Abstract

Comparative cognition is the interdisciplinary study of nonhuman animal cognition. It has been criticized for systematically underattributing sophisticated cognition to nonhuman animals, a problem that I refer to as the underattribution bias. In this paper, I show that philosophical treatments of this bias at the experimental level have emphasized one feature of the experimental-statistical methodology at the expense of neglecting another feature. In order to eliminate this bias, I propose a reformulation of the standard statistical framework in comparative cognition. My proposal identifies and removes a problematic reliance on the value of parsimony in the calibration of the null hypothesis, replacing it with relevant empirical and theoretical information. In so doing, I illustrate how epistemic and non-epistemic values can covertly enter scientific methodology through features of statistical models, potentially biasing the products of scientific research. Broadly construed, this paper calls for increased philosophical attention to the experimental methodology and statistical choices.