The discovery that a single gene might account for the RS factor, a gene that I called RS + (Annett, 1978b), was one of the greatest surprises of the RS story. For those who think theories are invented (out of the head of a thinker, or perhaps handed down as inspiration from sources on high) I would point out that this particular development of the RS theory depended on a research process that began some 17 years earlier. Some key steps in the process have been described already. To summarise briefly, my research in this field began with children who had unilateral epileptic foci, and the surprise that one-third of these children were mixed handed (Annett et al., 1961). Finding a similar proportion of mixed-handers in all subsequent samples, I suggested that uncertainties about relationships between handedness and cerebral dominance might be due to variability of hand preference and also variability of hemisphere speech, in this neglected subgroup of the population. My first genetic hypothesis (Annett, 1964) was that lateral asymmetries in mixed-handers might be due to variable expression in heterozygotes (RL genotypes, when R and L represented alleles for right- and left-handedness respectively). It was soon realised (1967) that this version of the classic genetic model could not be true when strictly interpreted (all heterozygotes mixed handed, and all mixed-handers heterozygote although, of course, variable penetrance could have rescued the model). Meanwhile, findings for hand preference and right minus left (R–L) peg moving skill suggested that both preference and skill are continuous variables, reliably related to each other. The binomial proportions of left-, mixed- and right-handedness, in all human samples that had been described in sufficient detail to be tested, and also in nonhuman samples although actual incidences were very different, raised some deep puzzles how these various phenomena could be related. The most difficult and pervasive problem, that incidences of left-handedness in the literature ranged from some 1 to 40%, was solvable if handedness was recognised to be a continuous variable that could be classified at any of a wide range of thresholds. The thresholds required to distinguish consistent left-, mixed-and consistent right-handers in humans, turned out to be identical with those required to distinguish these groups in nonhumans (when estimated at about 25, 50 and 25% under the curve of the normal distribution) if the human distribution was displaced to the right. Thus, it was possible to distinguish the shape of the distribution, continuous and unimodal in all species, and the location of the distribution, shifted to the right in humans only.