Someone who is highly trained in distinguishing different faces might still make occasional mistakes when presented with a particular face. How might this be explained by stimulus sampling theory?