During object-based sensorimotor tasks, humans look at target locations for subsequent hand actions. These anticipatory eye movements or guiding fixations seem to be necessary for a successful performance. By practicing such a sensorimotor task, humans become faster and perform fewer guiding fixations (Foerster and Schneider, In Prep; Foerster et al. in J Vis 11(7):9:1-16, 2011). We aimed at clarifying whether this decrease in guiding fixations is the cause or effect of faster task completion time. Participants may learn to use less visual input (fewer fixations) allowing shorter completion times. Alternatively, participants may speed up their hand movements (e.g., more efficient motor control) leaving less time for visual intake. The latter would imply that the number of fixations is directly connected to task speed. We investigated the relationship between the number of fixations and task speed in a computerized version of the number connection task (Foerster and Schneider in Ann N Y Acad Sci 2015. doi: 10.1111/nyas.12729 ). Eye movements were recorded while participants clicked in ascending order on nine numbered circles. In 90 learning trials, they clicked the sequence with a constant spatial configuration as fast as possible. In the subsequent experimental phase, they should perform 30 trials again under high-speed instruction and 30 trials under slow-speed instruction. During slow-speed instruction, fixation rates were lower with longer fixation durations and more fixations were performed than during high-speed instruction. The results suggest that the number of fixations depends on both the need for visual intake and task completion time. It seems that the decrease in anticipatory eye movements through sensorimotor learning is at the same time a result and a cause of faster task performance.