To study the temporal dynamics and capacity-limits of attentional selection and encoding, researchers often employ the attentional blink (AB) phenomenon: subjects' impaired ability to report the second of two targets in a rapid serial visual presentation (RSVP) stream that appear within 200–500 ms of one another. The AB has now been the subject of hundreds of scientific investigations, and a variety of different dual-target RSVP paradigms have been employed to study this failure of consciousness. The three most common are those where targets are defined categorically from distractors; those where target definition is based on featural information; and those where there is a set switch between T1 and T2, with the first target typically being featurally defined and T2 requiring a detection or discrimination judgment (probe task). An almost universally held assumption across all AB theories is that these three tasks measure the same deficit; however here, using an individual differences approach, we demonstrate that AB magnitude is only related across categorical and featural tasks. Thus, these paradigms appear to reflect a distinct cognitive limitation from that observed under set-switch conditions.