Abstract
Memory for a view of a scene tends to include surrounding layout that was never presented (i.e. boundary extension; BE). In previous research, providing prior knowledge that memory for the view's expanse would be tested, did not eliminate BE, but did attenuate it (Intraub & Bodamer, 1993). Why? Prior test knowledge might have caused an increase in fixations to regions near the picture's boundaries, thus enhancing representation of peripheral layout. To test this, we presented 12 close-up photographs for 4 s each (allowing multiple fixations) or 250 ms each (preventing multiple fixations). SOA was held constant at 5 s by interspersing a mask between pictures. Stimulus duration was crossed with test instruction. All observers (N = 144) were instructed to remember objects and backgrounds for a “memory test”; what differed was whether the specific test instruction preceded (test-informed groups) or followed (test-naïve groups) the sequence. At test, the same 12 pictures were presented again and rated on a 5-pt scale (“much-closer [−2]”, “slightly-closer [−1]”, “same [0]”, “slightly-wider [1]” or “much-wider [2]”). As expected, in the 4-s condition BE occurred in both instruction conditions but was greater in the test-naïve group (M = −.39) than the test-informed group (M = −.22). However, this was also the case when duration was only 250 ms (test naïve: M = −.37; test-informed: M = −.28). An ANOVA revealed a main effect of instruction F (1,140) = 5.5, p[[lt]].03, no effect of duration, F[[lt]]1 and no interaction, F[[lt]]1. Surprisingly multiple fixations did not affect memory for the layout's expanse. Given only a single fixation, the representation of layout also expanded beyond its physical limits, and was similarly moderated by prior knowledge of the test. Thus, BE attenuation appears to draw on a global process, rather than an increase in high acuity sampling.