RT Journal Article SR Electronic T1 Humans treat unreliable filled-in percepts as more real than veridical ones JF bioRxiv FD Cold Spring Harbor Laboratory SP 066217 DO 10.1101/066217 A1 Benedikt V. Ehinger A1 Katja Häusser A1 José Ossandón A1 Peter König YR 2017 UL http://biorxiv.org/content/early/2017/03/26/066217.abstract AB Humans often evaluate sensory signals according to their reliability for optimal decision-making. However, how do we evaluate percepts generated in the absence of direct input that are, therefore, completely unreliable? Here, we utilize the phenomenon of filling-in occurring at the physiological blind-spots to compare partially inferred and veridical percepts. Subjects chose between stimuli that elicit filling-in, and perceptually equivalent ones presented outside the blind-spots, looking for a Gabor stimulus without a small orthogonal inset. In ambiguous conditions, when the stimuli were physically identical and the inset was absent in both, subjects behaved opposite to optimal, preferring the blind-spot stimulus as the better example of a collinear stimulus, even though no relevant veridical information was available. Thus, a percept that is partially inferred is paradoxically considered more reliable than a percept based on external input. In other words: Humans treat filled-in inferred percepts as more real than veridical ones.