Rather than investigating the extent to which training can improve performance under experimental conditions ('what could be'), we ask about the origins of expertise as it exists in the world ('what is'). We used the twin method to investigate the genetic and environmental origins of exceptional performance in reading, a skill that is a major focus of educational training in the early school years. Selecting reading experts as the top 5% from a sample of 10,000 12-year-olds twins assessed on a battery of reading tests, three findings stand out. First, we found that genetic factors account for more than half of the difference in performance between expert and normal readers. Second, our results suggest that reading expertise is the quantitative extreme of the same genetic and environmental factors that affect reading performance for normal readers. Third, growing up in the same family and attending the same schools account for less than a fifth of the difference between expert and normal readers. We discuss implications and interpretations ('what is inherited is DNA sequence variation'; 'the abnormal is normal'). Finally, although there is no necessary relationship between 'what is' and 'what could be', the most far-reaching issues about the acquisition of expertise lie at the interface between them ('the nature of nurture: from a passive model of imposed environments to an active model of shaped experience').