## Abstract

Parental protection of eggs represents one of the most basic forms of parental care. Theory suggests that even such basic parental investment represents a trade-off between current offspring survival and future reproductive success. However, few studies have quantified the underlying costs and benefits of parental care for marked individuals across an entire lifetime. I marked and followed 370 females of Publilia concava (Hemiptera: Membracidae) that exhibited a range of guarding durations for their first clutch. Greater hatching success was correlated with longer guarding durations, and a removal experiment verified that female presence was responsible for a twofold increase in hatching success. On the other hand, females that remained to guard eggs had a lower number and size of future broods, suggesting that parental care may reduce lifetime fecundity. Marked females exhibited a bimodal distribution of guarding durations, reflecting the extreme tactics of immediate abandonment or remaining through hatching. Estimates of lifetime number of nymphs produced by females that abandon eggs early versus guard eggs through hatching revealed roughly equivalent levels of fitness. I discuss the conditions under which we might expect a female to adopt each of the alternative tactics, given the costs and benefits of parental care that were quantified in this study.

Parents that experience multiple reproductive bouts across their lifetime are expected to weigh the benefits of investing in current offspring against the costs to future reproduction (Clutton-Brock, 1991; Trivers, 1972; Williams, 1966). Several models have addressed the optimal duration of male parental care by focusing on alternative mating opportunities (Grafen and Sibly, 1978; Maynard Smith, 1977; Townsend, 1986; Whittingham et al., 1992). Models addressing the optimal duration of female parental care focus on opportunities for laying additional clutches (Andersson et al., 1980; Carlisle, 1982; Zink, 2001). In insects, the most basic form of parental care involves the physical protection of eggs from predators and parasites (Eickwort, 1981; Tallamy, 1984; Tallamy and Wood, 1986). Female egg guarding occurs in a wide variety of insects, including lace bugs (Tallamy and Denno, 1981), stink bugs (Eberhard, 1975), cockroaches (Nalepa and Bell, 1997), and treehoppers (Eberhard, 1986; Wood, 1974, 1976). Because egg guarding represents an investment in terms of time and energy, these egg-guarding females are expected to face a trade-off between offspring survival and future fecundity. A comprehensive understanding of egg guarding requires quantifying both the benefits (in terms of hatching success) and the costs (in terms of future fecundity) of parental care.

Removal experiments in some insect species have verified that egg guarding by adults acts to increase the hatching success of eggs. In a species of nymphalid butterfly in which females straddle eggs to exclude predators, the removal of females resulted in lower hatching success (Nafus and Schreiner, 1988). Similar removal experiments in cannibalistic thrips (Crespi, 1990) and other arthropods such as a foliage spider (Toyama, 1999) have further confirmed the adaptive significance of female egg guarding. In many of these insect species, a fraction of egg guarding females will abandon eggs without providing parental care (Eberhard, 1986; Kaitala and Mappes, 1997; Tallamy, 1982; Wood, 1977). One study revealed that up to 50% of females abandon eggs in an egg-guarding nymphalid butterfly (Nafus and Schreiner, 1988). In treehoppers, rates of egg abandonment by females can be as high as 36–47% (Eberhard, 1986; Olmstead and Wood, 1990b). These studies suggest that females are balancing the benefits of remaining with a clutch (in terms of hatching success) against the benefits of abandoning a clutch (in terms of additional clutches).

In this article, I report on a field study with the treehopper Publilia concava (Hemiptera: Membracidae) that specifically quantified these costs and benefits of egg guarding. Treehoppers commonly guard their eggs and nymphs for some time, and these behaviors reduce predation (Tallamy and Schaefer, 1997; Tallamy and Wood, 1986; Wood, 1974, 1977). Offspring protection sometimes involves elaborate mechanisms for deterring predators, such as kicking or wing fanning (Cocroft, 2002; Eberhard, 1986; Wood, 1976), as well as sophisticated communication through substrate vibrations (Cocroft, 1996, 1999). These guarding behaviors have proven to be important for the protection and survival of eggs. In the treehopper Entylia bactriana early abandonment of eggs was correlated with a 47% reduction in hatching success (Olmstead and Wood, 1990b; Wood, 1977). However, it is likely that egg protection represents a reduction in future fecundity because other studies have found a negative association between nymph guarding and the number of future clutches (Bristow, 1983; Del-Claro and Oliveira 2000). These fecundity costs are particularly relevant to species in which females lay multiple clutches over space and time. In treehopper species that lay only one clutch (semelparous), we expect the costs of remaining to be so minimal that females should always express extended parental care (Billick et al., 2001; Tallamy, 1999).

The treehopper Publilia concava is an excellent species for examining the costs and benefits of egg guarding because females allocate care across multiple clutches. In the first experiment, female removals tested the hypothesis that egg guarding causes an increase in hatching success. In a second field study, I followed the reproductive success of 370 marked adult females throughout their entire lifetimes. This work revealed the exact form of this relationship between female guarding duration and hatching success. I then tested the hypothesis that increases in female guarding duration (for the first brood) are associated with a decrease in future reproductive success. I used field data from marked individuals to correlate female guarding duration (for the first brood) with future brood number, future clutch size, and future guarding durations. Females exhibited a naturally bimodal distribution of female guarding durations centered around the extremes of abandoning immediately or remaining with eggs through hatching. For each female, I combined current and future reproductive success, in terms of lifetime nymphs produced, to examine the relative fitness for high and low levels of parental care. I discuss the conditions under which females may be expected to adopt each alternative tactic given the costs and benefits of egg guarding that were quantified in this study.

## METHODS

### Natural history study

The study species, Publilia concava (Hemiptera: Membracidae), is found throughout the eastern United States, where it engages in a mutualism with ant colonies (primarily of the genus Formica). Ants collect honeydew excreted by treehopper adults and nymphs in return for protection from predators and parasites; experiments have shown that the removal of ants results in high rates of predation and mortality for nymphs (McEvoy, 1977, 1979; Morales, 2000a,b). I studied two adjacent populations of Publilia concava in an abandoned agricultural field near Newfield, New York, USA. One population was clustered around a mound of the ant Formica exsectoides and the other around a mound of the ant Formica subsericea. All host plants (Solidagoaltissima) were found within 15 m of their respective ant mound (7.3 m on average).

In 1998, adult males and females emerged in mid-May. By the end of May, males had disappeared and females could be found clustered on individual plant stems in groups of up to 10 individuals. During this brief window of time, between mating and egg laying, I tagged all plants containing Publilia concava adults (28–29 May). On the same two evenings, I collected all adults, keeping females from each individual plant in a separate vial. Vials were refrigerated, and each female was randomly marked with a distinct color combination on her pronotum by using UniPaint™ fine line paint markers. The use of up to 10 colors and various orientations of four dots per female allowed me to distinguish among hundreds of individuals in the field. Females were returned to their original plants during the same evening that they were collected. On the day after marking, some females were found on adjacent plants; these females were returned to their original plant in an attempt to compensate for any disturbance caused by the previous night's markings.

Females began to lay eggs a few days after they were marked. Eggs were laid in a discrete clutch on the underside of an individual leaf and guarded by the ovipositing female, who usually positioned herself above or just to the side of the eggs. In 1998, 370 marked females initiated their own egg masses at the beginning of the season; an additional 108 marked females began by ovipositing into preexisting egg masses. Egg masses that were visited by these secondary females did not suffer reduced hatching success relative to unvisited egg masses (Zink, unpublished data). Therefore, I focused my analyses on the 370 females that initiated their own egg masses, monitoring them for the remainder of their lives across multiple initiated broods. Although a fraction of these females eventually oviposited into preexisting egg masses, these females were not more (or less) likely to guard their first clutch. Therefore, any hidden benefits of egg dumping were distributed evenly across different guarding durations. The behaviors of each female were recorded every 2 days and included the initiation of an egg mass, guarding of eggs, and feeding on other parts of the plant. Weekly searches revealed that a small fraction of the marked females eventually moved to nearby untagged plants. All untagged plants with marked females were immediately tagged and included in subsequent data collection. Data collection continued until 7 August, well after the final eggs had been laid and most females had disappeared.

Leaves with individual clutches of eggs were identified and tagged on the first day of egg laying and were monitored every 2 days thereafter. The length (in millimeters) of all egg masses were recorded every 2 days until nymphs first hatched, and the number of nymphs was recorded thereafter (for the first two instars) until large nymphal aggregations made it impossible to keep track of discrete clutches (11 July). Hatching success was calculated as the maximum number of nymphs observed divided by the number of eggs as estimated from the maximum length of the egg mass (in millimeters). In 1999, at the same site, 211 egg masses were collected and measured for their overall length, and their total egg number was counted under a dissecting scope. The length of an egg mass was found to be an excellent predictor of overall egg number (r2 =.87, N = 211, p <.0001) by using a linear model with a slope of 13.1 eggs per millimeter and an intercept of −7.8 eggs. A linear model was clearly the best fit for the relationship between length and egg number as all higher order terms were not significant (p >.10). The analyses of hatching success were limited to egg masses initiated in the first 10 days (1–10 June), resulting in a subset (n = 280) of the total number of broods initiated. Egg masses that were initiated after 10 June usually had some nymphs immigrating from previously hatched eggs on the same plant, obscuring accurate measures of hatching success.

To estimate lifetime fitness, I developed a simple equation that quantifies current and future reproduction. When females allocate parental investment across multiple clutches, their lifetime fitness can be represented as

Here P represents the number of eggs invested in a clutch, i, and ti represents the guarding duration in clutch i. The variable N represents the lifetime number of clutches, and the function g(t) represents the relationship between guarding duration and hatching success. I assume that the form of g(t) is equivalent across all clutches 1 to N. Focusing attention on an individual female's first clutch, Equation 1 becomes
By combining current and future reproduction, Equation 2 allowed me to compare the lifetime fitness for the alternative tactics of guarding versus abandoning eggs. All of the analyses in this paper were conducted with JMP Statistical Software (SAS Institute). All comparisons of mean (±SE) were nonparametric (Wilcoxin) owing to nonnormal or unequal variances. There was no effect of Formica species (i.e., site) on any of the analyses or results, so the data from the two adjacent populations were combined into one data set.

### Removal experiment

This experiment directly tested the effect of female guarding duration on the hatching success of eggs. In this population (near Enfield, New York) and year (1999), the majority of P. concava females began laying eggs on or before 31 May. On this day, I located and tagged all plants with P. concava aggregations within the population boundary (10 × 30 m), totaling 194 plants. I then marked all initiated egg masses and gave females 4 days to finish laying their clutches (previous work has shown that the majority of eggs are laid in 1–3 days). On the morning of 4 June, I randomly designated one of the marked broods from each of the 194 plants as a control, and a second brood from half of these plants as a treatment (for female removal). As a result, there were 194 broods available as controls (to monitor the natural progression of ant tending and female tenure), of which 97 were directly compared with a treatment brood on the same plant (to determine the effects of female removal). Females were removed from treatment broods on the morning of 4 June and placed on the upper portion of their same plant to maintain original female group size. If the occasional female returned to a treatment brood at a later date, she was removed again but counted for that day. Female presence was recorded for each brood late in the afternoon of 4 June and every day thereafter through 28 June. The number of ants (Formica subsericea) found on the leaf surface surrounding a brood (usually tending the adult) was recorded daily at the same time as female presence. I took this instantaneous assessment of ant tending (i.e., “snapshot” in time) to be representative of the daily average. In addition, egg predators and parasites were noted on broods and elsewhere on the plant. I also counted the total number of eggs laid by females in all control and treatment broods. When the first nymphs emerged on 22 June, nymphs were counted daily through 28 June. Hatching success was calculated as the maximum number of nymphs observed during this period divided by the total number of eggs laid.

## RESULTS

### Benefit of egg guarding: increased hatching success

The distribution of guarding periods showed that many females leave immediately after egg laying, but others remain well through hatching (Figure 1). For egg masses initiated in the first 10 days in 1998 (by 10 June), the average time between the first day of egg laying and the first day of nymph hatching was 21.55 (0.23) days (n = 240 that hatched); this mean is depicted by the vertically dashed line (Figure 1). In 1998, hatching success increased linearly with the maximum guarding duration of primary females (r =.42, n = 280, p <.0001) (Figure 1). In the linear model, the slope corresponds to an increase of 0.007 nymphs per egg per day of guarding, and the intercept equals 0.15 nymphs per egg. This means that a female staying through hatching doubled her hatching success (on average) relative to a female that abandoned immediately after laying eggs (0.30 nymphs per egg versus 0.15 nymphs per egg). The average hatching success of all egg masses (a) was 0.23 nymphs per egg (see horizontally dashed line; Figure 1), which is equivalent to the hatching success of a primary female that stayed half the time to hatching.

The removal experiment confirmed the causality in this relationship between female presence and hatching success. Among control broods in the removal experiment, there was also a positive correlation between the number of days that a female was present (out of a maximum of 25) and hatching success (r =.37, n = 194, p <.0001). The same was true for the proportion of days that a female was present during the egg-guarding period (through day 21; r =.35, n = 194, p <.0001). The removal experiment was successful in dramatically decreasing the number of days that a female was present on a treatment brood relative to a control brood within the same plant (18.6 [0.5] days for control females versus 5.1 [0.6] days for treatment females; z = 10, N = 97, p <.0001). Although there was no difference in the number of eggs laid by control and treatment females (z = 0.03, p =.98) (Figure 2), the maximum number of hatching nymphs was significantly lower in treatment broods with the female removed (z = 6.2, p <.0001) (Figure 2). Dividing the number of eggs by the maximum number of nymphs for each brood revealed a significant difference in hatching success, with control broods having higher survival (0.804 nymphs per egg versus 0.483 nymphs per egg, z = 5.9, p <.0001). The presence of females over eggs is likely to exclude a variety of predators (see Wood, 1974, 1977). In the 25 days of observation, for example, 57 plants were found to harbor an average of 3.3 predatory mites (Erythraeidae), 16 plants contained nabids, and 30 plants contained mirids. All of these arthropods are known predators of insect eggs; on at least 17 plants, adult mites were directly observed attacking eggs.

An examination of all control broods from the removal experiment (N = 194) revealed that the departure of females late in the guarding period coincided with the hatching of nymphs (Figure 3A). Ants began tending nymphs immediately after nymphs hatched, such that the average level of ant presence around a brood increased dramatically during days 22–28 (Figure 3B). The female removal experiment revealed that during the egg-guarding phase, ants were less likely to visit eggs when a female was absent. The removal of females was associated with a decrease in the average presence of ants around the egg mass for the days before hatching (1.9 [0.1] ants for control broods versus 0.7 [0.1] ants for treatment broods; z = 9.1, p <.0001). Within all controls there was also a high positive correlation between the percentage of days that a female was present on an egg mass (before hatching) and the average number of ants present around that egg mass (r =.62, n = 194, p <.0001). Thus, an additional benefit of remaining to guard eggs may be increased ant protection in the area around egg masses.

### Costs of egg guarding: decreased future reproduction

Early abandonment of eggs allowed females to increase the overall number of future broods initiated. In the natural history study (1998), females abandoning during days 0–4 initiated a greater number of future broods relative to females that remained (z = 6.6, p <.0001, n1 = 152, n2 = 218) (Figure 4A). Females that abandoned (versus remained) during days 5–10, however, did not exhibit an increase in the number of future broods initiated (z = 1.3, p =.20, n1 = 38, n2 = 180) (Figure 4A). When comparing females that abandoned versus stayed during days 11–20 in 1998, there was also no significant difference in the number of future broods initiated (z = 1.2, n1 = 81, n2 = 99, p =.23) (Figure 4A). It appears that early abandoning females were still able to lay a full clutch of eggs before leaving. There was no difference in the size of the initial egg mass (in millimeters) for females that abandoned versus stayed during days 0–4 (z = 1.2, p =.22, n1 = 152, n2 = 218), days 5–10 (z = 1.1, p =.26, n1 = 38, n2 = 180), or days 11–20 (z = 0.5, p =.61, n1 = 81, n2 = 99). However, females that abandoned initiated larger egg masses (on average) in the future relative to females that remained for days 0–4 (z = 4.9, p <.0001, n1 = 117, n2 = 107) (Figure 4B) and for days 5–10 (z = 3.9, p =.0001, n1 = 20, n2 = 87) (Figure 4B), but not for days 11–20 (z = 0.36, p =.72, n1 = 44, n2 = 43) (Figure 4B).

This increase in both the number and size of future egg masses translated into a higher lifetime fecundity for females that abandoned early. I added up the total sizes of all egg masses initiated over a female's lifetime, revealing that females that abandoned versus remained had a higher lifetime fecundity for days 0–4 (z = 5.5, p <.0001, n1 = 152, n2 = 218) and for days 5–10 (z = 2.6, p =.01, n1 = 38, n2 = 180), but not for days 11–20 (z = 0.15, p =.88, n1 = 81, n2 = 99). In addition, abandoning females guarded future broods for a longer period, potentially resulting in higher hatching success. Females that abandoned had longer guarding durations (on average) in the future relative to females that remained for days 0–4 (z = 4.1, p <.0001, n1 = 117, n2 = 107) (Figure 4C) and for days 5–10 (z = 1.8, p =.06, n1 = 20, n2 = 87) (Figure 4C), but for not days 11–20 (z = 0.55, p =.58, n1 = 44, n2 = 43) (Figure 4C). In general, the costs of abandoning diminished as hatching approached (when females are likely to abandon anyway). The overall relationships between future fitness female guarding duration are summarized in Table 1, within the context of Equation 1.

### Combining the costs and benefits for the two alternative tactics

Because the time to hatching (i.e., incubation) varied across egg masses, I subtracted the departure date from the hatching date for each female. The distribution of these values clearly shows that female durations fell in a bimodal distribution that matches an “early abandonment” tactic and an “abandon after hatching” tactic (Figure 5; here n = 311 as some broods did not hatch any nymphs). I divided female guarding durations into those in which females stayed either less than or at least 12 days after brood initiation; this division matched the two groups in Figure 5. For each individual female (n = 370), I estimated the lifetime number of nymphs by multiplying the length of the egg masses by the predicted hatching success. The predicted hatching success of each individual brood was determined from the guarding duration of the female (on that particular brood) and the linear model in Figure 1. I then summed these estimates across all broods for each female by using Equation 1, which provided an overall estimate of nymph production. Comparing w(t1) for the two groups in Figure 5 revealed roughly equivalent lifetime fitness for the alternative tactics of abandoning versus remaining for the first clutch (42.78 nymphs versus 39.72 nymphs, respectively; z = 0.70, p =.48).

## DISCUSSION

Egg guarding is considered the most basic and primitive form of parental care, and an understanding of its evolution can lend insight into the origin of sociality in insects (Tallamy, 1984; Tallamy and Wood, 1986). The adaptive significance of egg guarding in Publilia concava was revealed from the positive linear relationship between guarding duration and hatching success. Females that abandoned immediately after egg laying suffered a 50% reduction in hatching success relative to that of females that remained through hatching. The removal experiment verified that female abandonment causes this decrease in hatching success (as opposed to females abandoning damaged clutches). There was a 50% reduction in hatching success for females that were removed relative to controls. However, the absolute number of offspring surviving (for both guarding and abandoning females) was twice as high in the field where the removal experiment was conducted. Very few control females abandoned their egg masses in this field, possibly reflecting a response to conditions that make egg guarding more effective. For example, this experimental field had more than 20 ant mounds, which may have translated into higher overall ant presence relative to the second field. Overall, these results match those from the treehopper Entylia bactriana, in which early abandonment of eggs is also associated with nearly a 50% reduction in hatching success (Olmstead and Wood, 1990b; Wood, 1977). In the treehopper Umbonia crassicornis, Wood (1976) also found that fewer nymphs emerged from egg masses when females were removed. By counting the exact number of eggs in the masses of P. concava females, I was able to determine the proportion of eggs that hatched for broods with and without females removed. This comparison verified the causal relationship between female egg guarding and increases in hatching success.

Despite the benefits of egg guarding in terms of hatching success, there may also be large costs associated with decreases in future reproduction. Previous work with treehoppers has found that abandoning females have a higher probability of laying a second clutch (Bristow, 1983; Del-Claro and Oliveira, 2000; Olmstead and Wood, 1990b). P. concava females that abandoned had a greater number of future clutches, larger future clutches, and a greater lifetime fecundity. Abandoning females also exhibited longer guarding durations for future clutches, possibly reflecting a limitation on overall guarding time. It appears that any benefits of abandoning (in terms of future fecundity) are only present for P. concava females that abandon within the first few days after laying eggs. This may explain why the females that do not abandon early usually remain through hatching. Future work on the costs of egg guarding will require the experimental removal and subsequent monitoring of marked females at various stages of the egg guarding period.

Marked P. concava females exhibited a bimodal distribution of female guarding durations, with the peaks corresponding to the tactics of immediate abandonment or guarding through hatching. Separating females that fall into these two groups revealed that both tactics have equivalent estimates of lifetime fitness. This result raises the possibility that the alternative tactics of abandonment versus guarding are part of a mixed strategy that is maintained at equilibrium within the population. Another possibility is that females are adopting a conditional strategy in which the choice to abandon or guard depends on female phenotype and the conditions surrounding an egg mass. These alternative hypotheses are not necessarily mutually exclusive, and theoretical work suggests that frequency dependence can maintain a particular switchpoint for the choice of alternative tactics within a conditional strategy (Parker, 1984; Repka and Gross, 1995). Under conditions in which the benefits of remaining are large, females can be expected to adopt the tactic of guarding until hatching. For example, the parental care tactic is more suitable when oviposition sites are of high quality owing to plant nutrition, refuge from predators, or high ant presence. Under some conditions, female presence may also act to modify the host plant in a way that increases nutrients for offspring. In contrast, under conditions in which the costs of remaining are large, females can be expected to adopt the tactic of brood abandonment. This might occur if females have many more eggs to lay (high egg load) or if females have a limited amount of time to initiate and guard new clutches. Although there is no obvious physical component to egg protection in P. concava, other than a female positioning herself atop eggs, there may be unseen physiological costs to egg guarding. For example, guarding females are forced to feed on older leaves (containing eggs) rather than the nitrogen-rich leaves that are emerging at the top of the plant.

For P. concava females that remain to guard, the synchrony of female departure and hatching date suggests that female presence may be less important once nymphs have hatched. In a previous experiment that removed P. concava females from nymphs, female presence had no effect on nymph survival (McEvoy, 1979). A similar experiment that removed Publilia reticulata females from nymphs found the same result (Bristow, 1983). In both of these experiments, however, the removal of ants caused a significant reduction in nymph survival (regardless of female presence). In Publilia modesta the removal of females only resulted in lower nymph survival when ants were present, suggesting that one motivation for female guarding is to attract ants (Billick et al., 2001). Female egg guarding can also help to establish ant presence, increasing eventual ant to nymph ratios (Billick et al., 2001; Olmstead and Wood, 1990b). It is likely that these indirect effects of ants are important in Publilia concava because early-season removal of ants causes females to desert plants (Zink, unpublished data). Also, P. concava females actively avoid ovipositing on plants without ants (Morales, 2002). In the removal experiment with P. concava, I found that female presence dramatically increased ant levels in the area around eggs. It is possible that females remain with eggs to attract ants and that ants then reduce levels of predation and parasitism on eggs. Ant levels increased dramatically as P. concava nymphs emerged, reflecting the recruitment of ant workers that capitalize on honeydew production by the nymphs. Previous work suggests that by abandoning ant-tended nymphs (such as in this study), female treehoppers are essentially turning over parental care to ants (Bristow, 1983; Del-Claro and Oliveira, 2000).

Although species such as Umbonia crassicornis (Dowell and Johnson, 1986; Wood, 1976) and Polyglipta dispar (Eberhard, 1986) actively protect eggs using elaborate defensive behaviors, Publilia species do not exhibit overt defensive behaviors (Billick et al., 2001; Bristow, 1983; McEvoy, 1979). In a survey of New World membracid species, Wood (1993) found that ant-tended membracids show fewer defensive behaviors compared to non–ant-tended species. One possible explanation is that ant tending provides enemy-free space (Atsatt, 1981) through the indirect protection of eggs. As a result, ant-tended species may have lost many of the adaptations for direct offspring defense (Olmstead and Wood, 1990a). Additional studies with P. concava should address the question of whether ants actively protect eggs by comparing the hatching success of eggs with and without ants when females are both present and absent. Overall, the work presented here brings us much closer to understanding the fitness costs and benefits of parental care in treehoppers and will guide future work on the proximate mechanisms facilitating parental care in treehoppers.

Figure 1

Correlation between the hatching success of a brood and the total number of days that a guarding female was present in 1998. The horizontal dashed line represents the average hatching success for all broods in the figure (0.232 nymphs per egg), and the vertical line represents the average day of hatching for all broods in the figure (21.55 days)

Figure 1

Correlation between the hatching success of a brood and the total number of days that a guarding female was present in 1998. The horizontal dashed line represents the average hatching success for all broods in the figure (0.232 nymphs per egg), and the vertical line represents the average day of hatching for all broods in the figure (21.55 days)

Figure 2

Mean (+1 SE) of the total number of eggs laid in the removal experiment and the total number of nymphs that emerged for each brood

Figure 2

Mean (+1 SE) of the total number of eggs laid in the removal experiment and the total number of nymphs that emerged for each brood

Figure 3

Fraction of control broods in the removal experiment with a female present over time (open circles), as well as the average number of nymphs after hatching (closed circles, 22 June). Note that although a fraction of females had abandoned early, later female abandonment coincides exactly with nymph emergence. An increase in the average number of ants found around each brood also coincided with the emergence of nymphs

Figure 3

Fraction of control broods in the removal experiment with a female present over time (open circles), as well as the average number of nymphs after hatching (closed circles, 22 June). Note that although a fraction of females had abandoned early, later female abandonment coincides exactly with nymph emergence. An increase in the average number of ants found around each brood also coincided with the emergence of nymphs

Figure 4

Means (+1 SE) of the future number of broods initiated (A), the mean length of those future broods (B), and the mean future guarding duration (C) for females abandoning versus remaining for each of three stages of egg guarding. **p <.0001, *p <.01

Figure 4

Means (+1 SE) of the future number of broods initiated (A), the mean length of those future broods (B), and the mean future guarding duration (C) for females abandoning versus remaining for each of three stages of egg guarding. **p <.0001, *p <.01

Figure 5

The distribution of departure dates relative to the day of hatching for all females in their first brood initiated. Note that the two peaks correspond with 20–22 days before hatching (i.e., immediately after laying eggs) and the days immediately after hatching

Figure 5

The distribution of departure dates relative to the day of hatching for all females in their first brood initiated. Note that the two peaks correspond with 20–22 days before hatching (i.e., immediately after laying eggs) and the days immediately after hatching

Table 1

Correlation between the number of days spent guarding the first brood (t1) and future fitness.

Variable Explanation r p
$${{\sum}_{\mathit{i}{=}2}^{\mathit{N}}}$$
Pi
Total future fecundity (brood lengths) −.42 <.0001
$${{\sum}_{\mathit{i}{=}2}^{\mathit{N}}}$$
Pi/N
Average future brood size (length) −.39 <.0001
$${{\sum}_{\mathit{i}{=}2}^{\mathit{N}}}$$
Pi * g(ti)
Total future fitness (nymphs produced) −.45 <.0001
N Number of future broods −.34 <.0001
$${{\sum}_{\mathit{i}{=}2}^{\mathit{N}}}$$
Pi
$${{\sum}_{\mathit{i}{=}2}^{\mathit{N}}}$$
ti
Average future guarding duration −.26 .0001
Variable Explanation r p
$${{\sum}_{\mathit{i}{=}2}^{\mathit{N}}}$$
Pi
Total future fecundity (brood lengths) −.42 <.0001
$${{\sum}_{\mathit{i}{=}2}^{\mathit{N}}}$$
Pi/N
Average future brood size (length) −.39 <.0001
$${{\sum}_{\mathit{i}{=}2}^{\mathit{N}}}$$
Pi * g(ti)
Total future fitness (nymphs produced) −.45 <.0001
N Number of future broods −.34 <.0001
$${{\sum}_{\mathit{i}{=}2}^{\mathit{N}}}$$
Pi
$${{\sum}_{\mathit{i}{=}2}^{\mathit{N}}}$$
ti
Average future guarding duration −.26 .0001

Fitness variables are expressed as components of Equation 1.

I am grateful to R. Bergman and J. Basile for access to their land. B. Barbosa, C. McDonnell, and E. Pueschel provided excellent help with the field work. B. Smith and A. Wild kindly helped with identification of the mites and ants. I thank P. Buston, B. Danforth, M. Geber, M. Hauber, H. K. Reeve, R. Root, P. Sherman, D. Westneat, and two anonymous reviewers for providing excellent comments on earlier drafts. This work was supported by an Andrew W. Mellon Research Fellowship, the Ecology and Systematics Student Research Fund at Cornell University, and a National Science Foundation Graduate Research Fellowship.

## REFERENCES

Andersson M, Wicklund CG, Rundgren H,
1980
. Parental defence of offspring: a model and an example.
Anim Behav

28
:
536
-542.
Atsatt PR,
1981
. Lycaenid butterflies and ants: selection for enemy-free space.
Am Nat

118
:
638
-654.
Billick I, Weidmann M, Reithel J,
2001
. The relationship between ant-tending and maternal care in the treehopper Publilia modesta.
Behav Ecol Sociobiol

51
:
41
-46.
Bristow CM,
1983
. Treehoppers transfer parental care to ants: a new benefit of mutualism.
Science

220
:
532
-533.
Carlisle TR,
1982
. Brood success in variable environments: implications for parental care allocation.
Anim Behav

30
:
824
-836.
Clutton-Brock T,
1991
. The evolution of parental care. Princeton,. New Jersey: Princeton University Press.
Cocroft RB,
1996
. Insect vibrational defense signals.
Nature

382
:
679
-680.
Cocroft RB,
1999
. Parent-offspring communication inn response to predators in a subsocial treehopper (Hemiptera: Membracidae: Umbonia crassicornus).
Ethology

105
:
553
-568.
Cocroft RB,
2002
. Antipredator defense as a limiting resource: unequal predation risk in broods of an insect with maternal care.
Behav Ecol

13
:
125
-133.
Crespi BJ,
1990
. Subsociality and female reproductive success in mycophagous thrips: an observational and experimental analysis.
J Insect Behav

3
:
61
-74.
Del-Claro K, Oliveira PS,
2000
. Conditional outcomes in a neotropical treehopper-ant association: temporal and species-specific variation in ant protection and homopteran fecundity.
Oecologia

124
:
156
-165.
Dowell RV, Johnson M,
1986
. Polistes major (Hymenoptera: Vespidae) predation of the treehopper, Umbonia crassicornis (Homoptera: Membracidae).
Pan-Pacific Entomol

62
:
150
-152.
Eberhard WG,
1975
. The ecology and behavior of a subsocial pentatomid bug and two scelionid wasps: strategy and counterstrategy in a host and its parasitoids.
Smithson Contrib Zool

205
:
1
-39.
Eberhard WG,
1986
. Possible mutualism between females of the subsocial membracid Polyglypta dispar (Homoptera).
Behav Ecol Sociobiol

19
:
447
-453.
Eickwort GC,
1981
. Presocial insects. In: Social insects, vol. 2 (Hermann HR, ed). New York: Academic Press; 199–280.
Grafen A, Sibly R,
1978
. A model of mate desertion.
Anim Behav

26
:
645
-652.
Kaitala A, Mappes J,
1997
. Parental care and reproductive investment in shield bugs (Acanthosomatidae, Heteroptera).
Oikos

80
:
3
-7.
Maynard Smith J,
1977
. Parental investment: a prospective analysis.
Anim Behav

25
:
1
-9.
McEvoy PB,
1977
. Adaptive significance of clumped dispersion in a treehopper, Publilia concava (Homoptera: Membracidae) (PhD dissertation). Ithaca,. New York: Cornell University.
McEvoy PB,
1979
Misc Publ Entomol Soc Am

11
:
1
-13.
Morales MA,
2000
. Mechanisms and density dependence of benefit in an ant-membracid mutualism.
Ecology

81
:
482
-489.
Morales MA,
2000
. Survivorship of an ant-tended membracid as a function of ant recruitment.
Oikos

90
:
469
-476.
Morales MA,
2002
. Ant-dependent oviposition in the membracid Publilia concava.
Ecol Entomol

27
:
247
-250.
Nafus DM, Schreiner IH,
1988
. Parental care in a tropical nymphalid butterfly, Hypolimnas anomala.
Anim Behav

36
:
1425
-1431.
Nalepa CA, Bell WJ,
1997
. Postovulation parental investment and parental care in cockroaches. In: The evolution of social behavior in insects and arachnids (Choe JC, Crespi BJ, eds). Cambridge: Cambridge University Press; 26–51.
1990
. Altitudinal patterns in species richness of neotropical treehoppers (Homoptera: Membracidae): the role of ants.
Proc Entomol Soc Wash

92
:
552
-560.
1990
. The effect of clutch size and ant attendance on egg guarding by Entylia bactriana (Homoptera: Membracidae).
Psyche

97
:
111
-119.
Parker GA,
1984
. Evolutionarily stable strategies. In: Behavioural ecology: an evolutionary approach, 2nd ed (Krebs JR, Davies NB, eds). Oxford: Blackwell; 30–61.
Repka J, Gross MR,
1995
. The evolutionarily stable strategy under individual condition and tactic frequency.
J Theor Biol

176
:
27
-31.
Tallamy DW,
1982
. Age specific maternal defense in Gargaphia solani (Hemiptera: Tingidae).
Behav Ecol Sociobiol

11
:
7
-11.
Tallamy DW,
1984
. Insect parental care.
Bioscience

34
:
20
-24.
Tallamy DW,
1999
. Semelparity and the evolution of maternal care in insects.
Anim Behav

57
:
727
-730.
Tallamy DW, Denno RF,
1981
. Maternal care in Gargaphia solani.
Anim Behav

29
:
771
-778.
Tallamy DW, Schaefer C,
1997
. Maternal care in the Hemiptera: ancestry, alternatives, and current adaptive value. In: The evolution of social behavior in insects and arachnids (Choe JC, Crespi BJ, eds). Cambridge: Cambridge University Press; 94–115.
Tallamy DW, Wood TK,
1986
. Convergence patterns in subsocial insects.
Ann Rev Entomol

31
:
369
-390.
Townsend DS,
1986
. The costs of male parental care and its evolution in a neotropical frog.
Behav Ecol Sociobiol

19
:
187
-195.
Toyama M,
1999
. Adaptive advantages of maternal care and matriphagy in a foliage spider, Chiracanthium japonicum (Araneae: Clubionidae).
J Ethol

17
:
33
-39.
Trivers RL,
1972
. Parental investment and sexual selection. In: Sexual selection and the descent of man (Campbell B, ed). Chicago: Aldine Press; 136–179.
Whittingham LA, Taylor PD, Robertson RJ,
1992
. Confidence of paternity and male parental care.
Am Nat

139
:
1115
-1125.
Williams GC,
1966
. Natural selection, the cost of reproduction, and a refinement of Lack's principle.
Am Nat

100
:
687
-690.
Wood TK,
1974
. Aggregating behavior of Umbonia crassicornis.
Can Entomol

106
:
169
-173.
Wood TK,
1976
. Alarm behavior of brooding female Umbonia crassicornis (Homoptera: Membracidae).
Ann Entomol Soc Am

69
:
340
-344.
Wood TK,
1977
. Role of parent females and attendant ants in the maturation of the treehopper, Entylia bactriana (Homoptera: Membracidae).
Sociobiology

2
:
257
-272.
Wood TK,
1993
. Diversity in the new world Membracidae.
Ann Rev Entomol

38
:
409
-435.
Zink AG,
2001
. The optimal degree of parental care asymmetry among communal breeders.
Anim Behav

61
:
439
-446.