A computer implemented method for measuring a persons cognitive load comprises initially receiving 100 stroke data (FIG. 4, FIG. 5(a)) representative of hand-based strokes produced by a person 200 while performing a task. A processor 216 selects 104 a subset of the stroke data FIG. 5(c) that meets one or more predetermined stability criteria. A measure indicative of the persons cognitive load based on the subset of stroke data is determined 106. In this was the users cognitive load in an objective, uniform and non-intrusive manner by analyzing the users writing behavior. An analysis of all of a users writing strokes will bias the evaluation result. The accuracy of the cognitive load measurement is increased by applying stability criteria to select the best strokes for further analysis. By disregarding unstable strokes the computation costs for determining the users cognitive load is also improved.
