Reconsidering the evidence for learning in single cells