How Our Predictions do not Deceive Us: An Investigation of the Illusory Perception of Upside-down Letters

Abstract

: Predictive processing (PP) theories explain how the brain processes visual information. They have been used to explain why people presented with a display of letters have the experience of seeing more letters than they can report. The partial awareness hypothesis (PAH) suggests that predictions about the content enhance relatively poor visual representations letters in such displays, such that the richness is inferred rather than really experienced. Critical support comes from a study purportedly showing that upside-down letters can be illusorily perceived as upright under difficult viewing conditions (de Gardelle et al., 2009). Here, three experiments investigate whether this finding could be due to a response bias instead of illusory perception. In the first experiment, a replication of the original study showing illusory perception shows that upside-down letters are reported as upright as much as not-shown letters. However, this could be explained by only a few letters being represented in iconic memory (IM). The second experiment repeats this finding when more letters are maintained in IM, providing stronger evidence against the illusion effect. The third experiment, a full-power replication of the second, confirms that upside-down letters are not reported as upright any more than not-shown control letters. This suggests that the tendency to report letters as upright is a post-perceptual effect rather than a perceptual illusion, but can still be explained within the PP framework. The rejection of the illusory perception phenomenon revives the debate about whether visual consciousness overflows capacities of report. Keywords: iconic memory, illusion, partial awareness hypothesis, partial report, predictive processing