Abstract
Serial dependence refers to an assimilation to prior stimuli in current perceptual decisions. This assimilative bias is theorised to occur because the recent past can act as a useful guide to the current state of the world, and so may be used to smooth over meaningless noise in perception. For this process to be truly optimal it should take into account the level of uncertainty associated with stimuli in a way that roughly follows a process of Bayesian inference.In this thesis experiments are carried out to determine if serial dependence is sensitive to uncertainty in a way that might allow for a Bayesian mechanism to dictate the strength of this bias. A series of experiments using orientation stimuli demonstrate the sensitivity of serial dependence to confidence in the prior stimulus judgement, and objective uncertainty in the current stimulus. This is suggested to be because objective uncertainty represents the best available measure of current stimulus uncertainty at the time of a decision. In prior stimuli, confidence represents a better measure of uncertainty as it may be derived from further post-decisional evidence processing. The combined impact of prior confidence and current uncertainty might allow serial dependence to be regulated in a way that conforms to the predictions of a Bayesian model.
Experiments were also carried out to determine whether this bias is the result of a low-level perceptual effect or an attractive effect of prior decisions. These experiments suggest that serial dependence is best quantified as an attraction to the prior response. However, it remains unclear if this attraction to responses represents a bias towards the prior decision or the prior percept. The observed sensitivity of serial dependence to confidence and feedback might suggest a higher-level basis to this effect.
| Date of Award | 3 Oct 2023 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Christopher P Benton (Supervisor) & Casimir J H Ludwig (Supervisor) |