The attentional blink paradigm tests attention by overloading it: a list of stimuli is presented very rapidly one after another at the same location on a computer screen, each item overwriting the last, and participants monitor the list using two criteria [e.g. detect the target (red letter) and identify the probe (letter p)]. If the interval between the target and the probe is greater than about 500 ms, both are usually reported correctly, but, when the interval between the target and the probe is within 200-500 ms, report of the probe declines. This decline is the attentional blink, an interval of time when attention is supposedly switching from the first criterion to the second. The attentional blink paradigm should be difficult to perform correctly without vigilantly attending to the rapidly presented list. Vigilance tasks are often used to assess attention-deficit hyperactivity disorder (ADHD). Symptoms of the disorder include hyperactivity and attentional dysfunction; however, some people with ADHD also have difficulty maintaining gaze at a fixed location. We tested 15 adults with ADHD and their age- and sex-matched controls, measuring accuracy and gaze stability during the attentional blink task. ADHD participants reported fewer targets and probes, took longer to recover from the attentional blink, made more eye movements, and made identification errors consistent with non-perception of the letter list. In contrast, errors made by control participants were consistent with guessing (i.e., report of a letter immediately preceding or succeeding the correct letter). Excessive eye movements result in poorer performance for all participants; however, error patterns confirm that the weak performance of ADHD participants may be related to gaze instability as well as to attentional dysfunction.