Assortative mating by size is a common mating pattern that can be gene
rated by several different behavioural mechanisms, with different evol
utionary implications. Assortative mating is typically associated with
sexual selection and has been regarded as an attribute of populations
, species, mating systems or even higher order taxa. In most animal gr
oups, however, appropriate analyses of assortative mating at these dif
ferent levels are lacking and the causes and forms of assortative mati
ng are poorly understood. Here, we analyse 45 different population lev
el estimates of assortative mating and non-random mating by size in se
ven confamiliar species of water striders that share a common mating s
ystem. A hierarchical comparative analysis shows that virtually all th
e variance within the clade occurs among samples within species. We th
en employ meta-analysis to estimate the overall strength of assortativ
e mating, to determine the form of assortative mating and to further a
ssess potential differences among species as well as the probable caus
es of assortative mating in this group of insects. We found overall we
ak but highly significant positive assortative mating. We show that an
alyses of the degree of heteroscedasticity in plots of male versus fem
ale size are critical, since the evolutionary implications of 'true' a
nd 'apparent' assortative mating differ widely and conclude that the p
ositive assortative mating observed in water striders was of the 'true
' rather than the 'apparent' form. Further, within samples, mating ind
ividuals were significantly larger than non-mating individuals in both
males and females. All of these non-random mating patterns were consi
stent among species and we conclude that weak positive assortative mat
ing by size is a general characteristic of those water strider species
that share this mating system. We use our results to illustrate the i
mportance of distinguishing between different forms of assortative mat
ing, to discriminate between various behavioural causes of assortative
mating and to assess potential sources of interpopulational variance
in estimates of assortative mating, Finally, we discuss the value of u
sing meta-analytic techniques for detecting overall patterns in multip
le studies of non-random mating.