I first met autistic children as a trainee clinical psychologist, and I was captivated for life. I thought them hauntingly mysterious. How could they do jigsaw puzzles straight off, and yet never respond to my simple requests to play with them? What was going on? How could they be tested? Here was a challenge that cried out for basic research.

My mentors, Beate Hermelin and Neil O'Connor, knew how to do elegant experiments with children who hardly had any language and were more than a little wild. I was elated when they offered to supervise me, and I got my dream job in their lab after I finished my PhD. I was hooked on the experimental study of cognitive abilities and disabilities in young children with autism and I wanted to know how they differed from other children. One of the innovations that O'Connor and Hermelin had introduced me to was the mental-age match. They argued that comparing bright and intellectually impaired children would get us nowhere. The brighter would do better, and this told us nothing that we didn't know already. Instead, they compared, say, eight-year-old children who on psychometric tests had a mental age of four, with four-year-old, typically developing children with a mental age of four.

I was proud of one memory experiment I did during my apprenticeship as a PhD student. We observed that autistic children often had a remarkable facility in remembering words by rote. This allowed us to compare autistic and non-autistic children who had the same short-term memory span. What we found gave me a key insight: Typically developing children could remember many more words when these words were presented in the form of sentences than if the same words were presented in a jumbled up fashion, but autistic children failed to show this advantage. I followed up this finding in experiments with binary sequences with clear structure, such as abababab, versus those without, such as aababaaa. The results suggested that structure or “meaning” allowed stimuli to be packaged into bigger units and thereby extended memory span. Did autistic children not see meaning in the way other children did, I wondered? Did meaning not exert the same dynamic force in their information processing?