Abstract
In 1988 Pelli and Robson developed and produced a letter chart used to determine contrast sensitivity in human subjects. An interesting and unintended phenomenon arises when one gets to the lowest-contrast letters. While at first these letters do not appear to be ‘there’ at all, over an extended amount of time (15–20 seconds) these letters gradually fade into appearance. This extremely long time course is unlike that of any other known mechanism involved in visual recognition. What is the nature of this mechanism and what types of information might this mechanism be using? In a series of identification tasks subjects were presented with Sloan letters at extremely low contrasts and their response (presented letter) was collected. A pilot experiment suggested a probability-summation-like mechanism. Subsequent experiments suggest an alternative scenario. In one experiment, an initial series of trials were used to establish a baseline threshold for relatively rapidly presented (∼2 seconds) letters. After this baseline was established, a series of trials consisting of randomly selected letters whose contrast was in the vicinity of the subjects' threshold were presented for durations of 2–20 seconds. As would be expected, suprathreshold letters performed well above chance. Surprisingly, guessing was also above chance for letters notably below the subjects' threshold. These conditions were so much lower than threshold that, when asked, subjects claimed to see ‘nothing.’ In current research we attempt to uncover the stimulus information necessary for recognition in these impoverished conditions. We will present additional results from experiments designed to investigate feature-integration properties and low-level image properties such as phase information.