Discuss the mechanisms involved in sound localisation

March 2003: How does the auditory system identify the frequency composition of
sounds?

March 2005: Discuss the mechanisms involved in sound localisation.

June 2005: Discuss the structure and function of the cochlea.

March 2006: How does the cochlea respond to different frequencies of sound?

June 2006: Discuss the mechanisms that enable sound to be located in three dimensional
space.
 Discuss the mechanisms involved in sound localisation

In 1969, Blauert showed that the listener can be tricked into thinking the sound is
coming from in front when it is actually coming from behind, and vice versa, by
electrically modifying the sound. This deception is possible because sound localisation
relies on acoustical cues and the auditory system cannot distinguish whether such cues
are produced by the transition of sounds from the pinna to the tympanic membrane, or are
integral features of the sound. Although monaural cues contribute to sound localisation,
as shown by Belendiuk & Butler (1975), the most important contributions are from
binaural cues. When sound originates from one side of the head, sound waves will arrive
at the near ear before the other resulting in an interaural time arrival difference (ITD) and
producing an interaural phase difference. These acoustical cues are particularly
significant for low frequency sounds and the medial superior olive (MSO) which
processes this acoustical cue is particularly sensitive to low frequency sounds. Another
acoustical cue for horizontal sound localisation is interaural sound level difference (ILD)
because sound amplitude at the far ear will be attenuated by the acoustical shadowing of
the head. ILD is particularly significant for high frequency sounds and the cells in the
lateral superior olive (LSO) which process ILD are principally sensitive to these same
high frequency sounds. Therefore, horizontal localisation of low frequency sounds
depend on ITD and that of high frequency sounds depend on ILD. This is referred to as
the “duplex theory” which was first proposed by Rayleigh in 1907. However, sounds
originating from points in the median saggital plane do not cause ILD or ITD because the
source is equidistant from both ears and thus vertical sound localisation must rely on
some other acoustical cues. In 1968, Roffler & Butler showed that occlusions of the
pinnae prevented vertical sound localisation. It is now understood that sound reflections
in the pinnae produce ‘colouration’ of the sound spectrum which is related to the
elevation of sound source and thus provide the cue for vertical sound localisation.
Electrophysiological studies of neurons in the central nervous system, particularly those
in the superior olivary complex, have led to an understanding of how the complex neural
circuits process acoustical cues to localise sound sources.

In 1948, Jeffress proposed the “coincidence model” as the mechanism by which

cells in the MSO process ITD. The MSO is innervated by neurons from the anteroventral
cochlear nuclei via the trapezoid body. These neurons have a short latency and so
accurately transmit the timing of auditory stimuli. Physiologists discovered that the MSO
contains ‘coincidence’ neurons which produce a maximal response when nervous
impulses from both cochlear nuclei arrive simultaneously. However, the axons
innervating these ‘coincidence’ neurons have variable lengths so that there is a
characteristic delay between the arrivals of action potentials from the anteroventral
cochlear nuclei. Therefore ‘coincidence’ neurons fire maximally when the ITD of
auditory stimuli matches the characteristic delay of these neurons.

Diagram to show coincidence model p.493 Kandel & Swartz

 ITD can be artificially simulated in listeners with headphones by presenting
sounds to one ear before the other. The greater the delay between the two sounds, the
more the sound source is perceived as coming from the side of the ear which receives the
sound first.

Like the MSO, neurons in the LSO receive auditory information from both ears
about a particular frequency because the afferent neurons are tonotopically distributed. In
contrast however, neurons in the LSO receive excitatory contralateral input and inhibitory
ipsilateral input, and are sensitive to ILD instead of ITD.

Graph to show neurons detecting ILD.p.494 kandel and swartz

The rate of firing is related to the value of ILD and so processing by the LSO
contributes to horizontal sound localisation. It is possible to trick the brain into thinking
that sound is originating from one side of the head by presenting sounds with higher
amplitude to one ear than the other ear.

The external ear also plays an important role in sound localisation, particularly in
the median saggital axis where ITD and ILD does not occur. Its contribution was
demonstrated by Roffler & Butler who showed that occlusion of pinna chambers, thus
preventing sound reflections in the pinna, has a detrimental effect on vertical sound
localisation. It is now understood that ‘colouration’ of the sound spectrum, amplification
and attenuation of particular frequencies, by the resonance chambers in the pinna relates
to the elevation of the sound source and provides the cue for vertical sound localisation.
However, the auditory system is unable to distinguish whether the spectral cues are a
result of ‘colouration’ by the pinna or whether they’re features of the original sound. This
was clearly demonstrated by Blauert who recorded the sounds generated in the ear canal
after presenting sounds to the listener from in front and from behind. He then modified
the sounds so that they produce sound spectra associated with sounds originating from
the opposite source. When these modified sounds were presented to the listener, sounds
originating from in front were perceived to be coming from behind and vice versa.

Diagram of ascending auditory pathways

All the information resulting from the processing of acoustical cues is relayed to
the primary auditory cortex which further contributes to sound localisation. Neurons from
the cochlear nuclei and the superior olivary nuclei project to the inferior colliculi on both
sides via the lateral lemniscus. Note that binaual interactions occur in the superior olivary
nucleus and between the nuclei of the lateral lemniscus, through the Probst’s commisure,
which explains why unilateral lesions of the central auditory pathway do not cause
monaural hearing or affect sound localisation.Neurons in the inferior colliculus, arranged
tonotopically, relay auditory information to the ipsilateral medial geniculate body of the
thalamus. These neurons then project to the ipsilateral primary auditory cortex in the
superior temporal gyrus (Brodmann’s areas 41 and 42). The primary auditory cortex is
functionally organized into columns which relates to spatial maps of sound localization in
the cortex.
How to conclude? What about front-back sound localisation?
 First rough draft
Sound waves are subtly modified when they interact with the head and pinna
which provides acoustical cues to the auditory system for sound localisation.

When a sound source is located to one side of the head, the sound waves will
reach the nearer ear first resulting in an interaural difference in time of arrival (ITD) and
producing an interaural phase difference (IPD). The sound will also be attenuated at the
far ear due to the acoustical shadowing by the head resulting in interaural sound level
difference (ILD). These acoustical cues are processed by the superior olivary complex,
where inputs from both cochlear nuclei converge, for sound localisation in the horizontal
axis.

Fig. 6.8 on p.170 in Physiology of Hearing has good diagram of ascending auditory
pathwaysof brain stem.

The trapezoid body relays the cochlear input from anteroventral cochlear nucleus
to the medial superior olive (MSO) which contribute to sound localization by processing
ITD. In 1948, Jeffress proposed the coincidence model to explain how this process
functions. Binaural neurons in MSO fire maximally when action potentials from two
sources arrive simultaneously. However, the axons innervating these neurons have
variable lengths so that each has a characteristic delay thus different neurons respond to
different ITD and the MSO is able to localise sounds from different directions. In contrast
to the MSO, neurons in LSO detect ILD instead because they receive excitatory
contralateral input and inhibitory ipsilateral input. Thus these neurons do not respond
when auditory inputs from both sides are equal but they increase in firing if sound
amplitude is greater in one ear than the other. The discharge rate of these neurons relates
to ILD which allows mapping of the direction of sound sources. Are ILD and ITD the
only acoustical cues for horizontal sound localisation? Although binaural cues are the
most important, studies by Butler et al have shown that monaural cues contribute to
sound localisation. Belendiuk & Butler (1975) showed that for sounds with energy above
4kHz, subjects with one occluded ear can still reasonably localise sounds from one of
five loudspeakers situated at 15o intervals on a horizontal plane in the quadrant between
the unoccluded ear and straight ahead.

The superior olivary nuclei and the cochlear nuclei project via lateral lemniscus
on both sides to the inferior colliculi where the cells are arranged tonotopically. Some
fibres in the lateral lemniscus cross via the Probst’s commissure at the level of lateral
lemniscus nuclei. The transmission of auditory input bilaterally explains why unilateral
lesions of the auditory system do not cause monaural hearing and loss of sound
localisation. Neurons in the inferior colliculi relay auditory information to the ipsilateral
medial geniculate body of the thalamus which project to the ipsilateral primary auditory
cortex in the superior temporal gyrus (Brodmann’s areas 41 and 42). The primary
auditory cortex contains several tonotopic maps and is organised into alternating
summation and suppression columns. It is believed that this columnar structure is related
to spatial maps of sound localisation.
 However, how does the auditory system localise sounds in the median saggital
plane which does not produce ILD or ITD? In 1967, Batteau proposed that the pinna
provide acoustical cues for sound localisation in the vertical axis. This idea was
supported by Roffler & Butler in 1968 who showed that vertical sound localisation is
disrupted when the convolutions of the pinnae are occluded. Sound reflections in the
pinna attenuate or amplify different frequencies in the sound spectrum depending on the
location of the sound source in the vertical axis. The ‘colouration’ of the sound spectrum
by the pinna is the major cue for vertical sound localisation. This results in an interesting
situation because the auditory system would have no way of distinguishing between
spectral cues generated by sound reflections in the pinna and spectral features that are
present in the source spectrum. Indeed, experimental studies show that we can deceive
the auditory system into thinking the sound comes from a different location to its actual
source by modifying the source spectrum. In 1969, Blauert recorded sounds in the ear
canal when noise was presented from directly in front or behind the listener. He then
electrically modified the sound so that the spectrum from each source mimicked the
spectrum associated with the opposite source direction. Blauert found that the listener
localised sounds according to the spectral cues rather than its actual location.

As well as localising stationary sound sources, how well can the auditory system
detect moving sound sources? The current understanding as to how motion in sound
source is detected is described by the ‘snapshot theory’ where the sound source location
is measured at two different times and a change in the location is interpreted as
movement by the sound source.
 Notes
Paper: Sound localization by human listeners

Rayleigh reported (Rayleigh 1907) that for spatial information is derived at high
frequencies from ILDs and at low frequencies from IPDs -> referred to as “duplex”
theory of sound localization. When sound presented from one side, far ear would be
shadowed by head and interaural sound pressure level difference (ILD) would result.
Amount of shadowing depend on comparison between sound wavelength and dimension
of head. Middlebrooks et al 1989 showed that ILD can be as much as 35dB for high
frequencies. However, for frequencies below 1000Hz, ILD would be negligible. Rayleigh
reluctantly came to accept that the only alternative to the intensity theory is to suppose
that the judgment is founded upon the difference of phases at the two ears. Rayleigh
confirmed the sensitivity of listeners to IPDs by presenting to the two ears a pair of
tuning forks that were tuned to slightly different frequencies. The slightly mistuned forks
produced a steadily varying phase difference, which produced the sensation of an
auditory image that moved back and forth between the two ears. Rayleigh found that the
sensitivity to phases declined with increasing frequency, with an upper limit in his
experiments around 770 Hz -> duplex theory.

Vertical localization:
Batteau 1967, 1968, was one of the first to emphasize that the pinna could be a sourceof
spatial cues that might account for vertical localization. Supported by Fisher & Freedman
(1968). When one ear occluded, localization suffered only when a 10cm tube inserted in
the ear canal effectively bypassed the pinna and deprived the listeners of its sound
reflections.

Several lines of evidence indicate that the spectral shape cues are the major cues for
vertical localization. E.g. accurate vertical localization is ovserved only when the
stimulus has a broad bandwidth (Roffler & Butler 1968) and contains energy at high
frequencies. Vertical localization is prevented when the convolutions of the pinnae are
occluded (Roffler & Butler 1968). Vertical localization is almost as good when listening
with a single ear as with binaural listening (hebrank & Wright, 1974a; Oldfield and
Parker, 1986). Vertical localization is sensitive to manipulations of the source spectrum.

The apparent elevation of a sound source can be influenced by modifying the source
spectrum. In a widely quoted experiment, Blauert (1969/1970) recorded froma subject’s
ear canals while presenting a noise from two source locations, either directly in back or
directly in fron of the listener. He then electronically modified both sources, so that each
mimicked the spectrum associated with the opposite source position. The listener’s
judgements of the source’s locus were solely determined by the source spectrum and not
influenced by its actual location. Thus listener localizaed the source as behind if it had the
“back” spectrum, even if played from the front speaker.

Any model of localization based on spectral cues must acknowledge the importance of a
priori information regarding the source spectrum. That is, the presence of a particular
spectral feature at the tympanic membrane could be the result of the transfer
characteristic of the ear, or it could be a feature that was present in the source spectrum.
Experiments show that the auditory system cannot effectively distinguish the two.

Monaural localization:
Although most investigators agree that two ears are better than one in localizing sounds
in space, there is considerable evidence showing that some localization can be achieved
with a single ear. Evidence came from as ear as Angell & Fite (1901) and more recently,
Fisher & Freedman (1968).

Comment on controls in studies. High-quality earplug can produce 30-40 dB attenuation

at low frequencies and greater attenuation at higher frequencies. If sound level is greater
than 40 dB, interaural temporal comparison might be relatively unaffected. Thus critical
to conduct localization tests with source at low sensation level so occluded ear will be
completely non-functional.

Butler and colleagues emphasized that monaural cues contribute to localization in

azimuth. Belendiuk & Butler (1975) established when source contains energy above 4
kHz, subjects wearing plug in one ear can achieve better than chance localization of one
of five loudspeakers located in 15o intervals on a horizontal plane in the quadrant
between the unoccluded ear and straight ahead.

Distance perception:
Not very good, minimal research. Coleman (1963) best summary of potential distance
cues.

Motion detection:
Change in azimuth/elevation, not change in source distance which can be cued by a
change in sound pressure or by Doppler shift in stimulus frequency. No compelling
evidence for motion-senstive systems in auditory system. Problem is there’re two
interpretations for any observation of sensitivity to source motion. One interpretation is
that auditory system is sensitive to dynamic aspects of localization cues, e.g. changing
ILDs or IPDs, i.e. source velocity might be a ‘directly perceived attribute of moving
stimuli’ (Lappin et al 1975). Second interpretation is the ‘snapshot theory’ where nervous
system measures sound source location at two distinct times and interprets a change in
location as motion.

Dynamic cues for localization:

Sustained sounds are localized best when the head is turned to face the direction of
source.

Simulating external sources over headphones:

Wightman & Kistler 1989 a, b; Wightman et al 1987)
Wightman & Kistler compared sound localization in free field and under headphone
listening. First step, measure directional transfer functions for multiple sound source
locations around the listener. Give the catalog of directional transfer functions established
for each listener, Wightman & Kistler next compared the localization judgements given
by the same subject, in free-field and under headphone listening. ‘The data…from the
headphone condition are nearly identical to the data from the free-field condition. Thus
transfer functions contain all the information needed to simulate the localization of sound
sources in the free field.

Wightman & Kistler investigated whether this basic technique can be used to produce a
practical auditory display device which, by providing an acoustic signal under headphone
listening, maintains an apparently stable location even if the head is moved. The system,
being developed in collaboration with Wenzel and Foster, uses a head position indicator
(ISOTRAK) to determine the roll, pitch, and yaw angles of the listener’s head. The head
position information is used to determine which transfer function should be used in the
headphone presentation to maintain an apparently stable source location. Prototype built
and demonstrated at annual meeting (1988) of National Research Council’s Committee
on Bioacoustics and Biomechanics (CHABA).

Summary:
Two parallel processes:
Azimuth of source determined using differences in interaural time or interaural intensity.
Wightman and colleagues (1989) believe the low frequency temporal information is
dominant if both are present.
Elevation of source is determined from spectral shape cues. The received sound
spectrum, as modified by the pinna, is in effect compared with a stored set of directional
transfer functions. These are actually the spectra of a nearly flat source heard at various
elevations. The elevation that corresponds to the best-matching transfer function is
selected as the locus of the sound.

Head motion is not a critical part of localization process, only moderately more precise
when listener points directly toward source. No evidence to support view that auditory
motion perception is anything more than detection of changes in static location over time.

Physiology of Hearing:

Superior olive receives information from both cochelar nucleus thus plays a part in sound
localization.

Lateral superior olive:

Majority of binaural neurons were excited by ipsilateral stimuli and inhibited by
contralateral ones forming ‘EI’ neurons. Therefore imagine that LSO responds to
difference in intensity at the two ears. Use cue in sound localization. Intensity difference
largest at high frequencies where degree of diffraction around head will be small. In
accordance with this, most of LSO is devoted to high frequencies.
I don’t understand how EI neurons lead to detection of interaural sound pressure level
difference! See page 494 for fig. 32-14 in Kandel and Swartz!

Medial superior olive:

MSO receives a direct innervation from the anteroventral cochlear nucleus of both sides.
Cells of anteroventral cochlear nucleus have short latency…so cells of MSO are able to
receive temporally matched and temporally accurate signals from the two ears. In the
dog, which has a particularly large MSO, Goldverg and Brown found that almost all cells
were binaural. Three-quarters of cells were excited by both ears (EE cells), remainder
were EI cells. Characteristic delay different for each cell, result from differences in the
speed of transmission of signals from the two ears. If temporal difference matches the
characteristic delay of neurons in the MSO, large responses will be generated. The
different neurons possessing different characteristic delays will then be responsive to
sounds originating in different directions, and so could be said to map the direction of the
sound source.

Summary:
In LSO, most cells are EI cells so respond to intensity differences between the two ears.
Mojority of cells are high frequency cells, so have characteristic frequencies in range
where sound source to one side will produce significant differences in interaural
intensity.
By contrast, most cells in MSO are EE cells and so will not be responsive to interaural
intensity differences. Respond to direction on basis of interaural temporal cues and
ingeneral are most responsive to low frequencies in the range in which phase information
is preserved. Supported by anatomical evidence, animals with large heads and poor high
frequency hearing, who might be expected to favour temporal cues, have large MSO
nuclei and small LSO nuclei, whereas animals with small heads and good high frequency
hearing tend to have the reverse (Masterton and Diamond 1967).

Central nucleus of inferior colliculus:

Combine complex frequency analysis of dorsal cochlear nucleus with sound localizing
ability of superior olive.

Columns in auditory cortex?:

Imig and Adrian (1977) showed that in AI (primary auditory cortex), cells which were
excited by both ipsilateral and contralateral stimuli (EE cells) were located in discrete
radial columns separate from cells which were excited by one ear and inhibited by the
other (EI cells). Middlebrooks et al (1980) suggested that these patches were organized in
strips running at right angles to the iso-frequency contours.

Responses of single neurones: Sound localization

Many neurones in AI show binaural interactions (e.g. Brugge et al 1969, Brugge and
Merzenich, 1973; Benson and Teas 1976). Many cells were sensitive to interaural phase
or intensity differences.

Fig. 6.8 on p.170 in Physiology of Hearing has good diagram of ascending auditory
pathwaysof brain stem.