In this article, we assess an alternative account of a key experimental pattern thought to index top-down control. The list-wide proportion congruence effect is the well-documented pattern whereby the congruency effect (i.e., Stroop effect) is attenuated in lists containing mostly incongruent trials relative to lists containing mostly congruent trials. This pattern has typically been interpreted as a signature of a top-down control mechanism that modulates attention to the word dimension based on the global probability of encountering conflict between the word and color. However, Schmidt (2013a, 2013b) has proposed an alternative account that stresses relative temporal differences in responding between mostly incongruent and mostly congruent lists rather than relative differences in the control of attention. To assess this temporal learning account, we evaluate the evidence reported by Schmidt (2013a) and report new analyses of three previously published data sets in which a list-wide proportion congruence effect was observed after controlling for other potential confounds. These analyses targeted three key topics: effects of reaction time (RT) transformations, statistical support for temporal learning, and measurement of temporal rhythm. The evidence for the temporal learning account was neither strong nor consistent, and there was a highly significant list-wide proportion congruence effect that survived multiple attempts to control for temporal learning. Accordingly, we conclude that the temporal learning account is not currently a robust alternative to the global control account in explaining list-wide proportion congruence effects.