Video segments may be characterized by formal design features with respect to factors such as complexity of narration, mutual influence of picture and sound, use of superimposed texts, information load due to technical terms, and animation. The paper suggests ways to operationalize these factors and reports about an experiment in which the influence of these formal features was studied with respect to perceived information load of video segments by learners. The length of a segment was thereby counted as the number of information elements it contains. An information element is defined as one uninterrupted statement of the narrator about which one factual question can be presented. The experiment was carried out by using an experimental videodisc programme about cheesemaking that contains 252 information elements which form a connected discourse of 36 min if the programme is played linearly without stopping. Subjects had the task to divide this programme into segments by deciding for themselves how often they would stop before completing the programme. At every stopping place, the subjects were questioned about the content of the just completed segment. The results of the experiment suggest that: (a) subjects tend to perceive narrated sentences as one whole, regardless of the number of facts implied, (b) content-related technical terms seem not to determine difficulty level, (c) superimposed texts show a trend to help subjects who prefer short segments to choose longer segments and subjects who prefer long segments to choose shorter segments, (d) animation is positively (but moderately) correlated with mean self-chosen segment length.