Purchase this article with an account.
Mark B. Neider, Gregory J. Zelinsky; Searching for Camouflaged Real-World Objects. Journal of Vision 2004;4(8):335. doi: https://doi.org/10.1167/4.8.335.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Although effects of target-distractor similarity are well known to the search literature, realistic scenes have backgrounds that might also be more or less similar to a target. In three experiments we varied the target-background (T-B) similarity in a search task. In Exp 1 observers (n=20) searched for a target on either a black or a complex background across a wide range of set sizes (9, 19, 29, 39, and 49 items). The target and distracters were randomly selected from a set of 50 real-world objects, chiefly children's toys. The complex background was created by cutting a 20×20 pixel square from the center of the target item and then using this section to tile an 800×600 pixel canvas, resulting in a background sharing characteristics with the search target. RT × set size functions revealed elevated y-intercepts in the complex background conditions compared to the black background conditions, with the largest offsets found in target absent trials. Slopes were unaffected by either the background or set size manipulations. In Exp 2 we reduced the number of set sizes (19, 34, and 49) but varied the T-B similarity by using different target tile sizes (5, 15, 25, and 35 pixels) to create the backgrounds. When the background was very target-like (35-pixel tiles) intercepts were approximately 3,300 ms longer compared to when the background was more homogenous (5-pixel tiles). As in Exp 1, slopes were unaffected by the tile size manipulation. To better understand the effect of tile size found in Exp 2, in Exp 3 we monitored how observers shifted their gaze during search under low (15-pixel tiles) and high (35-pixel tile) T-B similarity conditions. We found a larger proportion of eye fixations to the background as T-B similarity increased, with this effect being clearest in target absent trials. We conclude that camouflaged targets are difficult to find because false targets are created in the target-like background, effectively increasing the search set size.
Army Research Office grant DAAD19-03-1-0039 and NIMH grant R01 MH63748
This PDF is available to Subscribers Only