Purchase this article with an account.
Anthony C. Sampanes, Bruce Bridgeman; Undetected transformation of one scene into another of the same gist. Journal of Vision 2005;5(8):791. doi: https://doi.org/10.1167/5.8.791.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Change blindness experiments suggest that rather than integrating the information in one visual fixation with the next fixation, only a few attended objects and a general gist are carried over from one fixation to the next. This result leads to a startling prediction: if perception of attended objects can be suppressed, it should be possible to exchange one image for a completely different one of the same gist without an observer noticing the change. We test this prediction with a new change blindness paradigm; rather than alternating back and forth from one image to another in the flicker paradigm, we progressively alter segments of an image until one image is completely changed into another, a ‘progressive transformation paradigm’. To avoid misalignments of parts of the old and the new images we superimpose a checkerboard of white squares on the image, then swap old image segments for new ones within each checker, so that the changed area is completely surrounded by white. It is possible to swap several checkers at a time without subjects detecting the change, with a complete swap in about 12 steps. Each stage is presented for 1 sec, with a 60 msec blank between images. We suppress detection of changes in details with a fixation target on the left of odd-numbered images in the sequence, and on the right in even-numbered images, while monitoring eye movements. We apply this technique with several pairs of same-gist images, such as two different beaches, two different bus stops, two different white cars parked in different lots, etc. In half the trials, the image is not changed. Subjects fail to detect the image swap in about 1/3 of change trials; when detection occurs, it tends to occur late in the sequence. Control change trials, transforming to an image of different gist, are detected at a 98% rate. We conclude that gist is stored as a general description, not as a detailed image.
This PDF is available to Subscribers Only