The earliest data on autism rates is severely biased and greatly underestimates the prevalence of autism. In fact, all intertemporal comparisons of autism rates are biased for a simple but important reason: the definition of autism has changed! (View Highlight)