I entered the field of psychology around 1968, initially studying behavior, neurophysiology, and perception as an undergraduate at Columbia University. I was first exposed to the study of human thought and language in 1970 – a field then called “cognitive psychology” – after entering the PhD program in the Psychology Department at the University of Pennsylvania.

At that time, the field was dominated by a way of thinking that essentially went like this. Knowledge was a set of propositions (structured lists of items), stored in “long term memory.” Some of these propositions were rules that guided thought; others were statements characterizing properties and relations among items. The result of a cognitive act was the creation of new propositions, through a series of discrete steps. For example, William Chase and Herbert Clark considered how we mentally represent a visual display like this:

The idea was that we would translate the display into a proposition, such as “the star is above the plus.” We could then compare that to the proposition derived from a sentence, such as “the plus is not above the star.”

I couldn't deny that people could construct propositions from looking at displays, but I wasn't completely satisfied with these ideas. I had learned that a neuron fires at a rate that depends on how closely an input corresponds to its preferred stimulus. I had learned that the rate at which a rat or a pigeon would respond to a stimulus was a continuous function of the similarity of the stimulus to previously rewarded stimuli. As I studied cognition, I learned that the probability of success in recognizing a visually presented word was a continuous function of its frequency of occurrence and of its brightness, clarity, and duration.

So I started to ask myself whether it could make sense to think of knowledge and cognition in continuous terms. Although I started down this path based on my own search for a new framework, the eventual result was the collective product of the work of many, some of whom I've been lucky enough to have as collaborators. In what follows, I'll describe the development of three key ideas as I experienced them.

Units and Activations. For the first idea, I worked more independently as I made the transition from graduate student to assistant professor.