The drift-diffusion model

    Assumptions (for 2-alternative choice tasks)

  • At a given moment, evidence is a random draw from a Gaussian distribution ~ (μ, 1);
  • μ is linearly related to stimulus strength Coh ;
  • Evidence is accumulated over time into a decision variable (DV);
  • There are two decision bounds associated with the two chioces (the distance between the two bounds is a);
  • If DV reaches the upper bound first, choice 1 is made; alternatively, if DV reaches the lower bound first, chioce 2 is made; in either case, accumulation is terminated;
  • The time between the start and end of accumulation is the decision time;
  • Reaction time is the sum of the decision time and non-decision-related processing time (e.g., sensory delay and movement initiation).

The dots task used for simulation

    Main Features

  • Subject looks at a random-dot kinematogram and decides on the global motion of the dots;
  • Subject can indicate the decision at any time;
  • Reaction time (RT) is measured from dots onset to initiation of the decision-indicating action;
  • Motion direction is constant in a trial; the left/right directions are randomly interleaved across trials;
  • Motion strength (expressed as Coh, fraction of dots moving in the same direction) is randomly picked from (0.512, 0.256, 0.128, 0.064, 0.032) for a trial; Motion strength is constant in a trial but the actual value at a given moment fluctuates;
  • Correct choices are rewarded; error trials are not rewarded.


  • Rightward motion is associated with positive Coh;
  • Rightward choice is associated with the upper bound;

© 2017 Long Ding