There is a growing body of evidence suggesting that many traditional instructional techniques may unnecessarily overload limited working memory and impede learning. Based on cognitive load theory, it was hypothesised that instructional design only takes on a crucial role when there is a high level of interaction between learning elements resulting in those elements having to be simultaneously held in working memory. When there is little, if any, interaction between individual learning elements, then the format of presentation should be inconsequential. These hypotheses were tested using a computer-aided design/computer-aided manufacture (CAD/CAM) package with trainees from a Sydney company. Results showed that when instructions involved high element interactivity, a self-contained manual that physically integrated disparate information and did not require the use of the computer hardware was vastly superior to instructional formats that involved continual interaction with the computer. No differences were found between instructional formats when the learning material entailed low element interactivity. Evidence that these findings were due to cognitive load rather than other factors came from secondary task analysis. In light of these and previous results, suggestions are made for cognitively guided instructional packages.