Abstract
Humans can estimate the global trend in dynamic events from a stream of information. To understand the mechanism underlying such estimation, we have previously examined human performance for judging the temporal average of orientation or motion direction of a Gabor patch whose orientation or phase smoothly fluctuated for several seconds (Sato et al., VSS 2013). Our reverse-correlation analysis revealed that observers' judgments on the temporal average strongly depends on stimulus features immediately before their decision making; a peak-end effect. In the present study, we applied this paradigm to numbers. In our procedure, 20 digits (0 to 9) were presented sequentially for 2 sec. During the presentation, the digits temporally varied pseudo-randomly following a Gaussian distribution with particular mean and variance. Observers were asked to judge whether the arithmetic average of digits was above 5 or below. The results showed that observers equally weigh information of entire stimulus presentation, indicating no peak-end effect. This was neither because digits varied discretely since judgments on the average of discretely changing orientation showed a clear peak-end effect, nor because digits were more complex visual patterns since judgments on the average of dynamic facial expression also showed a clear peak-end effect. These results indicate a possibility that perceptual and numerical estimations of temporal average are subserved by distinct mechanisms.
Meeting abstract presented at VSS 2016