A number of formalisms have been developed to describe the user behavior in human-computer interaction. Two basic approaches can be identified: (1) formal modeling, (2) statistical modeling. In formal modeling, the ideal user behavior is described in the form of a grammar. Several formalisms have been proposed, such as the well-known Backus-Naur Form (BNF), Task Action Grammar (TAG) , Goals/Operators/Selectors/Methods (GOMS) , and others. A more modern (executable) example is SOAR . These formalisms are either conceptual or operational simulation languages. The models can be developed theoretically without experiments when the task goals and the application constraints are sufficiently known. A weak point however, of the formal methods is their inability to handle or describe human error, and the differences in user styles (e.g., perceptual/cognitive/motorical styles). The second group of models is more of a statistical nature, using the post-hoc identification of State Transition Networks, Petri nets, or probabilistic grammar inference. Both method categories (formal vs statistical) are much more suited for symbolical and discrete-action human-computer (HOC) interaction than for describing details of continuous control in HOC interaction, such as the dragging of an object on screen.