During copulation, male redback spiders (Latrodectus hasselti: Theridiidae) position themselves above the female's jaws. This apparent male complicity in sexual cannibalism is favored by sexual selection because cannibalized spiders receive two paternity advantages. First, cannibalized males copulated longer and fertilized more eggs than those that survived copulation. Second, females were more likely to reject subsequent suitors after consuming their first mate. These results represent empirical evidence for male copulatory suicide as an adaptive behavior. D a r w i n (1) proposed sexual selection to e x~l a i n the evolution of traits that decrease the probability of survival but give an advantage in the struggle for reproduction. However, there has been resistance to the idea that sexual cannibalism-in which males are consumed by females during copulation-might be adaptive for the consumed male (2). Sexual cannibalism is pen-~, -erally considered to be the result of predatory females overcoming the defenses of weaker males (2), but in theory there are circumstances in which males could benefit from being eaten (3-5) and might therefore facilitate their own consumwtion. Recent behavioral observations have documented apparent male complicity in cannibalism (6, 7). For example, male redback spiders (Latrodectus hasselti) always perform a somersault behavior during sperm transfer, in which the dorsal surface of the abdomen is laced directlv over the female's mouth-;arts and remains there throughout copulation (6. 8). This somersault occurs a few ~, , seconds after the male intromittant organ is inserted and has not been reworted in anv other Latrodectus species (6). Although these reports lend some support to the "male suicide" hypothesis, the adaptive basis of such a sacrifice has never been demonstrated. Here I show that male redback s~i d e r s are sexuallv selected to facilitate skxual cannibalism because it results in two paternity advantages for males that are eaten. These results show that sexual cannibalism need not be the result of a sexual conflict of interest (5) but instead can be an adaptive male strategy.

Male redback spiders twist their abdomens onto the fangs of their mates during copulation and, if cannibalized (65% of matings), increase their paternity relative to males that are not cannibalized. The adaptive male sacrifice hypothesis proposes that this increased reproductive payoff from a single mating outweighs the residual reproductive value of a cannibalized male, because high mortality during mate searching restricts alternative mating opportunities. It has been reported that redback male residual reproductive value is low because males are functionally sterile after one mating-a putative intrinsic constraint that could arguably favor self-sacrifice in the absence of ecological restrictions on multiple mating. However, sterility and self-sacrifice may both arise as aspects of a terminal investment strategy if the probability of multiple mating is sufficiently low. Here I report field data that support the adaptive male sacrifice hypothesis. More than 80% of redback males die without finding a potential mate in nature. Data from two observational field studies and one release experiment suggest that in the absence of cannibalism, male redbacks would expect fewer than one mating opportunity in a lifetime. This expectation was not significantly higher for a large male or one in good condition. A simple quantitative analysis confirms that even if males are assumed to be fertile throughout life, the measured mortality rate during mate search in combination with previously documented paternity benefits of cannibalism is sufficient to ensure that self-sacrifice is adaptive for male redback spiders.

Our understanding of selection in nature stems mainly from whole-season and cross-sectional estimates of selection gradients.These estimates suggest that selection is relatively constant within, but fluctuates between seasons. However, the strength of selection depends on demographics, and because demographics can vary within seasons, there is a gap in our understanding regarding the extent to which seasonal fluctuations in demographics may cause variation in selection. Here we use two populations of the golden orb-web spider (Nephila plumipes) that differ in density to examine how demographics change within a season and whether there are correlated shifts in selection. We demonstrate that there is within-season variation in sex ratio and density at multiple spatial and temporal scales. This variation led to changes in the competitive challenges that males encountered at different times of the season and was correlated with significant variation in selection gradients on male size and weight between sampling periods. We highlight the importance of understanding the biology of the organism under study to correctly determine the relevant scale in which to examine selection. We also argue that studies may underestimate the true variation in selection by averaging values, leading to misinterpretation of the effect of selection on phenotypic evolution. K E Y W O R D S : Demographic variation, Nephila plumipes, selection gradients, sexual selection.The strength and direction of selection pressures acting on heritable traits can predict the evolution of phenotypic distributions (Fisher 1930;Lande and Arnold 1983;Kingsolver et al. 2001). The advent of statistical methods to quantify phenotypic selection (Arnold and Wade 1984a,b;Lande and Arnold 1983;Brodie et al. 1995) has led to a better understanding of how selection shapes phenotypes through time. By using estimates of selection gradients derived from cross-sectional sampling or average fitness (longitudinal estimates) across a breeding season, a number of studies have provided information on variation in the strength of selection on a variety of traits associated with fitness (see Kingsolver et al. 2001 for a review). Comparisons of spatially 4 Current address: