Revision of 7 Dec, 1998.
(A modified version of this manuscript will appear in Nature on March
4, 1999)
*Laboratoire de Psychologie Expérimentale,
CNRS, Université René Descartes
28 rue Serpente, 75006 Paris, France
http://nivea.psycho.univ-paris5.fr/
+Cambridge Basic Research, Nissan Research &
Development, Inc
4 Cambridge Center, Cambridge, MA 02142, USA
++Department of Electrical Engineering, McGill
University, Montreal, Quebec, Canada H3A-2A7
When a few small, high contrast shapes are briefly spattered over a picture, like mudsplashes on a car windshield, very large changes can simultaneously be made in the scene without these being noticed. This occurs even when the mudsplashes do not in any way cover or obscure the changes. The phenomenon is important in driving, surveillance or navigation, since it shows that possibly dangerous events occurring in full view may go unnoticed if they coincide with even very small, apparently innocuous disturbances. The phenomenon is also of theoretical importance in understanding the way the brain represents the visual world.
48 picture pairs consisting of an original and a modified picture (each displayed for 3 s) were presented cyclically with an 80 ms duration "mudsplash" (Fig.1) superimposed at the moment of the change. Note there was no disruption in visual continuity at the moment of the change. The observers' task was to press a button as soon as they identified the change. The change could consist in a large object or region of the picture shifting location, changing colour, or appearing or disappearing. The object or region that changed could either be what was termed a "Central Interest" or a "Marginal Interest" element (cf. Fig. 1). The two types of changes were equalised for size and salience and were perfectly visible under normal conditions without the mudsplash.
The results for ten observers show that Central Interest changes are most often detected as soon as they occur, whereas Marginal Interest changes are generally only seen on their second or later occurrences. Furthermore, in 13%-30% of the cases, depending on the type of change, Marginal Interest changes, though in full view, were never detected at all during the viewing period of 40 seconds that was allowed.
A growing literature on "change blindness"1,2 is showing similar results in other natural scene viewing conditions: large changes can be missed when they occur simultaneously with a brief visual disruption caused by an eye movement3,4, a flicker5, a blink6, or a camera cut in a film sequence7.
One part of the explanation for these effects is certainly that attention-grabbing luminance transitions caused all over the visual field by the brief visual disruptions prevent attention from homing in on the location of the change. But the question remains of why, once the extraneous transients have subsided, comparison between the current and previous views of the scene is still impossible.
Here the results for mudsplashes provide an important argument: Because the mudsplashes only provoke a minor disturbance, and in particular, because they do not cover the change location, it cannot be argued that the change blindness they cause is due to masking, or to erasure or re-setting of the information contained in an internal representation of the visual world.
On the contrary, and as confirmed also by the work using the other paradigms, it seems that change blindness occurs because the internal representation of the visual world is rather sparse and essentially only contains Central Interest information. A second experiment confirmed this.
Instead of mudsplashes, a single black-and-white textured rectangle briefly covered the change location at the moment of the change. This method allowed us to draw attention to the change location without giving away the exact nature of the change.
We found that when the change concerned a Central Interest element, observers were generally immediately able to tell what the change was. This shows two things: first it confirms again that the masking rectangle was not somehow wiping out the internal representation. Second, it shows that observers had coded the content of Central Interest elements.
When the change concerned a Marginal Interest element, observers were often unable to determine what it was. This shows that observers had not coded the content of the Marginal Interest elements.
The conclusion seems to be that humans' internal representation of the visual field is much sparser than the subjective experience of "seeing" suggests. Only the parts of the environment that observers attend to and encode as "interesting" are available for making comparisons. Conclusions similar to this have been incipient in the classic literature on sensory visual memory8 over the last decades. But these older experiments used simple, artificial stimuli, presented for only fractions of a second, whereas the present demonstrations involve natural visual scenes presented for several seconds, and involve very large changes. For this reason the new studies are particularly poignant demonstration of the paucity of our internal visual representations.
If only attended parts of the environment are represented in the brain, how can we have the impression of such richness and completeness in the visual world outside us? The answer may be that the visual world acts as an external memory.9 We have the impression of simultaneously seeing everything because any portion of the visual field that awakens our interest is immediately available for scrutiny via an unconscious flick of the eye or of attention. On the other hand, those parts of the scene which are not being currently processed (and in some sense not "seen") nevertheless constitute a background or setting, that, like the supporting cast of a film, vivifies our visual experience.
The idea of the world as an outside memory is receiving attention from scientists and philosophers interested in the problem of perceptual "filling in"10 of visual scotomas and illusory contours. Work in robotics using the concept of "active" vision is also adopting the notion that using the outside world to represent information may be more efficient than making an internal copy.
J. Kevin O'Regan*, Ronald A. Rensink+, James J. Clark++
*Laboratoire de Psychologie Expérimentale, CNRS, Université René Descartes
28 rue Serpente, 75006 Paris, France
http://nivea.psycho.univ-paris5.fr/
+Cambridge Basic Research, Nissan Research & Development, Inc
4 Cambridge Center, Cambridge, MA 02142, USA
++Department of Electrical Engineering, McGill
University, Montreal, Quebec, Canada H3A-2A7
1. Simons, D. J. & Levin, D. T. Change Blindness. Trends in Cognitive Sciences 1, 261-267 (1997).
2. Intraub, H. The representation of visual scenes. Trends in Cognitive Sciences 1, 217-221. (1997).
3. McConkie, G. W. & Currie, C. B. Visual stability across saccades while viewing complex pictures. Journal of Experimental Psychology: Human Perception & Performance 22, 563-581 (1996).
4. Hayhoe, M. M., Bensinger, D. G. & Ballard, D. H. Task constraints in visual working memory. Vision Research 38, 125-137 (1998).
5. Rensink, R. A., O'Regan, J. K. & Clark, J. J. To see or not to see: The need for attention to perceive changes in scenes. Psychological Science8, 368-373 (1997).
6. O'Regan, J. K., Deubel, H., Clark, J. J. & Rensink, R. A. Picture changes during blinks: looking without seeing and seeing without looking. Visual Cognition in press (1999).
7. Levin, D. T. & Simons, D. J. Failure to detect changes to attended objects in motion pictures. Psychonomic Bulletin and Review, 4, 501-506 (1997).
8. Haber, R. N. The impending demise of the icon: A critique of the concept of iconic storage in visual information processing. Behavioural and Brain Sciences 6, 1-54 (1983).
9. O'Regan, J. K. Solving the "real" mysteries of visual perception: The world as an outside memory. Canadian Journal of Psychology 46, 461-488 (1992).
10. Pessoa, L., Thompson, E. & Noë, A. Finding about
about filling in: a guide to perceptual completion for visual science and
the philosophy of mind.
Behavioral and brain sciences 21,
6, 723-802 (1998).