Purchase this article with an account.
Steve Haroz, William Prinzmetal, David Whitney; Serial Search Can Occur in Multiple Feature Dimensions at the Same Time. Journal of Vision 2014;14(10):925. https://doi.org/10.1167/14.10.925.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Some models of attention suggest that it acts as a single spotlight or actively selects only a single feature dimension at a time. However, we propose that inefficient attentional processes can operate simultaneously on multiple features. We ran five subjects in an experiment whose stimuli comprised an 8 x 8 grid of colored shapes. Each trial used 1 to 7 different colors and 1 to 7 different shapes. For half of the trials, a random cell was given a either a unique color or a unique shape. The task was to report whether an oddball was present or absent. The experiment included three blocks that asked subjects to report only an oddball color, only an oddball shape, or an oddball in either feature. For the color oddball task, increasing the number of colors resulted in slower response times (unsurprisingly), whereas the number of shapes had no significant effect on RT (surprisingly). We found the same results for the shape oddball task—adding more task-irrelevant, distracting colors did not affect RT. The implication is that search time is only serial as a function of the number of task-relevant feature variants (number of colors for color oddball / number of shapes for shape oddball). The multi-feature task, where the oddball could be in either feature dimension, produced a more surprising result. We again found no significant effect from the variant count of the non-oddball feature dimension. Only the variant count of the feature dimension with the oddball impacted performance even though subjects did not know which feature that would be. These results imply that two search tasks, which independently operate serially and inefficiently, are happening at the same time without a significant negative performance impact. Search in multiple features may operate as a race model which reports the first oddball detected in either feature.
Meeting abstract presented at VSS 2014
This PDF is available to Subscribers Only