Empirically testing the selective influence assumption of the Diffusion Model
Mon-B22-Talk II-01
Presented by: Katja Pollak
Cognitive modeling has been proven useful to explain basic cognitive processes such as decision-making. The drift diffusion model (Ratcliff, 1978) constitutes one such cognitive model that has been shown to add value over traditional behavioral analyses in binary reaction time tasks. Specifically, by postulating that the decision process is characterized by a noisy information accumulation process the diffusion model explains behavior in binary reaction time tasks using four main parameters. These four main parameters each refer to one information-processing aspect of a decision which makes the model valuable for researchers and practitioners alike. However, despite its usefulness in disentangling different cognitive processes underlying decision making in binary reaction time tasks, doubts have recently emerged questioning one assumption of this model. That is, recent studies have shown that the selective influence assumption (SIA), i.e., the assumption that a specific manipulation maps onto only one model parameter, is not always met. We have investigated the SIA on the non-decision time - a parameter assumed to reflect encoding as well as motoric processes - by manipulating the ease of stimulus encoding. Preliminary results speak in favor of the convergent validity of the non-decision time parameter. However, the manipulation did not only affect the non-decision time but also other parameters of the diffusion model (starting point and drift rate), possibly challenging the discriminant validity of these parameters and the SIA.
Keywords: Cognitive modeling, Diffusion Model, non-decision time, validity, stimulus encoding