Abstract
We have previously shown that statistical regularity impacts our ability to see briefly presented masked scenes. That is, scenes that were probable (Greene et al., 2015) or representative (Caddigan et al., 2017) of their category (good exemplars) were more likely to be detected than those that are improbable or less representative of their category (bad exemplars). Here we extend the concept of statistical regularity to familiarity, acquired over the lifetime (famous) and within the experiment (repeated), and to logos and isolated objects. We used the same intact scrambled task in which subjects must discriminate intact images from scrambled images. Famous logos were selected across different categories, such as food chains (McDonald’s, Starbucks, etc.), and technology (Apple, Google, etc). Novel logos were computer-generated logos for the same products. Objects were computer-generated household items rendered in canonical and noncanonical viewpoints. As predicted on the basis of statistical regularity, subjects (n = 26) had higher d’ (for intact vs scrambled) for the famous logos (d’ = 2.40) than computer-generated logos (d’ = 2.22; t(25) = 2.35, p = .027), indicating that statistical regularity influences detectability even for simple highly stylized stimuli. Short-term familiarity (acquired within the context of the experiment) did not affect detectability; repeated logos, whether famous (d’ = 2.43) or not (d’ = 2.16), had equivalent d’s to those presented for the first time in the experiment, d’ = 2.40 and t(25) = -.20, p = .84 for famous logos and d’ = 2.22 and t(25) = .56, p = .58 for novel logos. Similarly, subjects (n = 20) had higher d’ for canonical object viewpoints (d’ = 2.67) than noncanonical (d’ = 2.30; t(19) = 5.05, p < .001). Together, these data support the idea that statistical regularity influences the detection of simple highly stylized logos and isolated objects.