SUMMARYNumbers of cbiasmata in the production of each gamete are closely related to the proportion of gametes which fail to function (r = 081 for 24 species).Gamete redundancy and chiasma frequency fail to correlate with chromosome number, DNA weight/nucleus, or the weight or age of the animal at puberty, as closely as with each other. Implications and corollaries are considered, and it is concluded that most gametes are functionless by-products of a necessarily faulty production system. IN 1967 I suggested that a faulty process associated with chiasmata (crossingovers) might account for the large excess of gametes, particularly sperms. This might have gone undetected by geneticists because all zygotes examined (usually as adults) would show no evidence of such faults, if gametes with these faults never fertilise. So production as well as utilisation errors might lead to wastage of the majority of gametes (Cohen, 1967).Comparison of the logarithm of sperm" redundancy " with the number of chiasmata involved in the production of each sperm, for a variety of species, could test this idea. As chiasma number increases linearly, so the proportion of non-faulty sperms would fall exponentially, and so the total number required for each fertilisation would rise exponentially. In 1967 I had only located mean chiasma frequency (C) and sperm redundancy (R) for seven species. One unpublished chiasma frequency and a preliminary sperm number investigation gave nine points and a correlation coefficient r = 090.Since then I have found many more figures, and the picture they produce is a very significant one ( fig. 1). There are 33 points for 24 animal species, and correlation coefficient r = 0.87 (and Spearman rank correlation is 0.81) if different workers with the same species are considered separately, or 08 1 (and 0.77) if they are considered together as a mean for the species. For 28 degrees of freedom P <000001 (some points are not independent) and for 22 degrees of freedom P<0O001. There is therefore no question that mean chiasma frequency and sperm redundancy are very highly associated.If this association does reflect causality, if a chiasma-associated phenomenon requires multiplicity of sperms because only a fraction" work ", then the regression of Log R on C gives more information. The slope of the association is a measure of the increase of Log R for one cross-over, and suggests that a third (0.29) of chiasmata give unacceptable products. The height of the line, as judged by the Y intercept, suggests that utilisation problems on average require about fifty "good" sperms to be offered for one to fertilise. So for organisms with high C, like man, fig. 1 grains per seedfmean chiasma frequency for Prunus avium. X and Y are total egg production/mean chiasma frequency for Crasso.trea ("R" and C, Longwell and Stiles, 1968) and Mtilus ("R" and C, Ahmed and Sparks, 1970). there would be about a hundred of these in a normal human ejaculate.Correlations do not necessarily reflect direct causality; both R and C might be depende...