Performance and data from some cognitive models suggested that emotions, experienced during problem solving, should be taken into account. Moreover, it is proposed that the cognitive science approach using both theoretical and experimental data may lead to a better understanding of the phenomena. A closer investigation of ACT-R cognitive architecture (Anderson 1993) revealed some properties analogous to phenomena known from the activation theory of emotion. A model of the classical Yerkes-Dodson experiment was built to test the predictions. The study explained such psychological phenomena as arousal, motivation and confidence within the mathematical notation. The influence of changes in these motivational states, controlled by emotion, on information processing has been investigated and it is shown that the dynamics corresponds to the well-known optimisation methods, such as best-first search and simulated annealing.