Perceptual Masking

Abstract

Perceptual masking, which MeSH classifies under both auditory perception and visual perception, is the reduction or elimination of the perception of one stimulus, the target, by the presence of another, the masker. In hearing, a sound raises the threshold for detecting another sound close to it in frequency and time; Fletcher's critical-band model fixed the rule that only masker energy within a narrow band around the target contributes. In vision, a pattern presented just before or after a target can abolish its visibility even without spatial overlap, as in metacontrast and object-substitution masking. Across both senses the field distinguishes energetic masking, a swamping of the target at the receptor, from informational masking and central interruption, a failure of later processing to segregate target from masker. Masking thus measures the filters of perception and reveals the time course of awareness.

Keywords: perceptual masking, critical band, backward masking, metacontrast, object-substitution masking

What Perceptual Masking Is

Perceptual masking is the interference of one stimulus with the perception of another. The stimulus whose perception is impaired is the target; the interfering stimulus is the masker. Masking is defined operationally: a masker is present when the threshold for detecting or identifying the target is higher, or the target's perceived clarity lower, than when the target is presented alone. The size of that change — the elevation of threshold in decibels, or the drop in identification accuracy — is the measure of masking.

Masking occurs in every sensory modality, but it has been studied most intensively in hearing and vision, where it serves two distinct scientific purposes. As a tool, masking measures the resolution of a sensory system: the extent to which one stimulus masks another reveals how finely the system separates stimuli in frequency, space, or time. Fletcher used auditory masking to measure the width of the ear's frequency filters; vision researchers use masking to measure the temporal grain of visual processing. As a phenomenon, masking exposes the mechanisms of perception themselves — where in the processing stream a stimulus can still be disrupted, and therefore where its conscious representation is built.

The central theoretical division, recurring in both senses, is between masking that happens because the masker physically overwhelms the target at the sensory surface and masking that happens because later, central processing cannot segregate the two. In hearing these are called energetic and informational masking; in vision the parallel distinction is between integration, where target and mask fuse into a single percept, and interruption, where the mask curtails the processing of a target that was briefly represented in full. The rest of this article follows that division through the two modalities.

Auditory Masking and the Critical Band

Auditory masking is the elevation of the threshold of one sound (the target or probe) by another (the masker). Its quantitative study began with the discovery that masking is frequency-selective: a masker raises the threshold of probes close to it in frequency far more than distant ones, tracing out a masking pattern that peaks at the masker's frequency and falls away to either side.

Fletcher explained this pattern with the critical band. He proposed that in detecting a tone in noise, the ear acts as a bank of overlapping band-pass filters, and that only the noise energy falling within a narrow band centred on the tone contributes to masking it; energy outside that band is filtered out and does not interfere (Fletcher, 1940). The tone is detected when its power exceeds a fixed multiple of the noise power inside this critical band. This power-spectrum model makes a sharp prediction: as the bandwidth of a noise masker centred on the tone is widened, the tone's threshold rises — but only until the noise fills the critical band, after which further widening adds energy that the filter rejects and the threshold stops climbing. Measuring the bandwidth at which threshold stops rising therefore measures the critical band directly, a method Greenwood used to map critical bandwidth across the whole range of hearing and relate it to position along the cochlea (Greenwood, 1961).

Masking also depends on time. A masker raises the threshold of a probe that comes shortly before it (backward masking) or shortly after it (forward masking), with the effect decaying as the gap grows to a few tens of milliseconds. The modern synthesis of simultaneous and non-simultaneous masking, and of their basis in the mechanics of the cochlea and the coding of sound in the auditory nerve, is reviewed by Oxenham (2018).

Energetic and Informational Masking

The critical-band account is a theory of energetic masking: the masker and target compete for the same auditory filter, and the target is lost because its representation at the sensory surface is swamped. But listeners often fail to hear a target even when it is energetically detectable — when the target and masker fall in different frequency regions, or when the masker is a competing talker whose energy does not overlap the target's. This residual interference is informational masking: a failure of the central auditory system to segregate and select the target from a confusable background, rather than a failure of the peripheral filters (Durlach et al., 2003).

The distinction matters most for speech heard against other speech. A background of competing talkers masks a target voice through both routes at once: an energetic component, as overlapping spectro-temporal energy fills the target's critical bands, and an informational component, as the listener struggles to assign the right syllables to the right talker. Separating the two is an active methodological problem, because they call for different remedies — energetic masking is relieved by improving the signal-to-noise ratio, informational masking by anything that helps segregate the sources, such as spatial separation or a distinct voice. Recent work isolates the informational component by comparing speech maskers with energy- and modulation-matched noise across the lifespan, showing that the informational burden is heaviest for young children and older adults (Tuomainen et al., 2024).

Visual Masking: Integration and Interruption

Visual masking is the reduction of a target's visibility by a mask presented close to it in time and space. Because vision unfolds over time, the stimulus-onset asynchrony (SOA) between target and mask — not merely their overlap — is the controlling variable, and varying it reveals the time course of visual processing.

Kahneman (1968) organised the early literature into two broad accounts. In integration masking, target and mask are summed into a single composite percept because they arrive within the same brief integration window; the target is lost because it is diluted by the mask's energy, much as in energetic auditory masking. In interruption masking, the mask instead halts the ongoing processing of a target that had already been registered, so the target is lost not by fusion but by the premature termination of its analysis. Turvey (1973) sharpened this into a distinction between peripheral masking, which depends on the energy and eye of presentation and behaves like integration, and central masking, which survives presentation to opposite eyes (dichoptic presentation) and behaves like interruption — evidence that the two forms act at different levels of the visual system.

The signature of interruption is metacontrast: a mask that spatially surrounds but does not overlap a target can nonetheless render it invisible, and does so most strongly not at zero SOA but at an intermediate delay of roughly 50–100 milliseconds, giving a U-shaped function of visibility against SOA. This non-monotonic case is called type-B masking, to distinguish it from type-A masking, in which visibility is lowest at zero SOA and climbs steadily as the interval grows; type-A is the signature of peripheral integration, type-B of central interruption. Because the type-B mask follows the target yet erases it, metacontrast cannot be a peripheral summation of overlapping light; it must reflect the mask overtaking the target in central processing, which dual-channel models attribute to a fast transient response to the mask catching up with a slower sustained response to the target (Breitmeyer & Ogmen, 2000).

Object-Substitution Masking

A later discovery showed that masking need not require a surrounding contour or even a strong mask at all. In object-substitution masking (OSM), a target is surrounded by a sparse mask — classically four dots — that does not touch it and, presented alone, is barely visible. When the four dots appear with the target and then linger after the target disappears, and especially when attention is spread across many items, the target becomes much harder to identify, as though the trailing dots have replaced it (Di Lollo, Enns, & Rensink, 2000).

Di Lollo and colleagues explained OSM with reentrant processing. Perception is not a single feed-forward sweep but an iterative exchange in which higher areas send hypotheses back down to be checked against the still-arriving sensory evidence. When the display changes between the first sweep and the reentrant check — the target gone, only the four dots remaining — the returning signal finds the mask instead of the target and the mask's representation wins, substituting for the target. OSM is therefore distinguished from energy and contour masking by its dependence on the timing of attention and on the mask outlasting the target, not on spatial or energetic overlap (Enns & Di Lollo, 2000). A meta-analytic review of two decades of OSM work re-weighted the evidence toward an object-updating account, in which the trailing mask updates the perceptual object rather than discretely substituting for it (Goodhew, 2017).

Table 1

Principal Forms of Perceptual Masking by Modality and Locus

Note. The recurring division is between peripheral masking, where the masker swamps the target at the sensory surface, and central masking, where later processing fails to segregate or complete the target (Kahneman, 1968; Turvey, 1973; Durlach et al., 2003).

Worked Example

Consider a 1 kHz tone to be detected in a background of white noise whose spectrum level — its power per hertz — corresponds to 40 dB. The critical band at 1 kHz is about 160 Hz wide. According to Fletcher's power-spectrum model, only the noise energy inside this 160 Hz band masks the tone, and the tone reaches its masked threshold when its power equals that in-band noise power. What is the masked threshold?

The in-band noise power is the spectrum level plus the decibel equivalent of the bandwidth: L = 40 + 10 log₁₀(160). Since 10 log₁₀(160) = 22.0 dB, the masked threshold is 62.0 dB SPL. The tone must be raised to about 62 dB to be just audible in this noise, 22 dB above the noise's per-hertz spectrum level — the 22 dB being exactly the width of the critical band expressed in decibels.

The model's signature prediction follows from the same arithmetic. Suppose the critical band were twice as wide, 320 Hz. The masked threshold would rise to 40 + 10 log₁₀(320) = 65.0 dB — a 3 dB increase, because doubling the bandwidth doubles the in-band power and doubling power is by definition a 3 dB change. This is why widening a band-limited noise masker raises the tone's threshold at 3 dB per doubling — but only up to the critical bandwidth. Widen the noise beyond 160 Hz and the threshold stops climbing, because the extra energy falls outside the filter and is rejected. The bandwidth at which the threshold levels off is the critical band itself, which is how it is measured. These numbers match the Critical-Band Masking demonstration above.

Current Directions

The most active theoretical front is visual masking's role as a probe of consciousness. Because a mask can abolish the conscious visibility of a target that the visual system has demonstrably processed — as metacontrast and object-substitution masking both show — masking is a standard tool for separating processing that reaches awareness from processing that does not, and for timing when in the first few hundred milliseconds a percept becomes conscious (Bachmann, 2018). The reentrant account of object-substitution masking feeds directly into this programme, since it locates the failure of awareness not in the sensory input but in the iterative matching of hypothesis to evidence (Goodhew, 2017).

In hearing, the live problem is informational masking in realistic listening. Isolating the informational component from the energetic one, and tracing how each changes across development and ageing, bears directly on why understanding speech in a noisy room is so much harder for children and older adults even when the audiogram is normal (Tuomainen et al., 2024). Both frontiers share a theme: the limiting factor in perception is increasingly located not at the sensory surface but in the central processes that segregate, select, and complete a stimulus.

Discussion

Perceptual masking earns its central place in sensory psychology because it answers two questions at once. It measures the filters of perception — the critical band in hearing, the integration window in vision — by using one stimulus as a calibrated ruler against another. And it dissects the stages of perception, by showing where in the processing stream a stimulus can still be disrupted and therefore where its percept is built.

The unifying idea across the two modalities is the division between peripheral and central masking. Energetic auditory masking and visual integration masking are peripheral: the masker swamps the target at or near the receptor, and the target is lost by summation. Informational auditory masking, visual interruption, metacontrast, and object substitution are central: the target survives the sensory surface intact but is lost in later processing that cannot segregate it from the masker, is interrupted before it completes, or is overwritten when reentrant signals return to a changed display. That a single conceptual axis — swamping at the surface versus failure of central segregation — organises masking in both hearing and vision is why the phenomenon has remained a workhorse of perception research for nearly a century, and why it now serves as one of the principal experimental handles on visual awareness itself.

Common Misconceptions

A mask must physically overlap the target to hide it.

Metacontrast and object-substitution masking both abolish a target with a mask that never touches it. The mask that surrounds but does not overlap the target disrupts its central processing, which is why these forms cannot be explained by summation of overlapping light (Breitmeyer & Ogmen, 2000; Di Lollo et al., 2000).

Masking is strongest when target and mask coincide.

For metacontrast the opposite is often true: visibility is highest at zero SOA and lowest at an intermediate delay of 50–100 ms, giving a U-shaped function. The mask is most destructive when it arrives slightly after the target, not simultaneously with it (Kahneman, 1968).

If a sound is loud enough to measure, it cannot be masked.

A target can be energetically audible yet still masked informationally, when the listener cannot segregate it from a confusable background such as a competing talker. Informational masking is a central failure of selection, not a peripheral loss of energy (Durlach et al., 2003).

A wider noise masker always masks a tone more.

Only noise within the critical band around the tone contributes. Widening a band-limited masker raises the tone's threshold until the noise fills the critical band; beyond that, extra energy is rejected by the filter and the threshold stops rising — the very fact that lets masking measure the band (Fletcher, 1940).

Glossary

Auditory perception.

The perception of sound, one of the two faculties under which MeSH classifies perceptual masking; auditory masking is its study of how one sound hides another.

Backward masking.

Masking in which the masker follows the target in time yet still impairs it; its existence implies the masker disrupts ongoing processing rather than the sensory input.

Critical band.

The bandwidth of the auditory filter around a tone within which noise energy masks it; energy outside the band is rejected, so masking measures the band's width.

Dichoptic presentation.

Showing target and mask to opposite eyes; masking that survives this cannot be peripheral and is attributed to central processing.

Energetic masking.

Masking in which the masker swamps the target at the sensory surface, competing for the same auditory filter; relieved by improving the signal-to-noise ratio.

Forward masking.

Masking in which the masker precedes the target, raising its threshold for a few tens of milliseconds after the masker ends.

Informational masking.

Masking that remains when energetic masking is controlled, caused by the central system's failure to segregate a target from a confusable masker such as a competing talker.

Integration masking.

Visual masking in which target and mask fall within one integration window and fuse into a single percept, diluting the target; the visual analogue of energetic masking.

Interruption masking.

Visual masking in which the mask halts the processing of a target that was already registered, so the target is lost by premature termination rather than fusion.

Masker.

The stimulus that impairs perception of the target; the quantity of masking is the change in the target's threshold or clarity caused by its presence.

Metacontrast.

Backward masking by a stimulus that spatially surrounds but does not overlap the target, strongest at an intermediate SOA and yielding a U-shaped visibility function.

Object-substitution masking.

Masking by a sparse, non-overlapping mask that lingers after the target, strongest under divided attention; attributed to reentrant processing finding the mask in place of the target.

Power-spectrum model.

Fletcher's account in which a tone is detected when its power exceeds a fixed multiple of the noise power within the critical band around it.

Reentrant processing.

The iterative exchange in which higher visual areas send hypotheses back to lower areas for checking; the mechanism Di Lollo and colleagues proposed for object-substitution masking.

Stimulus-onset asynchrony (SOA).

The interval between the onsets of target and mask; the controlling variable in visual masking, whose variation traces the time course of visual processing.

Target.

The stimulus whose perception is to be measured and which the masker impairs; detection or identification of the target is the dependent measure in masking.

Visual perception.

The perception of sight, the second faculty under which MeSH classifies perceptual masking; visual masking is its study of how one pattern hides another over time.

Key Researchers

Talis Bachmann

(living). Experimental psychologist at the University of Tartu working on visual masking and the neural and microgenetic basis of perceptual awareness. His metacontrast and perceptual-retouch work ties masking to theories of consciousness. ORCID 0000-0001-9595-1594.

James T. Enns

(living). Professor of psychology at the University of British Columbia, co-author of the object-substitution-masking studies and the reviews that separated substitution masking from energy and contour masking. ORCID 0000-0002-3676-8316.

Harvey Fletcher

(1884–1981). American physicist at Bell Telephone Laboratories who introduced the critical-band concept and the power-spectrum model of auditory masking, the foundation of every later account of how one sound renders another inaudible. See his Wikipedia biography.

Stephanie C. Goodhew

(living). Professor at the Australian National University whose meta-analytic review re-weighted two decades of object-substitution-masking evidence toward object updating, the current reference point for the mechanism debate. ORCID 0000-0002-5066-8303.

Daniel Kahneman

(1934–2024). Israeli-American psychologist and Nobel laureate whose early review organised visual masking into integration and interruption accounts and set the field's methodology for a generation. See his Wikipedia biography.

Vincent Di Lollo

(living). Emeritus professor of psychology at Simon Fraser University who, with Enns and Rensink, defined object-substitution masking and the reentrant account of it. ORCID 0000-0003-1195-9842.

Andrew J. Oxenham

(living). Professor of psychology at the University of Minnesota whose review of hearing covers the cochlear filter and simultaneous and non-simultaneous auditory masking. ORCID 0000-0002-9365-1157.

Michael T. Turvey

(1942–2023). Distinguished Professor of experimental psychology at the University of Connecticut whose information-processing analysis separated peripheral energy-integration masking from central pattern-interruption masking. See his Wikipedia biography.

Frequently Asked Questions

What is perceptual masking?

Perceptual masking is the reduction or elimination of the perception of one stimulus, the target, by another, the masker. It is defined by measurement: masking is present when the threshold for detecting or identifying the target rises, or its clarity falls, relative to presenting the target alone. It occurs in every sense but is studied most in hearing and vision.

What is the difference between energetic and informational masking?

Energetic masking is a peripheral effect: the masker swamps the target at the sensory surface, competing for the same auditory filter, and is relieved by improving the signal-to-noise ratio. Informational masking is a central effect: the target is energetically detectable but the listener cannot segregate it from a confusable background, such as a competing talker, so it is relieved by anything that helps separate the sources.

What is the critical band?

The critical band is the bandwidth of the auditory filter centred on a tone within which noise energy contributes to masking it. Fletcher proposed that energy outside this band is rejected, so widening a noise masker raises a tone's threshold only until the noise fills the band. Measuring where the threshold stops rising measures the band's width.

What is metacontrast masking?

Metacontrast is a form of backward masking in which a mask that spatially surrounds but does not overlap a target renders it invisible. It is strongest not when target and mask coincide but at an intermediate stimulus-onset asynchrony of about 50–100 milliseconds, producing a U-shaped function of visibility against SOA, which shows the mask disrupts central processing rather than the input.

What is object-substitution masking?

Object-substitution masking occurs when a sparse mask, classically four dots that do not touch the target, lingers after the target disappears and, especially under divided attention, makes it hard to identify. It is explained by reentrant processing: when higher visual areas send signals back to check the target, they find the trailing mask in its place, and the mask's representation wins.

How is masking used to study consciousness?

Because a mask can abolish the conscious visibility of a target that the visual system has demonstrably processed, masking separates processing that reaches awareness from processing that does not, and times when in the first few hundred milliseconds a percept becomes conscious. Metacontrast and object-substitution masking are the main tools for this.

Does the masker have to overlap the target in space?

No. In energetic and integration masking the masker and target do overlap, but in metacontrast and object-substitution masking the mask surrounds or accompanies the target without touching it, and still erases it. These non-overlapping forms are evidence that masking can act centrally, on processing rather than on the sensory image.

Why is it harder to understand speech in a room full of talkers?

Because competing talkers mask a target voice through both routes at once: an energetic component as overlapping sound fills the target's critical bands, and an informational component as the listener struggles to assign the right syllables to the right talker. The informational burden is heaviest for young children and older adults even when hearing thresholds are normal.

References

Bachmann, T. (2018). Visual masking: Contributions from and comments on Bruce Bridgeman. Consciousness and Cognition, 64, 13–18. https://doi.org/10.1016/j.concog.2018.04.013

Breitmeyer, B. G., & Ogmen, H. (2000). Recent models and findings in visual backward masking: A comparison, review, and update. Perception & Psychophysics, 62(8), 1572–1595. https://doi.org/10.3758/BF03212157

Di Lollo, V., Enns, J. T., & Rensink, R. A. (2000). Competition for consciousness among visual events: The psychophysics of reentrant visual processes. Journal of Experimental Psychology: General, 129(4), 481–507. https://doi.org/10.1037/0096-3445.129.4.481

Durlach, N. I., Mason, C. R., Kidd, G., Arbogast, T. L., Colburn, H. S., & Shinn-Cunningham, B. G. (2003). Note on informational masking. The Journal of the Acoustical Society of America, 113(6), 2984–2987. https://doi.org/10.1121/1.1570435

Enns, J. T., & Di Lollo, V. (2000). What's new in visual masking? Trends in Cognitive Sciences, 4(9), 345–352. https://doi.org/10.1016/S1364-6613(00)01520-5

Fletcher, H. (1940). Auditory patterns. Reviews of Modern Physics, 12(1), 47–65. https://doi.org/10.1103/RevModPhys.12.47

Greenwood, D. D. (1961). Auditory masking and the critical band. The Journal of the Acoustical Society of America, 33(4), 484–502. https://doi.org/10.1121/1.1908699

Goodhew, S. C. (2017). What have we learned from two decades of object-substitution masking? Time to update: Object individuation prevails over substitution. Journal of Experimental Psychology: Human Perception and Performance, 43(6), 1249–1262. https://doi.org/10.1037/xhp0000395

Kahneman, D. (1968). Method, findings, and theory in studies of visual masking. Psychological Bulletin, 70(6, Pt.1), 404–425. https://doi.org/10.1037/h0026731

Oxenham, A. J. (2018). How we hear: The perception and neural coding of sound. Annual Review of Psychology, 69, 27–50. https://doi.org/10.1146/annurev-psych-122216-011635

Turvey, M. T. (1973). On peripheral and central processes in vision: Inferences from an information-processing analysis of masking with patterned stimuli. Psychological Review, 80(1), 1–52. https://doi.org/10.1037/h0033872

Tuomainen, O., Rosen, S., Taschenberger, L., & Hazan, V. (2024). The effects of informational and energetic/modulation masking on the efficiency and ease of speech communication across the lifespan. Speech Communication, 162, 103101. https://doi.org/10.1016/j.specom.2024.103101