Manual response times to peripherally presented targets have been reported to be greater in the presence of task-irrelevant pictorial faces at fixation which establish an eye contact with the observer. This effect is interpreted as evidence that direct-gaze faces hold attention. In three experiments, we investigated whether this attention-holding effect is also reflected in saccadic response times. Participants were asked to make a saccade towards a symbolic target that could appear rightwards or leftwards, in the presence of a task-irrelevant centrally placed face with either direct gaze or closed eyes. Unexpectedly, saccadic response times did not show any consistent response pattern as a function of whether the faces were presented with direct gaze vs. closed eyes. Interestingly, saccadic peak velocities were found to be lower in the presence of faces with direct gaze rather than closed eyes (Experiment 1). This effect emerged even in the presence of non-human primate faces (Experiment 2), and no differences between direct gaze and closed eyes emerged when the faces were presented inverted rather than upright (Experiment 3). Overall, these findings suggest that eye contact can have an impact on the saccadic generation system.