Data versus Phenomena

Psychological studies do not give theories their targets directly. They produce data: observations generated by particular people, tasks, instruments, and analysis choices. These observations may be reaction times, brain activations, or questionnaire ratings. From them, researchers infer phenomena: stable patterns that recur across studies and that theories aim to explain, such as the speed-accuracy trade-off, the power law of practice, or mood recovery after a setback. Bogen and Woodward's data-phenomena distinction (1988) marks the gap between these two levels. Phenomena give theories their explanatory targets, while data provide the evidence through which those targets are identified and tested. The current debate is about how these two levels should work together in evaluating theories.

Consider a reaction-time study in which people are 40 ms slower in one condition than another. A phenomenon-based test might ask whether a model can reproduce that mean difference. That is a meaningful question, but it may not be enough. The same average could reflect a small slowing for nearly everyone, a subgroup shifting from fast guesses to deliberate responses, slower evidence accumulation that stretches the right tail of the distribution, or a few lapses that pull the mean upward. These are different psychological mechanisms, even though they produce the same headline result. Looking at the full data structure can therefore reveal distinctions that the summarized phenomenon leaves open.

But fit to data is not the whole answer either. A flexible model might predict the reaction times well by learning who is generally fast, which items are slow, how responses change over the session, and which trials look unusual. That fit is useful, but it does not by itself explain what changed between conditions. The explanation still depends on a theoretical claim about the mechanism, such as response caution, evidence accumulation, strategy use, or attention. The debate, then, is not whether data or phenomena are the right target. It is about how they constrain each other. Phenomena specify what a theory is trying to explain, and data show whether the model captures the structure through which that phenomenon appears. The readings below trace this debate from Bogen and Woodward's distinction to recent work arguing for this integrated view.