Abstract

Two current models seek to explain reproduction of subordinates in social groups: incentives given by dominants for peacefully remaining in the group (reproductive skew model) or incomplete control by dominants. These models make different predictions concerning genetic relatedness between individuals for the distribution of reproduction and the stability of cooperative breeding associations. To test these models and to further explore the relationships between reproductive skew, genetic relatedness, and investment of each participant, we performed behavioral observations of female wood mice (Apodemus sylvaticus) raising pups communally. Our results do not support previous models. Differences in lifetime reproductive success were significantly greater within mother—daughter pairs than within pairs of sisters or unrelated females. Subordinate females of either breeding unit did not differ in their direct reproduction. Calculations of inclusive fitness based on our results lead to the following predictions: (1) Communal nests should occur only when ecological circumstances prevent solitary breeding. (2) Subordinate females gain the highest inclusive fitness joining their mothers; they also show the highest nursing investment. (3) Mothers should accept daughters, who have no opportunity for solitary breeding. (4) Dominant sisters and unrelated females should reject subordinate females because cooperative breeding reduces their reproductive success. However, breeding units of dominant sisters and unrelated females nevertheless occur and can be explained by our finding that such females significantly reduce nursing time, which may help them save energy for future breeding cycles. Our data demonstrate that both genetic relatedness and investment skew are important in the complex evolution of reproductive skew in cooperative breeding.

Communal breeding involves several contradictory driving forces of a genetic, behavioral, and ecological nature. Social groups in which a single breeder monopolizes reproduction are described as having high reproductive skew, whereas low-skew groups are those in which reproduction is distributed more equally over some or all group members (Vehrencamp, 1983). Several recent studies have attempted to understand what drives this reproductive skew (Bernasconi et al., 1997; Cant, 2000; Creel and Waser, 1997; Komdeur, 1996). The results have led to two different models. The (classical) skew model is based on the assumption that reproductive sharing within social groups is the result of reproductive incentives given by dominants to subordinates to keep them as peaceful helpers in the group (Keller and Reeve, 1994; Reeve and Keller, 1995; Reeve et al., 1998; Vehrencamp, 1983). Most examples fitting this hypothesis have been found in ants and other social insects (see Keller, 1991). An alternative model is based on the assumption that reproduction by subordinates is the result of incomplete control by dominants (Clutton-Brock, 1998; Reeve et al., 1998). This may be the case in the naked mole rat (Heterocephalus glaber), where dominant females are sometimes unable to suppress the reproductive activity of subordinates (Jarvis, 1991).

These two skew models make different predictions about the role of relatedness between dominants and subordinates and about the regulation of the subordinate's share of reproduction. The optimal skew model predicts that the subordinate's reproduction is inversely proportional to its relatedness to the dominant. Assuming that dominants are benefiting from the presence of subordinates, they should try to keep subordinates as helpers in the group. It follows that dominants should allow less related individuals to raise more of their own offspring as an inducement to remain in the cooperative group, whereas more related individuals might stay with less of their own reproduction because they can increase their inclusive fitness by helping relatives raise their offspring (Vehrencamp, 1983). The decision of subordinates to stay rather than to leave to breed alone—that is, the assessment of dispersal risk—is based on ecological factors such as food and nesting availability. The optimal skew model further predicts that reproductive skew is great in breeding units with asymmetrical relatedness such as in matrifilial associations (the daughter is more related to the offspring of her mother than vice versa) and that reproductive skew is small in units composed of individuals of the same generation in which relatedness to offspring is symmetrical (Keller and Reeve, 1994; Reeve and Keller, 1995). Assessing the degree of relatedness among communally breeding females is complicated when females mate with different males during one estrous period; this results in a decrease of relatedness among offspring and affects an important component of the model (Reeve and Keller, 1996). Increased frequency of multiple matings will tend to decrease reproductive skew in both matrifilial and sibling associations, but such that skews in the former still exceed skews in the latter (Reeve and Keller, 1996).

The incomplete control hypothesis predicts just the opposite effect: when relatedness is symmetrical among group members, reproductive skew should decrease with, or be insensitive to, increasing genetic relatedness among them because dominants should prefer to tolerate reproduction in more related individuals (Cant, 1998; Clutton-Brock, 1998; Reeve et al., 1998). The occurrence of subordinate reproduction in asymmetrical parent—offspring associations is more consistent with the optimal skew than with the incomplete control model where the subordinate is predicted to obtain no direct reproduction (Reeve et al., 1998).

Neither model predicts high levels of aggressive testing by subordinates and assertion of status by dominants, but the optimal skew model predicts higher aggression for parent—offspring associations than the incomplete control model (Reeve and Ratnieks, 1993; Reeve et al., 1998; see also review on queen—queen interaction in social insects by Heinze, 1993; but see also Cant and Johnstone, 2000). According to the incomplete control model both the dominant and the subordinate will exert decreasing aggression effort to enhance their shares of group reproduction as their relatedness increases. At all values of relatedness the subordinate's aggressive effort will exceed that of the dominant. This will increase the payoff for the subordinate's aggressive testing of dominants.

The two models do not make clear predictions about division of work and investment. Evidence from Polistes wasps suggests that foundress associations in species with high reproductive skew have a more sharply defined caste system than associations with low skew (Heinze, 1993; Reeve and Ratnieks, 1993). It is not clear whether this varies with genetic relatedness between females. According to the incomplete control model, dominants might tolerate less investment of related subordinates better than of unrelated. One of the best known examples in vertebrates is the cooperative breeding system of the pied kingfisher, where less related secondary helpers invest less in feeding hatchlings (Reyer, 1984).

To test which model best explains the relationships between reproductive skew, genetic relatedness (symmetrical and asymmetrical), and investment of effort, we performed behavioral observations on female wood mice (Apodemus sylvaticus). Wood mice are an excellent model organism due to their variable sociality. In addition to nests of solitary mothers, communal nests of two reproductive females pooling their offspring were frequently observed in a large outdoor enclosure (Musolf and Gerlach, in preparation) and telemetry studies revealed communal use of nest sites of reproductive females under natural conditions (Wilson and Montgomery, 1992). Concurrent studies of the genetic structure of wild-living wood mice show that strict family groups of related adult animals living together within one territory do not exist and that the probability of a female encountering related as well as unrelated individuals is high (Gerlach, in preparation). Therefore, related and unrelated females might be chosen as cooperative partners when raising offspring together. This provides an important and interesting difference with previous studies on cooperative behavior in species such as house mice (König, 1994a) that live in strict family units where access and choice of unrelated females is less probable. A high degree of multiple mating of female wood mice was found under natural conditions (Baker et al., 1999) and in an enclosure experiment (Bartmann and Gerlach, 2001), indicating a promiscuous mating system in wood mice. In enclosure experiments in which four females were living together with four males, the fraction of daughters to share the same father within one litter was 0.44, and, considering consecutive litters, only 20% of all daughters born by one mother shared the same father (Bartmann and Gerlach, 2001). Taking these multiple matings into account, the relatedness of daughters to the offspring of their mothers ranges between 0.25 and 0.5, and the relatedness of a female to the offspring of her daughter always equals 0.25. This causes an asymmetrical relatedness between mother—grandchild and daughter—sib/half-sib.

In this study of reproductive skew and cooperative breeding in wood mice, we wanted to test the impact of relatedness on reproductive success, nursing investment, infanticide, and aggressive interactions between females raising offspring together. We included for comparison the reproductive success of solitarily breeding females. To quantify all parameters, we conducted a laboratory study where relatedness and dispersal could be controlled and where behavioral interactions could be observed. As it is likely that in nature ecological constraints determine the probability of dispersal and thus reproduction outside the familiar group, we prevented dispersal to enhance the effects on reproductive skew.

MATERIALS AND METHODS

Experimental design

Experimental animals were the first and second filial generations of wood mice trapped from different places near Konstanz, southern Germany. All animals were kept at 20°C on a 10:14 h light:dark cycle (light on at 1000 h). Food and water were available ad libitum. At 4 weeks of age, experimental animals were separated from their parents and kept in groups of the same sex, with the exception of animals destined for mother—daughter (MD) breeding units. In that case, when a daughter was 4 weeks old, her father and littermates were removed and she was kept together with her mother for the next 4 weeks. The father had been separated before the mother gave birth to ensure that she was not becoming pregnant again. Other females were assigned to one of three additional experimental groups: one female alone (M), a breeding unit of two sisters of the same age who were never separated from each other (S), or a breeding unit of two unrelated, unfamiliar females of the same age (F).

At the beginning of each experiment, only mothers of experimental group MD were 16 weeks old and had given birth to one litter; all other females were 8 weeks old and sexually inexperienced. Both females of a breeding unit were marked individually by shaving fur in different places. One male (8-10 weeks old) who was unfamiliar and unrelated to the females was added to each of the experimental groups. In case of the MD associations, this was not the father of the daughter.

Experimental setup for groups MD, S, and F consisted of three cages in a line (each 42 × 26 × 15 cm) and two nest-boxes (each 15 × 15 × 15 cm) linked to either cage 1 or 3 by a tube; those for experimental group M consisted of two cages and one nest-box. Each experimental group was kept for 120 days, which is considered to equal the life span of wood mice under natural conditions (Gerlach, in preparation) and was checked daily for litters. We measured body weight of the young on the day after birth (corresponding to day 1 of lactation) and on day 28, when they were weaned and removed from the experimental group. Several times per week we conducted behavioral observations for 30 min, equally divided between day and night. Two activities of the females were recorded and analyzed: time spent nursing and number of aggressive interactions. We analyzed 595 observation periods; sample size for each breeding unit was M: n = 11, MD: n = 10, S: n = 10, and F: n = 9.

Data analysis

To compare differences in reproductive success (number of offspring and litters and body weight of offspring) among the experimental groups, we used ANOVA procedures of the program JMP (SAS Institute, 1995) with further contrast analysis.

Both females of a breeding unit were suckling not only their own offspring, but also the offspring of the other female. When evaluating whether the two females invested different amounts of time in nursing, the age of her own and the age of the other female's offspring had to be taken into account because offspring of the two females within a breeding unit were of different age. Thus, we calculated nursing times of dominant and subordinate females separately and evaluated nursing times according to the age categories of a female's pups. To examine the factors influencing the time a female spent nursing, we used an ANOVA, Procedure Mixed (SAS Institute, 1992) for mixed linear models (models with both fixed and random effects). In the test design, each female was considered a random effect to account for the repeated measurement design. We started the analysis with a comprehensive model of potentially relevant factors, preferring those models with a minimum Schwartz's Bayesian criterion in the model-fitting information. Schwartz's Bayesian criterion is a sensitive indicator of the goodness of fit that compensates for the tendency of many goodness of fit indicators to increase with an increasing number of factors.

When calculating direct and indirect reproductive success, we estimated the relatedness, r, of a mother to her own pups as r = 0.5; the relatedness of a mother to the offspring of her daughter as r = 0.25; a daughter to the offspring of her mother, due to different paternity in our experimental design, as r = 0.25; two sisters to each other's offspring as r = 0.25; and r = 0 for unrelated females to each other's pups.

RESULTS

Reproductive success

Females always put their young together in one nest and nursed them communally. But because their litters were of different age, it was easy to recognize to which mother the pups belonged. The reproductive success of a female was defined as the number of young reaching the age of 17 days within the experimental period of 120 days. (At 17 days, pups begin to eat solid food and are assumed to be capable of surviving on their own.) Figure 1 shows the number of offspring born and weaned in the four experimental groups. There were significant differences between experimental groups in number of offspring born (F = 9.29, df = 3, p =.0001) and weaned (F = 2.95, df = 3, p =.0455). On average, one mother alone (M) gave birth to 17.9 (± 1.5 SE) offspring and weaned 16.2 (± 1.3), whereas cooperatively breeding females gave birth to and weaned more joint offspring per breeding unit than solitary mothers (MD: born, 31.2 ± 1.9, weaned, 24.7 ± 2.1; S: born, 24.5 ± 1.9, weaned, 20.3 ± 2.0 SE; F: born, 28.5 ± 2.0, weaned, 20.6 ± 2.4; Figure 1). There was no statistical difference between the three breeding units in the number of pups born (F = 1.95, df = 2, p = 0.59) or weaned (F = 1.67, df = 2, p =.10).

Figure 1

Number of offspring born and weaned in different experimental groups (M: solitary mother, MD: mother—daughter, S: sisters, F: unrelated females). Differences among all experimental groups (M, MD, S, F) were statistically significant in number of offspring born (F = 9.29, df = 3, p =.0001) and weaned (F = 2.95, df = 3, p =.046).

Figure 1

Number of offspring born and weaned in different experimental groups (M: solitary mother, MD: mother—daughter, S: sisters, F: unrelated females). Differences among all experimental groups (M, MD, S, F) were statistically significant in number of offspring born (F = 9.29, df = 3, p =.0001) and weaned (F = 2.95, df = 3, p =.046).

However, within a breeding unit, one of the two females always weaned far more offspring than the other (on average twice, see Table 1 and Figure 1). We defined the female with the larger number of offspring as the dominant and the other as the subordinate female (e.g., MDdom, MDsub, etc.). To analyze whether variances in number of young produced were higher in breeding units of cooperating females than in females raising offspring solitarily, we randomly distributed solitary mothers into two groups. Differences in number of offspring between these two groups of solitary mothers were significantly smaller than between dominant and subordinate females of the breeding units (young born: t = 2.11, p =.043; young weaned: t = 2.92, p =.007). Therefore, we concluded that differences in reproductive success between subordinate and dominant females cannot be explained by normal variance between females.

Table 1

Parameters of reproductive success of solitary and cooperatively breeding females

  M   MDdom  MDsub  Sdom  Ssub  Fdom  Fsub  Statistical comparisona 
M: solitary mother, MD: mother—daughter, S: sisters, F: unrelated females; dom: dominant females, sub: subordinate (least square means and standard errors).  
aFor statistical comparisons, asterisks indicate significant differences; all others are nonsignificant. Number of litters: a, b-d (t = 3.07, p =.004); a-c (t = 2.23, p =.031); b, d-c (t = 2.79, p =.007). Number of pups born: a-b (t = 1.77, p =.084); a-d (t = 0.99, p =.32); a-c (t = 6.4, p = 2.2 × 10-7); b-d (t = 1.88, p =.067). Number of pups born/litter: a-b (F = 1.15, df = 3, p =.34); a-c (F = 9.93, df = 3, p <.0001); b-c (t = 3.42, p =.002). Number of pups weaned: a-b (t = 0.157, p =.87); a-d (t = 2.25, p =.031); a-c (t = 7.04, p = 7.04 × 10-7); b-d (t = 1.93, p =.061). Body weight of weaned offspring: a-b (F = 2.17, df = 3, p =.63); a-c (F = 0.52, df = 3, p =.67); b-c (F = 0.06, df = 2, p =.81). Number of dead pups: a-b (F = 0.91, df = 3, p =.48); a-c (F = 1.93, df = 3, p =.14). Total number of own and foreign pups was calculated according to Gadagkar (1991) by summing up number of own weaned pups (multiplied by r = 0.5) and number of weaned pups born by the female counterpart (multiplied by the degree of relatedness to those pups) (MDdom to MDsub's pups: r = 0.25, MDsub to MDdom's pups: r = 0.25, Sdom and Ssub to each other's pups: r = 0.25, Fdom and Fsub to each other's pups: r = 0; see methods for further explanation): a-b (t = 2.26, p =.028); a-c (t = 0.169, p =.86); a-d (t = 3.6, p =.0006); b-c, d (t = 4.07, p =.0001). Nursing time: a-b (t = 0.99, p =.33), a-c (t = 2.31, p =.027); a-d (t = 4.06, p =.0002); b, c-d (t = 5.8, p = 3×10-7). Nursing time/total reproduction was calculated by dividing mean values of time spent nursing by total reproduction (direct and indirect): a-c (t = 0.82, p =.41); a-d (t = 0.74, p =.46); a-e (t = 2.5, p =.0145); b-c (t = 0.9, p =.369)); d-e (t = 3.57, p =.0007). Probability values in the above tests were adjusted for multiple simultaneous tests using Bonferroni corrections.  
Litters   3.2 ± 0.2 (a)   3.4 ± 0.2 (b)   2.4 ± 0.3 (c)   2.6 ± 0.2 (d)   2.4 ± 0.3 (c)   3.2 ± 0.2 (b)   2.7 ± 0.3 (c)   a, b-d*; a- c; b, d-c* 
Pups born   17.9 ± 1.4 (a)   19.7 ± 1.5 (b)   10.1 ± 1.1 (c)   15.6 ± 1.4 (d)   8.9 ± 1.1 (c)   16.8 ± 1.5 (d)   11.8 ± 1.1 (c)   a-b; a-d; a-c*; b-d  
Pups born/litter   5.6 ± 0.4 (a)   5.6 ± 0.4 (b)   4.7 ± 0.3 (c)   6.1 ± 0.4 (b)   3.7 ± 0.3 (c)   5.1 ± 0.4 (b)   4.5 ± 0.3 (c)   a-b; a-c*; b-c* 
Pups weaned   16.2 ± 1.3 (a)   15.9 ± 1.4 (b)   7.3 ± 1.1 (c)   12.6 ± 1.3 (d)   7.2 ± 1.7 (c)   12.7 ± 1.4 (d)   7.9 ± 1.3 (c)   a-b; a-d; a-c*; b-d;  
Pup weight at weaning (g)   16.3 ± 0.27 (a)   15.9 ± 0.33 (b)   15.5 ± 0.40 (c)   15.9 ± 0.29 (b)   15.8 ± 0.32 (c)   16.0 ± 0.32 (b)   16.7 ± 0.40 (c)   a-b; a-c; b-c  
Pups dead   1.7 ± 1.1 (a)   3.8 ± 1.1 (b)   2.8 ± 0.9 (c)   3.0 ± 1.1 (b)   1.2 ± 0.8 (c)   4.1 ± 1.2 (b)   3.9 ± 0.9 (c)   a-b; a-c  
Total no. of own and foreign pups   8.1 ± 0.8 (a)   10.7 ± 0.9 (b)   8.6 ± 0.9 (c)   8.2 ± 0.8 (c)   7.0 ± 0.8 (c)   6.3 ± 0.9 (c)   3.9 ± 0.9 (d)   a-b; a-c; a-d*; b-c, d* 
Nursing time (min)   15.3 ± 1.2 (a)   16.4 ± 1.6 (b)   17.7 ± 1.6 (c)   9.7 ± 1.3 (d)   8.3 ± 1.3 (d)   9.9 ± 1.4 (d)   10.0 ± 1.4 (d)   a-b; a-c; a-d*; b, c-d* 
Nursing time/total no. of own and foreign pups   1.9 ± 0.5 (a)   1.8 ± 0.6 (b)   2.6 ± 0.6 (c)   1.3 ± 0.5 (d)   1.3 ± 0.5 (d)   1.8 ± 0.6 (d)   3.9 ± 0.6 (e)   a-c; a-d; a-e; b-c; d-e* 
  M   MDdom  MDsub  Sdom  Ssub  Fdom  Fsub  Statistical comparisona 
M: solitary mother, MD: mother—daughter, S: sisters, F: unrelated females; dom: dominant females, sub: subordinate (least square means and standard errors).  
aFor statistical comparisons, asterisks indicate significant differences; all others are nonsignificant. Number of litters: a, b-d (t = 3.07, p =.004); a-c (t = 2.23, p =.031); b, d-c (t = 2.79, p =.007). Number of pups born: a-b (t = 1.77, p =.084); a-d (t = 0.99, p =.32); a-c (t = 6.4, p = 2.2 × 10-7); b-d (t = 1.88, p =.067). Number of pups born/litter: a-b (F = 1.15, df = 3, p =.34); a-c (F = 9.93, df = 3, p <.0001); b-c (t = 3.42, p =.002). Number of pups weaned: a-b (t = 0.157, p =.87); a-d (t = 2.25, p =.031); a-c (t = 7.04, p = 7.04 × 10-7); b-d (t = 1.93, p =.061). Body weight of weaned offspring: a-b (F = 2.17, df = 3, p =.63); a-c (F = 0.52, df = 3, p =.67); b-c (F = 0.06, df = 2, p =.81). Number of dead pups: a-b (F = 0.91, df = 3, p =.48); a-c (F = 1.93, df = 3, p =.14). Total number of own and foreign pups was calculated according to Gadagkar (1991) by summing up number of own weaned pups (multiplied by r = 0.5) and number of weaned pups born by the female counterpart (multiplied by the degree of relatedness to those pups) (MDdom to MDsub's pups: r = 0.25, MDsub to MDdom's pups: r = 0.25, Sdom and Ssub to each other's pups: r = 0.25, Fdom and Fsub to each other's pups: r = 0; see methods for further explanation): a-b (t = 2.26, p =.028); a-c (t = 0.169, p =.86); a-d (t = 3.6, p =.0006); b-c, d (t = 4.07, p =.0001). Nursing time: a-b (t = 0.99, p =.33), a-c (t = 2.31, p =.027); a-d (t = 4.06, p =.0002); b, c-d (t = 5.8, p = 3×10-7). Nursing time/total reproduction was calculated by dividing mean values of time spent nursing by total reproduction (direct and indirect): a-c (t = 0.82, p =.41); a-d (t = 0.74, p =.46); a-e (t = 2.5, p =.0145); b-c (t = 0.9, p =.369)); d-e (t = 3.57, p =.0007). Probability values in the above tests were adjusted for multiple simultaneous tests using Bonferroni corrections.  
Litters   3.2 ± 0.2 (a)   3.4 ± 0.2 (b)   2.4 ± 0.3 (c)   2.6 ± 0.2 (d)   2.4 ± 0.3 (c)   3.2 ± 0.2 (b)   2.7 ± 0.3 (c)   a, b-d*; a- c; b, d-c* 
Pups born   17.9 ± 1.4 (a)   19.7 ± 1.5 (b)   10.1 ± 1.1 (c)   15.6 ± 1.4 (d)   8.9 ± 1.1 (c)   16.8 ± 1.5 (d)   11.8 ± 1.1 (c)   a-b; a-d; a-c*; b-d  
Pups born/litter   5.6 ± 0.4 (a)   5.6 ± 0.4 (b)   4.7 ± 0.3 (c)   6.1 ± 0.4 (b)   3.7 ± 0.3 (c)   5.1 ± 0.4 (b)   4.5 ± 0.3 (c)   a-b; a-c*; b-c* 
Pups weaned   16.2 ± 1.3 (a)   15.9 ± 1.4 (b)   7.3 ± 1.1 (c)   12.6 ± 1.3 (d)   7.2 ± 1.7 (c)   12.7 ± 1.4 (d)   7.9 ± 1.3 (c)   a-b; a-d; a-c*; b-d;  
Pup weight at weaning (g)   16.3 ± 0.27 (a)   15.9 ± 0.33 (b)   15.5 ± 0.40 (c)   15.9 ± 0.29 (b)   15.8 ± 0.32 (c)   16.0 ± 0.32 (b)   16.7 ± 0.40 (c)   a-b; a-c; b-c  
Pups dead   1.7 ± 1.1 (a)   3.8 ± 1.1 (b)   2.8 ± 0.9 (c)   3.0 ± 1.1 (b)   1.2 ± 0.8 (c)   4.1 ± 1.2 (b)   3.9 ± 0.9 (c)   a-b; a-c  
Total no. of own and foreign pups   8.1 ± 0.8 (a)   10.7 ± 0.9 (b)   8.6 ± 0.9 (c)   8.2 ± 0.8 (c)   7.0 ± 0.8 (c)   6.3 ± 0.9 (c)   3.9 ± 0.9 (d)   a-b; a-c; a-d*; b-c, d* 
Nursing time (min)   15.3 ± 1.2 (a)   16.4 ± 1.6 (b)   17.7 ± 1.6 (c)   9.7 ± 1.3 (d)   8.3 ± 1.3 (d)   9.9 ± 1.4 (d)   10.0 ± 1.4 (d)   a-b; a-c; a-d*; b, c-d* 
Nursing time/total no. of own and foreign pups   1.9 ± 0.5 (a)   1.8 ± 0.6 (b)   2.6 ± 0.6 (c)   1.3 ± 0.5 (d)   1.3 ± 0.5 (d)   1.8 ± 0.6 (d)   3.9 ± 0.6 (e)   a-c; a-d; a-e; b-c; d-e* 

Considering solitary mothers and dominant females in all breeding units, there was no difference in the number of offspring born, but both solitary mothers and MDdom weaned more offspring than Sdom or Fdom (Table 1). Considering solitary mothers and subordinate females, solitary mothers gave birth to more pups and weaned more pups than any of the subordinate females (Table 1). We did not compare number of offspring of dominants and subordinates because they were different by definition.

Reproductive skew

Differences between the number of offspring weaned by dominant and subordinate females (reproductive skew) were significantly greater in MD breeding units than in either of the other units (MD versus S, F breeding units: t= 2.15, p =.038; Figure 2). On average, MDsub weaned 7.9 (± 1.4 SE) fewer pups than MDdom, and Ssub had 4.9 (± 1.4) and Fsub 4.8 (± 1.4) pups less than their dominant counterpart. Therefore, we conclude that reproductive skew was greater within asymmetrical breeding units than in symmetrical units, but no difference was evident between breeding units consisting of sisters and unrelated females.

Figure 2

Reproductive skew (differences between number of weaned pups of the dominant minus the subordinate female) in different breeding units (MD: mother—daughter, S: sisters, F: unrelated females).

Figure 2

Reproductive skew (differences between number of weaned pups of the dominant minus the subordinate female) in different breeding units (MD: mother—daughter, S: sisters, F: unrelated females).

Total reproductive success

Broadly generalized, Hamilton's concept of inclusive fitness (Hamilton, 1964a,b) predicts that the evolution of cooperation is favored over solitary breeding if r × b - c > 0, where r is the coefficient of genetic relatedness between these individuals, b is the benefit of cooperation in terms of additional reproduction, and c is the cost in terms of a loss of reproduction (Grafen, 1984). We calculated inclusive fitness values for each female of a breeding unit according to this formula to evaluate whether subordinate females of either breeding unit should join the nest and whether dominant females should accept subordinate females in the nest (Table 2). It turned out that subordinate females should never join a nest when they have the chance to breed solitarily. But they should always join a nest, even with a nonrelated female, when they have no chance of finding a nest site for their own. Only MDdom should accept a daughter in their nest when the daughter had no chances to breed solitarily. According to this calculation, neither Fdom nor Sdom should accept subordinate females in their nest because the cost—benefit relation for Sdom (-1.6) and for Fdom (-8.3) is > 0.

Table 2

Consequences for reproductive success resulting from group breeding or solitary breeding: kin selection analysis

  Mother-daughter   Sisters   Unrelated females  
aTwo alternatives were considered: becoming a solitary breeding female or not breeding at all (Grafen, 1984).  
Decision to join a dominant female of different relatednessa 
Number of pups when breeding solitarily   16.2   16.2   16.2  
Number of pups of subordinate females   7.3   7.2   7.9  
Costs of joining instead of breeding solitarily   8.9   9.0   8.3  
Costs of joining rather than not breeding at all   -7.3   -7.2   -7.9  
Number of pups in communal nests   23.2   20.3   21.5  
Benefit in a communal nest (pupscom - pupssol)   7.0   4.1   5.3  
Joining or breeding solitarily? r × b - c > 0 (?)   0.5 × 7.0 - 8.9 > 0   0.5 × 4.1 - 9.0 > 0   0 × 5.3 - 8.3 > 0  
  -5.4   -7.0   -8.3  
When should a subordinate female join a communal nest rather than breed solitarily?   Never   Never   Never  
Joining or not breeding at all? r × b - c > 0   0.5 × 7.0 - (-7.3) > 0   0.5 × 4.1 - (-7.2) > 0   0 × 5.3 - (-7.9) > 0  
  10.8   9.3   7.9  
When should a subordinate female join a communal nest rather than not breed at all?   Always   Always   Always  
Decision of a dominant female to accept subordinate females of different relatedness  
Number of pups when breeding solitarily   16.2   16.2   16.2  
Number of pups of dominant females   15.9   12.6   12.7  
Costs of accepting subordinate females   0.3   3.6   3.5  
Number of pups in communal nests   23.2   20.3   21.5  
Benefit in a communal nest (pupscom - pupssol)   7.0   4.1   5.3  
Joining or breeding solitarily? r × b - c > 0   0.5 × 7.0 - 0.3 > 0   0.5 × 4.1 - 3.6 > 0   0 × 5.3 - 8.3 > 0  
  3.2   -1.6   -8.3  
When should a dominant female accept a subordinate in a nest rather than breed solitarily   Always   Never   Never  
  Mother-daughter   Sisters   Unrelated females  
aTwo alternatives were considered: becoming a solitary breeding female or not breeding at all (Grafen, 1984).  
Decision to join a dominant female of different relatednessa 
Number of pups when breeding solitarily   16.2   16.2   16.2  
Number of pups of subordinate females   7.3   7.2   7.9  
Costs of joining instead of breeding solitarily   8.9   9.0   8.3  
Costs of joining rather than not breeding at all   -7.3   -7.2   -7.9  
Number of pups in communal nests   23.2   20.3   21.5  
Benefit in a communal nest (pupscom - pupssol)   7.0   4.1   5.3  
Joining or breeding solitarily? r × b - c > 0 (?)   0.5 × 7.0 - 8.9 > 0   0.5 × 4.1 - 9.0 > 0   0 × 5.3 - 8.3 > 0  
  -5.4   -7.0   -8.3  
When should a subordinate female join a communal nest rather than breed solitarily?   Never   Never   Never  
Joining or not breeding at all? r × b - c > 0   0.5 × 7.0 - (-7.3) > 0   0.5 × 4.1 - (-7.2) > 0   0 × 5.3 - (-7.9) > 0  
  10.8   9.3   7.9  
When should a subordinate female join a communal nest rather than not breed at all?   Always   Always   Always  
Decision of a dominant female to accept subordinate females of different relatedness  
Number of pups when breeding solitarily   16.2   16.2   16.2  
Number of pups of dominant females   15.9   12.6   12.7  
Costs of accepting subordinate females   0.3   3.6   3.5  
Number of pups in communal nests   23.2   20.3   21.5  
Benefit in a communal nest (pupscom - pupssol)   7.0   4.1   5.3  
Joining or breeding solitarily? r × b - c > 0   0.5 × 7.0 - 0.3 > 0   0.5 × 4.1 - 3.6 > 0   0 × 5.3 - 8.3 > 0  
  3.2   -1.6   -8.3  
When should a dominant female accept a subordinate in a nest rather than breed solitarily   Always   Never   Never  

Parameters determining reproductive success

Number of litters produced

Solitary females and MDdom had more litters than the dominant females of the other two experimental groups (Table 1). MDdom produced more litters than MDsub. No difference in number of litters was found between dominants and subordinates in any of the other breeding units (Table 1). Thus, the MDdom mother appears to have a special reproductive advantage in producing litters.

Number of offspring per litter

Solitary mothers gave birth to an average of 5.6 pups per litter (Table 1), similar to dominant females of all breeding units, whereas subordinate females delivered significantly fewer pups. In the MD and S (but not F) breeding units, dominant females had a significantly larger number of pups per litter than subordinate females. It could be argued that dominant females of the MD unit had more offspring because they were older and more experienced at the beginning of the experiment; they had already given birth to a litter. However, in solitarily breeding mothers, litter size did not change with increasing number of litters (ANOVA, F = 0.089, df = 3, p =.96), and there was also no difference between the first and second litter (t = 0.62, p =.54).

Interbirth intervals

On average, a solitary female gave birth to her first litter 31.9 days (± 4.1 SE) after her first encounter with the male (Table 3). Solitary and dominant females in the different breeding units did not differ in the time span until the first litter was born. However, dominant females gave birth to their first litter significantly faster than subordinate females. MDsub had the longest interval until she gave birth to her first litter: 45 days (Table 3), but differences among breeding units for subordinate females were not statistically significant due to high variance.

Table 3

Interbirth intervals

  M   MDdom  MDsub  Sdom  Ssubfit.  Fdom  Fsub  Statistical comparison  
Time (least square means and standard errors) until birth of first litter and consecutive litters (interbirth interval) in the different experimental groups: M: solitary mother, MD: mother/daughter, S: sisters, F: unrelated females; dom: dominant females, sub: subordinate females. Time until first litter was born: a-b (F = 0.099, df = 3, p =.96); a-c (F = 0.67, df = 3, p =.57); b-c (F = 4.33, df = 2, p =.042). Intervals between first and consecutive litters: a-b (F = 1.49, df = 3, p =.23); a-c (F = 1.1, df = 3, p =.36); b-c (F = 4.99, df = 3, p =.029). All comparisons nonsignificant.  
Time until first litter (days)   31.9 ± 4.1   (a) 30.8 ± 5.6   (b) 45.2 ± 5.6   (c) 29.1 ± 5.9   (b) 37.6 ± 5.9   (c) 31.4 ± 5.9   (b) 38.0 ± 5.9   (c) a-b; a-c; b-c  
Interbirth interval (days)   26.6 ± 1.1   (a) 26.0 ± 1.5   (b) 31.3 ± 2.1   (c) 28.3 ± 1.8   (b) 27.8 ± 2.1   (c) 25.0 ± 1.7   (b) 30.5 ± 2.0   (c) a-b; a-c; b-c  
  M   MDdom  MDsub  Sdom  Ssubfit.  Fdom  Fsub  Statistical comparison  
Time (least square means and standard errors) until birth of first litter and consecutive litters (interbirth interval) in the different experimental groups: M: solitary mother, MD: mother/daughter, S: sisters, F: unrelated females; dom: dominant females, sub: subordinate females. Time until first litter was born: a-b (F = 0.099, df = 3, p =.96); a-c (F = 0.67, df = 3, p =.57); b-c (F = 4.33, df = 2, p =.042). Intervals between first and consecutive litters: a-b (F = 1.49, df = 3, p =.23); a-c (F = 1.1, df = 3, p =.36); b-c (F = 4.99, df = 3, p =.029). All comparisons nonsignificant.  
Time until first litter (days)   31.9 ± 4.1   (a) 30.8 ± 5.6   (b) 45.2 ± 5.6   (c) 29.1 ± 5.9   (b) 37.6 ± 5.9   (c) 31.4 ± 5.9   (b) 38.0 ± 5.9   (c) a-b; a-c; b-c  
Interbirth interval (days)   26.6 ± 1.1   (a) 26.0 ± 1.5   (b) 31.3 ± 2.1   (c) 28.3 ± 1.8   (b) 27.8 ± 2.1   (c) 25.0 ± 1.7   (b) 30.5 ± 2.0   (c) a-b; a-c; b-c  

We suspected that communal nursing would decrease the time intervals between litters because mothers might save energy by not nursing the pups alone. Therefore, interbirth intervals were analyzed (ANOVA) using experimental group and individual female nested in experimental group (as a random factor) as independent variables. No difference was evident in interbirth intervals between solitary mothers and dominant females or between solitary mothers and subordinate females. Subordinate females had longer interbirth intervals than dominant females, but due to Bonferroni corrections the difference was no longer significant (Table 3).

Body weight of offspring

Body weight of weaned offspring did not differ significantly among the experimental groups. Additionally, pups of subordinate females weighed as much as pups of the dominant or solitary mothers (Table 1).

Aggressive interactions

Aggressive interactions between females were rare. On average, females had 0.2 (± 0.06 SE) fights per observation period with other females, and they were more aggressive against the male (0.72 ± 0.07). Agonistic encounters were highest in F breeding units (0.3 ± 0.08), lower in S (0.15 ± 0.09), and lowest in MD (0.09 ± 0.13); the differences were not statistically significant due to high variance in the data. Dominant females tended to be more aggressive against subordinate females than vice versa (t = 1.87, p =.059). There was a trend that Fdom were more aggressive against subordinates than MDdom and Sdom (t = 1.753, p =.09), while subordinates of all breeding units should equal aggressiveness against dominants. In breeding units where dominant females were aggressive against subordinates, the reproductive success of the subordinate partner was significantly lower than when no aggression was observed (Fisher's Exact test, χ 2 = 17.901, p =.0002). Thus, overt aggression, rare as it appears, seems to be involved in maintaining (and perhaps determining) dominance.

Infanticidal behavior

To further evaluate how females in breeding units control each other's reproduction, we analyzed the amount of infanticide in the experimental groups (Table 1). On average, a solitary mother weaned 1.7 offspring fewer than she gave birth to within the experimental time period of 120 days. The difference in number of offspring born and weaned was not statistically different among experimental groups or between females of the same breeding unit (Table 1). On average, 4 (± 0.7 SE) pups of dominant females and 2.5 (± 0.7) pups of the subordinate did not survive until weaning. Deaths of pups of one female occurred shortly before the other female gave birth to her next litter and when her previous litter was already weaned. This would be a good strategy to avoid killing one's own pups by mistake.

Comparison of nursing time between solitary and communally breeding females

Solitary mothers spent similar amounts of time nursing (47%) as MDdom (52%; Figure 3a), which was significantly more than either Sdom (29%; Figure 3b) or Fdom (32%; Figure 3c). Although no longer statistically significant after Bonferroni corrections, MDsub spent more time suckling (57%) than solitary mothers; whereas Ssub (26%) and Fsub (29%) spent statistically significant far less time suckling than solitary mothers (47%).

Figure 3

Nursing time of solitary females compared to dominant and subordinate females of different breeding units. (a) Mother—daughter breeding units (MDdom and MDsub). Solitary mothers nursed for the same length of time as MDdom (M — MDdom; t = 1.06, p =.29) and less than MDsub (M — MDsub; t = 2.24, p =.031). (b) Sisters (Sdom and Ssub). Solitary mothers nursed more than Sdom (M — Sdom; t = 3.98, p =.0003), and more than Ssub (M — Ssub; t = 3.96, p =.0003). (c) Unrelated females (Fdom and Fsub). Solitary mothers nursed more than Fdom (M — Fdom; t = 3.26, p =.0024) and more than Fsub (M — Fsub; t = 3.08, p =.004).

Figure 3

Nursing time of solitary females compared to dominant and subordinate females of different breeding units. (a) Mother—daughter breeding units (MDdom and MDsub). Solitary mothers nursed for the same length of time as MDdom (M — MDdom; t = 1.06, p =.29) and less than MDsub (M — MDsub; t = 2.24, p =.031). (b) Sisters (Sdom and Ssub). Solitary mothers nursed more than Sdom (M — Sdom; t = 3.98, p =.0003), and more than Ssub (M — Ssub; t = 3.96, p =.0003). (c) Unrelated females (Fdom and Fsub). Solitary mothers nursed more than Fdom (M — Fdom; t = 3.26, p =.0024) and more than Fsub (M — Fsub; t = 3.08, p =.004).

The time a female spends nursing depends on the age of her offspring. For reference we used again solitary mothers, who spent about 63% of their time nursing until the pups were 19 days old; this time decreased significantly to 31% when the pups were 20-24 days old and to 20% when they were 25-28 days old. The number of offspring per litter had no influence on the nursing time of solitary mothers (ANOVA Procedure Mixed, F1,158 = 1.28, p =.25). We investigated whether nursing time of communally breeding females depends on the nursing effort of the female counterpart and how females allocate their nursing efforts when pups of different age (and thus parentage) were present. For this we calculated separately the nursing times of dominant and subordinate females according to the age categories of their own pups.

Nursing efforts of dominant females

In a first analysis the time a dominant female was nursing was used as the dependent variable and the category age of own young as an independent variable. Further independent variables were experimental group, time the subordinate female spent nursing young, age difference between pups of the dominant and subordinate female, and the interaction of experimental group*age difference between pups of the dominant and subordinate female (Table 4). The age of their own pups had a significant effect on the mother's nursing time: with increasing pup age, dominant females spent less time suckling them. Nursing time of the subordinate female had no influence on nursing time of the dominant. The experimental group had a significant effect: the MD breeding unit spent more time nursing than the S or F unit. With increasing age difference between her own pups and the pups of the subordinate female, the dominant female spent less time suckling than solitary females. Otherwise the results seem to indicate that nursing is a physiological process decreasing with pup age and unaffected by other pups and other nursing efforts (Table 4), indicating that during times when offspring of the subordinate female were smaller and needed more milk, the dominant female reduced her nursing time.

Table 4

Time spent nursing

  df  F ratio  p 
Shown are factors influencing time spent nursing pups in different breeding units (mother—daughter, sisters, unrelated females). Age of own pups and age difference are classified in six categories: 1: 0-4 days, 2: 5-9, 3:10-14, 4:15-19, 5:20-24, 6:25-28.  
Nursing time of dominant females: dependent variable     
Independent variable:     
Experimental group (breeding unit)   2, 26   14.1   .0001  
Age of own pups   5, 288   3.6   .0036  
Age difference between own and foreign pups   1, 288   7.91   .0053  
Nursing time of the subordinate female   1, 288   0.02   .899 (ns)  
Breeding unit*age difference   2, 288   0.02   .980 (ns)  
Nursing time of subordinate females: dependent variable     
Independent variable:     
Experimental group (breeding unit)   2, 26   11.88   .0002  
Age of own pups   5, 281   4.86   .003  
Age difference between own and foreign pups   1, 281   0.28   .600 (ns)  
Nursing time of the dominant female   1, 281   0.03   .853 (ns)  
Breeding unit*age difference   2, 281   5.22   .006  
  df  F ratio  p 
Shown are factors influencing time spent nursing pups in different breeding units (mother—daughter, sisters, unrelated females). Age of own pups and age difference are classified in six categories: 1: 0-4 days, 2: 5-9, 3:10-14, 4:15-19, 5:20-24, 6:25-28.  
Nursing time of dominant females: dependent variable     
Independent variable:     
Experimental group (breeding unit)   2, 26   14.1   .0001  
Age of own pups   5, 288   3.6   .0036  
Age difference between own and foreign pups   1, 288   7.91   .0053  
Nursing time of the subordinate female   1, 288   0.02   .899 (ns)  
Breeding unit*age difference   2, 288   0.02   .980 (ns)  
Nursing time of subordinate females: dependent variable     
Independent variable:     
Experimental group (breeding unit)   2, 26   11.88   .0002  
Age of own pups   5, 281   4.86   .003  
Age difference between own and foreign pups   1, 281   0.28   .600 (ns)  
Nursing time of the dominant female   1, 281   0.03   .853 (ns)  
Breeding unit*age difference   2, 281   5.22   .006  

Nursing effort of subordinate females

The same analysis was performed using time a subordinate female was nursing as the dependent variable. Subordinate females also decreased their nursing time with increasing age of their pups and were not influenced by the nursing efforts of the other female (Table 4). But, in contrast to dominant females, differences between the age of her own and the pups of the dominant mother had no significant effect. A significant difference appeared in the interaction of experimental group*age difference, indicating an increase of breeding effort in the MD unit compared to S and F units. MDsub increased their breeding effort by 4.2% per category of age difference. Thus, when younger pups of the dominant female were present, MDsub spent more time nursing than she would have done according to the age of her own pups.

The worst situation for communally breeding females was that in nine cases the whole litter of a subordinate female was killed. In six of those cases, the subordinate female stopped nursing completely, although there were pups of the dominant female present. This shows that begging behavior of pups is not the only stimulus that induces females to nurse; mothers also noticed whether their own pups were present or not. On average, such females decreased their breeding effort by 94%.

Investment in relation to the number of own and foreign pups

To compare breeding effort of each female according to the number of own and foreign pups, we divided mean values of time spent nursing (Figure 3) by the number of pups of the communal nests, taking into account their genetic relatedness to the offspring of the other female. We added the number of a female's own pups (multiplied by r = 0.5) and the number of pups of the other female (multiplied by the relatedness between either female and the pups). No differences could any longer be revealed among females, with the exception that Fsub showed the highest effort in relation to her reproductive success. The Fsub spent significantly more time nursing pups in relation to the number of own pups than Fdom. We concluded that the revealed highest absolute breeding effort of MDsub was balanced by a high number of related pups, even when we considered that the daughter was only r = 0.25 related to the pups of her mother.

DISCUSSION

Two main results have emerged from this study: (1) a striking relationship between differences in investment and reproductive success among cooperatively breeding females that vary in relatedness and (2) convincing evidence for litter discrimination in communal nests, which is not affecting nursing investment. We discuss these results below in light of previous theories on reproductive skew.

Evolution of cooperative breeding units

According to our calculations of inclusive fitness, communal nests of females raising pups together should not occur when ecological circumstances permit solitary breeding. The costs for subordinate females joining a nest always exceeded the benefits, regardless of their relatedness to the dominant female (Table 2). In contrast, when solitary breeding is impossible—as might be the case when population density is high and appropriate nest sites are limited—subordinate females should always join nests (Table 2). They would gain the highest inclusive fitness joining their mother. Our calculations show that the mother (MDdom) should accept a daughter, who has no opportunity for solitary breeding, whereas Sdom and particularly Fdom should always reject subordinate females (Table 2). Yet, this is not the case. Here, nursing investment adds an interesting new twist to the story.

Our results on differences in nursing investment strongly emphasize the importance of maternal investment and degree of relatedness when trying to understand the occurrence of breeding associations. In comparison to solitary mothers, Sdom and Fdom reduced their nursing effort by 38% and 32%. Also Ssub and Fsub invested significantly less time nursing pups than solitary mothers. This result may explain why communal nests founded by sisters and unrelated females can still be advantageous despite a decrease in reproductive success compared to solitary breeding. By spending significantly less time nursing offspring than solitary mothers, both partners in the symmetrical relationships Sdom, Ssub and Fdom, Fsub save energy for future reproductive effort while still having moderate reproductive success. Comparing nursing time and total (direct and indirect) reproduction in S and F breeding units, we showed that Sdom and Ssub invested equally, whereas Fsub invested significantly more time nursing than Fdom relative to their total reproduction. Therefore, a subordinate female has the greatest disadvantage when breeding cooperatively with an unrelated dominant female. MDdom spent the same time nursing as solitary mothers, indicating again that she has no additional costs but also no measurable direct benefits. Subordinate females of the MD breeding units nursed more than solitary mothers, but the extraordinary effort by subordinate daughters breeding with their mothers pays off in gaining the greatest increase of inclusive fitness.

Time spent nursing as an investment in pups

Investment of effort can take several forms. Besides direct nursing, time spent with offspring is likely to enhance nest temperature and protection of young. This could result in costs such as females having less time for activities such as foraging. Perhaps the more important question that arises is to what degree our behavioral measure “time spent nursing” is correlated with energetic expenditure. Although we cannot prove this point directly because we did not quantify milk transfer, we offer the following indirect evidence. In house mice, the amount of milk produced depends on the suckling stimulus of young (Hall and Williams, 1983) and increases with increasing litter size (König et al., 1988). Milk ejections are produced in the mother by pulsatile release of oxytocin from the pituitary, which is in turn stimulated by suckling (Wakerley and Lincoln, 1971). Therefore, we assume that MDsub not only nursed longer but as a result also gave more milk than any of the other females: nursing periods were longer and more frequent, resulting in stronger suckling stimulation by pups. In addition, we can speculate that the longer time span needed to produce the next litter is indirect evidence for higher energetic investment of MDsub because her longer lactation period delayed the next reproductive cycle. Fuchs (1981, 1982) showed this relationship in house mice. The fact that longer nursing periods in the MD units did not result in greater body weight of offspring than in other experimental groups was probably the result of the greater number of pups in the MD nests. Our results concerning body weight of offspring are different from previous studies. House mice (Sayler and Salmon, 1969) showed greater pup weight in communal nests. Greatest body weight of weaned offspring was found in females living with a sister compared to monogamous females or to females communally nursing with an unrelated partner (König, 1993).

Discrimination between own and foreign offspring

The second result from our study provides convincing evidence for litter discrimination in communal nests. In our experiments, a mother's nursing time was strongly related to the age of her own pups but not to the age of foreign pups. This was seen clearly where dominant females of all breeding units spent significantly less time suckling young when the offspring of the subordinate females needed more milk than their own. Only subordinate females of the MD breeding unit broke this rule and increased their nursing time when the offspring of dominant females were younger and needed more milk than their own litter. We interpret this higher investment as a result of the daughters' greater relatedness to the offspring of the mother. All these results demonstrate the ability of female wood mice to discriminate between pups of mothers, sisters, and unrelated females. In addition, in cases where the whole litter of a subordinate female was killed, she reduced her nursing time by 94% regardless of the presence of other pups, indicating recognition of her own litter. Thus, our experiments show that nursing is not simply a response caused by the presence of unweaned pups. This ability of pup discrimination appears not to be used in preferential suckling in mixed nests (e.g., by removing foreign pups from their nipples). Coincidentally, as in the present study on well-fed wood mice, female house mice under restricted feeding conditions killed the same number of their own and foreign pups (König, 1989).

Cooperative breeding units of wood mice in regard to reproductive skew theories

Our data demonstrate an unequal distribution of reproduction between cooperating females with a higher reproductive skew in asymmetrical MD than in symmetrical S and F breeding units. This higher skew was not caused by a smaller number of pups by MDsub but by a higher number of pups produced and weaned by MDdom in comparison to breeding units of dominant sisters and unrelated females. Indeed, number of pups did not differ between subordinate females of either breeding unit. Therefore, neither different incentives (Vehrencamp, 1983) nor different tolerance levels of the dominant female toward subordinates (Clutton-Brock, 1998; Reeve et al., 1998) differed according to genetic relatedness. Direct reproductive success of all three categories of subordinates was reduced equally, both by longer time intervals until and between litters and by a smaller number of offspring born per litter. Therefore, neither of the two models fit these data. Although there was a striking difference in reproductive success between asymmetrical and symmetrical associations, we found almost no difference between the two symmetrical associations considering direct reproductive success. This result contradicts the reproductive skew model, which predicts higher reproductive incentives for unrelated females than for sisters (Vehrencamp, 1983). And considering total reproductive success, Fsub was even significantly less successful and had a greater breeding effort in relation to total reproduction than Ssub. Our results differ from results obtained on communal nursing in house mice, where communally nursing sisters had higher direct lifetime reproductive success than unrelated or solitarily breeding females (König, 1994a). In house mice, reproductive success was equally divided between cooperating sisters (König, 1994b). These differences may reflect the different social systems of house mice and wood mice (Bartmann and Gerlach, 2001). Multiple and promiscuous matings of female wood mice within one estrous period are common. Therefore, under natural conditions most of the sisters within one litter are only maternal half-sibs, and the probability is quite high that unfamiliar females are also half-sibs (paternal) because the same males have also mated with their mothers. This might decrease behavioral and reproductive differences between these breeding units.

Neither of the two models we considered originally makes clear predictions about division of work and investment. The optimal skew model predicts that subordinates should be most likely to engage in risky or energetically costly tasks (Reeve and Ratnieks, 1993) but does not clearly specify whether this should be affected by genetic relatedness. The overall results of this study do not fit the predictions of either model in a satisfying way. Perhaps reproductive skew models have been developed mainly for social insects, and their application to mammals is difficult. In addition, the existing models do not sufficiently consider the role of differential investment in raising offspring.

The predicted effect of relatedness on dominance interactions is complex. At least, when it does not significantly reduce group productivity, optimal skew models predict that the greater the reproductive skew (with increased relatedness), the greater will be the subordinate's payoff for “ testing” dominant queens (Reeve and Ratnieks, 1993). Thus, the intensity and frequency of dominance interactions should increase with increasing reproductive skew and increasing relatedness (but see also Cant and Johnstone, 2000).

In our study, aggressive encounters were highest in associations of unrelated females, but this was not statistically significant. Dominant females, defined in our study as those with the greatest reproductive fitness, tended to be more aggressive than subordinate females, which might indicate that agonistic encounters were more likely dominant aggression to control subordinates than subordinates challenging the dominant's fighting abilities. Subordinates of all breeding units did not differ in their aggressiveness against dominants, but there was a trend that Fdom were more aggressive than MDdom and Sdom. In addition, higher aggressiveness of dominant females was correlated with decreasing reproduction of subordinates. This further supports the conclusion that this aggression may be a mechanism to suppress subordinates. In several ant species, dominant cofoundresses also grow intolerant of each other after emergence of the first workers when the time of benefiting from a subordinate cofoundress is past (see Heinze, 1993). Long-term studies of naked mole rats (Heterocephalus glaber) have shown that aggression by breeding females suppresses reproduction in other females, primarily daughters (Faulkes et al., 1991). These attacks might cause physiological stress for the pregnant subordinate female, reducing her litter size and fertility. Urine pheromones, which play an important role in female—female interactions in small mammals (Jemiolo et al., 1989; Labov, 1981; Lawton and Whitsett, 1979; Wolff, 1992), might cause similar physiological changes, leading to fewer and smaller litters.

A further strategy to manipulate reproductive success of the cooperative partner is infanticidal behavior (for review see Hrdy, 1979). In our study, infanticide occurred in litters of dominant and subordinate females. Similar behavior was demonstrated in pregnant house mice killing each other's offspring (Manning et al., 1995). This was interpreted as a manipulation to invest less in foreign offspring while raising pups communally. In our study, although not quite significant at the 5% level, infanticidal behavior was highest in unrelated females. It is not unexpected that unrelated females, both dominant and subordinate, might have the least interest in investing in offspring of the counterpart and engage in greatest infanticidal behavior. It is interesting that in MD and S breeding units, infanticidal behavior of subordinates was greater than that of dominants. This might be a subordinate strategy, particularly for MDsub, to balance investment by reducing the too numerous offspring of the dominant female. It could be excluded that males performed infanticide (Gerlach, unpublished data). These results might fit the incomplete control model better than the optimal skew model, which predicts higher infanticide or oophagy in dominants (Reeve and Ratnieks, 1993), demonstrated in the ponerine ant Pachycondyla apicali (Oliveira and Hölldobler, 1991).

Conclusion

Our results do not support previous models (Clutton-Brock, 1998; Reeve et al., 1998; Vehrencamp, 1983). We show that subordinates pay for being accepted rather than being paid (receiving incentives) for being helpers. Dominant females have no advantage in terms of increasing their reproductive success by breeding communally when ecological circumstances do allow solitary breeding. Recently, Johnstone and Cant (1998) added to the complexity of cooperative breeding the interesting behavioral observations that, in many species, dominant group members forcibly evict or exclude subordinates from the group, and subordinates are reluctant to leave (Gerlach, 1990, 1996; see review on rodents: Anderson, 1989). This is also true in wood mice (Gerlach, unpublished data), and it fits the predictions based on our calculations of inclusive fitness of either breeding unit. Ecological circumstances, such as nest site availability and food limitation, are supposed to be the crucial factors determining the occurrence of communal breeding units. If nest sites are limited, MDdom should accept a daughter who has no other chances for reproduction. If ecological conditions are so harsh that saving time and energy with nursing would pay off, Fdom, and even more Sdom, should accept subordinate females because this allows them to reduce their nursing investment. Subordinates should make the best of their situation. Daughters could and did increase their inclusive fitness by helping their mothers to wean offspring. Subordinate sisters or unrelated females did not increase their fitness so effectively. Their reduced investment might be an appropriate response.

We predict that under natural circumstances, mother—daughter breeding pairs should be more common and stable than those of sisters or unrelated females. Kokko and Johnstone's model (1998) showed that stable breeding units are possible even if they lead to a decrease in immediate reproductive success as long as subordinates can gain long-term success through inheritance of territory, which may be more efficient than attempting to establish a new territory. Recently, Queller et al. (2000) showed that subordinate, unrelated helpers in the social wasp Polistes dominulus gain through inheritance of a territory. This might be an additional factor for MDsub to accept her role because she might have a greater probability to take over the maternal territory than Ssub and Fsub, who are of the same age as their dominant counterparts. Our observations in a large outdoor enclosure give first evidence that this is indeed the case (Musolf and Gerlach, in preparation).

We express our gratitude to J. Atema, L. Keller, and two anonymous referees for valuable criticism of the manuscript and to W. Nagl for statistical advice. We also thank R. Hellmann for laboratory help and H. Markl for generously supporting the project. This study was supported by the University of Konstanz, Verband der Chemischen Industrie, and by the Deutsche Forschungsgemeinschaft DFG (Ge 842/1-3).

References

Anderson PK,
1989
.
Dispersal in rodents: a resident fitness hypothesis
 . Spec Publ Am Soc Mammal 9.
Baker RJ, Makova KD, Chesser RK,
1999
. Microsatellites indicate a high frequency of multiple paternity in Apodemus (Rodentia).
Mol Ecol
 
8
:
107
-111.
Bartmann S, Gerlach G,
2001
. Multiple paternity and similar variance in reproductive success of male and female wood mice (Apodemus sylvaticus).
Ethology
 
107
:
889
-899.
Bernasconi G, Krieger M, Keller L,
1997
. Unequal partitioning of reproduction and investment between cooperating queens in the fire ant, Solenopsis invicta, as revealed by microsatellites.
Proc R Soc Lond B
 
264
:
1331
-1336.
Cant MA,
1998
. A model for the evolution of reproductive skew without reproductive suppression.
Anim Behav
 
55
:
163
-169.
Cant MA,
2000
. Social control of reproduction in banded mongoose.
Anim Behav
 
59
:
147
-158.
Cant MA, Johnstone RA,
2000
. Power struggles, dominance testing, and reproductive skew.
Am Nat
 
155
:
406
-417.
Clutton-Brock TH,
1998
. Reproductive skew, concessions and limited control.
Trends Ecol Evol
 
13
:
288
-292.
Creel SR, Waser PM,
1997
. Variation in reproductive suppression among dwarf mongooses: Interplay between mechanisms and evolution. In:
Cooperative breeding in mammals
  (Solomon NG, French JA, eds). Cambridge: Cambridge University Press;
150
-170.
Faulkes CG, Abbott DH, Liddell CE, George LM, Jarvis JUM,
1991
. Hormonal and behavioral aspects of reproductive suppression in female naked mole-rats. In:
The biology of the naked mole-rat
  (Sherman PW, Jarvis JUM, Alexander RD, eds). Princeton, New Jersey: Princeton University Press;
426
-445.
Fuchs S,
1981
. Consequences of premature weaning on the reproduction of mothers and offspring in laboratory mice.
Z Tierpsychol
 
55
:
19
-32.
Fuchs S,
1982
. Optimality of parental investment: The influence of nursing on reproductive success of mother and female young mice.
Behav Ecol Sociobiol
 
10
:
39
-51.
Gadagkar R,
1991
. Belonogaster, Mischocyttarus, Parapolybia and independent-founding Ropalidia. In:
The social biology of wasps
  (Ross K, Matthews R, eds). Ithaca, New York: Cornell University Press;
149
-190.
Gerlach G,
1990
. Dispersal mechanisms in a captive wild house mice population (Mus domesticus Rutty).
Biol J Linn Soc
 
41
:
271
-277.
Gerlach G,
1996
. Emigration mechanisms in house mice—a laboratory investigation of the influence of social structure, population density, and aggression.
Behav Ecol Sociobiol
 
39
:
159
-170.
Grafen A,
1984
. Natural selection, kin selection and group selection. In:
Behavioral ecology: an evolutionary approach
  (Krebs JR, Davis NB, eds). Oxford: Blackwell Scientific;
62
-89.
Hall WG, Williams CL,
1983
. Suckling isn't feeding, or is it? A search for developmental continuities.
Adv Study Behav
 
13
:
219
-254.
Hamilton WD,
1964
. The genetical evolution of social behaviour I.
J Theor Biol
 
7
:
1
-16.
Hamilton WD,
1964
. The genetical theory of social behavior II.
J Theor Biol
 
7
:
17
-32.
Heinze J,
1993
. Queen-queen interactions in polygynous ants. In:
Queen number and sociality in insects
  (Keller L, ed). Oxford: Oxford University Press;
334
-361.
Hrdy SB,
1979
. Infanticide among animals. A review classification and examination of the implications for the reproductive strategies of females.
Ethol Sociobiol
 
1
:
13
-40.
Jarvis JUM,
1991
. Reproduction in naked mole-rats. In:
The Biology of the naked mole-rat
  (Sherman PW, Jarvis JUM, Alexander RD, eds). Princeton, New Jersey: Princeton University Press;
384
-425.
Jemiolo B, Andreolini F, Xie TM, Wiesler D, Novotny M,
1989
. Puberty-affecting synthetic analogs of urinary chemosignals in the house mouse, Mus domesticus.
Physiol Behav
 
46
:
293
-298.
Johnstone RA, Cant MA,
1998
. Reproductive skew and the threat of eviction: a new perspective.
Proc R Soc Lond B
 
266
:
275
-279.
Johnstone, RA,
2000
. Models of reproductive skew: a review and synthesis.
Ethology
 
106
:
5
-26.
Keller L,
1991
.
Queen number and sociality in insects
 . Oxford: Oxford University Press.
Keller L, Reeve HK,
1994
. Partitioning of reproduction in animal societies.
Trends Ecol Evol
 
9
:
98
-102.
Komdeur J,
1996
. Influence oh helping and breeding experience on reproductive performance in the Seychelles warbler: a translocation experiment.
Behav Ecol
 
7
:
326
-333.
König B,
1989
. Kin recognition and maternal care under restricted feeding in house mice (Mus domesticus).
Ethology
 
82
:
328
-343.
König B,
1993
. Maternal investment of communally nursing female house mice (Mus musculus domesticus).
Behav Processes
 
30
:
61
-74.
König B,
1994
. Components of lifetime reproductive success in communally and solitarily nursing house mice—a laboratory study.
Behav Ecol Sociobiol
 
34
:
275
-283.
König B,
1994
. Fitness effects of communal rearing in house mice: the role of relatedness versus familiarity.
Anim Behav
 
48
:
1449
-1457.
König B, Riester J, Markl H,
1988
. Maternal care in house mice (Mus musculus): II. The energy cost of lactation as a function of litter size.
J Zool
 
216
:
195
-210.
Kokko H, Johnstone RA,
1998
. Social queuing in animal societies: a dynamic model of reproductive skew.
Proc R Soc Lond B
 
266
:
571
-578.
Labov JB,
1981
. Pregnancy blocking in rodents. Adaptive advantages for females.
Am Nat
 
118
:
361
-371.
Lawton AD, Whitsett JM,
1979
. Inhibition of sexual maturation by a urinary pheromone in male prairie deer mice.
Horn Behav
 
13
:
128
-138.
Manning CJ, Dewsbury DA, Wakeland EK, Potts WK,
1995
. Communal nesting and communal nursing in house mice Mus musculus domesticus.
Anim Behav
 
50
:
741
-751.
Oliveira PS, Hölldobler B,
1991
. Dominance orders in the ponerine ant Pachycondyla apicalis (Hymenopera: Formicidae).
Behav Ecol Sociobiol
 
27
:
385
-393.
Queller DC, Zacchi F, Cervo R, Turillazzi S, Henshaw MT, Santorelli LA, Strassmann JE,
2000
. Unrelated helpers in a social insect.
Nature
 
405
:
784
-787.
Reeve HK, Emlen ST, Keller L,
1998
. Reproductive sharing in animal societies: reproductive incentives or incomplete control by dominant breeders?
Behav Ecol
 
9
:
267
-278.
Reeve HK, Keller L,
1995
. Partitioning of reproduction in mother-daughter versus sibling associations: a test of optimal skew theory.
Am Nat
 
145
:
119
-132.
Reeve HK, Keller L,
1996
. Relatedness asymmetry and reproductive sharing in animal societies.
Am Nat
 
148
:
764
-769.
Reeve HK, Ratnieks LW,
1993
. Queen-queen conflicts in polygynous societies: mutual tolerance and reproductive skew. In:
Queen number and sociality in insects
  (Keller L, ed). Oxford: Oxford University Press.
Reyer HU,
1984
. Investment and relatedness: a cost/benefit analysis of breeding and helping in the pied kingfisher.
Anim Behav
 
32
:
1163
-1178.
SAS Institute,
1995
.
JMP statistics and graphics guide
 , version 3.1. Cary, North Carolina: SAS Institute.
SAS Institute,
1992
.
SAS technical report P-229, SAS/STAT software: changes and enhancements
 . Cary, North Carolina: SAS Institute.
Sayler A, Salmon M,
1969
. Communal nursing in mice: influence of multiple mothers on the growth of the young.
Science
 
164
:
1309
-1310.
Vehrencamp SL,
1983
. A model for the evolution of despotic versus egalitarian societies.
Anim Behav
 
31
:
667
-682.
Wakerley JB, Lincoln DW,
1971
. Intermittent release of oxytocin during suckling in the rat.
Nature
 
233
:
180
-181.
Wilson WL, Montgomery WI,
1992
. Range use in female woodmice (Apodemus sylvaticus) in deciduous woodland. In:
Wildlife telemetry, remote monitoring and tracking of animals
  (Priede JG, Swift SM, eds). New York: Ellis Horwood.
Wolff JO,
1992
. Parents suppress reproduction and stimulate dispersal in opposite-sex juvenile white footed mice.
Nature
 
359
:
409
-410.