By exposing cells to different fluctuations, and observing their behaviors, we would like to infer cellular mechanisms. To do this systematically, we developed a microfluidics device that allows up to fifty populations to be simultaneously tracked at single-cell resolution for many days under controlled fluctuations. This talk describes our recent experiments on bacterial populations exposed to periodic sugar fluctuations. The ability to obtain detailed lineage data motivates a general theoretical analysis. Populations can be described as collections of individual histories observed over fluctuating conditions. The statistical properties of this ensemble provide a powerful tool to distinguish individual response mechanisms from population-level selection.