FEAT: Fixation control by Evidence Accumulation to Threshold

Models of eye movement control differ in the extent to which fixation duration is controlled directly by ongoing cognitive processing. Models developed in Potsdam contain at their heart a random timer, which accumulates "clock ticks" to a particular threshold. The properties of this timer are under limited control by ongoing cognitive processing. Models in the decision making literature are often based on the idea that evidence accumulates to a threshold. Applied to saccadic decisions, this class of models embodies a much stronger direct control assumption, because processing difficulty maps directly on to the rate of evidence accumulation. Both model classes then have an accumulation-to-threshold mechanism, but make different assumptions about the way processing difficulty affects this accumulation process. We are trying to distinguish between these model classes, using a large data set from a sequential scanning task in which processing difficulty can increase and decrease unexpectedly. The data we have are extremely challenging for any model to explain. I will show that for an evidence accumulation model, a sudden change in processing difficulty (either increasing or decreasing difficulty) requires a transient "interrupt signal" that slows evidence accumulation down briefly. This mechanism is somewhat similar to foveal inhibition in the mixed control models developed by the Potsdam group, but is triggered under different circumstances. Moreover, this mechanism interacts with changes in control settings over longer timescales in order to account for the complex pattern of adaptation in fixation duration over the course of several fixations.

Invited by Hans Trukenbrod und Ralf Engbert