Evo/utwn, 42(4), 1988, pp. 736-749

FIELD MEASUREMENTS OF NATURAL AND SEXUAL SELECTION IN THE FUNGUS BEETLE, BOLITOTHERUS CORNUTUS

JEFFREY CONNER 1 Section of Neurobiology and Behavior, Cornell University, Ithaca, NY 14853

Abstract. -Selection on three phenotypic traits was estimated in a natural population of a fungus beetle, Bolitotherus cornutus. Lifetime fitness of a group of males in this population was estimated, and partitioned into five components: lifespan, attendance at the mating area, number of females courted, number of copulations attempted, and number of females inseminated. Three phenotypic characters were measured-elytral length, horn length, and weight; there were strong positive correlations among the three characters. Selection was estimated by regressing each component of fitness on the phenotypic traits. Of the three traits, only horn length was under significant direct selection. This selection was for longer horns and was due mainly to differences in lifespan and access to females. The positive selection on horn length combined with the positive correlations between horn length and the other two characters resulted in positive total selection on all three characters.

Received November 10, 1986. Accepted February 5, 1988

An important goal of current evolution­ selection for an increase in the value of char­ ary biology is to study selection occurring acter x, then an individual with a high value in natural populations. Many studies have of x will tend to have high fitness. If char­ demonstrated selection in nature, but very acter y is positively correlated with char­ few of these studies have measured lifetime acter x, then that same individual will have fitness, separated direct selection acting on a high value of y, thus creating indirect se­ a character from indirect selection on cor­ lection for an increase in y. The total selec­ related characters, or identified selective tion acting on a character is the sum of direct agents (Endler, 1986). The first shortcoming and indirect selection. can only be overcome by long-term field Second, identifying selective agents can studies; in animals, lifetime reproductive be facilitated by dividing lifetime fitness into success of individuals has just begun to be biologically meaningful components (e.g., estimated (Clutton-Brock et al., 1982; lifespan, access to females, insemination Fincke, 1982; Banks and Thompson, 1985; success) that can be more easily linked to Hafemik and Garrison, 1986; Koenig and the agents responsible for fitness differences Albano, 1987). Two recent theoretical ad­ (e.g., predation, male competition, female vances make the second and third problems choice) (Arnold and Wade, 1984a; Koenig easier to resolve. and Albano, 1986). Thus, by estimating as First, the methods of Lande and Arnold many components of fitness as possible over (1983) allow one to estimate selection acting the lifetime of an organism and relating on several correlated traits simultaneously, variation in these components of fitness to separating direct from indirect selection. several different characters simultaneously, Direct selection for an increase in character the specific agents that cause direct selection value (i.e., magnitude of a quantitative on individual target characters can often be character) results from a positive relation­ identified. ship between that character and fitness, when In the study reported here, I estimated other correlated characters are held con­ direct and total selection on three pheno­ stant. A character under direct selection can typic traits during the adult lifetime of male be called the "target" of selection (Price, Bolitotherus cornutus. B. cornutus is a te­ 1984; Price et al., 1984). Indirect selection nebrionid beetle that lives throughout its can be explained as follows. If there is direct life cycle on polypore shelffungi (principally Ganoderma applanatum). Eggs are laid sin­ gly on the surface of the fungi, and larvae 1 Current address: Department of Entomology, Cor­ feed by tunneling through the interior (Liles, nell University, Ithaca, NY 14853. 1956). After pupation (about three months 736 NATURAL AND SEXUAL SELECTION IN A BEETLE 737

after oviposition), adults emerge and are ac­ mate (Conner, 1987). Thus, female coop­ tive at night on the surface of the fungi, eration is necessary for mating to occur, be­ spending the daylight hours behind bark or cause males cannot force entry. After a suc­ in holes in the fungi. Eggs are laid from late cessful mating, a male guards the female for May through August. Adults may emerge two to five hours; this distinct behavior is late in the summer from eggs laid in the a reliable indicator of successful mating and spring, but most larvae overwinter and spermatophore transfer (Conner, 1987). emerge as adults the next year (Liles, 1956). A large natural population of B. cornutus Adults can live for several years (Pace, 1967). near Ithaca, NY, was studied for two mating Individual beetles do not come out on the seasons (1984 and 1985). This species surface of the fungi (where virtually all mat­ proved to be an excellent subject for studies ings occur) every night; thus only a subset of selection, since all of the biologically rel­ of males are active on any given evening. evant components of adult male mating The remainder spend the night behind the success were measurable (see below). Since bark or in holes in the fungi. Males appar­ growth does not occur during the adult stage ently come out on the surface of the fungi of B. cornutus, there were no complications solely for mating, since they cannot feed on due to correlations between the characters the upper surface of the fungi and do noth­ measured and age. One potential problem ing on the surface other than mating and was that generations overlap in this species. aggressive behavior (pers. observ.). Male Cross-sectional (short-term) studies on pop­ B. cornutus possess pronotal horns, which ulations with overlapping generations that vary allometrically with body size (Brown are increasing or decreasing in size can lead and Siegfried, 1983). The horns are used in to erroneous estimates of the magnitude and aggressive encounters between males. Fe­ even the direction of selection (Travis and males lack horns and show little aggressive Heinrich, 1986). In the present study, how­ behavior. ever, fitness was measured over all or most Courtship begins when a male climbs onto of the adult lifespan, and the population size the back of a female. Females appear to was stable throughout the study (see below). have little control over which males court them. A male's courtship frequency is de­ MATERIALS AND METHODS termined at least in part by male competi­ Field Observations tion, since males often aggressively chase The study population of Bolitotherus cor­ each other away from fungi containing fe­ nutus was located on a dead tree with 15 males (pers. observ.). After ten minutes to living fungi (Ganoderma applanatum). Bee­ several hours of courtship, the male at­ tles were collected when they were first seen tempts to mate. The female has very little and taken to the laboratory for marking and control over this stage, since she is rarely measuring; every adult beetle in the popu­ able to prevent a courting male from at­ lation was marked. The lengths of the tempting to copulate. The number of at­ horns and elytra (wing covers) of all males tempts a male makes is determined by at in this population were measured (to ±0.2 least two factors: male competition (since mm) with an ocular grid on a stereo micro­ males disrupt the courtships of other males scope. All beetles were also weighed (to ±0.2 before the courting male can attempt to mg) with a Mettler® balance and given an mate) and, possibly, male choice (since some individual paint mark on the elytra (with courtships are abandoned by males without Testors ® white enamel). This treatment did an attempt [pers. observ.; Pace, 1967; Brown not seem to disturb the beetles, and 90% of et al., 1985]). marked beetles were observed at least once While females probably have little con­ after release. trol over which males attempt to copulate The population was first marked in late with them, they have complete control over July 1983. To estimate lifespans, the beetles which males successfully mate and transfer were censused from one to three times per a spermatophore. Females have a heavy night on a total of 190 nights. Censuses were plate at the abdominal tip that acts as a trap performed on 3 5 nights from 22 July to 21 door; it must be opened before a male can September 1983, 77 nights between 19 May 738 JEFFREY CONNER and 28 August 1984, 59 nights between 16 A total of 134 males were present during May and 27 August 1985, and 19 nights the 1984-1985 observation period; thus, the between 27 May and 28 July 1986. study cohort represents half of the beetles To estimate mating success, the popula­ present. The other 67 males were first ob­ tion was observed for approximately 1,000 served either before 25 September 1983 or hours on 136 nights throughout the 1984 after 6 June 1985. The direct-selection anal­ and 1985 mating seasons (the same nights yses were performed separately on the co­ on which censuses were conducted in those hort and on all 134 males; both sets of anal­ two years; see above). Since the summer of yses are presented here. The population size 1985 was colder than 1984 (mean nightly was stable from 1984 to 1985; of the 134 low temperature 10.9°C vs. 12.3°C, t = 2. l, total males, 94 were present in 1984 and 93 P < 0.025) and beetle activity was sharply in 1985, with 53 males present in both sea­ reduced at temperatures below 10°C (pers. sons. observ.), the beetles were observed on fewer In 1985, migration between trees in the nights in 19 8 5. Observations were con­ study population was measured by locating ducted on about 90% of the nights when all other trees (N = 17) supporting G. ap­ weather conditions were suitable for beetle planatum in the approximately 5-hectare activity. Most observations were made dur­ patch of woods containing the main study ing the time when the beetles were most tree. These 17 trees were approximately 200- active, between midnight and 7:00 A.M. A 400 meters from the main study tree. All headlamp, equipped with an infrared filter beetles on these trees were measured and (Kodak Wratten #89B) that transmits some marked as above; a total of 7 5 males were far-red visible light, was used for illumi­ marked at the 17 trees. These trees were nation. The beetles remained active all night censused every night that observations were under illumination from this light, in con­ conducted during the first two-thirds of the trast to the complete disruption of b.eetle 1985 mating season. During the last third behavior that was caused by a less filtered of the 1985 season, the four trees closest to red light. All courtship, copulation, and male the main study tree (200-250 m away) were fighting behavior was recorded. Matings did censused three times per night, while the not occur behind the bark or in holes in the other trees were censused twice weekly. fungi (where the beetles that were not on the fungal surface were), because there was Components of Fitness in Male not enough room for mating to take place B. comutus in these areas (this was verified by obser­ Lifetime male mating success in B. cor­ vations using a fiber-optics viewer). nutus was partitioned into five components. Selection was estimated using the 67 males The first, adult lifespan (W 1), was estimated that first appeared on the tree between 25 as the number of nights between the first September 1983 and 6 June 1985. Thus, the and last sightings of a beetle, not including study animals made up a "cohort" of bee­ nights on which observations were not con­ tles that emerged after most mating activity ducted. Since the beetles were observed on was over in 1983 and before the onset of about 90% of the nights when they were most mating in 1985. The beetles from late active, this method of estimating lifespan 1983 and early 1985 were included because reflects the relative amount of time avail­ a beetle that emerges as an adult in the late able to the beetles for reproduction. fall and overwinters is not biologically dif­ Three different estimates oflifespan were ferent from one that overwinters as a last­ made. The first estimate was lifespan during instar larva and emerges in the spring; they the 1984 and 1985 seasons only (the symbol hatch from eggs that were laid at about the for this estimate is W 1). This was not the same time, and their reproductive lives start best estimate of individual lifespans that in the spring. Indeed, a beetle that is first could be made, however, because 17 of the seen in the spring may have emerged as an 67 males (25%) were still alive at the begin­ adult in the fall and stayed inside the fungus ning of 1986, after the detailed observations until spring. Six beetles from 1983 and 13 made in the 1984 and 1985 seasons had from 1985 were included in the cohort. ended. The measured lifespans of these NATURAL AND SEXUAL SELECTION IN A BEETLE 739 males were truncated, because the obser­ pletely by the female, this component of vations ended before the males died. There­ fitness may be determined by female choice. fore, the second lifespan estimate ( W 1 ') was It is possible, however, that males can sense made using the additional census data from when a female is receptive (Conner, 1987); 1986. In the partitioning of fitness into com­ thus, this component of fitness could also ponents, W 1 was used instead of Wi', since be determined partly by male competition the fitness components other than lifespan for access to receptive females. were not measured in 1986. Of the 17 males Notice that each component is a prereq­ that were alive in 1986, eight (12%) were uisite for the next; a male must be alive still alive at the end ofJuly 1986, when the before he can attend the mating area, he population censuses were terminated. For must be in attendance before he can court this reason, the third lifespan estimate was a female, he must court before he can at­ made. The true mean lifespan of the entire tempt to copulate, and he must make an cohort was estimated using the LIFETEST attempt ifhe is to inseminate a female suc­ procedure (SAS Institute, 1982), which took cessfully. Also note that these components into account the fact that the lifespans for naturally separate the biologically relevant eight of the males were truncated. LIFE­ components oflifetime mating success, e.g.,

TEST cannot provide lifespan estimates for longevity ( W 1), attendance at the mating area individuals, so this estimate was used only (W 2), access to mates due to male compe­ to see how well the other two estimates rep­ tition (W 3, W4 , and [possibly] W 5), and fe­ resented the true mean lifespan of the co­ male choice of mates ( W 5). As required by hort. Arnold and Wade's (1984a) model, the The second fitness component is atten­ product of these components equals total dance at the mating area ( W2), which was fitness (Table 1). The denominator of each estimated as the number of nights a beetle component is the numerator of the com­ was on the fungi divided by its estimated ponent before it; thus, when the compo­ lifespan, i.e., the percentage of nights during nents are multiplied, all the elements cancel a male's life that it spent in the mating area. except number of females inseminated, Since there was no apparent reason for the which I have defined as total fitness. males to be out on the fungi other than mat­ Clearly, this omits the last component of ing, this is a component of sexual selection adult male fitness, which is the number of and is the first component of male access to eggs that are fathered per female insemi­ females. nated. This component of fitness could not The third fitness component is the num­ be directly estimated; however, related ber offemales courted (W 3). This is the sec­ studies of B. cornutus suggest that the num­ ond component of male access to females, ber of females inseminated is probably a and it was estimated by dividing the num­ good estimate of the number of eggs fa­ ber of females courted by attendance, which thered (Conner, 1987). Female B. cornutus equalled the average number of females were inseminated about every nine days on courted per night for each male. average in this population, whereas the mean The fourth fitness component is the num­ interval between ovipositions in the labo­ ber of copulations attempted by a male (W 4). ratory was about three days. Thus, females This is the third component of male access tend to lay at least one egg per insemination. to females, and it was estimated as the num­ The last male to inseminate a female usually ber of copulations attempted divided by the fathers most of the subsequent offspring number of females courted; in other words, (Conner, 1987). Therefore, an insemination this component is the proportion of a male's is very likely to lead to offspring, and the courtships that end in a mating attempt. number of females inseminated should be The fifth and final fitness component is related to the number of eggs fathered. There the number of females a male inseminates are several sources of variability in how

(W 5), which was estimated as the propor­ many eggs a male sires per insemination, tion of male mating attempts that result in however; the possible effects of this vari­ transfer of a spermatophore. Since the suc­ ability on the estimates of selection are dis­ cess ofa mating attempt is controlled com- cussed below. 740 JEFFREY CONNER

components. Five of the males had zero fit­ Measures of Selection ness in the third component (i.e., they court­ The methods employed in this study es­ ed no females), and nine had zero fitness in timate selection on phenotypes without the fourth component (i.e., they did not at­ considering inheritance; that is, they sepa­ tempt to copulate). Selection gradients for rate selection occurring within a generation total fitness were calculated by regressing from the evolutionary or genetic response relative total fitness (number of females in­ to selection across generations. Also, only seminated) against the three phenotypic directional selection was estimated in this characters for all 67 males. Since the gra­ study. dients for each component were calculated Selection Gradients. -A selection gradi­ with separate multiple regressions and the ent (fl) measures the relationship between gradients for the last two components do fitness and an individual character, holding not include all 67 males, the component other correlated characters constant (Lande gradients do not necessarily sum to the total and Arnold, 1983). Selection gradients selection gradient (see Arnold and Wade therefore estimate the direct selection on [1984a], Kalisz [1986], and Koenig and Al­ each character after the effects of indirect bano [1987] for discussions of this issue). selection on all correlated characters have One problem with this approach is that, been removed (assuming that all correlated if the correlations between the x variables characters have been measured). Selection are large, it is difficult to demonstrate the gradients presented in this paper are the significance of any of the selection gradients. standardized partial regression coefficients In the formula for the standard error of a for each character, which were generated in partial regression coefficient (the selection a multiple regression of relative fitness on gradient here), the squared correlations the three phenotypic characters measured among the x variables are subtracted from (Lande and Arnold, 1983). (Relative fitness the denominator (C. McCulloch, pers. is an individual's fitness divided by the pop­ comm.; Snedecor and Cochran, 1967). If ulation mean fitness.) Each of the pheno­ these correlations are large, the standard typic character values was first standardized error becomes large, and it is harder to show so that the mean equalled Oand the variance that the coefficient is different from zero. equalled 1 (zscores; Sokal and Rohlf, 1981). Therefore, there is a high probability of type For example, a selection gradient of0.5 for II error (i.e., of not detecting selection when character x means that, on average, an in­ it does exist). The high standard errors also crease of one standard deviation in char­ mean that the actual values of the selection acter x was associated with an increase in gradients should be interpreted with cau­ relative fitness of 50%. Since selection gra­ tion. In addition, it becomes more difficult dients are partial regression coefficients, this to show significance of the entire regression analysis provided significance levels for the model with increasing numbers of correlat­ gradients and the total variance in fitness ed x variables. This is because each addi­ that is accounted for by all the characters tional x variable increases the degrees of combined (the r2 value for the regression). freedom, without increasing the explained Selection gradients for each component sums of squares in proportion to the in­ were generated with a separate multiple crease in degrees of freedom (since the added regression. Males with zero fitness for a giv­ x variable is correlated with the x variables en component were not included in the already included in the model). regressions for subsequent components, Selection Differentials. -Selection differ­ since these males were "eliminated" by se­ entials estimate the shift in the population lection acting in the previous component. mean value of a quantitative character due For example, males that did not court fe­ to total selection, both direct and indirect. males (W 3 = 0) could not attempt to cop­ Total selection is the sum of the direct se­ ulate; thus to include these males in the lection on a character and the direct selec­ fourth component would not separate se­ tion on all other correlated characters, lection occurring in the third and fourth weighted by the phenotypic correlation be- NATURAL AND SEXUAL SELECTION IN A BEETLE 74i

TABLEI. Mean male fitness values and standard de­ that these differentials are equivalent to the viations for each fitness component. w, is the mean standardized univariate selection gradient lifespan for the main study period only (1984-1985), and the selection intensity, as defined by and W1' is the mean lifespan including the census data Falconer (1981). The differentials for each from 1986; both of these estimates include individuals whose lifespans were truncated. Wi" is the estimated fitness component were calculated by a sep­ total population mean lifespan, taking the eight trun­ arate regression and thus are similar, but cated data points from W1' into account (see text for not exactly equal, to the partitioned selec­ details). W2-s is total sexual selection (the product of tion intensities of Arnold and Wade (1984a). components 2-5). The five multiplicative fitness com­ The differentials used here have the advan­ ponents are in boldface; the product of these equals tage of being simply related to the gradients total lifetime inseminations. ((3)by Equation (1) but, like these gradients, may not sum to the differential for total Symbol Fitness component Mean SD fitness. Another advantage of these differ­ W1 Lifespan (days) 56.2 34.5 entials is that they can be assigned standard W1' Lifespan including 1986 65.5 46.1 (days) errors and be tested for significance like any W1" Estimated total lifespan 69.0 53.2 regression coefficient. Since these are simple (days) regressions, the standard errors are not in­ W2 Attendance/lifespan 0.65 0.19 flated by the covariances between the x vari­ W3 Courtships/attendance 1.01 0.47 ables as in multiple regressions (see above), W4 Copulation attempts/court- 0.60 0.25 ships so that small selection differentials are more Ws Inseminations/ copulation 0.24 0.22 likely to be significant than small selection attempts gradients. W2_5 Inseminations/lifespan 0.10 0.08 All calculations, regression analyses, and Total lifetime inseminations 5.31 6.20 statistical tests were performed with the Minitab (Ryan et al., 1982) and SAS (SAS Institute, 1985) statistical packages. tween each character and the character of RESULTS interest. In the case of three correlated char­ Migration acters, Eight males were observed to make ten ii = f31 + f32r12 + (33r13 (1) migrations among trees during the 1985 season (two males moved twice). Eight of where i 1 is the standardized selection dif­ these (80%) were short-distance moves of ferential on character 1, (31 is the standard­ ized selection gradient for character 1, and less than 25 m. Four of the 18 total trees with fungi in the woodlot (22%) contained r12 is the phenotypic correlation between only dead fungi. Six of the ten emigrations characters I and 2. (32r12 represents the in­ direct selection on character I that is due were from these four trees; thus, an average to direct selection on character 2. Note that of 1.5 males (6/4) were observed leaving the Equation (1) represents the standard rela­ trees with dead fungi in 1985, while 0.29 tionship between simple and multiple males (4/14) emigrated from trees with live regression coefficients. fungi. One male migrated to the main study The selection differentials given in this tree in 1985, while none was observed to paper are the regression coefficients from a emigrate from this population. simple regression of relative fitness on the standard score of each phenotypic charac­ Fitness Components ter. The selection differential is defined as The mean lifespan of the 67 males during the covariance between relative fitness and the 1984 and 1985 mating seasons (includ­ the character (Lande and Arnold, 1983). The ing the truncated lifespans) is listed in the regression coefficient equals the covariance first line of Table I (W 1). The mean lifespan divided by the variance of the character; for 1984-1986 (including the eight truncat­ with standard scores, the variance is 1, so ed lifespans) is given in the second line of the coefficient equals the differential. Note Table I ( Wt'). The estimate of the true mean 742 JEFFREY CONNER

TABLE 2. Statistics of the phenotypic characters (N = each of a suite of correlated characters and 67 males) in this study before selection. A) Means and are therefore an appropriate measure of se­ standard deviations. B) Phenotypic correlations among lection for these characters. The directional the characters. All coefficients are significant at P < selection gradients with their standard errors 0.001. for the five fitness components are given in

A. Character Mean SD Table 3. Note that the results for lifespan using the data through 1986 (Wi') are vir­ Elytral length (mm) 6.4 0.5 Horn length (mm) 2.0 0.9 tually indistinguishable from the results Live weight (g) 0.10 0.02 from the main study period only (W 1). This suggests that the data from the main study B. Character period accurately represent selection acting Character Horn length Elytral length over the lifetimes of these males. The total 2 Elytral length 0.91 r values show that the three phenotypic Live weight 0.89 0.85 characters together explained significant proportions of the variance in the third fit­ ness component (number of females court­ ed) and in the total sexual-selection com­ lifespan of the entire cohort, taking the eight ponent (components 2-5 combined). About truncated lifespans into account, is listed in 21 % of the variance in total fitness was ex­ the third line of Table 1 (Wi''). Note the plained by the three characters. The only overlapping standard errors of these last two character under significant direct selection estimates; this suggests that the 1984-1986 was horn length. The total fitness gradient estimate ( Wi') is not affected much by the of0.94 indicates that, with other characters truncated observations. Also note that the equal, males with horn lengths one pheno­ 1984-1985 mean lifespan estimate (W 1) is typic standard deviation greater than the 81 % of the estimated total lifespan. The population mean inseminated an average of possible effects of using this underestimate 94% more females than the population mean oflifespan on the measures of selection will insemination success. This selection for be discussed. Given the migration results longer horns was principally sexual selec­ above, it is not likely that these estimates tion, with the last three components (3-5) were affected greatly by migration; however, being about equally important. None of the possible effects of migration on the selection selection gradients for elytral length or estimates will also be discussed. weight was statistically significant. The mean fitnesses for the sexual selec­ The selection gradients for all 134 males tion components (Table 1) show that, on present during the 1984 and 1985 mating average, males attended the mating area on seasons are given in Table 4. Note the strong about two-thirds of the nights that they were similarity between these data and those for alive (W 2) and courted about one female per the cohort of 67 males in Table 3; since the night that they were at the mating area ( W 3). sample size is doubled in Table 4, more of Sixty percent of these courtships actually the coefficients and r2 values are significant. ended in a copulation attempt (W 4), and The gradients for the first (lifespan) com­ 24% of these attempts were accepted by fe­ ponent of fitness and that for total fitness males (W 5). The last two lines of Table 1 indicate that there is selection for shorter show that males inseminate one female elytra that is principally due to natural (not every 10 days of their lives (W 2_ 5), and the sexual) selection. As in Table 3, the selec­ average male inseminates just over five fe­ tion on horn length is significantly positive, males in his lifetime. although somewhat weaker. The differences between the coefficients in Tables 3 and 4 Selection on Phenotypic Traits are not significant, either statistically or bi­ Selection Gradients. - The three pheno­ ologically. typic traits measured (weight, elytral length, Selection Differentials. - Standardized and horn length) were positively correlated selection differentials, which measure total with each other (Table 2). Selection gradi­ directional selection (both direct and indi­ ents estimate the direct selection acting on rect) on a character, are given in Table 5 for NATURAL AND SEXUAL SELECTION IN A BEETLE 743

TABLE3. Standardized selection gradients (with standard errors) for the three phenotypic characters measured. Gradients were calculated by a multiple regression of relative fitness on the three characters. Each line of the table gives the results from a separate multiple regression, one line for each of the five components of fitness, one for lifespan including data from 1986, one for total sexual selection (components 2-5 combined or insem­ inations per night that the male was alive), and one for total fitness (the total number of females inseminated in a male's lifetime). N = 67 males for all rows except W4 (N = 62) and Ws (N = 58). Gradients for the five components that multiply to equal total fitness are in boldface. Significance levels of the total r2 are from the F ratio (regression MS/residual MS). Significance levels for the gradients (the regression coefficients for each character) are from the t ratio (coefficient/standard error of coefficient).

Characters Symbol Fitness component Elytrallength Homlength Live weight Totalr 2 W1 Lifespan -0.24 (0.18) 0.26 (0.21) 0.12 (0.17) 8.5 W1' Lifespan including 1986 -0.22 (0.21) 0.30 (0.24) 0.10 (0.19) 8.7 W2 Attendance/lifespan 0.04 (0.09) 0.03 (0.10) -0.03 (0.08) 3.6 W3 Courtships/attendance -0.01 (0.13) 0.19 (0.15) -0.ot (0.12) 13.8* W4 Copulation attempts/ courtships -0.06 (0.14) 0.30* (0.15) -0.20 (0.11) 8.8 Ws Inseminations/ copulations -0.05 (0.32) 0.26 (0.33) 0.00 (0.25) 6.0 W2-s Sexual selection (inseminations/lifespan) -0.09 (0.23) 0.59* (0.26) -0.06 (0.21) 27.3*** Total fitness (number of inseminations) -0.33 (0.32) 0.94** (0.36) -0.16 (0.29) 20.7**

* P < 0.05; ** P < 0.01; *** P < 0.001. the cohort of 67 males. The selection dif­ itive direct selection on horn length (Table ferentials for total fitness are significant and 3), coupled with the large positive correla­ positive for all characters. For example, the tions between horn length and the other two total fitness differential of 0.49 for horn characters, resulted in increases in elytral length means that selection increased the length and live weight. The total selection population mean horn length by about half on elytral length and on weight, however, of a standard deviation, or from 2.0 mm to is slightly weaker than that on horn length. 2.45 mm. The differential is less than the This is due to the weak negative direct se­ direct selection gradient (0.94), due to the lection on these two characters and the fact negative direct selection on elytral length that the correlation between horn length and and weight, which are positively correlated the other characters is less than 1. with horn length [Table 2; Equation (1) shows how direct selection on correlated DISCUSSION characters combine to produce total selec­ This study shows how direct and indirect tion]. Note that total selection increased ely­ selection on correlated characters interact tral length and weight, despite the negative to produce the total selection on a given direct selection (Table 3) on each of these character. The high correlations among the characters. Indirect selection accounts for characters measured in this study suggest this apparent contradiction; the strong pos- that all three traits are indicators of general

TABLE4. Standardized selection gradients (with standard errors) for all 134 males that were present at some time during the 1984 and 1985 mating seasons. For explanation, see Table 3.

Characters Symbol Fitness component Elytrallength Homlength Live weight Total?- W1 Lifespan -0.37*** (0.10) 0.23 (0.14) 0.19 (0.11) 11.6** W2 Attendance/lifespan -0.08 (0.05) 0.o7 (0.07) 0.06 (0.05) 6.5* W3 Courtships/ attendance -0,07 (0.08) 0.12 (0.10) 0.o7 (0.08) 8.5** W4 Copulation attempts/ courtships -0,07 (0.07) 0.17 (0.09) 0.01 (0.07) 8.7** Ws Inseminations/ copulations 0.18 (0.14) 0.03 (0.18) 0.00 (0.14) 7.5* W2-s Sexual selection (inseminations/lifespan) -0.05 (0.12) 0.32* (0.16) 0.11 (0.12) 23.4*** Total fitness (number of inseminations) -0.47** (0.18) 0.62** (0.23) 0.22 (0.18) 18.7***

* P < 0.05; ** P < O.Ol;*** P < 0.001. 744 JEFFREY CONNER

TABLE5. Standardized selection differentials, which measure the total selection acting on the three phenotypic characters. Differentials are the slope of a univariate regression ofrelative fitness on each character individually. In all cases, the standard errors of the three differentials within each fitness component were the same, so they are given in one column at the end of each line. Sample sizes equal 67 males for all rows except W4 (N = 62) and W5 (N = 58). For explanation of the fitness components, see Table 3.

Characters Symbol Fitness component Elytrallength Horn length Live weight SE W1 Lifespan 0.10 0.15* 0.14 0.07 W2 Attendance/lifespan 0.05 0.05 0.04 0.04 W3 Courtships/ attendance 0.15** 0.17** 0.15** 0.05 W4 Copulation attempts/ courtships 0.04 0.07 0.01 0.05 Ws Inseminations/ copulations 0.19 0.22 0.18 0.11 W2-s Sexual selection (inseminations/lifespan) 0.39*** 0.45*** 0.38*** 0.09 Total fitness (number of inseminations) 0.38** 0.49*** 0.39** 0.13 * P < 0.05; •• P < 0.01; ••• P < 0.001. body size. The fact that the selection gra­ variance in fitness with little selection oc­ dient for horn length differs in sign from the curring. other two, however, suggests that there is Another possible use of the opportunity selection for increased horn length indepen­ for selection is to help explain an observed dent of body size. Elytral length and live lack of selection; however, similar infor­ weight are probably both good indicators of mation is available directly from a multiple body size; thus, selection gradients for these regression analysis without calculating the characters may reflect direct selection for opportunity for selection. There are two body size, at least in part. The nonsignifi­ possible biological reasons for a lack of se­ cant gradients for these two characters in lection on a character. First, there may be the cohort (Table 3) do not indicate that ample fitness variance, but the relationship either character is a target of selection; the between fitness and the phenotype may be analysis of the expanded data set (Table 4) mostly random (Fig. IA). Second, the vari­ suggests that elytral length may be slightly ance in fitness may be very small (Fig. 1B); larger than optimal in terms oflifespan. The since the variance in fitness sets an upper positive selection differentials on these bound on the amount a character can be characters, however, show that there is changed by selection (Arnold and Wade, overall selection for increased adult body 1984a; Houck et al., 1985; Arnold, 1986), size due to the selection on horn size. the lack of fitness variance explains the lack Most other studies that have partitioned of selection in Figure lB. One can distin­ fitness have used variance in relative fitness guish between the two possible reasons for or the "opportunity for selection" as the lack of selection using the r2 value for the only measure of selection (e.g., McCauley, regression. The regression slopes in Figure 1983; Banks and Thompson, 1985; Fincke, IA and 1B are the same (0.034)-that is, 1986; Hafernik and Garrison, 1986; but see selection is weak in both examples. The r 2 Arnold and Wade [1984b] and Koenig and values of the two regressions differ greatly, Albano [ 1987]). Fitness variance is neces­ however; since there is little variance in Fig­ sary for selection to occur, but variance alone ure lB more of this variance is explained is not sufficient evidence that selection has by the same slope. Therefore, if the results occurred. This is because fitness variance of a regression analysis indicate that the provides no information on the relationship slope(s) are small but the r2 value is large, between fitness and phenotypic characters then this suggests that selection may have and, consequently, does not measure selec­ been limited by a lack of fitness variance. tion acting on the phenotype. A large pro­ Conversely, if the r2 value is small, then portion of fitness variance may be random selection could not have been limited. with respect to the phenotype (Sutherland, The r 2 values presented in this paper 1985a, 1985b); therefore, there can be great clearly indicate that selection on the three NATURAL AND SEXUAL SELECTION IN A BEETLE 745

A. 2 Slope = 0.034 R-squared = 0.005 • • • • • ,- • Ill •• Ill •• Q) C ;;:: -Q) > :;:; CG Q) • • a: • • • • • • • •

0 -2 -1 0 2

B. 2 Slope= 0.034

R-squared = 0.79

N Ill 1/l Q) C ;;::- Q) • > :;:; CG Q) a:

0 +-----.---r---..-----r--..-----.---~--, -2 -1 0 2

Phenotypic trait

Fro. I. Plots ofrelative fitness versus a phenotypic trait for two hypothetical populations, designed to illustrate the two possible reasons for a lack of selection on a character. The populations differ only in the variance in relative fitness; the mean relative fitness and the mean and variance of the phenotypic trait are the same in both populations. Note that selection is weak in both populations (both regression slopes equal 0.034), but the greater r2 value in (B) suggests that selection may have been limited by a lack of fitness variance. traits measured was not constrained by a variance in total fitness. What accounts for lack of variance in any component of fitness. the other 79% of the fitness variance? First, The three characters measured in this study there are obviously many characters that together explained only about 21 % of the influence fitness other than the three mea- 746 JEFFREY CONNER

sured. For example, behavioral traits could be important; characteristics of the males' Evolution of Phenotypic Traits long and complex (for a beetle) courtship Based on the large positive selection dif­ behavior could well influence the percent­ ferentials for all the characters found in this age of a male's copulation attempts that are study (Table 5), one might predict that male accepted by a female ( W 5). Second, much B. cornutus are evolving longer horns and of the variation in fitness in the five com­ larger body sizes. The accuracy of this pre­ ponents may be random with respect to the diction, however, depends on two factors phenotype. not addressed in this study. The first is the genetics of the traits. Clearly, if horn and Selective Agents and Targets body size are not heritable in this popula­ of Selection tion, then there would be no response to the The results reported in this study illus­ selection found in this study, and there trate how partitioning lifetime fitness into would be no increase in size across gener­ its components (Arnold and Wade, 1984a, ations. Results from laboratory studies of 1984b) can separate the effects of natural other species of insects suggest that body and sexual selection on phenotypic char­ size is generally heritable (e.g., Bell and Bur­ acters and help identify the selective agents ris, 1973; Palmer and Dingle, 1986), but that are important (see Banks and Thomp­ these estimates are not directly applicable son [1985] and Koenig and Albano [1986] to B. cornutus. A related question involves for discussions of this problem). For the ex­ the degree to which the phenotypic corre­ panded data set (but not for the cohort data lation between horn length and body size is alone) the gradients for the lifespan com­ caused by a genetic correlation (which de­ ponent indicated that there was selection for pends in part upon the degree to which the shorter elytra. Since this study did not deal two characters are heritable [Falconer, with sources of mortality, the agents of se­ 1981]). If there is a significant positive ge­ lection can only be speculated upon. Per­ netic correlation, then the population will haps males with longer elytra were more evolve larger bodies as a correlated response visible to predators. B. cornutus has three to the selection for longer horns, as is sug­ separate defense mechanisms that are gested by the selection differentials (Table adapted to different types of predators (Con­ 5). ner et al., 1985), suggesting that predation The other important factor in predicting has been important in the evolutionary his­ the evolution of these traits is the extent to tory of this beetle. The alternative expla­ which the selection differentials presented nation of increased metabolic costs oflarger in this paper reflect the total selection acting size is not likely, because the direct selection on these traits. The first possible problem on weight was positive. is the truncated lifespans and their effect on Most of the selection for longer horns, the selection differentials. It seems clear that however, clearly involves sexual selection, the truncated lifespans do not affect the se­ and most of this is due to male competition lection estimates for the lifespan compo­ for access to females. Recall that access to nent, since there is no difference between females (components 2-4) is determined the gradients for the main data set and the principally by male competition and male set that includes 1986 data. The estimates choice. One would expect this result for sev­ of sexual selection presented here are valid eral reasons: horns are a sexually dimorphic unless the mating rate (i.e., inseminations trait, horns are used in fights over females per day) of the 17 surviving males changed (Conner, 1987), and laboratory studies of greatly in 1986. B. cornutus (Brown, 1980; Brown and Sieg­ The second potential problem is the effect fried, 1983; Brown et al., 1985; Brown and of migration on the estimates of lifespan. Bartalon, 1986) and other horned beetles This is probably not a major problem, since (Eberhard, 1977, 1979, 1981;Palmer, 1978) only one male migrated from the study tree also suggest that beetle horns have evolved to one of the other 17 trees in 1985. It is for use in male competition. possible that undetected migration oc- NATURAL AND SEXUAL SELECTION IN A BEETLE 747 curred, either within the woodlot studied or the likely effects of this unmeasured selec­ between the study tree and trees outside this tion on the three phenotypic characters? The patch of woods. It is unlikely that unde­ character under the strongest direct selec­ tected migration occurred in large enough tion in this study was horn length. The se­ numbers to affect the results significantly, lection gradients for horn length presented however, since most migrations were be­ here probably represent all or most of the tween trees that were less than 25 m apart. direct selection on horn length, because lar­ The distance from the study tree to the clos­ vae and females do not have horns and male est fungus-infested tree was more than 200 horn length is not likely to affect male fe­ m, and the nearest woods outside the study cundity directly, unless females choose to patch was more than 500 m away. It is pos­ lay more eggs after mating with large-horned sible that there was greater migration in 1984 males. due to the warmer temperatures that year. If male B. cornutus with larger bodies have This is unlikely to be a major problem for more offspring per insemination, then this two reasons. First, if large numbers of bee­ could lessen the negative direct selection on tles emigrated from the study tree in 1984, elytra length or create positive selection on many of them would have been found at weight. In this case the positive total selec­ the other trees in 1985; only one 1984 em­ tion found on all three characters in this igrant was found. Second, in a five-year study study would be an underestimate. A labo­ of B. cornutus migration in a five-hectare ratory study of a cerambycid beetle showed study area in Virginia, L. Brown (pers. that females produced more eggs per cop­ comm.) also found that migrations of more ulation when they were mated to larger males than 100 meters were extremely rare. than when they were mated to smaller males Another possibility is that B. cornutus (Hughes and Hughes, 1985). tends to emigrate immediately after pupa­ If elytra length and weight in males are tion, so that adults do not appear on the genetically correlated with the same char­ fungi until after migration. Given the in­ acters in females, then selection on these tensity of observation and censusing at the characters in females would cause a corre­ study site, this seems unlikely, but it cannot lated response to selection in males. Since be ruled out. If this early migration does males and females do not differ in elytral occur, then selection due to differential mi­ length (Conner, unpubl.), it seems reason­ gration success, which was not measured in able to expect selection for shorter elytra in this study, could be important. This would the lifespan component in females, as was not affect the measurements oflifetime mat­ found in males (Tables 3, 4). This selection ing success within the main study popula­ for smaller bodies would not be opposed by tion, however, because an immigrant's re­ selection for longer horns; however, there productive life would not begin until it could be selection for larger-bodied females appears on the fungi. Thus, it is likely that iffecundity is correlated with size in B. cor­ the beetle's first appearance on the tree is nutus as it is in other insects (e.g., Parker, close to the beginning of its reproductive 1970; Juliano, 1985). life. It is also likely that the last observation There can be no direct selection on horn of a beetle accurately represents the beetle's length, elytra length, or adult weight at the death and not emigration, since males were larval stage, since these characters are not much more likely to leave a tree after all expressed in larvae. However, indirect se­ fungi died (all 15 fungi on the study tree lection due to selection on correlated char­ were alive after censuses were terminated). acters in larvae (e.g., development time, The third potential problem is that this growth rate, etc.) could be very important. study only estimates selection due to vari­ For example, selection for shorter devel­ ability in insemination success of adult opmental times to reduce the risk of larval males. Therefore selection on larvae and fe­ mortality would lead to smaller adult size, males has not been measured, nor has se­ given a fixed larval growth rate. Thus, there lection due to variance in male fecundity could be directional selection for smaller (eggs fathered per insemination). What are size at the larval stage opposing the selection 748 JEFFREY CONNER

for larger adult size. This directional selec~ quirements for the Ph.D. degree. I thank G. tion in opposite directions at different life Eickwort, T. Eisner, S. Emlen, J. Endler, S. stages could give rise to overall stabilizing Levin, C. McCulloch, P. Sherman, J. Tra­ selection for body size (Endler, 1986; J. Tra­ vis, S. Via, M. Wade, D. Winkler, two anon­ vis, pers. comm.). ymous reviewers, and especially D. Maddox Finally, to understand the evolution of and B. Silverman for insightful discussion these traits in B. cornutus, temporal and and/or comments on this manuscript. L. spatial heterogeneity of selection must be Bernard, C. Frissora, P. Renzullo, M. Shi­ considered. While this study covered two manura, and especially I. Kaatz contributed years, longer-term temporal fluctuations in invaluable assistance in the field. This work selection are possible (e.g., selection due to was supported by the T. Roosevelt Fund of differential migration ability). This study the American Museum of Natural History, suggests that there is little emigration while Sigma Xi Grants-in-Aid (National and Cor­ the fungi are still alive, and since individual nell chapter) and NIH Grant #AI-02908 (to G. applanatum can live as long as 10 years T. Eisner). The author was supported by an (White, 1919), most individuals probably NSF predoctoral fellowship and an A. D. remain on the same tree throughout their White fellowship from Cornell University. lives. If success in colonization of new fun­ gus-infested trees is related to horn length, LITERATURECITED elytral length, or weight, then this would ARNOLD,S. J. 1986. Limits on stabilizing, disruptive, and correlational selection set by the opportunity cause periodic selection on these characters. for selection. Amer. Natur. 128:143-146. In his five-year study of migration, how­ ARNOLD,s. J ., ANDM. J. WADE. 1984a. On the mea­ ever, L. Brown (pers. comm.) found no re­ surement of natural and sexual selection: Theory. lationship between horn or elytral length and Evolution 38:709-719. --. 1984b. On the measurement of natural and successful migration within a 5-hectare study sexual selection: Applications. Evolution 38:720- site. 734. This study does not address interpopu­ BANKS,M. J., ANDD. J. THOMPSON. 1985. Lifetime lational (spatial) differences in selection, mating success in the damselfly Coenagrion puella. since only one population was studied. In a Anim. Behav. 33:1175-1183. BELL,A. E., ANDM. J. BURRIS. 1973. Simultaneous study of two other lower-density popula­ selection for two correlated traits in Tribolium. Ge­ tions, however, even stronger selection on net. Res. 21 :29-46. male horn size was found (Conner, 1987), BROWN,L. 1980. Aggression and mating success in suggesting that the selection gradients re­ males of the forked fungus beetle, Bolitotherus cor­ nutus (Panzer) (Coleoptera: Tenebrionidae). Proc. ported here may underestimate selection on Entomol. Soc. Wash. 82:430-434. horn length over all populations of B. cor­ BROWN,L., ANDJ. BARTALON.1986. Behavioral cor­ nutus. relates of male morphology in a homed beetle. Amer. In summary, it seems clear that there is Natur. 127:565-570. strong selection for longer horns in adult BROWN,L., J. MACDONELL,AND V. J. FITZGERALD. 1985. Courtship and female choice in the homed male B. cornutus and that this selection is beetle Bolitotherus cornutus (Coleoptera: Teneb­ not likely to be changed much by selection rionidae). Ann. Entomol. Soc. Amer. 78:423-427. on females or by spatial and temporal vari­ BROWN,L., AND B. D. SIEGFRIED. 1983. Effects of ation in selection. It seems quite likely, male horn size on courtship activity in the forked fungus beetle, Bolitotherus cornutus (Coleoptera: however, that this selection might be op­ Tenebrionidae). Ann. Entomol. Soc. Amer. 76:253- posed by negative selection on correlated 255. characters in larvae. Whether or not this CLUTTON-BROCK,T. H., F. E. GUINNESS,AND S. D. species is in the process of evolving longer ALBON. 1982. Red Deer. Univ. Chicago Press, horns depends upon the magnitude of this Chicago, IL. CONNER,J. 1987. Natural and sexual selection in a opposing indirect selection and the extent fungus beetle. Ph.D. Diss. Cornell Univ., Ithaca, to which horn length is heritable. NY. CONNER,J., S. CAMAZINE,D. ANESHANSLEY,AND T. ACKNOWLEDGMENTS EISNER. 1985. Mammalian breath: Trigger of de­ fensive chemical response in a tenebrionid beetle This paper is part of a thesis presented to (Bolitotherus cornutus). Behav. Ecol. Sociobiol. 16: the faculty of the graduate school of Cornell 115-118. University in partial fulfillment of the re- EBERHARD,W. G. 1977. Fighting behavior of male NATURAL AND SEXUAL SELECTION IN A BEETLE 749

Golofa porteri beetles (Scarabeidae: Dynastinae). LILES,M. P. 1956. A study of the life history of the Psyche 83:292-298. forked fungus beetle, Bolitotherus cornutus (Panzer) ---. 1979. The function of horns in Podischnus (Coleoptera: Tenebrionidae). Ohio J. Sci. 56:329- agenor (Dynastinae) and other beetles, pp. 231- 337. 258. In M. Blum and N. Blum (eds.), Sexual Se­ McCAULEY,D. E. 1983. An estimate of the relative lection and Reproductive Competition in Insects. opportunities for natural and sexual selection in a Academic Press, N.Y. population of milkweed beetles. Evolution 37:701- --. 1981. The natural history of Doryphora sp. 707. (Coleoptera, Chrysomelidae) and the function of its PACE,A. 1967. Life history and behavior ofa fungus sternal horn. Ann. Entomol. Soc. Amer. 74:445- beetle, Bolitotherus cornutus (Tenebrionidae). Oc­ 448. cas. Pap. Mus. Zool. Univ. Mich. 653:1-15. ENDLER,J. A. 1986. Natural Selection in the Wild. PALMER,J. 0., AND H. DINGLE. 1986. Direct and Princeton Univ. Press, Princeton, NJ. correlated responses to selection among life-history FALCONER,D. S. 1981. Introduction to Quantitative traits in milkweed bugs (Oncopeltusfasciatus). Evo­ Genetics, 2nd Ed. Longman, N.Y. lution 40:767-777. FINCKE,0. M. 1982. Lifetime mating success in a PALMER,T. J. 1978. A horned beetle which fights. natural population of the damselfly, Enallagma Nature 274:583-584. hageni (Walsh) (Odonata: Coenagrionidae). Behav. PARKER,G. A. 1970. Sperm competition and its evo­ Ecol. Sociobiol. 10:293-302. lutionary effect on copula duration in the fly Sca­ ---. 1986. Lifetime reproductive success and the tophaga stercoraria.J. Insect Physiol. 16: 1301-1328. opportunity for selection in a nonterritorial dam­ PRICE, T. D. 1984. Sexual selection on body size, selfly (Odonata: Coenagrionidae). Evolution 40:791- territory and plumage variables in a population of 803. Darwin's finches. Evolution 38:327-341. HAFERNIK,J. E., JR., AND R. W. GARRISON. 1986. PRICE, T. D., P. R. GRANT, H. L. GIBBS,AND P. T. Mating success and survival rate in a population of BoAG. 1984. Recurrent patterns of natural selec­ damselflies: Results at variance with theory? Amer. tion in a population of Darwin's finches. Nature Natur. 128:353-365. 309:787-789. HOUCK,L. D., S. J. ARNOLD,AND R. A. THISTED. 1985. RYAN,T. A., JR., B. L. JOINER,AND B. F. RYAN. 1982. A statistical study of mate choice: Sexual selection Minitab Reference Manual. Minitab, Inc., State in a plethodontid salamander (Desmognathus och­ College, PA. rophaeus). Evolution 39:370-386. SAS INSTITUTE.1985. Users Guide: Statistics, Ver­ HUGHES,A. L., ANDM. K. HUGHES. 1985. Female sion 5 Ed. SAS Inst., Inc., Cary, NC. choice of mates in a polygynous insect, the white­ SNEDECOR,G. W., ANDW. G. CocHRAN. 1967. Sta­ spotted sawyer Monochamus scutellatus. Behav. tistical Methods. Iowa State Univ. Press, Ames. Ecol. Sociobiol. 17:385-387. SOKAL,R.R., ANDF. J. ROHLF. 1981. Biometry, 2nd JULIANO,S. A. 1985. The effects of body size on Ed. Freeman, San Francisco, CA. mating and reproduction in Brachinus lateralis (Co­ SUTHERLAND,W.J. 1985a. Chancecanproduceasex leoptera: Carabidae). Ecol. Entomol. 10:271-280. difference in variance in mating success and explain KAusz, S. 1986. Variable selection on the timing of Bateman's data. Anim. Behav. 33:1349-1352. germination in Collinsia verna (Scrophulariaceae). ---. 1985b. Measures of sexual selection. Oxford Evolution 40:4 79-491. Surv. Evol. Biol., 2:90-101. KOENIG,w. D., ANDS. s. ALBANO. 1986. On the TRAVIS,J., ANDs. HEINRICH. 1986. Some problems measurement of sexual selection. Amer. Natur. 127: in estimating the intensity of selection through fer­ 403-409. tility differences in natural and experimental pop­ ---. 1987. Lifetime reproductive success, selec­ ulations. Evolution 40:786- 790. tion, and the opportunity for selection in the white­ WHITE, J. H. 1919. On the biology of Fornes ap­ tailed skimmer Plathemis lydia (Odonata: Libel­ planatus. Trans. Roy. Can. Inst. Toronto 12: I 33- lulidae). Evolution 41:22-36. 174. LANDE,R., ANDS. J. ARNOLD. 1983. The measure­ ment of selection on correlated characters. Evolu­ Corresponding Editor: J. A. Endler tion 37:1210-1226.