Abstract
Signal detection theory typically considers a single observer's response to a given signal. However, signals can also be received by multiple observers simultaneously, allowing for aggregate group-level decisions to be made. Recent experiments have shown that the sensitivity of a group of observers making a group-level decision about the presence of a redundant signal tends to exceed that of any individual observer (Bahrami et al., 2010). In these previous experiments, each observer received an identical copy of the stimulus and the group as a whole made a decision about the signal’s presence. Here, we take this approach a step further by asking whether group-level sensitivity is further enhanced by making the signal non-redundant—i.e., by splitting the stimulus up across observers within a group. To test this possibility, 12 groups of 4 subjects were given the task of detecting the presence or absence of a Gaussian bump embedded within Gaussian noise. The stimulus consisted of a square region, with one bump appearing in the center of one of the 4 quadrants of the square. Groups were randomly assigned to one of two conditions: In the ‘full’ condition, all subjects viewed all four quadrants. In the ‘partial’ condition, each subject was shown just a single fixed quadrant. Following each stimulus presentation, the group was allowed to engage in free discussion, after which one individual was randomly chosen to be the final ‘decider’ for the group as a whole. We found that groups in the ‘full’ condition significantly outperformed groups in the ‘partial’ condition, indicating that, in the case of signal detection, sharing redundant information produces better performance than dividing information across the members of a group.
Meeting abstract presented at VSS 2012