A study assessed the variable amounts of attention that are required for a viewer to process two kinds of interruptions that commonly occur in television: the shift from one message to a different, unexpected message; and the reference to previously presented material that follows an interruption. Twenty-six subjects recruited from an upper-division course in communication research methods viewed a 30-minute videotape composed of 24 different sequences of programs. Each sequence consisted of an initial segment that introduced one program, an interrupt segment that presented a second program, and a re-orienting segment that presented material from the first program. Attention was measured using reaction times to audio tones that were located 1 second and 6 seconds after the onset of the interrupt and re-orient segments. Results indicated (as predicted) that 30-second initial sequences produced longer response times (more attention) during the interrupt segments. Results also indicated that for the re-orient segments, 30-second initial segments required more attention (contrary to the hypothesis), and 30-second interrupt segments also required more attention (confirming the hypothesis). The results reconciled previously conflicting results of research on attention to television over time, and indicated the importance of defining message units in the cognitive processing of television. Future research should focus on integrating natural aspects of the television stimulus with an experimental agenda that treats television viewing as a cognitive process. (Three figures are included; 25 references are attached.) (PRA)