Bridging the gap between neurally principled models of choice RT and EEG data

In speeded choice response time experiments, participants are asked to make a series of fast choices in response to a series of stimuli as they are presented (“is the object shown on the screen red or green?”). The last few decades, diffusion models have been adopted as the golden standard for describing these data. As the name suggests, diffusion models propose that the process leading up to the choice is inherently noisy: It is said to be a noisy accumulator of evidence. More recently, researchers started to look for neural correlates of this noisy evidence accumulation during choice RT tasks. However, the precise link between the concept of evidence accumulation and specific neural measurements remains for the most part unclear. In this research, we try to elucidate the link between EEG and a recent generation of diffusion models that have a more solid grounding in neural principles. Although these models are by construction closer to the actual neurological reality and promising in that respect, using them poses considerable statistical and computational challenges. Additionally, EEG data contains contributions of many processes, most of them not related to the actual decision making. To correctly compare a decision model to its supposed EEG fingerprint, it is key to filter out the non-relevant signal. In this research, we will try to address these challenges and build a bridge between these promising neurally principled models of choice RT and actual EEG data.