Summary. Cognitive load theory suggests that many conventional instructional formats are ineffective as they involve extraneous cognitive activities, which interfere with learning. The split-attention effect provides one example of the consequences of inappropriate cognitive activities caused by poor instructional design. Learners are often forced to split their attention between and mentally integrate disparate sources of information (e.g., text and diagrams) before the instructional material can be rendered intelligible. This preliminary process of mental integration, while an essential precursor to learning, is likely to impose a heavy extraneous cognitive load. Physical integration (e.g., combining text and diagrams) may reduce cognitive load and so facilitate learning. This study reports findings from two experiments investigating the split-attention effect. Using an engineering programming language (Numerical Control programming), the first experiment investigated the possible advantage of physically integrating text and diagrams. In a normal training environment, the integrated instructions group outperformed the conventional group. Experiment 2 was designed to see if the split-attention effect would generalise to an area where mutually referring segments of text are conventionally separated, namely, empirical reports in psychology and education. In a laboratory study, Experiment 2 showed that students in an integrated group spent less time processing instructions yet still outperformed students in a conventional group on test questions. The consequences of these results for cognitive load theory and for instruction design are discussed.