Signal and noise in perceptual learning

Performance in perceptual tasks often improves with practice.This effect is known as 'perceptual leaming', and it has been the source of great interest and debate over the course of the last century.Aithough much is known about how percephal leaming changes the properties of cortical circuitry, littie is known about how these changes manifest themselves at the level of behavior.in this thesis, the behaviorai effects of percephiai leaming are considered within the context of signal detection theory.Within a signal detection fnunework, an observer's sensitivity in a percepnial task is defined by the ratio of signai-to-noise within the systern.Thus, accordhg to signal detection theory, the improvements that take place with perceptual leaming can be due to increases in internai signai srrength or decreases in h t e d noise.These quantities m o t be measured drectly.Instead, psychophysical techniques must be used to infer their magnitudes.in this thesis, two psychophysical techniques were used to discriminate between the effects of signal and noise as observes leamed to identay sets of unfamiliar visual patterns.Noise mmkxng was med to me-observers' equivaient input noire and iu c a k u k h ~t efficiencrF quantitb thatcorrespond ta intemal noise and intemal sigpal strength, respectively, within the context of a simple black-box mode1 of the visual system.Equivalent input noise only reflects the effects of an internal noise whose magnitude is independent of the magnitude of the stimulus.Response conslstency was used io estimate the effect of learning on internal noise that depends on the magnitude of the stimulus.Calculation efficiency improved by as much as a factor of four across learning sessions for two very different pattern identification tasks (face and texture ideniification).However.neither form of internai noise changed signifxcantl y with lezrning.These resdts were used to test the prediction that an observer's cdculation should become more similar to the calculation of an ideal discriminator with practice.Response ciassification was w d io estimate observers' linear templates as learning twk place, and showed observer's calculations became significaritly more correlated with the ideal template with practice.Taken together, these results place new theoretical constraints on models of percephial leaming.First and forernost, I wodd ke to thank Patrick J. B e ~e

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.