Recombination values and their errors.

Two four-point testcrosses comprising 87,000 tomato plants were grown and the data collected from 28 subgroups. Each subgroup consisted of 2,000 or 5,000 plants and should give a valid estimate of the three recombination values. The 28 values for each interval give more outlyers (23% are outside the 95% limits set by the standard deviation calculated by the binomial formula square root of p q/n) than would be expected by chance. If each subgroup was regarded as the control and the other groups tested against this, then 42% of the time the two subgroups would be significantly different. It is suggested that there are many cases in the literature where this comparison has been made and the significant difference wrongly ascribed to treatment. While the causes of these changes in recombination value are unknown and therefore uncontrollable, they must be anticipated in all such studies. Control and treatment must be replicated enough that chance extreme values will not be attributed to treatment.