Abstract
In order to relate neural dynamics to cognitive computations, it is critical to determine whether subjects are employing behavioral strategies that actually use those computations, as opposed to simpler heuristics. Here, I describe the problem of adjudicating between degenerate behavioral algorithms and degenerate neural implementations in perceptual and value-based decision-making tasks. Multi-dimensional behavioral measures allow researchers to build arguments from numerous lines of evidence in favor of or against hypotheses about algorithmic strategies. Neural perturbations and individual differences can help resolve degenerate neural implementations. Rigorously contending with these issues is critical for leveraging rodent models to reveal mechanisms of cognition.