Purchase this article with an account.
Carlos Cabrera, Zhong-Lin Lu, Barbara Dosher; Separating decision noise and encoding noise in perceptual decision making. Journal of Vision 2011;11(11):805. doi: https://doi.org/10.1167/11.11.805.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Describing the brain's representation of stimuli at various levels of perceptual and cognitive processing remains a challenging task in cognitive research. While Signal Detection Theory (SDT) has provided investigators with measurements of sensitivity and bias at the decision level, it also formally assumes a static decision criterion by absorbing criterion noise into perceptual noise (Wickelgren & Norman, 1966). However, over half a century of research has provided a great deal of evidence contradicting the assumption of a static decision criterion, and correct interpretation of sensory processes requires that criterion noise should be segregated from perceptual noise. A dramatic reassessment of previous research findings based on traditional SDT is suggested by Benjamin et al. (2009) and Rosner & Kochanski (2009). Although extensions of SDT accounting for decision noise have appeared recently, efforts to obtain separate estimates of decision and encoding noise during binary stimulus tasks have relied on significant assumptions about the relationship between various noise components. Here, we develop a new approach based on the “double pass” procedure (Burgess & Colborne, 1988). We first derive and compare representative models capable of independent estimates of the noise components at the decision stage. We then combine a rating procedure with an external noise paradigm employing multiple passes through identical stimulus sets. That is, subjects respond to multiple presentations of identical sets of signal and external noise samples. By providing estimates of observer consistency across passes, these “n-pass” procedures generate sufficiently rich data sets to constrain extended SDT models without making specific assumptions about the relationship between noise components. This combination of theoretical and experimental approaches allows for independent measurements of noise components at several criterion boundaries at the decision stage. Lastly, the analytical framework can also be used to investigate the decision protocol that subjects adopt for a given task.
This PDF is available to Subscribers Only