategy to satisfy those
demands (recall the problem of stereotypes in perceiving the sex of a walker or clapper). The
analytic attitude, in some sense, undermines their naturalness in detecting the appropriate
information. When participants are hurried or less focused on explicit judgments they are, in fact,
more accurate (Heft, 1993). Making the perceptual task part of an on-going activity would be the
ideal situation from a theoretical as well as methodological perspective. This is easier to achieve
in the realm of interceptive behaviors, to which we now turn.
Acoustic information about impending collisions
Perceiving time to contact in the visual domain has been a topic of much investigation (for
reviews, see Lee, 1980, 1990; Tresilian, 1993). The basic premise is that, given a collision course
32
between a perceiver and a surface, the perceiver needs to see when the surface will be reached
before the collision actually occurs. This is so whether the perceiver is moving towards a surface
or a surface is moving towards the perceiver. A single event illustrates both of these. A base
runner-as-perceiver churns towards the catcher-as-upcoming surface; the catcher-as-perceiver
prepares to make the tag on that runner-as-upcoming-surface. In the simplest case of rectilinear
forward motion with constant velocity, the perceiver who would control that collision
prospectively needs to know, If current conditions persist, when will contact occur?
The optical variable that specifies time to contact (under conditions of constant velocity
and rectilinear motion) is a quantity, termed (Lee, 1976, 1980), which is given by the inverse of
the relative rate of dilation of an optical angle. In the preceding example, an optical angle defined
with the catcher as its base and the runners point of observation as its apex expands as the
runner gets closer to the catcher (Figure 9a). That angle increases gradually while the runner is
still far away but virtually explodes when contact is imminent (Figure 9b). Lee (1990) has argued
that is quite general and applies to any time-varying array variable. This includes other aspects
of optical structure (e.g., the relative rate of constriction of an optical gap might specify contact
between two objects; Bootsma & Oudejans, 1993) as well as any of a variety of acoustic array
a b
Time to Contact (s)
100 0
t1 t2
Angle(deg)
75
-2
50
-4
25
0 -6
0 25 50 75 100
Distance (ft)
Figure 9. (a) As a base runner approaches the catcher, the optical angle subtended by the catcher expands.
(b) The expansion (circles) changes most dramatically right before contact. For constant velocity, time to
contact changes at a constant rate (pluses).
33
variables that change over time as a sound source nears a listener. For example, as a sound
source approaches from the right, its loudness increases, its pitch rises in a pattern known as the
Doppler shift, and the onset of sound at the right ear precedes the onset at the left ear.
The amplitude changes are most like the traditional understanding of optical . Indeed, a
formal derivation of acoustic for rectilinear approach verifies an intensity-based structure
specific to time to contact (Shaw, McGowan, & Turvey, 1991):
acoustic = 2I/dIdt (2)
where I is given by the inverse square law and velocity is constant. Plots of intensity changes
during approach to a sound source look very much like Figure 9b. A variation on this theme has
also been derived for pass-by: the negative of the intensity relative to the time derivative of
intensity (Erwin, 1995).
This means that acoustic structure specific to time to contact exists. But can listeners use it?
Some listeners, at least, can perceive time to contact by ear as well as observers judge it by eye.
Blind adults listened to the sound tracks of filmed events that sighted adults viewed: approach
trajectories (either direct or near-misses less than 5 from midline) of 12 vehicles and 2 people
filmed at varying speeds (Schiff & Oldak, 1990). Vehicle sounds included engine and tire noises;
people sounds were primarily continuous talking. Film clips lasted from 4 to 6 sec after which
the approaching sound source vanished 1.5 to 6.5 sec before it would have reached the plane of
the participants shoulders. Participants pressed a key to indicate when the vehicle or walker
would have reached them had it continued at the same speed. Six congenitally or early-blind
adult listeners performed a least as well on the audio-alone sequences as did 20 sighted observers
on the video-alone portions (the small sample size of blind participants precluded a statistical
evaluation). Performance by both groups deteriorated for long times-to-arrival.

34
Two more groups of sighted observers (20 in each condition) listened to the audio alone or
were presented with the complete audio + video sequences. During those first 3-4 sec, their
performance did not differ from their video-alone counterparts nor did the combination of audio
and video improve performance (and all were underestimates). These groups also experienced
more difficulty for long times-to-arrival, with the deterioration being especially dramatic for the
audio-alone, sighted listeners. Not surprisingly, kinematic analyses of acoustic structure (and,
apparently, optical structure as well) for events in this velocity range indicate little differentiation
in intensity, Doppler, or inter-aural patterns beyond 3 sec (Wightman & Jenison, 1995). In other
words, listeners and observers have difficulty in a range where the available structure is not
informative. It could be argued that this range is beyond the need for fine-grained prospective
control (e.g., Shaw et al., 1991). Alternatively, one could take the perspective that the acoustical
structure in question, while not appropriate for the timing of interceptive behavior, is appropriate
for orienting the perceiver toward the approaching object (Guski, 1992). When appropriately
oriented, the optical can then be adaptively exploited. The preceding orienting hypothesis
highlights that audition and vision, like all perceptual systems, are typically used together. In the
setting under discussion, vision and audition coordinate in localizing an object and determining
time to arrival (Guski, 1992).
The preceding results suggest that when contact is less than 4 sec away, acoustic structure
is equivalent to optical structure in specifying time to contact. A finer scaling might be provided,
however, by an auditory analog of a standard visual experiment that asks observers to decide
which of two approaching objects would hit them first. When the objects are computer
simulations of approaching squares, viewers achieve a 75% success criterion as long as the
arrival times of the two objects differ by at least 50 msec (Todd, 1981). When the simulations are
35
of two sounds approaching a listener, one from the left and one from the right, preliminary data
indicate that the 75% criterion is achieved as long as the arrival time differs by at least 300 msec
(Wightman & Jenison, 1995).
The kinematic analyses by Wightman and Jenison (1995) show less dramatic changes for a
transverse trajectory compared to oblique trajectories. We might expect listeners to have even
more difficulties in such situations. The contribution of intensity, Doppler, and interaural
patterns to perceiving pass-by on a transverse trajectory has been examined experimentally with
a simulated European ambulance siren presented over earphones (Rosenblum, Carello, &
Pastore, 1987). These variables change coherently in naturally occurring events. In simulations,
however, they can be provided singly or jointly and, if the latter, they can be made consistent or
be put in competition. Listeners were to indicate, by pressing a key, when the simulated siren
passed right in front of them. Each type of change on its own was sufficient to support perception
of the moment of pass-by but listeners were most accurate when all three types of change were
available and consistent. When placed in competition, that is, when the three variables simulated
different times of passage, listeners were biased in favor of amplitude change. Doppler shift
fared least well in competition, perhaps because this variable is least generalit is not detectable
for sounds that lack pitch change (clicks, squeaks, and slow velocities).
Unlike the rectilinear approaches, the transverse events included the moment of pass-by.
The importance of this structure has been assessed directly (Rosenblum, Wuestefeld, & Saldaa,
1993). Recordings of an approaching car at two speeds, 15 mph and 25 mph, were divided into
thirds and combinations of these thirds were presented to six listeners who were provided with
practice trials but no feedback. Judgments of time to arrival were as accurate when the acoustic
signal contained the actual time of passage as when it did not, showing the importance of
36
acoustic structure before the moment of arrival. However, systematic occlusion or removal of
portions of the acoustic signal interferes with listeners ability to judge time to arrival
(Rosenblum, 1993). Thus, if the time-varying aspect of the acoustic signal provides information
about time to arrival, performance declines as this information is degraded or impoverished
(Wuestefled & Rosenblum, 1996).
A direct evaluation of one of the derived quantities, the intensity-based for pass-by,
found mixed results (Erwin, 1995). Simulated pass-by events used intensity modulated sine
waves (1000 Hz) to vary speed while keeping offset distance from the listener constant, or to
vary offset distance while keeping speed constant. Events were truncated before the object
passed in front of the listener who pressed a key to indicate when that pass-by would have
occurred. The constant error of those judgments varied as a function of the moment of inflection
of the -functions but with opposite signs for the two manipulations.
What do these results tell us about the information for imminent contact? In many
treatments of similar phenomena, each of the acoustic variables and portions of the signal might
be labeled a source of information (e.g., Cutting, 1986) and the event would be said to be
multiply specified. To our way of thinking, however, there is only one source and that is the
moving, sound-emitting object. Amplitude changes, Doppler shift and interaural differences are
simply variables that accompany that event. They probably combine in a non-linear fashion to
form a higher order variable (cf. Michaels & Carello, 1981) specific to time to arrival.
Controlling collisions is actually a rather general phenomenon (Kugler, Turvey, Carello, &
Shaw, 1985). An obvious example includes braking before crashing. Less obvious examples
include muscle activation preparatory to landing after jumping from a height (Sidaway, McNitt-
Gray, & Davis, 1989) or making postural adjustments to stay with the region of reversibility for
37
balance (Carello, Turvey, & Kugler, 1985; Riccio, 1993). Guiding the filling of a vessel to the
brim with liquid also qualifies as a controlled collision and this one has been examined
acoustically (Cabe & Pittenger, 2000). For a closed cylindrical tube, the fundamental resonant
frequency is influenced by the height of the air column in the tube, the radius of the tube, and
the speed of sound in air. As the tube is filled, the column of air shortens and increases but as
the tube is drained, the column of air lengthens and decreases. If liquid is released at the same
rate it is introduced, should not change. Acoustic structure, therefore, distinguishes these three
events. Monophonic tape recordings of 3 filling events, 3 emptying events, and 3 maintenance
events were created with water flowing from a spigot into a plastic tube (30 cm long with a
radius of 5 cm). Randomized blocks of the nine events were presented nine times to nine
listeners over a loudspeaker without practice or feedback. Given three possible events, chance
performance was 33%. Listeners exceeded this in all cases: 67% correct for filling, 87% for
emptying, and 67% for maintaining.
Subsequent experiments had more of the flavor of visual experiments in that the listeners
actively controlled the flow of water through the spigot and arrested the stream to effect one of
two outcomes: full to the brim or to preferred drinking level. Their level of accuracy was
established with full informationholding a 17.5 cm long tube while watching and hearing it
fill. For audition alone, their eyes were closed and they did not hold the tube. Ten listeners in
each information condition were instructed to turn the water on and off only once during each of
30 trials. The tubes were filled to 96% capacity under full information and to 88% for auditory
information alone. For preferred drinking level, they were filled to 86% and 70%, respectively.
Although the particular levels achieved under auditory control differed from the particular levels
achieved under full information control, auditory control nonetheless allowed listeners to
38
distinguish the two levels of fullness.
More variation in the acoustic signal was introduced by the use of three different vessel
volumes and two different flow rates (with 10 trials of each combination). Ten blindfolded
listeners were instructed to fill each vessel to the brim with one opening and closing of the
spigot. They were not told of the variations in vessel size or flow rate. Although there was a
tendency to underfill vessels (at the fast rate, small vessels were filled to 95% capacity, medium
to 86%, and large to 77%; at the fast rate, small vessels were filled to 94% capacity, medium to
85%, and large to 74%), generally the larger the vessel, the higher the fill level. Moreover, ten
blind listeners who performed the same task produced the same pattern of results. Finally, a
counterpart to the truncated approach and pass-by events described earlier (i.e., indicate when the
object would have hit you had it continued) required listeners to anticipate the end point of the
filling event in the absence of continuing acoustic structure. Nine listeners heard filling events
controlled by the Experimenter. Three flow rates were used to fill one vessel to 1/4, 1/2, or 3/4
full. Once the spigot was closed, listeners waited to respond until they thought the vessel would
have been full had that flow rate continued. Estimated time to fill tracked actual time to fill very
well (with 80-90% of the variance being accounted for).
The problem of synthesis
The Warren and Verbrugge (1984) study provides a model for doing ecological acoustics
research: Establish the ability, identify candidate sound variables, and produce synthetic events
on this basis. The process of synthesis is a tricky one, however. It can be motivated by the
structure of natural events or it can be motivated by the idealized structure expected from the
physical equations for the events. One problem with the latter is that the events are simple and
39
idealized, with putatively messy structure eliminated, quite likely at the cost of informativeness.
Our survey of experimental investigations of auditory source perception strongly suggests that
listeners are good when they have a lot of acoustical structure, even if that acoustical structure is
not readily quantified by the scientists. In contrast, it seems to be the case that when events are
constrained so that the scientist has a better understanding of the sound structure, perceiving the
source becomes harder for the listener.
The issues are readily illustrated by the methods and arguments presented in an
investigation of auditory discrimination of the material composition of struck, clamped bars
(Lutfi & Oh, 1997). The sounds were synthesized according to principles of theoretical acoustics,
with material composition as the only difference between bars. Each material was uniquely
identified by nominal values of frequency, amplitude, and decay (which were chosen to be in a
range that typically allowed discrimination performance in the range of 70-85% correct). The
sounds were presented over headphones in a sound attenuation chamber. Their six listeners were
musically trained, with extensive practice and feedback in each condition (e.g., they were told
that the sounds would differ in pitch, loudness, and decay). The task was a 2 AFC: which tone
was the iron bar (where the alternatives were silver, steel, or copper) or which tone was he glass
bar (where the alternatives were crystal, quartz, or aluminum). There were 4000 trials per listener
for the iron bar as target and 4000 trials per listener with the glass bar as target. The basic result
was that all listeners depended excessively on frequency, with a reduction in performance
efficiency that sometimes approached 80%.
The investigators characterized their setting as a best case scenario (p. 3647)they had
made the task for the listener as straightforward as possibleyet performance was less than
ideal. Consequently, they suggested that the optimistic view that sources are perceived on the
40
basis of available information (Gibson, 1966/1983; Fowler, 1990) was not supported because
listeners did not optimize decisions based on appropriate combinations of frequency, amplitude,
and decay (p. 3655). Instead, listeners seemed to make inferences based on only a single cue
(cf. Wildes & Richards, 1988). While acknowledging the constraints on their experimental
setting, these investigators thought that inexperienced listeners and multiply varying sound
sources ought to make the task even harder. As already discussed, this expectation is belied by
recent findings: Listeners exposed to actual impact events that included variation of shape were
nearly perfect in the identification of material (Kunkler-Peck & Turvey, 2000). Far from causing
listeners difficulties, the added natural variation seems to have made the relevant time-varying
structure more readily apparent.
Synthesizing events is not an inherently bad thing for ecologically oriented psychologists.
The issue is simply one of what is being synthesized. Gibson himself put the issue in terms of
synthesizing information (see also Gaver, 1993b). Work on dynamic touch (that based in the
states of muscles and tendons) provides a model. The description of an object with respect to its
inertia tensor is a description that is relevant to the role of touch in controlling movements. This
particular role can be understood as the proper function (cf. Millikan, 1984, 1993) of dynamic
touch. This is the function that dynamic touch has served throughout evolution that has
encouraged its survival. Dynamic touch is not so much a property-analyzing system as it is a
system for controlling movements of limbs and hand-held objects. When scientists manipulate
the inertial characteristics of an experimental object, they are synthesizing information about its
movableness not about it mass or length or diameter.
The parallel for audition seems more likely to be found in sound effects on old-time radio
(e.g., alternately slapping coconut halves to simulate a galloping horse), which are really more in
41
the spirit of synthesizing information than the technically more sophisticated simulations
permitted by theoretical acoustics. The success of auditory metaphors in cinema, exploited in the
musical scoring that accompanies lurking madmen or sea battles or madcap exploits, points to
the abstractness of the information that is unlikely to be found in nominal values of physical
parameters of frequency, amplitude, and decay.

42
References
Anderson, K., Carello, C., & Kunkler-Peck, A. J. (May, 1996). Rotational inertia constrains
perception of object length by sound. Poster presented at the North American Society for
Ecological Psychology. Hartford, CT.
Bingham, G. P. (1993). Perceiving the size of trees: Forms as information about scale. Journal of
Experimental Psychology: Human Perception & Performance, 19, 1139-1161.
Bingham, G. P., Rosenblum, L. D., & Schmidt, R. C. (1995). Dynamics and the orientation of
kinematic forms in visual event recognition. Journal of Experimental Psychology: Human

Perception and Performance, 21, 1473-1493.
Bootsma, R. J., & Oudejans, R. R. D. (1993). Visual information about time-to-collision between
two objects. Journal of Experimental Psychology: Human Perception and Performance, 19,
1041-1052.
Cabe, P., & Pittenger, J. B. (2000). Human sensitivity to acoustic information from vessel filling.
Journal of Experimental Psychology: Human Perception and Performance, 26, 313-324.
Carello, C., Anderson, K. L., & Peck, A. (1998). Perception of object length by sound.
Psychological Science, 9, 211-214.
Carello, C., Grosofsky, A., Reichel, F. D., Solomon, H. Y., & Turvey, M. T. (1989). Visually
perceiving what is reachable. Ecological Psychology, 1, 27-54.

Carello, C., Turvey, M. T., & Kugler, P.N. (1985). The informational support for upright stance.
Behavioral and Brain Sciences, 8, 151-152.
Chomsky, C. (1986). Analytic study of the Tadoma method: Language abilities of three deaf-
blind subjects. Journal of Speech and Hearing Research, 29, 332-347.
Clifton, R., Perris, E., & Bullinger, A. (1991). Infants perception of auditory space.
Developmental Psychology, 27, 187-197.
Cutting, J. E. (1986). Perception with an eye for motion. Cambridge, MA: MIT Press.
Diehl, R., & Walsh, M. (1989). An auditory basis for the stimulus-length effect in the perception
43
of stops and glides. Journal of the Acoustical Society of America, 85, 2154-2164.
Erwin, T. (1995). Prospective acoustic information for object pass by. In B. G. Bardy, R. J.
Bootsma, & Y. Guiard (Eds.), Studies in perception and action III (pp. 325-328). Mahwah,
NJ: Erlbaum.
Fitzpatrick, P., Carello, C., Schmidt, R. C., & Corey, D. (1994). Haptic and visual perception of
an affordance for upright posture. Ecological Psychology, 6, 265-287.
Fitzpatrick, P., Carello, C., & Turvey, M. T. (1994). Eigenvalues of the inertia tensor and
exteroception by the muscular sense. Neuroscience, 60, 551-568.

Fletcher, N. H., & Rossing, T. D. (1991). The physics of musical instruments. New York:
Springer-Verlag.
Fowler, C. A. (1986). An event approach to the study of speech perception from a direct-realist
perspective. Journal of Phonetics, 14, 3-28.
Fowler, C. A. (1990). Sound-producing sources as object of perception: Rate normalization and
nonspeech perception. Journal of the Acoustical Society of America, 88, 1236-1249.
Fowler, C. A. (1991). Auditory perception is not special: We see the world, we feel the world,
we hear the world. Journal of the Acoustical Society of America, 89, 2910-2915.
Fowler, C. A., & Dekle, D. J. (1991). Listening with eye and hand: Cross-modal contributions to
speech production. Journal of Experimental Psychology: Human Perception and

Performance, 17, 816-828.
Freed, D. J. (1990). Auditory correlates of perceived mallet hardness for a set of recorded
percussive events. Journal of the Acoustical Society of America, 87, 311-322.
Friberg, A., & Sundberg, J. (1999). Does music performance allude to locomotion? A model of
final ritardandi derived from measurements of stopping runners. Journal of the Acoustical
Society of America, 105, 1469-1484.
Gaver, W. W. (1988). Everyday listening and auditory icons. Unpublished doctoral dissertation,
University of California, San Diego.

44
Gaver, W. W. (1993a). What in the world do we hear? An ecological approach to auditory event
perception. Ecological Psychology, 5, 1-29.
Gaver, W. W. (1993b). How do we hear in the world? Explorations in ecological acoustics.
Ecological Psychology, 5, 285-313.
Gibson, E. J. (1994). Has psychology a future? Psychological Science, 5, 69-76.
Gibson, J. J. (1983). The senses considered as perceptual systems. Prospect Heights, IL:
Wavelan. (Original work published in 1966)
Gibson, J. J. (1979). The ecological approach to visual perception. Boston: Houghton Mifflin.
Gordon, C., & Webb, D. (1996). You cant hear the shape of a drum. American Scientist, 84, 46-
55.
GrandPr, D., & Carello, C. (2001). Elasticity and length perception by dynamic touch. In G.
Burton & R. C. Schmidt (Eds.), Studies in perception and action, VI (pp. 21-24). Mahwah,
NJ: Erlbaum.
Guski, R. (1992). Acoustic tau: an easy analog to visual tau? Ecological Psychology, 4, 189-197.
Heft, H. (1993). A methodological note on overestimates of reaching distance: Distinguishing
between perceptual and analytic judgments. Ecological Psychology, 5, 255-271.
Jenkins, J. J. (1985). Acoustic information for places, objects, and events. In W. H. Warren & R.
E. Shaw (Eds.), Persistence and change: Proceedings of the first international conference on
event perception (pp. 115-138). Hillsdale, NJ: Erlbaum.
Johansson, G. (1973). Visual perception of biological motion and a model for its analysis.
Perception & Psychophysics, 14, 201-211
Kac, M. (1966). Can one hear the shape of a drum? Mathematics Monthly, 13, 21-23.
Keller, H. (1903). The story of my life. New York: Doubleday.
Kugler, P.N., Turvey, M. T., Carello, C., & Shaw, R. E. (1985). The physics of controlled
collisions: A reverie about locomotion. In W. H. Warren & R. E. Shaw (Eds.), Persistence
and Change: Proceedings of the First International Conference on Event Perception (pp.
45
195-222). Hillsdale, NJ: Erlbaum.
Kunkler-Peck, A., & Turvey, M. T. (2000). Hearing shape. Journal of Experimental Psychology:
Human Perception and Performance, 1, 279-294.
Lakatos, S., McAdams, S., & Causs, R. (1997). The representation of auditory source
characteristics: Simple geometric form. Perception & Psychophysics, 59, 1180-1190.
Lee, D. N. (1976). A theory of visual control of braking based on information about time-to-
collision. Perception, 5, 437-459.

Lee, D. N. (1980). Visuo-motor coordination in space-time. In G. Stelmach, & J. Requin (Eds.),
Tutorials in motor behavior (pp. 281-295). Amsterdam: North Holland Press.
Lee, D. N. (1990). Getting around with light or sound. In R. Warren & A. Wertheim, (eds.).
Perception and control of self-motion (pp. 487-505). Hillsdale, NJ: Erlbaum.
Li, X., Logan, R. J., & Pastore, R. E. (1991). Perception of acoustic source characteristics:
Walking sounds. Journal of the Acoustical Society of America, 90, 3036-3049.
Liberman, A. M. (1982). On finding that speech is special. American Psychologist, 37, 148-167.
Liberman, A. M., Cooper, F., Shankweiler, D., & Studdert-Kennedy, M. (1967). Perception of
the speech code. Psychological Review, 74, 431-461.
Liberman, A. M., Ingemann, F., Lisker, L., Delattre, P., & Cooper, F. (1959). Minimal rules for
synthesizing speech. Journal of the Acoustical Society of America, 31, 1490-1499.

Liberman, A. M., & Mattingly, I. (1985). The motor theory revised. Cognition, 21, 1-36.
Liberman, A. M., & Mattingly, I. (1989). A specialization for speech perception. Science, 243,
489-494.
Lichte, W. H. (1941). Attributes of complex tones. Journal of Experimental Psychology, 28, 455-
480.
Lutfi, R. A., & Oh, E. (1997). Auditory discrimination of material changes in a struck-clamped
bar. Journal of the Acoustical Society of America, 102, 3647-3656.
MacDonald, J. W., & McGurk, H. (1978). Visual influences on speech perception processes.
46
Perception & Psychophysics, 24, 253-257.
McGurk, H. & MacDonald, J. W. (1976). Hearing lips and seeing voices. Nature, 264, 746-748.
Massaro, D. (1987). Speech perception by ear and eye: A paradigm for psychological inquiry.
Hillsdale, NJ: Erlbaum.
Massaro, D. (1998). Perceiving talking faces. Cambridge, MA: MIT Press.
Michaels, C. F. (1986). An ecological analysis of binocular vision. Psychological Research, 48,
1-22.
Michaels, C. F., & Carello, C. (1981). Direct perception. New York: Prentice Hall.
Michaels, C. F., Prindle, S., & Turvey, M. T. (1985). A note on the natural basis of action
categories: The catching distance of mantids. Journal of Motor Behavior, 17, 255-264.
Miller, J., & Liberman, A. M. (1979). Some effects of later-occurring information on the
perception of stop consonant and semivowel. Perception & Psychophysics, 25, 457-465.
Millikan, R. (1984). Language, thought and other biological categories. Cambridge, MA: MIT
Press.
Millikan, R. (1993). White queen psychology and other essays for Alice. Cambridge, MA: MIT
Press.
Norman, J. F., Todd, J. T., Perotti, V. J., & Tittle, J. S., (1996). The visual perception of three-
dimensional length. Journal of Experimental Psychology: Human Perception &

Perris, E., & Clifton, R. K. (1988). Reaching in the dark toward sound as a measure of auditory
localization in infants. Infant Behavior and Development,11, 473-491.
Prinz, W. (1997). Perception-action planning. European Journal of Cognitive Psychology, 9,
129-154.
Repp, B. H. (1987). The sound of two hands clapping: An exploratory study. Journal of the
Acoustical Society of America, 81, 1100-1110.
Riccio, G. E. (1993). Information in movement variability about the qualitative dynamics of

47
posture and orientation. In K. M. Newell & D. M. Corcos (Eds.) Variability and motor
control (pp. 317-358). Champaign, IL: Human Kinetics.
Rosenblum, L. D. (1993). Acoustical information for controlled collisions. In A. Schick (Ed.),
Contributions to psychological acoustics (pp. 303-322). Oldenberg, Germany: Bibliotheks.
Rosenblum, L. D., & Saldaa, H. (1996). An audiovisual test of kinematic primitives for visual
speech perception. Journal of Experimental Psychology: Human Perception and

Rosenblum, L. D., & Saldaa, H. (1998). Time-varying information for visual speech perception.
In R. Campell, B. Dodd, & D. Burnham (Eds.). Hearing by eye II: Advances in the
psychology of speech reading and auditory-visual speech (pp. 61-81). Hillsdale, NJ:
Erlbaum.
Rosenblum, L. D., Carello, C., & Pastore, R. E. (1987). Relative effectiveness of three stimulus
variables for locating a moving sound source. Perception, 16, 175-186.
Rosenblum, L. D., Johnson, J., & Saldaa, H. (1996). Point-light facial displays enhance
comprehension of speech in noise. Journal of Speech and Hearing Research, 29, 1159-1170.
Rosenblum, L. D., Wuestefeld, A., & Anderson, K. (1996). Auditory reachability: an affordance
approach to the perception of distance. Ecological Psychology, 8, 1-24.
Rosenblum, L. D., Wuestefeld, A., & Saldaa, H. (1993). Auditory looming perception:
influences on anticipator judgments. Perception, 22, 1467-1482.

Rossing, T. D., & Fletcher, N. H. (1995). Principles of vibrations and sound. New York:
Springer-Verlag.
Russell, M. & Turvey, M. T. (1999). Auditory perception of unimpeded passage. Ecological
Psychology, 11, 175-188.
Schiff, W., & Oldak, R. (1990). Accuracy of judging time to arrival: Effects of modality,
trajectory, and gender. Journal of Experimental Psychology: Human Perception and

48
Schubert, E. D. (1975). The role of auditory perception in language processing. In M. Rawson &
D. D. Duane (Eds.), Language, perception and reading (pp. 97-130). Baltimore, MD: York.
Shaw, B. K., McGowan, R., & Turvey, M. T. (1991). An acoustic variable specifying time to
contact. Ecological Psychology, 3, 253-261.
Shockley, K., Grocki, M., Carello, C., & Turvey, M. T. (in press). Heaviness perception
implicates an ecological form of the rigid body laws. Experimental Brain Research.
Sidaway, B., McNitt-Gray, J., & Davis, G. (1989). Visual timing of muscle preactivation in
preparation for landing. Ecological Psychology, 1, 253-264.

Snyder, W. S., Cook, M. J., Nasset, E. S., Karhausen, L. R., Howells, G. P., & Tipton, I. H.
(1974). Report of the task group on reference man. Oxford, England: Pergamon
Stoffregen, T. A., & Pittenger, J. B. (1995). Human echolocation as a basic form of perception
and action. Ecological Psychology, 7, 181-216.
Studdert-Kennedy, M. (1985). Perceiving phonetic events. In R. Shaw & W. H. Warren (Eds.),
Persistence and change: Proceedings of the first international conference on event
perception (pp. 139-156). Hillsdale, NJ: Erlbaum.
Sumby, W. H, & Pollack, I. (1954). Visual contribution to speech intelligibility in noise. Journal
of the Acoustical Society of America, 26, 212-215.
Summerfield, A. Q. (1987). Some preliminaries to a comprehensive account of audiovisual
speech perception. In B. Dodd & R. Campbell (Eds.), Hearing by eye: The psychology of
lip-reading (pp. 3-51). London: Erlbaum.
Todd, J. T. (1981). Visual information about moving objects. Journal of Experimental
Psychology: Human Perception and Performance, 7, 795-810.
Tousman, S. A., & Pastore, R. E. (1989). Source characteristics: A study of hand clapping.
Journal of the Acoustical Society of America Supplement 1, 85, S53.
Tresilian, J. R. (1993). Four questions of time to contact: A critical examination of research on
interceptive timing. Perception, 22, 653-680.

49
Turvey, M. T. (1992). Affordances and prospective control: An outline of the ontology.
Ecological Psychology, 4, 173-187.
Turvey, M. T. (1996). Dynamic touch. American Psychologist, 51, 1134-1152.
Turvey, M. T., & Carello, C. (1995). Dynamic touch. In W. Epstein & S. Rogers (Eds.),
Handbook of perception and cognition, Vol. 5. Perception of space and motion (pp. 401-
490). New York: Academic Press.
Turvey, M. T., Shockley, K., & Carello, C. (1999). Affordance, proper function, and the physical
basis of perceived heaviness. Cognition, 17, B17-B26.

Venderveer, N. J. (1979). Ecological acoustics: Human perception of environmental sounds.
Dissertation Abstracts International. 40/09B, 4543. (University Microfilms No. 8004002).
Warren, W. H., & Verbrugge, R. R. (1984). Auditory perception of breaking and bouncing
events: A case study in ecological acoustics. Journal of Experimental Psychology: Human
Perception and Performance, 10, 704-712.
Warren, W. H., & Whang, S. (1987). Visual guidance of walking through apertures: Body-scaled
information for affordances. Journal of Experimental Psychology: Human Perception and
Wightman, F. L., & Jenison, R. (1995). Auditory spatial layout. In W. Epstein & S. J. Rogers
(Eds.) Perception of space and motion (pp. 365-400). San Diego: Academic Press.
Wildes, R., & Richards, W. (1988). Recovering material properties from sound. In W. Richards
(Ed.), Natural computation (pp. 356-363). Cambridge, MA: MIT Press.
Wuestefeld, A., & Rosenblum, L. (1996). Perceptual learning in an auditory time to arrival task.
Technical Report UCR.96.011. University of California, Riverside.

50
Authors Notes
Preparation of this manuscript was supported by grants from the National Science Foundation
(SBR 97-09678 awarded to M. T. Turvey and Claudia Carello) and the National Institutes of
Health (HD 01994) awarded to Haskins Laboratories.

Acoustic Specification of Object Properties

Uploaded by

Acoustic Specification of Object Properties

Uploaded by

Ecological acoustics

Acoustic Specification of Object Properties

Claudia Carello,1,2 Jeffrey B. Wagman,1 and M. T. Turvey1,2

A general statement of the ecological approach to perception is that a particular animal-

is reliable, surprisingly accurate, and constrained by dynamical law.

Acoustic Specification of Object Properties

Orthodoxy, of course, is likely to inspire heterodoxy. For perception, a contrary treatment

permit auditory perception of geometric properties. Finally, we summarize research that

documents listeners success.

The ecological approach to perception

The ecological approach to perception is a metatheory. As such, it describes a particular

conceptualization of how we can be aware of our surroundings. In general, metatheories endorse

its source directly, without elaboration by internal mechanisms.

event structured energy distribution.

The inclusion of something as coarse-grained as an event highlights the ecological commitment

to discovering the appropriate level of description of an animal-environment setting. The object

indifferent to the medium.2

Protocol studies and the possibilities of perceiving on the basis of sound

It is not uncommon for experimenters in a traditional psychoacoustics experiment to ask

it means for them (Gaver, 1993a; Schubert, 1975).

scratching), pointing at the relevance of higher-order structure.3 Moreover, even without an

that systematic physical analyses are possible and are needed.

type of interaction, whether it is scraping or splashing or slapping, affects the time-varying

object? What is it doing? Where is it doing it?

to be labeled ecological acoustics was an examination of listeners ability to distinguish breaking

Figure 2. (left) A bounce produces a single, damped, quasi-periodic pulse train

appropriate amounts of silence between subsequent bounces so as to preserve the single,

breaking with 87% accuracy (whether augmented by a noise burst or not).

The foregoing is an exemplary study in ecological acoustics. Although such a pedigree

traditionally oriented psychoacousticians have contributed to this literature. Two studies of

different hand configurations (palm-to-palm, finger-to-palm, or something in between), a new

sample of sound-producers may be idiosyncratic, either in physical dimensions (height and

without indicating their appreciation of relative differences.

Perceiving mechanical events on the basis of sound

investigation focused on percussive sound eventsa single impact of a mallet on a hollow

identification of acoustic parameters that would be useful as predictors of timbre. A combination

of the spectrum) represents the darkening over time.5

An inevitable implication of source-oriented auditory perception is that information about

Still audible, for example, was the balls revolution speed.

available in acoustic structure.

rank-ordering, can be assessed. That is to say, perceived increments in a dimension of interest

recent investigations of shape and size perception on the basis of sound.

A first step toward metrical precision in perceiving mechanical events

depends on how an object is struck. This so-called exciter-resonator relationship is often

Figure 4. A metal bar tends to vibrate (a) transversely, (b) longitudinally,

succession, accompanied by two response alternatives presented visually. These alternatives

tend to vary along three perpendicular axes.

reverberation, loudness is equalized, only clear recordings are usedshould we be impressed

Metrical precision in perceiving struck plates

illustrating the challenge to be faced by the then-new science of spectroscopy:

by two manifolds that differ geometricallyare to blame. But, of course, mathematicians

there were no demonstrations or practice trials. Perceiving material was absolutely

rectangles. For the squares, those reports were the same.

91.6 cm) with a mean reliability of 5.5%.

solutions to that equation are given by

mn = 0.453 h E m+12 + n+12 (1)

do the same thing functionally.

Metrical precision in perceiving the lengths of dropped rods

Perceived Length (cm)

accounts for only 18% of the variance in perceived length.

manipulating a higher-order property, an objects mass distribution. This property is quantified

experiments. Manipulations of material provide an interesting case, however. Material density

displaced by an external forcing function (such as accompanies dropping to a surface). Indeed,

As a matter of pedagogical convenience, we have been characterizing the achievements of

ecological approach. But it is incomplete. Perceptions raison dtre is guiding activity.

Affordances: Perceiving behavioral possibilities

properties such as length in terms of activity-relevant properties such as reach-with-able is a

pertains to a particular affordance. What pattern of structure in an ambient energy array is

were the same when scaled intrinsically (Figure 7c).

Attempts to characterize the informational support for auditory perception of reachability

humans (Michaels, 1986).

Just as questions of perceiving what is reachable ecologize the problem of distance

& Turvey, 1999; cf. Fitzpatrick, Carello, Schmidt,

60 times the listeners shoulder width which

the wall with that shoulder during ordinary

apertures that were judged not passable all of the time.