This outline summarizes major points covered in lecture. It is not intended to replace your own lecture notes.
Localization is a critical function of the auditory system.
Can have life-or-death consequences for predators, preys, and mates (reproduction).
Important in helping us distinguish signals from noise: Cocktail party effect.
Sound localization allows us to attend to specific signals within a noisy environment.
Remember that sound may have a source, but sound itself does not have a position.
o Lateralization is our ability to locate sound to either side relative to our midpoint.
o Azimuth is the horizontal plane.
o Elevation is the vertical plane.
o Radius is the distance from our head.
Sound Localization Cues
Azimuthal Localization (primarily binaural cues)
o Interaural time differences (ITDs)
o Interaural level differences (ILDs)
ITD cues are processed in the Medial Superior Olive (MSO) in brainstem
ILD cues are processed in the Lateral Superior Olive (LSO) in brainstem
Elevational Localization (primarily monaural cues)
o Pinna transformations of sound spectrum
o Head movements
o Interaural spectral differences
o Absolute SPL
o Excess attenuation effects
o Reverberation changes
o Changes in spectral balance
Measured ITD and ILD Cues
ITD cues vary with sound source location (azimuth) and remain relatively constant across frequency.
o Max ITD cues = 0.6-0.8 ms (depending on head size)
ILD cues are more complex and vary with sound source azimuth and elevation.
ILD cues increase with frequency.
Onset Time as ITD cue
Onset time provides unambiguous cue for localization.
Phase differences alone can produce ambiguity in source localization.
o E.g. If waveforms arriving at L and R ears are in-phase, a listener would wrongly perceive
sound source at the midline.
ITD does not change with frequency, and therefore is useful at both high and low frequencies.
Interaural Phase Difference (IPD)
IPD can be used only with lower frequencies. Lack of phase locking by ANFs at freq’s > ~4kHz.
Ambiguity due to head size above ~1.5kHz
IPD changes with frequency.
The information we get from IPD depends on the onset timing.
Jeffress Coincidence Detection Model
In order to work, we need phase locking, coincidence detectors, and delay lines.
A group of coincidence detector neurons are used to detect and encode ITD information.
Delay lines are used to retard timing of neural information so that spikes from near (ipsilateral) side
of head reach at a particular detector neuron at a time roughly compensated for by the acoustic
delay required to reach opposite (contralateral) ear
Psych 3A03 02 December 2011
Week 13 Dr. Paul A. Faure The notion is that an array or bank of coincidence detectors neurons exist and that they are tuned to
the range of acoustic (ITD) delays commonly generated based on size of head.
Coincidence detector neuron fires spikes only when contralateral acoustic ITD delay corresponds to
ipsilateral neural spike transmission delay so that a coincidence spikes from each side of head
arrives to a small population of coincidence detector neurons
The Barn Owl and Mammals
The barn owl is known to have strong phase locking in ANFs.
The barn owl also has neural delay lines and coincidence detector neurons.
Interestingly though, hearing in reptiles and birds evolved independently of mammals, so even
though the barn owl may use a Jeffress-like model for azimuthal sound localization, this doesn’t
necessarily mean that mammals use the same mechanism.
Mammals do not rely solely on an array of ITD-tuned neurons that receive excitation from both ears
(i.e. E-E neurons), but require inhibition to sharpen ITD tuning.
It turns out that barn owls also seem to use some inhibition, so the Jeffress model is being
questioned (Jeffress model does not use inhibition, just excitation).
Precedence (or Haas) Effect
Experiments have shown that the first wavefront arriving at the ears dominates in establishing
location of sound source
We suppress information coming from later arriving wavefronts (echoes or reflections)
This phenomenon is called Law of First Wavefront or Precedence (Haas) effect.
Azimuthal Sound Localization Performance
Listeners make errors localizing pure tones near 1-3 kHz.
Listeners make much fewer errors when localizing broadband noise signals.
Elevational Sound Localization Performance
Listeners make errors localizing pure tones near 2 kHz.
Listeners make fewer errors when localizing broadband noise signals.
Listeners use interaural spectral difference as a cue for elevational localization.
Maximal ITD for Humans
Consider head size in humans and speed of sound in air.
If head is 21 cm wide, and sound travels around 345 m/s, then:
o t = d / v ; 0.21 m / 345 m/s = 0.0006 s = 0.6 ms
When phase difference is ambiguous, you can check the values. Anything above maximal ITD is
incorrect, so our auditory system “deduces” that it must be a value below maximal ITD for humans.
When the frequency is too high and you can have different values below the maximal ITD for
humans, you cannot easily disambiguate the phase difference.
ILD Cues Increase With Sound Frequency
Geometric spreading (inverse square law) does not create appreciable ILD cues.
ILD is more prominent at high frequencies
Size of object relative to wavelength determines magnitude of ILD and sound shadow.
Lateral Superior Olive (LSO)
LSO cells receive converging excitatory (E) and inhibitory (I) inputs: E-I neurons
LSO cells receive binaural inputs and are sensitive to interaural level difference (ILD).
Minimum Audible Angle (MAA)
Testing for the minimum audible angle
o No visual cues
o Free-field auditory test with human subjects
A human’s minimum audible angle is around 1° (degree) near the midline.
MAA’s become larger the further away from the midpoint the sound source is located.
Listeners had larger MAA when loudspeakers were located off to the side than when located directly
in front (i.e. poor angular discrimination at side)
Psych 3A03 02 December 2011
Week 13 Dr. Paul A. Faure Over certain frequency ranges, we have additional increases in our MAA.
MAA changes across the frequency spectrum (i.e. more errors in mid-frequency region)
Duplex Theory of Sound Localization
Applies only to azimuthal (horizontal) localization
Humans perform best below 1.5 kHz and above 5 kHz, with worst performance around 2-3 kHz.
First proposed by John William Strutt (Lord Rayleigh) (1842-1919)
o Low frequency (<1500 Hz) pure tones localized by interaural time difference (ITD) cues
o High frequency (>1500 Hz) pure tones localized by interaural level difference (ILD) cues
Back to Duplex Theory
The reason for the different thresholds obtained when looking at minimum audible angle is that for
low frequency inputs, we can use the phase difference detecting mechanisms, whereas for high
frequency inputs we need to use pressure difference detecting mechanisms.
Lord Rayleigh (1842-1919) was correct.
Cone of Confusion
Both ITD and ILD cues are important for sound localization
For listeners that do not move their heads, there are a number of locations in space that provide the
same ITD and ILD cues (e.g. mid-sagittal plane) and this creates a cone of confusion
Localization errors usually made along cone of confusion.
Small head movements can resolve spatial location of sound.
Human ear remarkably sensitive to broad range of sound intensities (ca. 120 dB dynamic range).
Loudness: subjective impression of stimulus magnitude (i.e., stimulus intensity or sound pressure).
Difficult to measure quantitatively and objectively; strongly affected by bias and context effects.
How do we compare the loudness of different stimuli?
Just as with dB, we use a reference sound to compare loudness.
Loudness reference sound frequency = 1000 Hz sinusoid (pure tone).
Loudness lbevel defined as the level of a 1000 Hz sinusoid that is equal in loudness to test sound.
Auditory system does not respond equally to sounds of equal sound pressure level.
Perceived loudness of a sound can be predicted by audiogram.
Human auditory system is most sensitive to sounds ca. 500 Hz to 5000 Hz.
Frequency dependent sensitivity in loudness is represented by equal loudness contours (ELC).
Equal Loudness Contours (ELC)
Equal loudness contours tell us how Ss perceive loudness across frequency spectrum.
ELC measured using matching procedures.
Reference sound: level of a 1000 Hz tone in dB SPL.
Phon scale tells us which sounds are perceptually equally loud to 1000 Hz tone.
Phon = 1 dB SPL @ 1000 Hz
60 phons = the loudness of any signal that is equally loud to a 60 dB SPL ,1000 Hz tone.
Subjective loudness on phon scale is not directly proportional to its loudness level in phons.
o e.g. 80 phons is not twice as loud as a sound of 40 phons.
40 dB ELC also used to define loudness level in dB that approximates human hearing – dBA scale
dBA measure is the total amount of sound power that passes through a filter with cutoffs and
attenuation rates that match the 40 dB ELC
dBA takes into account that sounds in frequency region where humans are most sensitive are most
important in loudness determination