Résumé : This article introduces the Complexity of Interaction Sequences model (CIS). CIS describes the structure of interaction techniques and the SimCIS simulator uses these descriptions to predict their performance in the context of an interaction sequence. The model defines the complexity of an interaction technique as a measure of its effectiveness within a given context. We tested CIS to compare three interaction techniques: fixed unimanual palettes, fixed bimanual palettes and toolglasses. The model predicts that the complexity of both palettes depends on interaction sequences, while toolglasses are less context-dependent. CIS also predicts that fixed bimanual palettes outperform the other two techniques. Predictions were tested empirically with a controlled experiment and confirmed the hypotheses. We argue that, in order to be generalizable, experimental comparisons of interaction techniques should include the concept of context sensitivity. CIS is a step in this direction as it helps predict the performance of interaction techniques according to the context of use.