NEUROSCIENCE UConn Neuroscience in Salamanca, Spain May, 2009

Masking, The Critical Band

and Frequency Selectivity
Enrique A. Lopez-Poveda, Ph.D.
Neuroscience Institute of Castilla y León
University of Salamanca
ealopezpoveda@usal.es
Aims
• Define the different forms of sound masking:
simultaneous, forward and backward.
• Define auditory filter, tuning curve and critical band.
• Describe common psychoacoustical methods to
measure auditory filters.
• Define masking pattern and describe its common
interpretation.
• Describe the neurophysiological basis of masking.
• Define frequency selectivity and describe its main
characteristics in normal-hearing listeners.
• Analyze the main consecuences of hearing impairement
on auditory frequency selectivity.
Acknowledgements
• This presentation is inspired on a lecture prepared by
author to be taught as part of the international Audiology
Expert degree of the University of Salamanca (Spain).

• Some figures and examples were adapted from several

sources (see references).
What is (sound) masking?
Just like a face mask hides the identity of its
wearer, a sound can mask another sound
making its perception (or detection) more
difficult.
In psycoacoustics

Masking is the process by which the detection threshold of a sound (called

‘the signal’) is increased by the presence of another sound (called ‘the
masker’).

The amount of masking is defined as the increase (in decibels) in the

detection threshold of a sound (signal) due to the presence of a masker
sound.
Types

Simultaneous
masking

Forward (post-
stimulatory)
masking

Backward (pre- Signal

Masker sound
stimulatory) Sound time
masking
Simultaneous masking: Auditory
filters

time
 The “critical band”
Adapted from Schooneveldt & Moore (1989)
Stimulus spectrum
Spectral level

2 kHz
Critical bandwidth = 400 Hz
Signal detection threshold
increases with increasing
masking noise bandwidth up to a
point after which signal threshold
becomes independent of masker
bandwidth.

Schooneveldt GP, Moore BCJ. (1989). Comodulation masking release for various monaural and binaural combinations of the signal, on-frequency, and
flanking bands. J Acoust Soc Am. 85(1):262-272.
 Auditory filters

Fc Frequency

To explain the previous result, Fletcher (1940) suggested that the auditory
system behaves like a bank of overlapping bandpass filters. These filters are
termed “auditory filters”.
Fletcher H. (1940). Auditory patterns. Rev. Mod. Phys. 12, 47-65.
 An explanation of the critical band

The amount of masking

increases with increasing the
noise (masker) energy that
gets through the filter.

Up to a point…!

Further increases in noise

bandwidth do not increase
the masker energy through
the filter.
2 kHz
Frequency
Equivalent rectangular bandwidth
(ERB)

Fc Frequency
ERB

An auditory filter (yellow area) and its ERB filter (green area).
Both have different shapes but equal height and total area. That
is, both let the same energy through.
The masking threshold according to
Fletcher’s power spectrum model
Fletcher (1940) proposed that the
masking threshold occurs when the
acoustic power of the signal (S) at
the filter output is proportional to the
acoustic power of the masker (M) at
the filter output:

S/M = k, with k being a

proportionality constant.
For a noise masker with constant
spectral density (N), and a critical
band W, masking threshold occurs
when:

S/(WxN) = k, hence W = S/(kxN).

Thus measuring S and knowing N,

we can infer W.
Caution!

Fletcher’s model is useful to explain several

auditory phenomena, but is just a model!

More details to come...

How to estimate the shape of a filter?

Iso-stimulus curve Iso-response (tuning) curve

Same input amplitude, Same output amplitude,

Different input frequency. Different input frequency.
Measure output amplitude. Adjust input amplitude.
Output amplitude

Input amplitude
Frequency Frequency
How to estimate the shape of an
auditory filter?

Iso-stimulus curve Iso-response (tuning) curve

Same masker level, Same signal level,

Measure signal detection threshold. Measure masker level at signal
detection threshold.

Masker level at signal

threshold for a fixed

detection threshold
Signal detection

masker level

Signal frequency Signal frequency

Psychoacoustical tuning curves

Method B produces
PSYCHOACOUSTICAL
TUNING CURVES (PTCs).

Psychoacoustical tuning
curves for normal-hearing and
hearing-impaired listeners.
Signal was a 1-kHz pure tone
at 10 dB SL. Masker was
narrowband noise (Moore &
Glasberg, 1986).

Moore BCJ, Glasberg, BR (1986). Comparisons of frequency selectivity in simultaneous and forward masking for subjects
with unilateral cochlear impairments. J. Acoust. Soc. Am. 80, 93-107.
Physiological and psychoacoustical tuning curves

Auditory nerve fiber

tuning curves (Palmer,
1987).

Psychoacoustical
tuning curves (Vogten,
1974).

Palmer AR (1987). Physiology of the cochlear nerve and cochlear nucleus, in Hearing, edited by M.P. Haggard y E.F. Evans (Churchill Livingstone, Edinburgh).
Vogten, L.L.M. (1974). Pure-tone masking: A new result from a new method, in Facts and Models in Hearing, edited by E. Zwicker and E. Terhardt (Springer-
Verlag, Berlin).
Off-frequency listening
Off-frequency listening is said to occur when the signal is
detected through a filter different from the one with a center
frequency (Fc) equal to the signal frequency.

Fc Frequency

Off-frequency listening occurs because auditory filters are

asymmetric and have steep high-frequency slopes (Moore,
1998).

Moore BCJ (1998). Cochlear Hearing Loss, Whurr Publishers, London.

The notch-noise method of Patterson
(1976)
W

Power (linear scale)

Frequency (linear scale)

Notch-noise spectrum (green area) versus auditory filter

shape (yellow area).
The notch noise method
W1

Signal detection trhreshold

W2

W1 W2 W3
Notch bandwidth, W

W3

Signal detection threshold decreases with

increasing notch bandwidth. Filter shape is
the integral of the red curve (Patterson,
1976). This method reduces off-frequency
listening.
Patterson RD. (1976). “Auditory filter shapes derived with noise stimuli.” J. Acoust. Soc. Am. 59, 640-654.
Auditory filters obtained with the notch-
noise method

From: Baker S, Baker RJ. (2006). Auditory filter nonliearity across frequency using simultaneous
notched-noise masking. J. Acoust. Soc. Am. 119, 454-462.
Simultaneous masking: Masking
patterns

Time
What is a masking pattern?

Signal frequency
FIXED VARIABLE

FIXED MASKING
Masker frequency

MASKING
HARDLY USEFUL PATTERN
PATTERN
VARIABLE

AUDITORY FILTER ?

A masking pattern is a masked audiogram, i.e., a graphical

representation of the audiogram measured while pure tones of different
frequencies are presented in the presence of a masker sound (with any
spectrum).
 The masking pattern of a pure tone

From: Egan JP, Hake HW. (1950). On the masking pattern of a simple auditory stimulus. J. Acoust. Soc. Am. 22, 622-630.
 The interpretation of a masking pattern

Adapted from Moore (2003)

It is assumed that the signal is
c detected through the auditory
filter giving the greatest signa-
to-masker ratio at the output.
Relative gain (dB)

d
Therefore, a different detection
auditory filter is used as the
e signal frequency changes. The
b frequency of the masker tone
is fixed.
a

Frequency

From: Moore BCJ. (2003). An introduction to the psychology of hearing. 4 Ed. Academic Press, London.
 The interpretation of a masking pattern

Adapted from Moore (2003) The illustration shows five

c detection auditory filters with
different center frequencies.
Each filter has 0-dB gain at its
Relative gain (dB)

d tip. The vertical line illustrates

the pure tone signal whose
excitation pattern is to be
e measured. The dots illustrate the
b excitation of each filter in
response to the masker tone.
a

Frequency

Moore BCJ. (2003). An introduction to the psychology of hearing. 4 Ed. Academic Press, London.
 The excitation pattern

Adapted from Moore (2003)

c
The excitation pattern (red
curve) would be a plot of the
Relative gain (dB)

masker output from each

d filter to the masker tone.

Therefore, it represents
e something akin to the internal
b excitation pattern of the
a masker spectrum.
Frequency
Indeed, it is possible to measure a masking patter
for any masker…

…and the result is thought to represent approximately the ‘internal’ excitation

evoked by the masker.
 For example: The excitation pattern of a vowel
Excitation patterns for three different
subjects at different sound levels

Espectra of synthesized /i/ /æ/

/i/ & /æ/ vowels

From Moore BCJ, Glasberg BR. (1983). “Masking patterns for synthetic vowels in simultaneous and forward masking,” J. Acoust. Soc. Am. 73,
906-917.
The neurophysiological bases of simultaneous
masking

Neural response
swamping

Simultaneous
Simultaneous
Neural suppression Other
masking
masking
Simultaneous masking may reflect
swamping of neural responses
Neural response to Neural response to pure
A) B)
masking noise tone (signal)

Characteristic
Characteristic

frequency
frequency

Time Time

Neural response to Neural response to noise

C) D)
noise+signal + signal

Characteristic
Characteristic

frequency
frequency

Time Time
Simultaneous masking may reflect
suppression of neural activity to the signal

Neural response to Neural response to a

A) B)
masking noise pure tone signal

Characteristic
Characteristic

frequency
frequency

Time Time

Neural response to the Neural response to

C) D)
masker+signal masker + signal

Characteristic
Characteristic

frequency
frequency

Time Time
Most probably, it is a combination of those
two plus other phenomena

Swamping
Suppression

Other
+
= Masking
The neurophysiological bases of (psychophysical)
auditory filters

Basilar membrane
frequency response

Auditoryfilters
filters Inner hair cell
Other Auditory frequency response

Lateral inhibition
Auditory filters almost certainly reflect cochlear
tuning

From: Ruggero MA, Rich NC, Recio A, Narayan SS, Robles L. (1997). “Basilar-membrane responses to tones at
the base of the chinchilla cochlea,” J Acoust Soc Am. 101(4):2151-63.
But the inner hair cell may also contribute!
Psychoacoustics Auditory nerve
(Lopez-Poveda et al., 2006) (Rose et al., 1971)
BASE (CF = 2100 Hz)

Discharge rate (spikes/s)

Masker level (dB SPL)

BASE (CF = 6000 Hz)

APEX (CF = 125 Hz) APEX (CF = 200 Hz)

Frequency (Hz) Frequency (Hz)

Lateral inhibition may also contribute to “sharpen”
auditory filters
10 10
Stimulus spectrum
5 5

Neurons 5 5 10 10 10 5 5

-0.2×

1×

3 2 7 6 7 2 3
Output signal
amplitude

7 6 7

Output spectrum 2 2
5 5

Yost, W. A. (2000). Fundamentals of hearing. Academic Press, San Diego.

Post-stimulatory (forward)
masking

tiempo
A sound may be masked by a preceeding sound

Left panels illustrate the amount of

masking as a function of the time gap
between the masker offset and the signal
onset. The masker was a narrow-band
noise. The signal was a pure tone. Each
symbol is for a different masker level (in
decibels).

Right panels illustrate the amount of

masking as a function of the masker
spectral level. Each symbol is for a
different time gap (in ms).

Each row shows results for a different

signal frequency (1, 2 and 4 kHz).

From: Moore BCJ, Glasberg BR (1983). Growth of masking for

sinusoidal and noise maskers as a function of signal delay:
implications for suppression in noise. J. Acoust. Soc. Am. 73, 1249-
1259.
Psychoacoustical tuning curves measured with
forward masking

detection threshold (dB

Masker level at signal

SPL)

Masker frequency (Hz)

From: Lopez-Poveda, E. A., Barrios, L. F., Alves-Pinto, A. (2007). "Psychophysical estimates of level-dependent best-frequency shifts
in the apical region of the human basilar membrane," J. Acoust. Soc. Am. 121(6), 3646-3654.
The neurophysiological bases of forward masking

‘Ringing’ of basilar
membrane
responses

Forwardmasking
masking Auditory nerve
Central inhibition Forward adaptation

Neural response
persistence
Persistence of basilar membrane responses after masker
offset (‘ringing’)

estímulo The response of the basilar membrane

does not end immediately after the stimulus
offset. Instead, it persists over a period of
time (as shown in the left figure). This
‘ringing’ effect may make the detection of
the following signal more difficult.

From: Recio A, Rich NC, Narayan SS, Ruggero MA. (1998).

“Basilar-membrane responses to clicks at the base of the chinchilla
cochlea,” J. Acoust. Soc. Am. 103, 1972-1989.
Auditory nerve fiber adaptation

Stimulus Nerve response

signal
masker

Tiempo Tiempo

From: Meddis R, O’Mard LP. (2005). A computer model of the auditory-nerve response to forward-
masking stimuli. J. Acoust. Soc. Am. 117, 3787-3798.
Persistence of neural activity

Acoustic
stimulus signal
masker

time

Neural
activity

time

The persistence of neural activity may impair

the detection of the signal.
Central inhibition

Acoustic excitation Neural

1 2
stimulus response

delay
inhibition
Acoustic
stimulus signal
masker

Activity of neuron
1

delay
Inhibition induced by the masker
Activity of neuron
2

time
Pre-stimulatory (backward)
masking

Signal
Masker sound
sound Time
Backward masking
• Little is known about it.
• Hardly observed in well-trained subjects.
• Possibly, listeners misinterpret the brief signal
with the start of the masker.
Frequency selectivity

Enrique A. López-Poveda
Neuroscience Institute of Castilla y León
University of Salamanca
ealopezpoveda@usal.es
What is frequency selectivity?

500 Hz
+ 100 Hz

It is the ability to
perceive separately
multiple frequency
components of a
complex sound
How does it occur?
Low-frequency sound

Base Apex

High-frequency sound

Base Apex
It depends on the functional state of the
cochlea

Normal cochlea Damaged cochlea

Psychoacoustical measures of
frequency selectivity
Masking

Simultaneous signal masker

Forward masker signal

Frequency selectivity may be measured using masking

techniques like those previously described.
Psychophysical tuning curves are a
measure of frequency selectivity

Psychoacoustical tuning curves

have different shapes depending
on the masking method employed
to measure them.

Curves measured with forward

masking appear more tuned than
those measured with simultaneous
masking.

From: Moore BCJ. (1998). Cochlear Hearing Loss.

Whurr Publishers, London.
Cochlear suppression affects
(psychoacoustical) frequency selectivity

Auditory nerve
Stimulus
response

amplitude
amplitude

Signal only

tiempo time

amplitude
amplitude

Masker + signal

tiempo time

The response to the signal is

lower in the presence of the
masker as a result of cochlear
suppression.
Frequency selectivity in normal-
hearing listeners
Filter bandwidth varies with center frequency

Psychoacoustical estimates for Auditory nerve data for cat

normal-hearing listeners

From: Moore BCJ (1998). Cochlear Hearing Loss. From: Pickles JO (1988). An Introduction to the
Whurr Publishers, London. Psychology of Hearing. Academic Press, London.
Filter tuning varies with sound level

Psychoacoustical estimates for Guinea-pig basilar membrane

normal-hearing listeners response
CF = 1 kHz CF = 10 kHz

90 dB SPL

20

From: Moore BCJ (1998). Cochlear Hearing From: Ruggero MA, Rich NC, Recio A, Narayan S,
Loss. Whurr Publishers, London. Robles L. (1997). Basilar membrane responses to tones
at the base of the chinchilla cochlea. J. Acoust. Soc. Am.
101, 2151-2163.
 Tuning also varies with sound level: Tuning curves

Psychoacoustical estimates for Guinea-pig basilar membrane

normal-hearing listeners response
CF = 4 kHz CF = 10 kHz
100
Masker level (dB SPL)

80

60

40

20
800 2400 4000 5600 7200
Frequency (Hz)

From: Lopez-Poveda, EA, Plack, CJ, and Meddis, R. From: Ruggero MA, Rich NC, Recio A, Narayan S, Robles
(2003). “Cochlear nonlinearity between 500 and 8000 Hz L. (1997). Basilar membrane responses to tones at the
in normal-hearing listeners,” J. Acoust. Soc. Am. 113, 951- base of the chinchilla cochlea. J. Acoust. Soc. Am. 101,
960. 2151-2163.
Auditory filters

Adaptado de Baker y Rosen (2006)

Frequency selectivity in the auditory
nerve: intact and damaged cochleae
Total outer hair cell (OHC) damage
Cochlear status Tuning curves
Intact
IHC

damaged

Total normal
OHC
damage

Total OHC damage. Intact IHC.

From: Liberman MC, Dodds LW, Learson DA. (1986). “Structure-function correlation in
noise-damaged ears: a light and electrone-microscopy study.” in RJ Salvi, D Henderson,
RP Hamernik, V Colletti. Basic and applied aspects of noise-induced hearing loss.
(Plenum Publishing Corp, 1986).
Total OHC damage (cont.)

1. Reduced
sensitivity, raised
response threshold.
Total OHC damage (cont.)

2. Broader tuning,
reduced frequency
selectivity.
Total OHC damage (cont.)

3. Lower
characteristic
frequency (CF).
In vivo and post-mortem basilar membrane
responses

The effects described before are similar

to those observed when comparing in
vivo and post-mortem basilar
membrane tuning curves for the same
cochlear region (left figure).

Consequently, it is generally thought

that auditory nerve tuning reflects
basilar membrane tuning.

From: Sellick PM, Patuzzi R, Johnstone BM. (1982). Measurements of basilar membrane
motion in ght guinea pig using the Mössbauser technique. J. Acoust. Soc. Am. 72, 131-141.
 Partial inner hair cell (IHC) damage
Cochlear status Tuning curves

Partial IHC
damage
damaged

normal
Normal
OHCs

Partial IHC damage (arrow).

Intact OHCs
From: Liberman MC, Dodds LW, Learson DA. (1986). “Structure-function correlation in
noise-damaged ears: a light and electrone-microscopy study.” in RJ Salvi, D Henderson,
RP Hamernik, V Colletti. Basic and applied aspects of noise-induced hearing loss.
(Plenum Publishing Corp, 1986).
Severe OHC and IHC damage
Cochlear status Tuning curves

Severe
damaged
IHC
damage

normal
Severe
OHC
damage

From: Liberman MC, Dodds LW, Learson DA. (1986). “Structure-function correlation in
noise-damaged ears: a light and electrone-microscopy study.” in RJ Salvi, D Henderson,
RP Hamernik, V Colletti. Basic and applied aspects of noise-induced hearing loss.
(Plenum Publishing Corp, 1986).
Partial (combined) OHC and IHC damage
Cochlear status Tuning curves

Partial
IHC
damage
damaged

normal
Partial
OHC
damage

Partial IHC damage

Partial OHC damage
From: Liberman MC, Dodds LW, Learson DA. (1986). “Structure-function correlation in
noise-damaged ears: a light and electrone-microscopy study.” in RJ Salvi, D Henderson,
RP Hamernik, V Colletti. Basic and applied aspects of noise-induced hearing loss.
(Plenum Publishing Corp, 1986).
Frequency selectivity in normal-
hearing listeners vs. listeners with
cochlear hearing loss
Difficult comparison because sound level is
different!
Filter shape varies with sound level and sound level is necessarily
higher for hearing-impaired listeners.
Absolute hearing threshold

Normal hearing Hearing impaired

Difficult comparison because of off-frequency
listening!

The signal should produce the

maximum excitation at this point on the
basilar membrane…

However, this will be actually

amplitude

the point of maximum excitation

Basilar membrane
travelling wave

base apex
2
1

Therefore, the signal is detected

through auditory nerve fiber 2 and not 1
(despite the CF of the latter equals the
signal frequency).
Psychoacoustical tuning curves

The signal was a pure tone of

1 kHz at 10 dB SL. The
masker was a narrowband
noise. From Moore & Glasberg
(1986).

PTCs are broader for hearing-

impaired listeners!
Does frequency selectivity decrease with amoung
of hearing loss?

Generally yes, but not always!

The figure compares
psychoacoustical tuning curves
for two hearing impaired
listeners with similar losses at 4
kHz. Their tuning curves are
very different (one of them is
almost normal).

From: Lopez-Poveda, EA, Plack, CJ, Meddis, R, and Blanco, JL. (2005). "Cochlear compression between 500
and 8000 Hz in listeners with moderate sensorineural hearing loss," Hearing Res. 205, 172-183.
Why is this?

Possibly, the hearing loss of

listener DHA is due to outer hair
cell dysfunction…

…while that of listener ESR is due

to inner hair cell dysfunction.
Dead (cochlear) regions

Psychoacoustical tuning curve of the

same patient for a 2-kHz signal.

Audiogram of a patient with a

cochlear dead region (in red) around
2 kHz.

From: Moore BCJ (1998). Cochlear

Hearing Loss. Whurr Publishers, London.

IHC stereocilia in healthy

(green) and dead (red)
cochlear regions.
Auditory filters in listeners with cochlear hearing loss

Cochlear hearing loss is

typically (but not always)
accompanied by auditory
filters that are broader than
normal.

The figure illustrates 1-kHz

filters for a collection of
listeners with unilaterla
cochlear hearing loss.

Moore BCJ (1998). Cochlear Hearing

Loss. Whurr Publishers, London.
Filter bandwidth increases on
average with increasing absolute
threshold (thus the amount of
hearing loss).

The wide spread of values

indicates, however, that
filter bandwidth cannot be
predicted based on
absolute threshold.
Moore BCJ (1998). Cochlear Hearing Loss. Whurr
Publishers, London.
Impaired speech perception with less frequency
selectivity
(HEARLOSS demo)
Thank you!