In a previous short communication on the subject of human assortative mating, we concluded that ‘an open verdict should be recorded concerning the claims and counterclaims surrounding genetic similarity theory, pending more precise theorizing and more analytic investigations’ (Russell & Wells 1994, page 464). Rushton (1995) has disputed the arguments that led to this conclusion. We question his rebuttal of the three points that we originally made, and present some further questions that genetic similarity theory needs to address satisfactorily. Our first point was that if genes produce variation in, for example, height, and in preference for a mate with a particular height, then an individual of a particular height will be chosen by a mate with the relevant preference, and their offspring will tend to inherit both the attribute and the preference for it. Thus, we stated, ‘people will inherit the tendency to find attractive in others the attributes they themselves possess’ (page 463). Rushton does not accept this point, saying that our ‘analysis then stops . . . and is thereby incomplete for it misses the next vital step. Unless the chance configuration is adaptive, it will break up in later generations’ (page 547). As a step towards resolving this issue, we ran a computer simulation. For the sake of simplicity, we assumed that mate choice was confined to preference for a single attribute. The program presumed a population of 200 ‘people’, 100 of each sex. Each person had 20 pairs of chromosomes. The first 10 pairs each carried a ‘gene for height’, which received the value 0 or 1 with a probability of 0·5 at the outset. The second 10 pairs each carried a ‘gene for preferred height’, again taking the value 0 or 1 with a probability of 0·5. We determined the height of each individual simply by summing across the first 10 chromosome pairs. Thus, height varied according to the binomial distribution, taking a value between 0 and 20. Preferred height was determined in the same way, except that it was computed from the genetic material in the last 10 chromosome pairs. The rule for mate choice worked as follows. Each female sequentially chose the male whose height most closely resembled her preferred height, from among the pool of those eligible. Once a male had been chosen, he was swapped to the location of the female, to remove him from consideration by females who had yet to make their choice. When a female found more than one male tying for first place in her affections, she chose the last of those encountered. Thus, the first female had 100 males to choose from, and the last had no choice at all. At this point in the program, the assortative mating coefficient for height was calculated. Then, each couple produced one male and one female offspring. Each offspring’s first chromosome (in each pair) was selected from the father. Whether the gene on a chromosome came from the first or second of each pair of the father’s chromosomes was separately determined at random in each instance. The gene on the second chromosome of each pair was selected in the same way from the mother. Thus, each offspring received a genetic makeup that came equally from the father and the mother, and equiprobably from each of the four grandparents. Although males were able to inherit height preferences, these preferences were not expressed, but could be partly passed on to sons or daughters. After ‘reproduction’, the offspring became the parental generation, and the cycle started again. We ran this simulation for 100 generations, and replicated the whole process 50 times. Assortative mating coefficients were averaged across the 50 replications (Fig. 1). On generation zero, the assortative mating coefficient is just under zero. In 10 generations it climbed to around 0·5 and gradually reached a plateau at around 0·7. There is little evidence for Rushton’s assertion that, in the absence of adaptive mechanisms, ‘the chance configuration . . . will break up in later