Purchase this article with an account.
Jochen Triesch, Brian T. Sullivan, Mary M. Hayhoe, Dana H. Ballard; Transient visual representations: a change blindness approach. Journal of Vision 2002;2(7):244. https://doi.org/10.1167/2.7.244.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Experiments on change blindness, or the inability to notice scene changes that occur during saccades or other transients, reveal the limited extent of visual representations. It is thought that only the attended parts of a scene are represented and retained in visual memory. The present study reveals that even attended information may not be maintained in memory, and that what is represented at any moment may be only a very limited subset of the visual properties of an attended object that depends on the current task demands. We used a Virtual Reality environment, where subjects pick up and place virtual objects using two haptic force-feedback devices. Subjects moved bricks of two different sizes onto two different conveyor belts, according to different criteria that manipulated the relevance of brick size. Between the pick-up and put-down, brick size sometimes changed by 20%. When subjects pick up the large bricks first and were required to place them on a specific conveyor belt, they were quite likely to notice the change (53%). However, if all blocks had to be placed on the same belt regardless of size, then only 24% of changes were detected. One way to explain the result is that size information might not be maintained after picking up the brick in this condition. In the first condition size remains relevant at put-down, so observers must either maintain this information or recompute it at put down. Finally, when bricks could be placed in any order, even fewer changes were noticed (17%). Sometimes changes went unnoticed even though the subjects visually tracked the object as they moved it toward the belt. In conclusion, subjects often fail to notice changes even when attending to an object if the changed stimulus feature is not task relevant at that moment. These results underscore the dynamic and task specific nature of visual computations, and the fragmented nature of attention.
This PDF is available to Subscribers Only