Abstract
Parallel processing is a basic organization principle of the brain. For perceptual decision making, this suggests that multiple decisions can be made simultaneously. Each decision involves that evidence for an external noisy signal is accumulated over time until a criterion is reached. Critically, evidence accumulation is also subject to internal noise which is, however, difficult to measure directly. Here, we investigated internal noise in parallel perceptual decision making using simple detection tasks. We continuously presented audio-visual noise (Gaussian noise band-pass filtered between 262–330 Hz and 200 randomly moving dots, respectively). We asked human participants to indicate the onset of auditory (294 Hz tones) and visual signals (50% of the dots moved coherently) and measured response latencies. We first show that latencies in single decision conditions with only the auditory or only the visual signal were negatively correlated to the recent trial history. Second, this knowledge allowed us to predict the exact latency distribution in a dual decision condition, in which both signals were presented simultaneously so that detection of either signal was sufficient for a correct response (OR coupling). By comparing the empirical distribution with the prediction, we concluded that the internal noise level must have been increased in dual decision conditions. Third, we tested this conclusion introducing a new dual decision condition, in which conjunctions of auditory and visual signals were targets (AND coupling). If our noise hypothesis is correct, the latency distribution in the AND condition is fully predicted by the correlation and noise estimates derived in the OR condition. The empirical distribution followed this prediction closely. Thus, separate decisions are flexibly coupled by AND/OR decision gates depending on task demands. Critically, our findings imply that a significant fraction of the internal noise is produced by decision processes themselves, which sets a fundamental capacity limit for parallel perceptual decision making.
The research leading to these results has received funding from the European Community's Seventh Framework Program FP7/2007-2013 under grant agreement number 214728-2.