Predicting temporal errors in complex task environments: A computational and experimental approach

Management in complex environments requires knowledge about temporal contingencies. Expectations about durations enable us to prepare for important events in good time, but also to detect irregularities. Unfortunately, time perception is not invariant. Situational aspects as well as features of the task at hand may dramatically change our sense of time. Particularly under varying workload conditions, temporal distortions may lead to performance errors. A valid and reliable model of time perception must account for these characteristics.Based on the cognitive architecture ACT-R (Anderson et al., 2004), we developed a computational model in line with this requirement. Specific emphasis was placed on mechanisms of coordinative working memory which seem to influence time encoding and perception.The model’s assumptions were tested in three steps. First, the model was applied to account for time distortions ‘a posteriori’. Effects of varying working memory demands reported by Dutke (2005) were replicated and explained by simulations of the model. Second, the model was used for predicting effects ‘a priori’. Augmenting Dutke’s (2005) approach by switching between different degrees of memory demands, predictions of time distortions were derived from the model. These predictions were compared with experimental data. Central assumptions of the model were supported, but there were also some deviations that the model had not captured. Based on the conclusions from the results of the experiment, a second a priori testing addressed temporal expectations in a complex task using a micro-world scenario. The results support the interpretation of the previous experiment and provide new insights for modelling time perception.In summary, our results indicate that coordinative working memory – in contrast to general attention – causes differences in timing performance. This characteristic is captured by our approach. The model we propose heavily relies on mechanisms of working memory and can be applied to explain effects for different time intervals, under a variety of experimental conditions and in different task environments.