To study the decision process during rating tasks, PC cursor trajectories were recorded and analyzed. The trajectories were often successions of rapid saccadic-like movements that are called strokes in this paper. The analysis of strokes revealed that the distribution of strokes differed across tasks as a function of task difficulty. A simple number matching task elicited fewer strokes, shorter response times, and velocity patterns resembling simple ballistic reaching movements. A personality rating task tended to elicit multiple strokes and longer RTs, which caused a typical inverted-U RT effect. The shape and speed of tangential velocity of trajectories may reflect participantfs internal states, especially when cognitive loads are high.