Studying robust learning means hypothesizing the point at which when said learning occurs. Unfortunately, there is no way to directly observe these learning events, or points in time / shifts in knowledge representations (?) when students are said to learn.
As researchers, however, we do have the ability to observe instructional events, which are external to the student and involve the presentation of information or elicitation of a behavioral response.
The instructional event, therefore, is used as evidence to infer student learning, or to presume the existence of an unobservable learning event.