Masks That Steal an Object’s Borders Are the Ones That Hide It Best

Why does a flashed object sometimes vanish from awareness even though it has already been processed in high-level visual areas of the brain? A new study published in Attention, Perception, & Psychophysics by Doris E. Dijksterhuis, Nina Vreugdenhil, Pieter R. Roelfsema and Matthew W. Self offers a strikingly simple answer with deep implications: the object disappears when a mask steals ownership of its borders. The team, working at the Netherlands Institute for Neuroscience and the University of Glasgow, designed a series of psychophysical experiments showing that metacontrast masking, a classic form of visual masking known for more than half a century, owes much of its power to a rapid reversal in border-ownership at the boundary between the target and the mask.

Border-ownership, or BO, refers to the assignment of a visual contour to one object rather than another. We live in a three-dimensional world in which nearby objects partially occlude distant ones, and the brain must decide which side of each edge belongs to the foreground. This process, first explored by the Gestalt psychologists in illusions such as Rubin’s face-vase figure, determines whether a patch of the image is perceived as an object or as background. Regions that lose possession of their borders become background, and background regions are recognised more poorly and encoded less strongly into memory. In extreme cases, such as the cigar illusion described in the paper, an object embedded in a brick wall can remain completely invisible because its borders are incorrectly assigned to the surrounding bricks, only becoming visible once an observer is told where to look.

At the neural level, border-ownership is encoded by specialised neurons found in large numbers in mid-tier visual areas such as V2, V3 and V4. These cells respond not merely to the presence of a contour but to which side of it belongs to a figure, a property first demonstrated in the monkey visual cortex by Zhou and colleagues in 2000 and confirmed in many subsequent studies. The authors of the new work propose that sustained, recurrent interactions between these border-ownership cells and neurons in primary visual cortex are essential for building a stable percept of an object. Metacontrast masking, they reasoned, might work precisely because the mask disrupts this recurrent dialogue by flipping the direction of ownership at the target’s borders, instantly demoting the target region from figure to background.

Metacontrast masking occurs when a briefly presented target is followed by a mask whose contours closely abut those of the target, such as two flanking bars surrounding a central target bar. In the standard paradigm, the borders of the black target bar are initially owned by the target, but when the mask appears the ownership direction reverses: the flanking bars take over the shared contour and the former target region becomes background. The researchers noted that this gain-of-ownership by the mask is common to many metacontrast designs, yet had never been isolated as a causal factor. They therefore built new masking paradigms in which the mask either reversed ownership of the target’s borders or left the ownership direction intact, holding everything else as constant as possible.

In Experiment 1a, seven participants judged which of two briefly flashed vertical bars, one on each side of a fixation cross, was longer. Target duration varied from 33 to 133 milliseconds, and the target was always immediately followed by one of two masks. The BO-reverses mask was a traditional pair of flanking black bars. The BO-stable mask contained the same bars but added extra black regions so that a central grey area formed the shape of the letter I, which participants perceived as a coherent object thanks to its central position, convex corners and the familiarity of letter shapes. The psychometric fits were excellent, averaging a coefficient of determination of 0.95, and the slopes did not differ between conditions, suggesting that decision uncertainty and sensory noise were unaffected. Crucially, the target discrimination threshold was 78 milliseconds in the BO-reverses condition versus 67 milliseconds in the BO-stable condition, a statistically significant difference that shows participants needed more viewing time when the mask reversed border ownership.

Experiment 1b refined the design by holding target duration constant at 67 milliseconds while independently varying the stimulus onset asynchrony, or SOA, between target and mask from 0 to 150 milliseconds, with sixteen participants and a larger, more convincing I-shaped mask. Accuracy showed the classic U-shaped relationship with SOA characteristic of Type B masking, with performance dropping sharply at intermediate SOAs in the 30-to-100-millisecond range. Masking was consistently stronger with the BO-reverses mask: the difference between mask types emerged at around 33 milliseconds and became highly significant, after Bonferroni correction, at the 50, 67 and 83 millisecond SOAs. The difference in the fitted minima of individual masking functions was significant across the group, and a generalised linear mixed-effects model including mask type fitted the data far better than a model without it. In six of the fifteen participants the individual effect was itself statistically significant.

The strongest test came in Experiment 2, which confronted a fundamental limitation of the first experiments: the two mask types differed physically in their amount of black area and hence in luminance flux. To eliminate this confound as far as possible, the team exploited the transparent-bar illusion developed by the group of Rudiger von der Heydt. A configuration of light and dark grey rectangles is normally perceived as two overlapping transparent bars, but rounding the inner corners by a few pixels flips the percept into four separate rectangles. The two masks were nearly pixel-identical yet produced opposite border-ownership assignments at the target boundary. The task also changed: participants compared the contrast of a checkerboard target, presented where the mask would appear, with that of a reference checkerboard in the opposite visual field, allowing the researchers to measure perceived contrast directly rather than inferring visibility from a length judgement.

The results were unambiguous. The four-rectangle, BO-reversing mask reduced the perceived contrast of the target significantly more than the transparent, BO-stable mask at the 50, 67 and 83 millisecond SOAs, while the two masks behaved identically at the shortest SOAs. Because the physical contrast of the BO-reversing mask was, if anything, slightly lower due to the rounded corners, energy-based theories of masking would predict the opposite result. A separate Type A style effect, in which perceived contrast dropped at the shortest SOAs regardless of mask type, was attributed to surround suppression, a suppressive interaction between concurrent stimuli that does not depend on perceptual organisation. The masking effect in this experiment was modest, reducing perceived contrast by a few percent rather than rendering the target invisible, but the principle was clearly demonstrated: subtle manipulations of perceptual organisation can substantially weaken a metacontrast mask.

The findings are difficult to reconcile with classical accounts of metacontrast that rely on inhibitory interactions between luminance contours or on transient-onset channels interrupting sustained target activity, since the contours and transients were closely matched across conditions in both experiments. Instead, the results align with theories in which masking disrupts reentrant, feedback processing and figure-ground assignment. The authors propose that a reversing mask rapidly curtails the activity of border-ownership cells that signal the target’s figural status, cutting short the recurrent interactions with early visual areas that are thought to be critical for conscious perception. A stable mask, by contrast, keeps the same population of border-ownership cells active, allowing prolonged recurrent processing and a stronger percept. The framework also shares conceptual ground with object substitution theory, in which a mismatch between reentrant target information and new feedforward mask information degrades perception, but here the mismatch is specified in the well-characterised language of border-ownership circuits.

The broader message is that conscious vision depends not just on detecting contours but on the sustained and consistent assignment of those contours to objects. The timing of the effects fits neatly with known physiology: recurrent signals take roughly 50 milliseconds to return to primary visual cortex from higher areas, and it is precisely in the 30-to-100-millisecond window that border-ownership reversal wreaks its havoc, with the strongest difference between mask types observed at the 50 millisecond SOA where the mask appeared exactly as the target vanished. Earlier work by O’Herron and von der Heydt showed that border-ownership signals can persist briefly after a stimulus disappears and are rapidly reversed by new objects that claim ownership, which may explain why the effect survived interstimulus intervals between target and mask. The authors note open questions, including whether the reversal primarily affects surface properties such as contrast and colour while sparing contour detection, and whether masks that preserve contrast polarity as well as ownership direction would be weaker still. What is already clear is that the brain’s internal model of the visual world must be built and maintained over time, and that the moment a mask steals an object’s borders, the object’s claim on perception collapses.

Subject of Research: The role of border-ownership reversals in metacontrast masking and visual perception

Article Title: Border-ownership reversals determine the strength of metacontrast masking

Article References: Dijksterhuis, D. E., Vreugdenhil, N., Roelfsema, P. R., & Self, M. W. (2026). Border-ownership reversals determine the strength of metacontrast masking. Attention, Perception, & Psychophysics, 88(7), Article 193. https://doi.org/10.3758/s13414-026-03339-z

Image Credits: AI Generated

DOI: 10.3758/s13414-026-03339-z

Keywords: border-ownership, metacontrast masking, visual perception, figure-ground segregation, psychophysics, V2, recurrent processing, visual masking, conscious perception, border-ownership cells, object perception, stimulus onset asynchrony