When there is a single causal factor that is creating measurable differences, usually in a controlled experiment. For example, if we take two large groups of people matched for all relevant background and skills and then give one group a standard interface (system A) and the other group a new interface (system B), the change in interface is a single cause for any differences in the average behaviour.
Having a single cause makes analysis easier, but is difficult to achieve in HCI. This is partly because we often want to perform in the wild research or other forms of more ecologically valid study where we cannot control the context. Even where we do have a relatively controlled situation, such as the system A vs B study, the complex nature of interface design means that there will be many differences between the two, making it difficult to pin down precisely which factors give rise to an observed effect.
Used on page 133