Purchase this article with an account.
Filip Dechterenko, Jiri Lukavsky; Remembering overlapping scenes: higher false alarm rates for unseen parts of scenes. Journal of Vision 2017;17(10):556. doi: 10.1167/17.10.556.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Our visual experience is continuous; we do not see individual images and the remembered views overlap. We are able to remember large amount of individual scenes with high accuracy and low false alarm rates (Konkle et. al, 2010). Here, we studied how recognition accuracy changes with repeated presentation of identical vs overlapping scenes. We selected images and divided each into three segments. Our stimuli consisted of two adjacent segments - thus we could present one image twice or present two images sharing 50% of the area. We showed for 220 images each for 3 s. We either presented images once (110 images), as two identical views (55x) or two overlapping views (55x). To assess working memory performance, we asked subjects after each 5 trials whether they have seen an image patch (43% x 43%). The patches were selected from the 1) presented views, 2) unseen views from the remaining image segment, 3) could be distractors. After the experiment, we tested the recognition of the patches from the views seen twice (2x patches), once (1x), from unseen views (0x) and from the new images (distractors). Results (28 subjects) showed that in the working memory test, subjects were able to recognize patches from the presented images (percentage correct: 73%) and identify distractors (false alarm, FA 19%), but they showed high FA rate for unseen parts (41%). For the long term memory, the recognition increased with repeated presentation (2x patches: 69%, 1x patches: 58%). The FA rate was higher (41%). People struggled to identify the patches from unseen parts (FA 49%). The results showed that while people can learn a large number of photographs, they have difficulties when queried about images coming from the unseen part of same scene. The visual and semantic similarity together with photography style are likely factors behind this effect.
Meeting abstract presented at VSS 2017
This PDF is available to Subscribers Only