When two masked targets (T1 and T2) are presented within approximately 500 ms of each other, subjects are often unable to report the second of the two targets (T2) accurately, even though the first has been reported correctly. In contrast, subjects can report T2 accurately when instructed to ignore T1, or when T1 and T2 are separated by more than 500 ms. The above pattern of results has been labelled the attentional blink (AB). Experiments have revealed that the AB is not the result of perceptual, memory or response output limitations. In general, the various theories advanced to account for the AB, although they differ in the specific mechanisms purported to be responsible, assume that allocating attention to T1 leaves less attention for T2, rendering T2 vulnerable to decay or substitution. The present report attempts to bring together these various accounts by proposing a unifying theory. This report also highlights recent attempts to determine if the AB exists across stimulus modalities and points to applications of AB methods in understanding deficits of visual neglect. We conclude by suggesting that investigations of the AB argue in favour of the view that attention may be thought of as a necessary (but not sufficient) condition for enabling consciousness.