In research design, retrospective typically refers to that both the predictor and outcome are being collected after both have already occurred, in contrast to prospective data, in which the predictor data is collected initially and used to predict incidence of the outcome over time (so, new diagnoses of diabetes in the follow up period). Even in prospective studies, people are often asked to summarize behavior over some amount of time if self-report is used at all. Even studies using ecological momentary assessment or daily diary, in which people are asked once or multiple times a day about what they are doing, often ask people to report on experiences/behavior over the last period of assessment, not just what is happening right now. What’s different is that the data is used to predict future outcomes, not to try to explain what already happened. So it’s worth noting that just because a study is prospective, it is typically still correlational and not causal, but it at least establishes temporal precedence (i.e., the predictor came first), which is one of several requirements for causality. Experimental research is the only type of research that is causal in nature (and then still, only to the extent that it is done well and rarely definitively so), and many of the questions we are interested in do not lend themselves well to experimental research, because doing randomized controlled trials of lifestyle factors is often not feasible.