Importantly, increasing the occurrence frequency of the +0.5° stimulus to 75% of the trials during a testing block (80 trials) leads observers to being more likely to provide the “+” response independent of the stimulus presented, leading to P(+|+) = 0.81 and P(+|−) = 0.44. Here, the sum shows 1.09, slightly above 1, indicating a small, statistically non-significant, bias in favor of the “+” response ( t (6) = 1.59, P = 0.16, two-tailed t test). The bias in the task, which is the preference for responding “+” over “−”, independent of stimulus orientation, can be quantified by comparing the sum of these two conditional probabilities to 1 (see Fig. For example, the “+” stimulus was sometimes reported as having a “−” orientation, showing \( + | - ) = 0.36\) (mean, SEM ≤ 0.04 N = 7 observers). This is a challenging task, in the sense that our observers provided the correct answer on only ~70% of the trials. For example, consider a task involving a fine discrimination between two oriented objects (Gabor patches, σ = 0.42°, λ = 0.3°, see Methods), slightly tilted from vertical, clockwise (CW), or counter-clockwise (CCW) (+0.5° or −0.5°), briefly presented (50 ms). When faced with a difficult visual discrimination task in which one of the objects is more probable, observers tend to choose the more frequent alternative when uncertain. Our experimental results confirm this prediction, and suggest that time-dependent bias is due to temporal accumulation of sensory evidence and noise. The prediction observed and investigated here is that perceptual biases are strong with fast decisions and are much reduced, possibly eliminated, with slow decisions, regardless of stimulus duration. The resulting response bias is expected to decrease with decision time, due to accumulation of evidence and noise. In such integrators, the initial state of accumulation is set by prior evidence favoring (biasing) one decision outcome over others 12, implementing an approximation of Bayes’ rule 13. Theories adhering to this principle, such as drift diffusion models (DDM 10), offer remarkable explanatory power, notably predicting human reaction-time in decision tasks 11, and accounting for neuronal activity in brain regions correlated with decision-making 8. Here, we consider an alternative, single process, account.Ī powerful idea in the neurosciences is that decision makers, brains included, integrate noisy evidence over time to improve performance 8, 9. Experiments show these biases to be reduced with longer exposure duration 5, 6, intuitively explained by a hierarchy of processing levels, with the slower processes, computationally more powerful, providing a more veridical perception 5, 6, 7. In visual perception, contextual effects, and prior experience lead to systematic biases in the judgment of objects’ properties such as orientation, size, and color 1, 2, 3, 4.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |