Mission: Impossible? Modelling the Verbal Estimation of Duration
Three participants produced a large number of verbal estimates of tone durations in the range of 77–1183 ms. Data from this task were simulated by an ‘attractor model’, which used the idea of competition between ‘attractors’ (‘quantized’ values output as verbal estimates) which differed in weight, and distance from the stimulus duration to be estimated. To produce an estimate, all attractors competed for priority as output values, with the final value being decided probabilistically. The model embodied underlying scalar representations of time, in the form of mean accuracy and constant coefficient of variation. The model was able to reconcile such scalar properties of time with deviations from scalar properties often found in verbal estimation data, such as declining coefficients of variation with increasing duration value. The model furthermore showed that multiplicative and additive changes in underlying time representations should be translated veridically into behaviour, although the attractor competition process could distort patterns and absolute values of underlying variance.