A theoretical framework for experience-driven critical period effects in learning

Critical period effects (CPEs) in learning refer to the general phenomenon that learning appears to be easier during a particular window of time, often early in life. These windows, or critical periods, are valuable targets for learning interventions and best practices. However, making sense of the vast body of empirical findings is difficult. We lack clear theoretical foundations of critical periods and the effects they generate, which breeds disagreement about how to conceptualize and study them, as well as how to turn empirical findings into pedagogical practice. Here, we develop a computationally-motivated framework for explaining how experience can contribute to the emergence of CPEs. In building our framework, we offer a precise definition, a methodology for identification, and geometric intuitions for the influence of experience on CPEs. We identify three qualitatively different ways in which experience-driven CPEs can emerge, and use simple computational models to illustrate each of these possibilities. With this work, we hope to provide more structure and intuition to the study of CPEs in a way that is useful across neuroscience, psychology, cognitive science, and machine learning.