Abstract
Imagine you are having your eyesight tested. The optometrist asks you to look at a line on a Snellen chart. The letters look a bit blurry, but you can make out the first as "R." To what extent does your knowledge of what something looks like (in this case, the form of the letter "R") actually help you see it? This stunningly simple question is surprisingly difficult to answer. If people report that they see familiar objects more clearly than less familiar ones, they may be biased to report familiar objects as being easier to see. If people turn out to be better at recognizing more familiar objects, perhaps they are just better at guessing the identity of familiar objects (e.g., guessing a blurry R to be an "R"). To avoid these pitfalls, we used a novel task that tested people's ability to detect contrast changes (blurring and sharpening) of marginally-legible letter strings. Across four experiments (N=90), we measured the difference in people's ability to detect these changes in (a) meaningful vs. scrambled sentences and words (Exps. 1-2), (b) high-frequency vs. low-frequency words (Exp. 3), and (c) the effect of being told what one was looking at (Exp. 4). In all cases, prior knowledge overwhelmingly increased people's judgments of legibility. But in addition, knowledge improved objective performance in detecting visual changes. For example, people's detection of sharpening in real sentences was 80% compared to 70% when viewing scrambled sentences containing all the same letters (z=4.2, p< .0001). Performance was also better for more frequent relative to less frequent words, and when people heard the word prior to seeing it. This novel demonstration of knowledge affecting on-line visual perception provides strong counter-evidence to recent claims that visual perception is not modulated by prior knowledge and expectations.
Meeting abstract presented at VSS 2016