I first heard of connectionism in 1982, when I began studying cognitive psychology. I had read Edward deBono's pioneering work, The Mechanism of Mind, twenty years earlier, and I had found it fascinating, but at that time the term “connectionism” had not yet been invented. When I learned about semantic networks, in which concepts were represented as points connected by links of various sorts, it seemed to me that concepts were much too rich to be described as mere points. Instead, I imagined them as long tangled threads meandering around in several dimensions, and I imagined the links between the concepts as the points where these threads met.

When I described this image to my cognitive psychology professor, Benny Shanon, he said, “That's the new theory everyone is talking about – it's called connectionism.” He had just ordered the brand-new book on the topic, Hinton and Anderson's collection of papers, Parallel Models of Associative Memory, and was waiting for it to arrive. When the book came we spent a lot of time arguing over who should get to read it first. Each of us would take it home for a week or two and try to read a few pages, then give it to the other for the next week or two. On the one hand, the new ideas were fascinating, but on the other, they were very difficult to grasp.

Over the years since then I have read a great many papers on connectionism, but none of them was easy enough to recommend to a beginner.