Boundary extension
We remember having seen more of a scene than was actually shown.
What it means
Boundary extension is the tendency to remember a viewed scene as if it extended beyond the boundaries of the original image, so people falsely recall background and layout that were never visible. The favored explanation is that perception is predictive: the visual system represents a view within its likely larger spatial context, and this extrapolated layout is later misremembered as perceived. The error appears within a fraction of a second, even across a single saccade. It is reliable for close-up, object-centered views, but its direction is not fixed: for wide-angle scenes with small, dispersed objects, memory can instead contract inward, and whether a view extends or contracts tracks its apparent distance and framing. Either way, the finding shows that scene memory is reconstructive and schema-guided at the perceptual level, not merely for semantic content.
How it works
The standard account, the multisource model, holds that perceiving a view is not just registering what is in frame but constructing a spatial model of the surrounding place. Objects cropped by the edge are amodally completed, and surfaces are continued past the border into the layout they imply. Because this extrapolated region is built from the same scene schema that guides perception, the mind never tags it as inferred; moments later it is retrieved as if it had been seen. The error is thus a by-product of a useful function: stitching successive glimpses of a continuous world into one stable representation across eye movements. Predictive-processing versions frame the same idea as the visual system anticipating what probably lies just outside the current view and filling it in ahead of the evidence.
How fast it happens
Boundary extension is not a slow reconstruction that creeps in as a picture fades from memory. Intraub and Dickinson (2008) interrupted a scene with a mask lasting only forty-two milliseconds, less than the time it takes to blink, and observers already remembered the pre-mask view as wider than it was. The effect survives across saccades, the moments when the eyes jump and the world briefly vanishes, which is exactly the interval where a predictive spatial model would be doing its work. This speed is the strongest argument that the distortion begins in perception rather than in later verbal or schematic reconstruction: there is simply no time for deliberate inference. That has made the phenomenon a favoured tool for probing where perception ends and memory begins.
When it reverses to contraction
For decades the effect looked universal, but the stimuli were mostly close-up photographs of a single object. Testing a thousand varied images, Bainbridge and Baker (2020) found boundary contraction, remembering a narrower view, was about as common as extension: close-up, object-centred shots extended, while wide-angle scenes with small, dispersed objects contracted, and the apparent distance to the main object predicted the direction. Gandolfo and colleagues (2023) added that a photograph's depth of field matters, with naturalistic focus yielding more extension. Intraub replies that contraction reflects a separate process, normalization toward a remembered average, not the scene-construction mechanism behind true extension. The theoretical dispute is unresolved, but the practical lesson is settled: whether a remembered view grows or shrinks depends on how the scene was framed.
The hippocampal link
Boundary extension has been tied to the brain's scene-construction machinery. Mullally, Intraub and Maguire (2012) found that amnesic patients with hippocampal damage show markedly reduced extension, and, paradoxically, this leaves their memory for the true image boundaries more accurate than healthy controls', because they add less that was never there. Damage to the ventromedial prefrontal cortex attenuates the effect as well. The interpretation is that the hippocampus builds the surrounding spatial context that intact viewers then misremember as seen, linking the illusion to the same network implicated in imagining novel scenes and future events. This connects a small laboratory memory error to a large claim: that remembering, perceiving and imagining a place may all draw on one constructive, scene-building system rather than on separate faculties.
Examples
Shown a close-up photo of a stop sign, viewers later 'remember' more of the surrounding street than the picture contained.
Back from holiday you are oddly disappointed by your own photo of a cathedral: memory had framed it wider, with more of the square in shot than the lens ever caught.
A witness shown a tight CCTV still later describes the pavement and shopfront on either side with confidence, having filled in a street the camera never recorded.
A designer crops a product photo tight for a thumbnail, then recalls it as roomier and re-crops too wide, restoring margins the original shot never had.
Asked to sketch a remembered film shot, viewers draw the actor smaller and further back, adding stretches of wall and floor that the close framing never showed.
First described in Helene Intraub & Michael Richardson (1989).
Key references
- Hubbard, T. L. (2025). Setting the scene for boundary extension: Methods, findings, connections, and theories. Psychonomic Bulletin & Review, 32(1), 97-138. doi.org/10.3758/s13423-024-02545-w
- Gandolfo, M., Nagele, H., & Peelen, M. V. (2023). Predictive processing of scene layout depends on naturalistic depth of field. Psychological Science, 34(3), 394-405. doi.org/10.1177/09567976221140341
- Bainbridge, W. A., & Baker, C. I. (2020). Boundaries extend and contract in scene memory depending on image properties. Current Biology, 30(3), 537-543.e3. doi.org/10.1016/j.cub.2019.12.004
- Mullally, S. L., Intraub, H., & Maguire, E. A. (2012). Attenuated boundary extension produces a paradoxical memory advantage in amnesic patients. Current Biology, 22(4), 261-268. doi.org/10.1016/j.cub.2012.01.001
- Intraub, H., & Dickinson, C. A. (2008). False memory 1/20th of a second later: What the early onset of boundary extension reveals about perception. Psychological Science, 19(10), 1007-1014. pmc.ncbi.nlm.nih.gov/articles/PMC2792630/
- Intraub, H., & Richardson, M. (1989). Wide-angle memories of close-up scenes. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15(2), 179-187. doi.org/10.1037/0278-7393.15.2.179