Forgetting of stimulus attributes: methodological implications for assessing associative phenomena.