Subjects typically display superior reproduction of good (redundant, symmetrical) visual patterns compared with poor ones. This pattern goodness effect could conceivably involve encoding processes, short-term memory processes, or response processes. The present experiments explored the time course of wholistic encoding of Garner dot patterns as a function of tachistoscopic exposure time, delay of backward masking, and post-mask shadowing. Within the specific framework of additive factors theory, Experiment I showed: (a) equal rates of encoding for all patterns since comparable slopes were obtained for the recall X processing time functions; and (b) superior absolute recall for good patterns since different intercepts were obtained. Experiment II demonstrated that when degree of encoding was initially equalized for all patterns, the rate of extraction of further information remained constant over available processing time and was unaffected by pattern goodness, slopes and intercepts for good versus poor patterns then being equal. Experiment III confirmed that, given some fixed duration of available processing time, information is abstracted at the same rate for all pattern regardless of the ratio stimulus display time to delay of mask onset. Experiment IV indicated that maintenance rehearsal normally occurs in the present experimental situation, and that very good patterns are somewhat less disrupted by shadowing over a three-second interval. While STM is thus implicated in the pattern goodness effect it does not follow that STM constitutes a complete explanation of the intercept differences reported here. Empirical evidence of response bias toward production of good patterns, however, was not found. It was shown that very good patterns are highly familiar and nameable, and proposed that they do consequently have an early encoding advantage.
Examining the Time Course of Ultra-Rapid Visual Categorization with Backward Masking
