Gordon H. Bower had a profound influence on my eventual academic fate, well before I actually met him. When I arrived at Stanford in 1970, Gordon was on sabbatical (teaching in Austria, as I recall). One of the other faculty members gave me a preprint of an article he wrote, “Mental Imagery and Associative Learning,” later to appear in the volume edited by Gregg (1972). In that paper, a single line struck me right between the eyes (I vividly remember reading it, late at night in the original Stanford coffee house). He said something to the effect that “If images are like pictures, and can be scanned and otherwise inspected … .” This throwaway thought from Gordon (which he valued so little that he deleted it from the final version) instantly led me to have two ideas. First, I realized thatexactly analogous to the now-famous “mental rotation” experiments of Lynn Cooper, Jackie Metzler, and Roger Shepard (see Shepard & Cooper, 1982)—I could measure how much time people required to respond when they had to scan different distances across a drawing they were visualizing. If Gordon were correct, more time should be required to scan greater distances across an imaged object. Second, I also realized that I could use such response times more generally, as a kind of “mental tape measure,” to assess structural properties of the underlying representation. The thought was that if image representations (the short-term memory representations, which somehow give rise to the experience of “seeing with the mind’s eye”) were in some sense pictorial, then space in the representation should embody actual space. If so, I conjectured, then the time to

scan from one point to another on the image should directly reflect the distance between the points on the corresponding object.