Two experiments examined how TV news viewers divide attention between the audio and video messages of news stories. The experiment tested the “belongingness” hypothesis which asserts that two distinct perceptual stimuli will be attended to as if they were a single stimulus when they appear to belong together. The experiments extended the belongingness hypothesis by manipulating semantic units (i.e., audio and video messages) rather than perceptual units. Auditory-visual redundancy was used to manipulate the belongingness variable. It was hypothesized that dissonant audio and video would be viewed as conveying two different messages, with the result that attentional capacity would be exceeded. Conversely, redundant stories would be viewed as conveying one message, with the results that attentional capacity would not be exceeded. Using secondary task methodology—reaction time tasks in Experiment 1 and memory preloads in Experiment 2—the belongingness hypothesis was supported.