During the identification of a target embedded in a rapid serial visual presentation sequence, observers frequently miss a target when it appears early in the sequence as compared to when it appears later. This phenomenon, called attentional awakening, reflects a gradual modulation of temporal attention for a rapid sequence. In this study, we inserted a temporal gap (0, 100, 500, or 1000ms) in the sequence and measured the critical conditions that destroyed the modulated state of temporal attention. The results showed that the correct identifications of the target were impaired immediately after the 1000-ms gap (Experiments 1). However, this impairment was eliminated by superimposing a continuous sequence of random dots so that the sequences temporally separated by the gap could be encapsulated as a single event (Experiment 2). These findings suggested that the temporally modulated attention could be maintained within a single event.