Journal of Vision (2007) 7(1):3, 1–15 http://journalofvision.

org/7/1/3/ 1

Some observations on the pedestal effect

The Sensory Research Unit, Department of Experimental
G. Bruce Henning Psychology, Oxford University, Oxford, United Kingdom

Max Planck Institute for Biological Cybernetics,

Felix A. Wichmann Tübingen, Germany

The pedestal or dipper effect is the large improvement in the detectability of a sinusoidal grating observed when it is added
to a masking or pedestal grating of the same spatial frequency, orientation, and phase. We measured the pedestal effect
in both broadband and notched noiseVnoise from which a 1.5-octave band centered on the signal frequency had been
removed. Although the pedestal effect persists in broadband noise, it almost disappears in the notched noise. Furthermore,
the pedestal effect is substantial when either high- or low-pass masking noise is used. We conclude that the pedestal effect
in the absence of notched noise results principally from the use of information derived from channels with peak sensitivities
at spatial frequencies different from that of the signal and the pedestal. We speculate that the spatial-frequency components
of the notched noise above and below the spatial frequency of the signal and the pedestal prevent ‘‘off-frequency looking,’’
that is, prevent the use of information about changes in contrast carried in channels tuned to spatial frequencies that are
very much different from that of the signal and the pedestal. Thus, the pedestal or dipper effect measured without notched
noise appears not to be a characteristic of individual spatial-frequency-tuned channels.
Keywords: contrast discrimination, off-frequency looking, contrast gain control, nonlinear transducer function, dipper effect,
masking, spatial vision
Citation: Henning, G. B., & Wichmann, F. A. (2007). Some observations on the pedestal effect. Journal of Vision, 7(1):3,
1–15, http://journalofvision.org/7/1/3/, doi:10.1167/7.1.3.

Dannemiller & Stephens, 1998; Foley, 1994; Foley &

Introduction Boynton, 1993; Foley & Chen, 1997; Foley & Legge,
1981; Georgeson & Georgeson, 1987; Gorea & Sagi, 2001;
Behavioral evidence from measurements of detection Kontsevich, Chen, & Tyler, 2002; Legge, 1981; Legge &
thresholds suggests that the early visual system is Foley, 1980; Legge, Kersten, & Burgess, 1987; Nachmias
composed of many spatial-frequency-selective channels & Sansbury, 1974; Ross & Speed, 1991; Wichmann, 1999;
(Blakemore & Campbell, 1969; Campbell & Robson, Yang & Makous, 1995).
1968; DeValois & DeValois, 1988; Graham & Nachmias, The most prominent finding of studies of sinusoidal
1971; Henning, 1988; Henning, Hertz, & Hinton, 1981). contrast discrimination is the dipper-shaped threshold-
Although the notion of linear and independent chan- versus-contrast (TvC) function in which the contrast of the
nels is probably not viable (Albrecht & DeValois, 1981; signal or increment at Bthreshold[ is plotted against ped-
Derrington & Henning, 1989; Henning, Hertz, & Broadbent, estal contrast. In the case we consider, where the pedestal
1975; Wichmann, 2004; Wichmann & Tollin, 1997a, 1997b), is a sinusoidal grating with the same spatial frequency,
the multichannel model still captures many crucial aspects orientation, phase, and duration as the grating to be de-
of early spatial vision, and even in nonlinear systems, the tected, the signal and increment contrasts are identical.
determination of the linear component of the system typ- For a limited range of low pedestal contrasts, contrast dis-
ically remains important. However, it is the knowledge of crimination is possible with signal contrasts that are too
the visual system’s operation at suprathreshold contrasts small to be seen without the pedestal; this is the pedestal or
that is a prerequisite for virtually any useful model of Bdipper[ effect. In vision, three main explanations of the
spatial vision. effect have been suggested: first, a nonlinear transducer
Sinusoidal contrast discrimination provides one useful (Foley & Legge, 1981; Legge, 1981; Legge & Foley, 1980;
way to gain insight into the characterization of contrast Yang & Makous, 1995), second, contrast gain control
transduction and the contrast gain control mechanisms (Foley, 1994; Foley & Chen, 1997; Wichmann, 1999), and
thought to operate at suprathreshold contrasts within the third, stimulus uncertainty (Pelli, 1985).
single channels inferred from detection thresholds. It is Amplitude discrimination in hearing is formally equiv-
hardly surprising, then, that sinusoidal contrast discrim- alent to contrast discrimination in vision, and indeed, a
ination has been studied extensively since Nachmias and dipper-shaped BTvC[ curve is also found in hearing.
Sansbury’s influential article in 1974 (Bradley & Ohzawa, Models in hearing combine an energy transducer with a
1986; Burton, 1981; Cannon & Fullenkamp, 1991; noise whose standard deviation is, in effect, proportional
doi: 1 0. 11 67 / 7 . 1 . 3 Received April 14, 2006; published January 11, 2007 ISSN 1534-7362 * ARVO

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 2

to the mean signal strength (Henning, 1967, 1969). Only present in one of the two observation intervals. The ob-
very recently in vision has there been debate about servers’ task was to choose the interval in which the signal
whether, as in hearing, the internal sources of noise had been presented by pressing one of two keys. The
variance depend on signal strength (Gorea & Sagi, 2001; probability of the signal’s being in the first interval was
Henning, Bird, & Wichmann, 2002; Kontsevich et al., 0.5 on every trial. The contrast of the signal was fixed for
2002; Wichmann, 1999). A further source of interest in blocks of 50 trials and then changed to determine four to
hearing is the nature of the mechanisms underlying the eight points on the psychometric function relating the
dynamic range (106) over which amplitude discrimination proportion of correct responses to signal contrast. The
(governed approximately by Weber’s law) is possible experiments were then repeated in a different order to
(Plack & Viemeister, 1993; Viemeister, 1972; Zwicker, obtain at least 500 observations per psychometric function
1956, 1970). One of the mechanisms considered in hear- for each observer. (In preliminary detection experiments,
ing is off-frequency listening (Patterson, 1976; Zwicker, 100 observations per point were only obtained at two points
1970)Vby analogy, off-frequency looking (Henning et al., immediately above and below the value corresponding to
1981; Losada & Mullen, 1995). 75% correct to speed up exploration with the more ex-
In this article, we use spectrally flat noise, no noise, and perienced observer.) Including training, the results reported
a flat noise from which a 1.5-octave notch centered on the in this study are based on 65,650 trials for observer N.A.L.,
signal frequency has been removed. We used a signal grat- 23,335 for observer T.C.C., and 28,765 for observer G.B.H.,
ing of 4 c/deg, and were our observers to base their dis- who is one of the authors.
crimination solely on the output of a linear channel tuned to Different types of masking stimuli were presented in
4 c/deg, we would not expect any differences in the shape both observation intervals. One type of masking stimulus
and the depth of the dipper (after normalization) between consisted of a sinusoid of the same orientation, spatial
any of the three experimental conditions reported above. If frequency, phase, and duration as the signal. The contrast
we consider divisive contrast gain control mechanisms that of this masker, or pedestal as such a masker is sometimes
integrate stimuli across a broad range of spatial frequencies called, was fixed, and 50-observation-per-point psycho-
(Carandini & Heeger, 1994; Carandini, Heeger, & Movshon, metric functions were obtained by varying the signal con-
1997; Heeger, 1992; Schwartz & Simoncelli, 2001), the trast. Then, the pedestal contrast was changed and the
dipper is predicted to be less pronounced in the broadband- process was repeated for pedestal contrasts increasing from
noise condition than with no noise, as can be seen from, and 0% to 32%. The process was repeated with decreasing
as confirmed by, our data. However, for this class of model, pedestal contrast so that, in the end, psychometric func-
a notch in the noise should result in a release from masking, tions with at least 500 observations were obtained. The
that is, a dipper more closely resembling the no-noise observers’ task was to choose the observation interval in
condition. This is clearly not what our results will show. which the signal had been added to the pedestal. Because
Our experiments cast doubt on previous attempts to the pedestal had the same spatial frequency, duration, ori-
model the TvC curve that shows a big dipper effect in entation, and phase as the signal, the task became one of
spatial vision as arising from characteristics of a single contrast discrimination. Only a 4-c/deg sinusoidal signal
spatial-frequency-tuned channel. Rather, our experimental was used for this experiment.
results with no noise and with broadband noise are con- A second type of masker consisted of one-dimensional
sistent with the dipper’s arising from off-frequency look- Gaussian noise of the same (horizontal) orientation as the
ing: As pedestal contrast increases, observers appear to signal. The noise, when it was used, was presented in both
shift the channel through which they perform the discrim- observation intervals for the same 86-ms duration as the
ination away from those tuned to the spatial frequency of signal, and the observers’ task was again to indicate the
the signal. Thus, the TvC curve attributable to the opera- interval in which the sinusoidal signal was present. In
tion of a single channel is that derived from the notched- some experiments, both the sinusoidal masker (pedestal)
noise experiment. We base this conclusion on results from and the noise masker were used.
the three masking conditions with sinusoidal gratings where The stimuli were generated digitally and displayed care-
we have carefully manipulated the spectral properties of fully linearized displays, either on monochrome Clinton
the masking noise. Monoray CRT displaysVmodified Richardson Electronics
MR2000HB-MED CRT’s with fast DP104 phosphorV
(observers T.C.C. and G.B.H.) or on a Sony GDM-520 in
Methods monochrome mode (observer N.A.L.) using Cambridge
Research Systems VSG 2/5 cards. Identical systems in
Oxford and Tübingen were used. The stimuli were pro-
Several two-alternative forced-choice detection and dis- duced using a two-field frame (75-Hz frame rate for the
crimination experiments were performed. On each trial of Clinton displays, 70-Hz frame rate for the Sony display)
all the experiments, there were two 86-ms-long temporal with the masking noise, when present, produced in alter-
intervals, separated by a 250-ms pause. The signal to be nate fields. The signal, in the observation interval in which
detected, a horizontally orientated sinusoidal grating, was it was presented, as well as the pedestal, when present,

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 3

was produced in the other field. In the nonsignal interval, noise power reduces the amount of clipping but leaves both a
and when neither masking noise nor pedestal was pres- less effective masker and fewer bits with which to represent
ent, uniform fields replaced the signal, the noise, or the the details of the filtered noise on which the characteristics of
pedestal appropriately. The addition of neither the signal, the notch depend. We generated a large number of noise
nor the pedestal, nor the noise had any effect on the ap- samples and only kept those that had the following:
proximately 50 cd/m2 mean luminance of the displays.
The signal and the two masker types were all presented 1. A suitably high noise-power density with a mean value
inside a common circularly symmetrical spatial Hanning (across the ensemble) equivalent to a Michelson contrast
window, the diameter of which subtended 6- of visual of approximately 3.4% at each spatial frequency in the
angle at the viewing distance of 1.6 m; the 86-ms temporal discrete representation of the noise spectrumVthe
window was rectangular. broadband noise with this mean spectral density
The first preliminary experiment measured contrast sen- raises the detection threshold for a 4-c/deg grating by
sitivity as a function of the spatial frequency of the signal a factor of about 8.
without any sinusoidal masker (pedestal) and was repeated 2. A suitably deep notch; the notched noises we kept
in two noise conditionsVone in which the noise-power had a width of at least 1.5 octaves and effectively
density spectrum was flat to an 42.7-c/deg upper bound lacked components between 2.67 and 7.5 c/deg.
and one in which a nominal 2-octave notch centered geo- 3. An attenuation in the notch that was at least 35 dB
metrically on 4 c/deg was produced by adding a spectrally below the noise power on either side of the notch.
flat noise that had been low-pass filtered to remove com- 4. A flat noise-power density spectrum in most of the
ponents nominally above 2 c/deg to the same noise that passbands above and below the notch.
had been high-pass filtered to remove components nom-
inally below 8 c/deg. Filtering was performed in the fre- Figure 1 shows the noise-power density averaged over
quency domain, and the noises then transformed to the the 100 noises we used. The figure shows the spectrum
space domain, suitably windowed (Rabiner & Gold, 1975), after windowing and rounding. The standard deviation over
and rounded to the 8-bit dynamic range of our video the noise samples at each spatial frequency was below
memory (VRAM). Because of the finite dynamic range of 5 dB. All stimuli were generated as 512  512 pixel arrays,
the visual display system and the finite size of the stimuli, which, at the viewing distance of 1.6 m, gave the diameter
generation of notched noises is not trivial. In particular, of the Hanning window within which the stimuli were
Gaussian noise samples inevitably call for luminance values viewed an angular subtense of 6.0-.
that exceed the dynamic range of the display system. Figure 2 illustrates the stimuli. The upper panels show
Truncation at the boundaries of the dynamic range leads to noisesVthe notched noise on the left and the broadband
clipping, which, if excessive, removes the notch. Reducing noise on the right. The middle panels are both copies of

Figure 1. The curve shows average noise-power density (in decibels; mean level corresponding to a Michelson contrast of 3.4% for each
component in the discrete representation of the noise spectrum) as a function of spatial frequency (logarithmic scale). The data are for
100 notched noises with the notch centered geometrically on 4 c/deg. The noise-power density in the stop band of the notch is at least
35 dB below that in the passbands.

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 4

Figure 2. The top panels illustrate notched noise (left) and broadband noise (right). The center panels illustrate the 4-c/deg sinusoidal
signal, and the bottom panels show the result when the signal is added to the noise.

the sinusoidal grating to be detected, and the bottom the contrast corresponding to 75% correct responses as a
panels show the sums of the signal and noises. The signal function of spatial frequency, and both axes are logarithmic.
in the notched noise is more clearly visible than that in the The filled circles show the results obtained in the no-noise
broadband noise. condition; the downward-pointing triangles indicate the re-
sults obtained with the spectrally flat noise; the stars show
the results obtained with the notched noise, the geometrical
center of which is indicated by the extended vertical red
Preliminary experiments line. The error bars, where shown and when larger than the
symbols, indicate a range of approximately T1 SD obtained
In preliminary experiments, we measured the contrast sen- by a bootstrap method from cumulative Gaussian fits to
sitivity of sinusoidal gratings for spatial frequencies rang- the underlying psychometric functions (Wichmann & Hill,
ing from 1 to 12 c/deg both with no masking noise and then 2001a, 2001b). Similar results were obtained from the
with both the broadband and notched-noise maskers. other observers.
We also measured the effect of the addition of a sinu- First, consider the results in the condition with no mask-
soidal masker (pedestal) with contrasts ranging from 0% ing noise (circles). The reciprocal of the thresholds in this
to 32% on the detectability of a 4-c/deg sinusoidal signal condition is just the standard contrast-sensitivity function
both with and without the broadband-noise masker. (CSF). Because the underlying psychometric functions are
roughly parallel on semilogarithmic coordinates, the shape
of the CSF does not depend much on the performance level
Results and discussion chosen for the threshold. Here, we use the conventional
75% correct level. The results are similar to those obtained
The results of the preliminary detection experiment with with other observers using stimuli of duration similar to
three different masking noises and no pedestal are shown our 86-ms stimuli (Bird, Henning, & Wichmann, 2002;
for one observer, G.B.H., in Figure 3. The figure shows Kelly, 1979a, 1979b; Robson, 1966). Performance, except

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 5

noise compared with 0.9 log units for all the observers
with the broadband noise. With the notched noise, the
amount of masking increases as the spatial frequency of
the signal falls either below or above 4 c/deg, but there is
a visible difference until the signal frequency is about an
octave above or below the center of the notch.
These results are not inconsistent with some measure-
ments of the shape of the spatial-frequency-tuned
Bchannels[ through which the signals are assumed to be
detected (Blakemore & Campbell, 1969; Campbell &
Robson, 1968; DeValois & DeValois, 1988; Graham &
Nachmias, 1971; Henning, 1988; Henning et al., 1981;
Stromeyer & Julesz, 1972). Behavioral measurements of
channel shape using masking techniques suggest an
asymmetric characteristic with the channel skirts falling
about 0.7 log units per octave below the spatial frequency
Figure 3. The data show the signal contrast at 75% correct to which the channel responds best and only about 0.4 log
responses as a function of spatial frequency for G.B.H. Both axes units per octave above that frequency (Henning et al.,
are logarithmic. Stars show performance in the notched noiseV 1981; Stromeyer & Julesz, 1972). Thus, it is the noise with
a 1.5-octave-wide notch centered on 4 c/deg. Circles show perfor- spatial-frequency components that are within an octave or
mance in the absence of masking noise, whereas downward- so of the signal frequency that has measurable effects on the
pointing triangles indicate performance in the presence of the detectability of the signal. The notch in our noise is
broadband masking noise. Error bars, where visible, show approximately 1.5 octaves wide, but the 4-c/deg channel is
estimates of T1 SD. so broad that there is some masking from the components
of the noise that are more than an octave away.
possibly at the lowest spatial frequency, is a roughly Figure 4 shows the signal contrast at 60%, 75%, and
monotonic decreasing function of spatial frequency. 90% correct as a function of the pedestal contrast in the
When spectrally flat broadband masking noise is added absence of any added masking noise for the three ob-
(triangles), contrast sensitivity is lower by at least a fac- servers. The performance levels are determined by fitting
tor of 8 across all three observers and depends much less Gumbel functions to the psychometric functions that re-
on spatial frequency. The flat noise produces considerable late the proportion of correct responses to the logarithm of
masking at all the spatial frequencies we used. With the signal contrast and then converting the log thresholds back
broadband noise, there is some loss in sensitivity with in- to contrasts (Wichmann & Hill, 2001a, 2001b). Both axes
creasing spatial frequency, but the noise has a flat spec- are logarithmic. An indication of the slope of the psycho-
trum. The increase in masking is probably a result of metric functions (and, hence, the variability around each
the increasing bandwidth (approximately T1 octave) of the data point) can be obtained from the vertical separation
spatial-frequency-tuned channels through which the sig- of the contours at each pedestal level, but the error bars
nals are thought to be detected (Blakemore & Campbell, again show approximately T1 SD. The results for the ob-
1969; Campbell & Robson, 1968; DeValois & DeValois, servers take the usual form (Bird et al., 2002; Bradley &
1988; Graham & Nachmias, 1971; Henning, 1988; Henning Ohzawa, 1986; Cannon & Fullenkamp, 1991; Dannemiller
et al., 1981; Stromeyer & Julesz, 1972). Increases in band- & Stephens, 1998; Foley, 1994; Foley & Boynton, 1993;
width that were exactly proportional to frequency would Foley & Chen, 1997; Foley & Legge, 1981; Georgeson &
result in increases in masking (measured by the contrast of Georgeson, 1987; Gorea & Sagi, 2001; Kontsevich et al.,
the signal at 75% correct) at a rate of 1 log unit per decade 2002; Legge, 1981; Legge & Foley, 1980; Legge et al.,
of spatial frequencyVnot far from the slope of 0.9 log 1987; Wichmann, 1999; Yang & Makous, 1995): As the
units per decade that we observe if the lowest spatial fre- pedestal contrast increases from zero, performance im-
quency is excluded (approximately constant masking across proves until the pedestal contrast is about twice (90% cor-
spatial frequency may be obtained with Bpink[ noise, see rect) or 10 times (60% correct) that of the corresponding
Henning et al., 2002). threshold and then deteriorates in a way that corresponds
The amount of masking (logarithmic units) is shown by roughly to Weber’s law (Bird et al., 2002; Wichmann, 1999).
the difference between the circles and the other symbols. The form of the curves is sometimes called the dipper
The amount of masking produced by the notched noise function (Foley, 1994; Foley & Legge, 1981; Georgeson &
(stars) and by the broadband noise (triangles) is similar Georgeson, 1987; Legge & Foley, 1980), but the shape
only at the very low and very high spatial frequencies depends on the performance level that determines the
as might be expected. The least masking produced by contour (Bird et al., 2002; Wichmann, 1999). The improve-
the notched noise occurs in the center of the notch near ment is much greater for the 60% contour than for the 90%
4 c/degVabout 0.4 log units of masking for the notched contour, which implies shallower psychometric functions in

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 6

The results of the preliminary experiments serve to

illustrate the usual form of the dipper or pedestal effect
with our observers and stimulus parameters (Figure 4).
They also indicate the way in which contrast sensitivity
depends on spatial frequency both with and without broad-
band, flat Gaussian masking noise and in the presence of
the notched noise (Figure 3).
We now wish to measure the pedestal effect when the
signal (as well as pedestal) is presented at 4 c/deg in a
notched noise with the notch centered on that spatial
frequency.

Main experiment: Results and

discussion

Figures 5, 6, and 7 show, separately for each observer,

signal contrast as a function of pedestal contrast at each
of three performance levels: 60% in the top panels, 75%
in the center panels, and 90% in the bottom panels. Each
panel shows three curves: (1) the curve for broadband
masking noise (downward-pointing triangles), (2) the curve
for the notched masking noise (stars), and (3) the curve for
the absence of masking noise (circles). Figure 5 shows
the results for observer G.B.H. that were obtained when the
noise-power density in the passbands was the same. The
masking is so large with the broadband noise that the dip-
per effect is shifted to such high contrasts as to be almost
obscured because sufficient pedestal contrast cannot be pro-
duced to generate the rising part of the TvC function. (As
signal and pedestal are interleaved with the noise, the signal
plus pedestal are limited to a maximum of 50% contrast.)
A clearer picture emerges in Figures 6 and 7 for the other
observers. To generate these data, we scaled down the
noise-power density for the broadband noise to produce
the same amount of masking for the 4-c/deg signal at the
75% level as is produced by the notched noise at the same
spatial frequency. Under these conditions, the small dipper
effect with broadband noise is clearly larger than that with
the notched noise.
To compare the shape of the functions closely, we show
in Figures 8, 9, and 10, again separately for each observer,
normalized signal contrast as a function of normalized
Figure 4. The data show signal contrast as a function of pedestal pedestal contrast. The signal and the pedestal contrasts for
contrast for N.A.L. (a), G.B.H. (b), and T.C.C. (c). Both axes are all three curves in all three panels of Figures 8, 9, and 10
logarithmic, and contours corresponding to 90%, 75%, and 60% have been normalized by the threshold contrast obtained
correct responses are shown. Detection data obtained in the ab- in the absence of a pedestal (data shown at the extreme
sence of a pedestal are shown at the extreme left end for com- left end of all three panels of the figures). The signal fre-
parison (pedestal contrast = 0). Error bars, where visible, show quency was 4 c/deg, and both axes are logarithmic. Con-
estimates of T1 SD. tours for 60%, 75%, and 90% correct responses are shown
in separate panels of each figure. This procedure, with a
the vicinity of the best performance. The shallower slope is pedestal contrast of zero, sets the normalized signal con-
sometimes taken to indicate that it is the reduction in trast at threshold to one.
signal uncertainty due to the presence of the pedestal that Inspection of any set of figures shows that data from the
causes the improved performance (Pelli, 1985). notched-noise condition are different in that the pedestal

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 7

Figure 6. Same as Figure 5 except that this figure shows the results
Figure 5. Each panel shows signal contrast for observer G.B.H. as a
for observer N.A.L. and that the spectrum level of the broadband
function of pedestal contrast on double logarithmic coordinates.
masking noise was adjusted to produce similar masking in de-
Contours corresponding to 60% correct, 75% correct, and 90%
tection (no pedestal) to that of the notched noise at 4 c/deg.
correct are shown in Panels a–c, respectively. Circles show
results obtained with no masking noise (replotted from Figure 4).
Downward-pointing triangles show results obtained with broadband
masking noise at a noise-power density that was the maximum effect is very much reduced at every performance level. For
attainable on our display system; that is, both notched noise and observer G.B.H., for example, when the notched-noise
broadband noise had the same noise-power density below 4 c/deg condition is compared with the no-noise condition at the
and above 16 c/deg. Stars indicate the results obtained with the 60% performance level where the pedestal effect is
notched noise. Detection data obtained in the absence of a ped- greatest, we find that the 1.0 log unit improvement in
estal (pedestal contrast = 0) are shown at the extreme left end for detectability is reduced to a 0.3 log unit improvement (a
comparison. Error bars, where visible, show estimates of T1 SD. factor of 5.00 less enhancement); for the 75% correct

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 8

would not expect any differences in the shape and the

depth of the dipper (after normalization) between any of
the three experimental conditions reported above.
If we allowed for divisive contrast gain control
mechanisms that integrate stimuli across a broad range
of spatial frequencies (Carandini & Heeger, 1994;

Figure 7. Same as Figure 6 except that this figure shows the

results for observer T.C.C.

level, the improvement drops from 0.70 to 0.19 log units

(a factor of 3.25 less enhancement); at 90%, the improve-
ment drops from 0.55 to 0.09 log units (a factor of 2.86
less enhancement). Similarly, when the notched-noise
condition is compared with the broadband-noise condi- Figure 8. Each panel shows normalized signal contrast for G.B.H.
tion, we find that the depth of the dipper is reduced by a as a function of normalized pedestal contrast on double loga-
factor of 1.33 (at 90% correct) to 2.00 (at 60% correct). rithmic coordinates. Contours corresponding to 60% correct, 75%
Were our observers to base their discrimination solely correct, and 90% correct are shown in Panels a–c, respectively.
on the output of a linear channel tuned to 4 c/deg, we The data are from Figure 5.

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 9

General discussion
In detection experiments, it is a reasonable assumption
that the response of only one or at least a very limited
number of spatial-frequency-tuned channels determines
the observers’ behavior. Many neurons may contribute
to a single channel, but, in detection, the contrast even
at the top of the psychometric function is so low that

Figure 9. Same as Figure 8 except that this figure shows the

results for observer N.A.L. and that the data are from Figure 6.

Carandini et al., 1997; Heeger, 1992; Schwartz &

Simoncelli, 2001), the dipper would be predicted to be
less pronounced in the broadband-noise condition than
that with no noise, as, indeed, borne out by our data. How-
ever, for this class of model, a notch in the noise should
result in a release from masking, that is, a dipper more
closely resembling the no-noise condition. This is clearly Figure 10. Same as Figure 8 except that this figure shows the
not what our results show. results for observer T.C.C. and that the data are from Figure 7.

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 10

channels tuned to spatial frequencies other than that of The effect of the notched noise is to limit the number of
the signal are unlikely to contribute much to the overall spatial-frequency-tuned channels that can be of use to the
performance. observers in discriminating the interval containing the
In pedestal experiments, however, the assumption that a signal plus pedestal from the pedestal alone. The dipper or
single channel determines the observer’s behavior is more pedestal effect does not disappear when the observers
difficult to justify. As the pedestal contrast increases, the perform the discrimination task in a notched noise, but
response of the neurons underlying the channel tuned to the magnitude of the improvement is less by a factor of
the spatial frequency of the signal has driven up their re- between 2.86 and 5.00 depending on the performance
sponse function toward saturation (DeValois & DeValois, contour being considered. Further, the range of pedestal
1988; Geisler & Albrecht, 1997b; Geisler, Albrecht, Salvi, contrasts over which Bsubthreshold[ performance occurs
& Saunders, 1991; Henning, 2004). In addition, good is greatly reduced in the notched noise. Inspection of Pan-
discrimination performance hinges on the neural response els a of Figures 5 to 10Vthe 60% contoursVfor example,
that rests on the steep part of the function relating rate and shows that the range of pedestal contrasts over which
contrast. Under these conditions, moving to a channel that subthreshold performance occurs is very much reduced in
will respond to the signal (and pedestal) frequency but the notched-noise condition. That the pedestal effect does
does not respond best to that frequency might have the not completely disappear in the notched-noise condition
advantage of lowering the neural response rate to a steeper may be because of the following: our 1.5-octave notch
part of the function and, thus, improving discrimination permits some, albeit limited, off-frequency looking; there
performance. This is called Boff-frequency[ looking. is some off-orientation looking within the notch; or there
Off-frequency looking would be of little benefit when is a small pedestal effect within a single channel (Geisler
the pedestal level is very highVbecause the output of all & Albrecht, 1997b; Geisler et al., 1991). If the latter is
the relatively broadly tuned channels would be driven up true, then it is the very small pedestal effect seen in the
above their optimal operating point and because even at presence of notched noise that models of contrast gain
moderate pedestal levels, the outputs of the channels control (Heeger, 1992, 1994; Geisler & Albrecht, 1992,
appear to become highly correlated (Henning et al., 2002) 1995, 1997a; Yang & Makous, 1995) need to fit rather
Vor very lowVbecause at low levels, the channels that than the large pedestal effects seen in the absence of
are not tuned to the signal frequency would be responding, noise.
if at all, well below their optimal operating point. Thus, One obvious implication of the above analysis is that
there is only a limited range of pedestal contrast over the pedestal effect, which virtually disappears in notched
which off-frequency looking could improve performance. noise, should return if the masking noise either only below
A further consequence of off-frequency looking stems or only above the signal frequency is removed because, in
from the fact that there is a sizeable range of contrastsV either case, off-frequency looking becomes possible. We
the central, approximately linear range of the rate versus tested this prediction in a supplementary experiment using
contrast functionsVover which good discrimination per- either high- or low-pass noise.
formance is possible (Geisler & Albrecht, 1997b). This In the supplementary experiment, the experimental
means that over a range of pedestal contrasts, performance conditions were as described in the Methods section save
can be improved by combining information from different that masking noise was changed. For the low-pass case,
channels. It seems likely that it is the combination of in- the components of the broadband noise above the 4-c/deg
formation from different channels that produces the usual signal frequency were removed, and for the high-pass
dipper or pedestal effect. case, the components below 4 c/deg were removed. In
Using information from off-frequency channels would both cases, the noise-power density in the passband again
increase the range of pedestal levels over which good dis- had a Michelson contrast of 3.4% at each spatial fre-
crimination performance is possible. As the pedestal level quency, and two of the previous observers participated in
increases, the observers should continue to use informa- the experiment.
tion from channels with peak sensitivities that are farther Figure 11 shows the results with low-pass noise
and farther from the signal frequency until, ultimately, they separately for the two observers in the same format as
run out of channels. Discrimination performance would Figure 3Vsignal contrast corresponding to 60%, 75%,
then deteriorate in a fashion determined, presumably, by and 90% correct responses is shown as a function of ped-
the characteristics of last spatial-frequency channel in use estal contrast. There is a clear dipper effect for both ob-
or because with increasing pedestal levels, probability sum- servers. The magnitude of the dip is lesser with no noise
mation becomes weak because of increasing correlation but is larger with the notch. The average drop at 60% is
among the channels’ output (Bird et al., 2002). One con- 3.5; at 75%, it is 2.2; and at 90%, it is 1.65. This compares
sequence of such a procedure is that the dipper functions to the average drop in the notched noise of 1.6, 1.5, and
for signals at all spatial frequencies should ultimately co- 1.3 at 60%, 75%, and 90%, respectively.
incide as pedestal levels increase. That is what is found Figure 12 shows a similar result with high-pass noise.
(Bird et al., 2002; Bradley & Ohzawa, 1986). Here, the average drop in the dip is 3.6, 2.2, and 1.6 at

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 11

Figure 11. The panels show separately for each observer the signal contrast as a function of the pedestal contrast in the presence of low-
pass noise with a 4-c/deg cutoff. Both axes are logarithmic, and contours corresponding to 60%, 75%, and 90% correct responses are
shown. Detection data obtained in the absence of a pedestal are shown at the extreme left end for comparison (pedestal contrast = 0).
Error bars, where visible, show estimates of T1 SD.

60%, 75%, and 90%, respectively. Thus, the dipper-shaped the channels being equally sensitive and uncorrelated, and
TvC reappears. That the dip occurs at higher pedestal con- the extent to which the channels are correlated is not clear
trasts in high-pass noise is also consistent with the loss of (Henning et al., 2002, 1975). It is also difficult to estimate
peak sensitivity of channels tuned to higher spatial fre- z values at low<performance levels. Nonetheless, the
quencies. Both results are consistent with the notion of off- numbers are of some interest. To determine the number
frequency looking because in both cases, removing part of of independent channels that would be needed to move
the noise allows useful off-frequency looking in parts of the performance at its best in the notched-noise condition to
spatial-frequency spectrum. its best in the no-noise condition, we use the 75% contour
One final consideration is the number of independent, in the no-noise condition and note the z value associated
different, and equally sensitive spatial-frequency channels with the corresponding signal contrast in the notched-
that would need to be combined to produce the biggest noise condition. The ratio of the z values is then an
improvement in performance. Such a number is, of course, estimate of the square root of the number of independent
only a crude indicator because the calculation hinges on channels needed to move performance from that in the

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 12

Figure 12. Same as Figure 11 except that the masking noise was high pass with a 4-c/deg cutoff.

notched-noise condition to that in the no-noise condition. detection threshold in the absence of a pedestal. The effect
The number is about 8Vnot an unreasonably large is largest in the absence of any additional external mask-
number. ing noise but is still present if broadband masking noise
is added.
The improvement in detectability is almost abolished if
Summary the signal and the pedestal are presented in a notched
noise such that the only components of the masking noise
are remote in frequency from the spatial frequency of the
There is a large improvement in the detectabilty of a signal and the pedestal with no components within about
sinusoidal grating when the signal is added to a masking an octave of the spatial frequency of the signal and the
or pedestal grating with the same spatial frequency, ori- pedestal. The improvement in detectability, which is al-
entation, and phase. The contrast at which the largest most abolished in notched noise, is present with either
improvement occurs depends on the performance level low- or high-pass noise as well as with broadband noise.
(Wichmann, 1999)Vfor the 75% correct level, the largest We interpret our results to indicate that the notched noise
improvement occurs when the pedestal contrast is in the prevents observers’ using information from spatial-frequency-
vicinity of twice the contrast corresponding to the 75% tuned channels with peak sensitivities that are above or below

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 13

the spatial frequency of the signal and the pedestal; that is, Burton, G. J. (1981). Contrast discrimination by the
the notched noise prevents the use of information about human visual system. Biological Cybernetics, 40,
changes in contrast carried in off-frequency channels. 27–38. [PubMed]
We conclude that the pedestal effect in the absence of Campbell, F. W., & Robson, J. G. (1968). Application of
notched noise is principally a result of the use of Fourier analysis to the visibility of gratings. The
information derived from channels with peak sensitivities Journal of Physiology, 197, 551–556. [PubMed]
at spatial frequencies that are different from that of the [Article]
signal and the pedestal and, thus, that the pedestal or
dipper effect is not a characteristic of individual spatial- Cannon, M. W., & Fullenkamp, S. C. (1991). A transducer
frequency-tuned channels. Thus, the pedestal effect is not model for contrast perception. Vision Research, 31,
caused by nonlinear transduction, contrast gain control, or 983–998. [PubMed]
uncertainty about the signal but, instead, mainly stems Carandini, M., & Heeger, D. J. (1994). Summation and
from off-frequency looking. division by neurons in primate visual cortex. Science,
264, 1333–1336. [PubMed]
Carandini, M., Heeger, D. J., & Movshon, J. A. (1997).
Acknowledgments Linearity and normalization in simple cells of the
macaque primary visual cortex. The Journal of
This research was supported by the Max Planck Society Neuroscience, 17, 8621–8644. [PubMed] [Article]
(F.A.W.) and the Wellcome Trust (G.B.H.) and was started Dannemiller, J. L., & Stephens, B. R. (1998). Contrast
while G.B.H. was visiting Prof. Bernhard Schölkopf’s Em- gain control in psychophysical contrast discrimina-
pirical Inference Department at the MPI in Tübingen. We tion, Perception & Psychophysics, 60, 1153–1163.
gratefully acknowledge Prof. Schölkopf’s encouragement [PubMed]
and support. We would also like to thank Miss N. Althaus Derrington, A. M., & Henning, G. B. (1989). Some ob-
and Miss T. C. Cristescu for their dedicated observing as servations on the masking effects of two-dimensional
well as two anonymous reviewers and Prof. W. S. Geisler stimuli. Vision Research, 29, 241–246. [PubMed]
for helpful criticism. Part of this work was presented at the
German Experimental Psychology Society’s 2006 Annual DeValois, R. L., & DeValois, K. K. (1988). Spatial vision.
Meeting (TeaP) in Mainz (Wichmann & Henning, 2006) Oxford: Oxford University Press.
and at the Vision Sciences Society 2006 Annual Meeting Foley, J. M. (1994). Human luminance pattern-vision
(VSS) in Sarasota. mechanisms: Masking experiments require a new model.
nal of the Optical Society of America A, Optics, Image
Commercial relationships: none. Science, and Vision, 11, 1710–1719. [PubMed]
Corresponding author: Felix A. Wichmann. Foley, J. M., & Boynton, G. M. (1993). Forward pattern
Email: felix@tuebingen.mpg.de. masking and adaptation: Effects of duration, intersti-
Address: MPI for Biological Cybernetics, Spemannstr. 38, mulus interval, contrast, and spatial and temporal
72076 Tübingen, Germany. frequency. Vision Research, 33, 959–980. [PubMed]
Foley, J. M., & Chen, C. C. (1997). Analysis of the effect
References of pattern adaptation on pattern pedestal effects: A
two-process model. Vision Research, 37, 2779–2788.
[PubMed]
Albrecht, D. G., & DeValois, R. L. (1981). Striate cortex
responses to periodic patterns with and without the Foley, J. M., & Legge, G. E. (1981). Contrast detection
fundamental harmonics. The Journal of Physiology, and near-threshold discrimination in human vision.
319, 497–514. [PubMed] [Article] Vision Research, 21, 1041–1053. [PubMed]
Bird, C. M., Henning, G. B., & Wichmann, F. A. (2002). Geisler, W. S., & Albrecht, D. G. (1992). Cortical
Contrast discrimination with sinusoidal gratings of dif- neurons: Isolation of contrast gain control. Vision
ferent spatial frequency. Journal of the Optical Society Research, 32, 1409–1410. [PubMed]
of America A, Optics, Image Science, and Vision, 19, Geisler, W. S., & Albrecht, D. G. (1995). Bayesian analysis
1267–1273. [PubMed] of identification performance in monkey visual cortex:
Blakemore, C., & Campbell, F. W. (1969). Adaptation Nonlinear mechanisms and stimulus certainty. Vision
to spatial stimuli. The Journal of Physiology, 200, Research, 35, 2723–2730. [PubMed]
11P–13P. [PubMed] Geisler, W. S., & Albrecht, D. G. (1997a). Identification
Bradley, A., & Ohzawa, I. (1986). A comparison of contrast performance of neurons in monkey visual cortex for
detection and discrimination. Vision Research, 26, simple and complex images. Investigative Ophthal-
991–997. [PubMed] mology and Visual Science, 38, S734.

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 14

Geisler, W. S., & Albrecht, D. G. (1997b). Visual cortex on the shape of spatial-frequency filters inferred from
neurons in monkeys and cats: Detection, discrimination, masking experiments: I. Noise masks. Journal of the
and identification. Visual Neuroscience, 14, 897–919. Optical Society of America, 71, 574–581. [PubMed]
[PubMed] Kelly, D. H. (1979a). Motion and vision: I. Stabilized
Geisler, W. S., Albrecht, D. G., Salvi, R. J., & Saunders, S. S. images of stationary gratings. Journal of the Optical
(1991). Discrimination performance of single neurons: Society of America, 69, 1266–1274. [PubMed]
Rate and temporal-pattern information. Journal of Kelly, D. H. (1979b). Motion and vision: II. Stabilized
Neurophysiology, 66, 334–362. [PubMed] spatio-temporal threshold surface. Journal of the
Georgeson, M. A., & Georgeson, J. M. (1987). Facilitation Optical Society of America, 69, 1340 –1349.
and masking of briefly presented gratings: Time- [PubMed]
course and contrast dependence. Vision Research, 27, Kontsevich, L. L., Chen, C. C., & Tyler, C. W. (2002).
369–379. [PubMed] Separating the effects of response nonlinearity and
Gorea, A., & Sagi, D. (2001). Disentangling signal from internal noise psychophysically. Vision Research, 42,
noise in visual contrast discrimination. Nature Neuro- 1771–1784. [PubMed]
science, 4, 1146–1150. [PubMed] [Article] Legge, G. E. (1981). A power law for contrast discrim-
Graham, N., & Nachmias, J. (1971). Detection of grating ination. Vision Research, 21, 457–467. [PubMed]
patterns containing two spatial frequencies: A com- Legge, G. E., & Foley, J. M. (1980). Contrast masking in
parison of single channel and multichannel models. human vision. Journal of the Optical Society of
Vision Research, 11, 251–259. [PubMed] America, 70, 1458–1471. [PubMed]
Heeger, D. J. (1992). Normalization of cell responses in Legge, G. E., Kersten, D., & Burgess, A. E. (1987).
cat striate cortex. Visual Neuroscience, 9, 181–197. Contrast discrimination in noise. Journal of the
[PubMed] Optical Society of America A, Optics and Image
Heeger, D. J. (1994). The representation of visual stimuli Science, 4, 391–404. [PubMed]
in primary visual cortex. Current Directions in Losada, M. A., & Mullen, K. T. (1995). Color and
Psychological Science, 3, 159–163. luminance spatial tuning estimated by noise masking
Henning, G. B. (1967). A model for auditory discrim- in the absence of off-frequency looking. Journal of
ination and detection. Journal of the Acoustical the Optical Society of America A, Optics, Image
Society of America, 42, 1325–1334. [PubMed] Science, and Vision, 12, 250–260. [PubMed]
Henning, G. B. (1969). Amplitude discrimination in noise, Nachmias, J., & Sansbury, R. V. (1974). Grating contrast:
pedestal experiments, and additivity of masking. Jour- Discrimination may be better than detection. Vision
nal of the Acoustical Society of America, 45, 426–435. Research, 14, 1039–1042. [PubMed]
[PubMed] Patterson, R. D. (1976). Auditory filter shape derived with
Henning, G. B. (1988). Spatial-frequency tuning as a func- noise stimuli. The Journal of the Acoustical Society of
tion of temporal frequency and stimulus motion. Jour- America, 59, 640–654. [PubMed]
nal of the Optical Society of America A, Optics and Pelli, D. G. (1985). Uncertainty explains many aspects of
Image Science, 5, 1362–1373. [PubMed] visual contrast detection and discrimination. Journal
Henning, G. B. (2004). Masking effects of low-frequency of the Optical Society of America A, Optics and
sinusoidal gratings on the detection of contrast mod- Image Science, 2, 1508–1532. [PubMed]
ulation in high-frequency carriers. Journal of the Plack, C. J., & Viemeister, N. F. (1993). Suppression and
Optical Society of America A, Optics, Image Science, the dynamic range of hearing. Journal of the
and Vision, 21, 486–490. [PubMed] Acoustical Society of America, 93, 976–982.
Henning, G. B., Bird, C. M., & Wichmann, F. A. (2002). [PubMed]
Contrast discrimination with pulse trains in pink noise. Rabiner, L., & Gold, B. (1975). Theory and application
Journal of the Optical Society of America A, Optics, of digital signal processing. New York: Prentice-Hall
Image Science, and Vision, 19, 1259–1266. [PubMed] International.
Henning, G. B., Hertz, B. G., & Broadbent, D. E. (1975). Robson, J. G. (1966). Spatial and temporal contrast-
Some experiments bearing on the hypothesis that the sensitivity functions of the visual system. Journal of
visual system analyzes patterns in independent bands the Optical Society of America, 56, 1141–1142.
of spatial frequency. Vision Research, 15, 887–897. Ross, J., & Speed, H. D. (1991). Contrast adaptation and
[PubMed] contrast masking in human vision. Proceedings of the
Henning, G. B., Hertz, B. G., & Hinton, J. L., (1981). Royal Society B: Biological Sciences, 246, 61–69.
Effects of different hypothetical detection mechanisms [PubMed]

Downloaded from jov.arvojournals.org on 03/23/2024

 Journal of Vision (2007) 7(1):3, 1–15 Henning & Wichmann 15

Schwartz, O., & Simoncelli, E. P. (2001). Natural signal sta- Wichmann, F. A., & Hill, N. J. (2001a). The psychometric
tistics and sensory gain control. Nature Neuroscience, function: I. Fitting, sampling, and goodness-of-fit.
4, 819–825. [PubMed] [Article] Perception & Psychophysics, 63, 1293–1313.
Stromeyer, C. F., III, & Julesz, B. (1972). Spatial- [PubMed] [Article]
frequency masking in vision: Critical bands and Wichmann, F. A., & Hill, N. J. (2001b). The psycho-
spread of masking. Journal of the Optical Society of metric function: II. Bootstrap-based confidence inter-
America, 62, 1221–1232. [PubMed] vals and sampling. Perception & Psychophysics, 63,
1314–1329. [PubMed] [Article]
Viemeister, N. F. (1972). Intensity discrimination of
pulsed sinusoids: The effects of filtered noise. Journal Wichmann, F. A., & Tollin, D. J. (1997a). Masking
of the Acoustical Society of America, 51, 1265–1269. by plaid patterns is not explained by adaptation,
[PubMed] simple contrast gain-control or distortion products.
Investigative Ophthalmology and Visual Science,
Wichmann, F. A. (1999). Some aspects of modelling 38, S631.
human spatial vision: Contrast discrimination.
Unpublished doctoral dissertation, The University of Wichmann, F. A., & Tollin, D. J. (1997b). Masking by
Oxford, Oxford, UK. plaid patterns: Spatial frequency tuning and contrast
dependency. OSA Conference Program, 97.
Wichmann, F. A. (2004). Masking by plaid patterns
revisited. In D. Kerzel, V. Franz, & K. R. Gegenfurtner Yang, J., & Makous, W. (1995). Modeling pedestal ex-
(Eds.), Experimentelle psychologie. Beiträge zur 46. periments with amplitude instead of contrast. Vision
Tagung experimentell arbeiten der Psychologen Research, 35, 1979–1989. [PubMed]
(p. 285). Lengerich: Pabst Science Publishers. Zwicker, E. (1956). Die Elementaren Grundlagen zur
Wichmann, F. A., & Henning, G. B. (2006). The Bestimmung der Informationskapazita ¨ t des Gehörs.
pedestal effect is caused by off-frequency looking, Acustica, 6, 365–381.
not nonlinear transduction or contrast gain-control. Zwicker, E., (1970). Masking and psychological excitation
In H. Hecht, S. Berti, G. Meinhardt, & M. Gamer as consequence of the ear’s frequency analysis. In
(Eds.), Experimentelle psychologie. Beiträge zur 48. R. Plomp & G. Smoorenburg (Eds.), Frequency analysis
Tagung experimentell arbeitender Psychologen (p. 205). and periodicity detection in hearing (pp. 376–394).
Lengerich: Pabst Science Publishers. Leiden: A. W. Sijthoff.

Downloaded from jov.arvojournals.org on 03/23/2024