‘Sherrington … argued that the nervous system responsible for this simplest class of determinate decision making [harm avoidance] could be … composed of three critical elements: a selective sensory element, … a detector for … events in the outside world…; a selective motor element … that led to the activation of muscles; and a point of contact between these two systems (… the integrative element).’ A sensor fires when it detect something, however, since they are imperfect (both natural or man-made) an additional element is required to collect evidence over time (integrate) to remove errors and noise from the signal – as in a sequential-sampling model.
These models serve as the basis for more complex discriminating behavioural diffusion models, such as directional motion tracking. Consider a juvenile primate watching its mother walk through long grass: as the adult moves rightward it displaces grass, but the juvenile will also see leftward, upward, and downward movement. To tack its mother, the juvenile must discern which of these multiple movements is the primary movement.
Diffusion models add an integration step to collect the relative difference between two options and trigger an action only when the difference exceeds a threshold. Diffusion models have accurately accounted for the observed reaction times and accuracy in experiments compared to accumulator and race models. Diffusion models can also account for the fast erroneous actions which have been experimental observed. Diffusion models using the relative difference implicitly treat evidence for one alternative as evidence against another alternative preventing the errors of multiple conclusions and race-conditions possible in accumulator and counter models which do not have mutual inhibition.