Abstract
There is now a substantial literature documenting the ability to use strong-inference theory-based experimental methodologies to determine the fundamental characteristics of human information processing. The meta-theory known as systems factorial technology (SFT, Townsend & Nozawa, 1995), applied by way of an experimental task known as the double-factorial paradigm (DFP), has allowed for insights in tasks ranging from simple detection to complex face processing. However, much less is known about the effects of factors such as shifts in decisional bias on the stability and interpretablity of results. We present an investigation in which we induced shifts in decisional bias within observers. Participants performed a DFP task, based on the Hering illusion; stimuli were composed of two sets of vertical lines positioned equidistant from center, from which projected radial lines. Participants gave a positive response if they judged either or both sets of vertical lines to be curved outward, otherwise giving a negative response. Participants learned and performed the task with neutral payoffs; half of the participants were then switched to a positive and half were switched to a negative bias condition. Participants shifted to a positive bias showed reliable decreases in mean reaction times (RTs), increases in false alarm (FA) rates, and negative shifts in measures of response bias. All of these participants showed evidence for parallel self-terminating processing before and after the shift. Participants who were shifted to a negative response bias showed reliable increases in mean RTs, decreases in hit and FA rates, and positive shifts in response bias. All of these participants showed evidence for parallel self-terminating processing before the shift; there was evidence for both parallel exhaustive and serial processing after. The results suggest the robustness of the SFT/DFP approach and motivate the need to further develop theoretical models that can incorporate both RT- and accuracy-based metrics.
Meeting abstract presented at VSS 2016