Paternity and parental effort in dunnocks Prunella modularis: how good are male chick-feeding rules?

Anim. Behav., 1992, 43, 729-745

Paternity and parental effort in dunnocks Prunella modularis: how good are male chick-feeding rules? N. B. D A V I E S * , B. J. HATCHWELL*:~, T. ROBSON'[" & T. B U R K E t *Department of Zoology, University of Cambridge, Downing Street, Cambridge CB2 3EJ, U.K. t Department of Zoology, University of Leicester, Leicester LE1 7RH, U.K.

(Received19 March 1991; initial acceptance3 May 1991; final aceeptance10 September1991;MS. number3742)

Abstract. In polyandrous dunnocks, natural variation in alpha versus beta male share of matings was correlated with their paternity share (assessed by D N A fingerprinting) and with their share of work in chick feeding. A polyandrous male's share of mating access was a better predictor of his parental effort than was his total amount of mating access, suggesting that a male might monitor his paternity share by comparing his mating success with that of his rival. Males were removed temporarily at various stages of the mating cycle to create experimental variation in mating success. The effects of this on paternity and male parental effort were compared, to test how well a male's chick-feeding behaviour promoted his own reproductive success. Fingerprinting revealed that replacement males sired most of the eggs fertilized during the removal period. Removed males fed chicks only if they had gained matings during egg laying. This behaviour was adaptive because of the greater loss of paternity to replacement males earlier on in the mating cycle, but it led males (1) to undervalue matings achieved before laying, which could fertilize the first eggs in the clutch, and (2) to overvalue later matings, achieved at a time when most or even all of the clutch was fertilized. The removals confirmed that the alpha:beta share of parental effort in polyandry was determined by their share of matings, not by their dominance rank per se. By contrast, experimental manipulation of a monogamous male's mating access did not influence his parental effort, despite paternity loss. Why responses differ in monogamy and polyandry, and why dunnocks do not pass on a paternity marker as a better guide to parental effort than these crude, indirect, cues based on mating access, are discussed.

The theme underlying most research in behavioural ecology is that individuals are expected to behave in ways that maximize their reproductive success. However, relating behaviour to reproductive success is difficult, especially in species with complex social systems, because of the problems of measuring maternity and paternity. In recent years D N A fingerprinting, discovered by Jeffreys et al. (1985), has enabled unambiguous determination of parentage in animal populations (Burke & Bruford 1987; Wetton et al. 1987). Here we use this technique to test how well an individual's parental effort promotes its own reproductive success. Do animals themselves have some equivalent of D N A fingerprinting to assess their paternity? If not, what rules do they use to guide their parental effort? How good are those rules? :~Present address: Edward Grey Institute, Department of Zoology, South Parks Road, Oxford, OX1 3PS, U.K. 0003-3472/92/050729+17 $03.00/0

Our study species is the dunnock, a small passerine bird with a variable mating system (Davies 1990). Some female territories are defended by one male (monogamy), others by two, unrelated, males (polyandry). Sometimes two unrelated males jointly defend the territories of two or more adjacent females (polygynandry). Where two males share the defence of one or more female territories, one (alpha) is clearly dominant and able to displace the other (beta) from the female's vicinity. We have shown previously, using D N A fingerprinting (Burke et al. 1989), that there was no intraspecific brood parasitism so the female who attends a nest is the mother of the whole brood. The males resident on a territory sired almost all the offspring (132 of 133 young in 1988) with monogamous males gaining full paternity. On territories with two males, when the alpha male succeeded in monopolizing the matings he gained full paternity and only he helped the female to provision the

9 1992 The Association for the Study of Animal Behaviour 729

Animal Behaviour, 43, 5

730

brood. By contrast, when both alpha and beta males obtained matings paternity was often shared and they both helped to feed the young. There was no indication that males could recognize their own offspring. If they mated with the female they fed the young even in cases where they had no paternity, and in multiply sired broods they did not favour their own offspring. Males therefore used mating access as an indirect cue for paternity (Burke et al. 1989). In this paper we examine the relationship between amale's parental effort and his share of paternity of a brood. First, we show how natural variation in alpha and beta male parental care is related to their share of mating access. Second, we use temporary removals at different stages of the mating period to test whether there is a causal link between share of matings and parental effort. Finally, we use DNA fingerprintingto assess how natural and experimental variation in mating access influences paternity. Our aim is to use male chick-feeding responses to tell us how the males themselves assess the link between matings and paternity, and to use the DNA fingerprinting to tell us how mating access actually influences paternity. By comparing the two, we can then test how good a male's chick-feeding rules are at promoting his own reproductive success.

GENERAL METHODS Field Work A colour-ringed population of about 80 breeding dunnocks has been studied in the Cambridge University Botanic Garden since 1981 (see Davies 1990 for a review). Adult survival from the start of one breeding season (1 April) to the next is about 50% and up to the end of the 1990 season 221 males and 206 females had bred on the study site. The data reported here are from the 7 years 1981-1984 and 1988-1990, when detailed data were obtained on both mating behaviour and parental care. During the mating period (from time of nest completion to the onset of incubation, a period of ca 5-10 days) females were observed for on average a total of 4 h each and continuous recordings were made of which males were with them (within 10m). Observation times on a given day varied from 15 rain to 3 h. Total observation times were sufficient to give a reasonable measure of the relative mating access of alpha and beta males because

in 85% of the cases where beta males were seen to gain some access to the female, they did so within the first hour of observation (Burke et al. 1989). Exclusive mating access was scored as time when a male was the only male less than 10 m from the female, time when he could potentially copulate with her uninterrupted by other males (Davies 1985). Male mating success was scored as access time rather than number of copulations simply because the birds often hid away in dense vegetation so copulations could not always be seen. However, access time certainly gave a good measure of copulation success because there was no difference in alpha versus beta male copulation rate during periods of exclusive access to the female (Hatchwell & Davies 1992a). Provisioning of young was recorded when they were 5-11-day-old nestlings, when provisioning rates were highest. Nests were watched for periods of 0,5-2 h for an average total of 4-35 h per nest. This was sufficient to gain a good measure of male provisioning effort because in 90% of cases where we recorded a brood as provisioned by two males and a female, all three adults were seen feeding the young within the first hour of observation (Burke et al. t989). In 1989 and 1990, some males were caught in mist nets and removed temporarily from their territories for a period of 3 days in order to vary mating access experimentally. They were kept in aviaries away from the study site and then released back on their territories. All maintained good weight in captivity and all quickly settled back on their original territories when released (Hatchwell & Davies 1992b). Further details are given in the relevant section below.

DNA Fingerprinting Our DNA fingerprinting methods for the dunnocks were as described in detail in Burke et al. (1989), which gives illustrated examples, except that nylon (Hybond-N) membranes and a modified hybridization buffer were used (see Birkhead et al. 1990 for a detailed description). DNA was extracted from blood (40-100 ~1 from brachial veins of adults and 6-7-day-old nestlings) and fingerprints were obtained with probes 33'15 and 33-6 (Jeffreys et al. 1985). As in our previous study (1) cases of a single novel band per offspring (N= 5) were presumed to be the result of novel mutation; (2) there were no instances of intraspecific brood

Davies et al.: Paternity and parental effort parasitism (N= 146 for 1989 and 1990) and (3) there were at least five paternally derived bands per offspring, conservatively implying that the probability that an offspring assigned to a particular male (i.e. one who had all the paternal bands) was in fact fathered by another unrelated male in the population was always less than 8 • 10-4, usually by many orders of magnitude (Burke et al. 1989). Thus we can be confident that paternity was always assigned correctly. In a small number of cases (N=6) paternity could not be assigned because not all the potential replacement or extra-pair males were fingerprinted. However, for each of these the offspring contained at least five bands not present in the male who was monopolizing exclusive mating access, and we were therefore confident that a replacement male was the father; typical exclusion probability for 33'15 alone, where x = mean probability of band sharing between individuals, q = mean allele frequency and n = mean number of resolved bands per fingerprint, =xn(l-x)/(2-q)~'lO -4 (from data in Burke et al. 1989). In the case of three broods containing a total of nine offspring, DNA was not available from the female and relatedness with the males was tested by comparing the number of bands shared with each nestling with that expected for (1) first order relatives and (2) non-relatives (Burke et al. 1989; Birkhead et al. 1990). In each case one male was excluded from being a non-relative, and so inferred to be the father (G~ = 7" 1-29.7, P < 0-01; mean G~= 18.7, N = 9, P < 0-0001), and each of the other males was excluded from being a close relative (G I = 5.81-35.0, P<0"025; mean G~=15-7, N = 1 3 , P<0-0001). The blood sampling data were recorded independently from the experimental data and the fingerprints were scored in Leicester without knowledge of which individuals were used in the removal experiments. MATING ACCESS, PATERNITY AND CHICK FEEDING In monogamy, the resident male guarded the female closely during the mating period, neighbouring males rarely gained any access, and monogamous males provisioned the young at the same rate as the female, bringing on average 50% of the feeds to the brood (Hatchwell & Davies 1990, 1992a).

731

Where two males shared a territory, however, there was intense competition for matings and the outcome was very variable. Sometimes alpha males guarded females closely throughout and beta males were never seen to gain access. In other cases access was shared between alpha and beta males but the share varied widely; although alpha males usually gained more, occasionally beta males did better, especially on territories where vegetation was dense and where alpha males found it difficult to maintain close contact with the female (Davies 1985). In some cases both males lost the female for long periods and she hid away feeding alone quietly, while in others one, or both, of the males escorted her almost all of the time. In Fig. 1 we have used this natural variation to test the relationship between various measures of a beta male's association with a female during the mating period and his share of male effort in the provisioning of her brood. In polygynandry, a male's provisioning of the young at one nest may be influenced by the activities of other females in the system (Davies & Hatchwell, in press), so cases where other females were laying or had young simultaneously were excluded. Figure 1 includes only those cases where beta males gained some exclusive mating access and shows, for all three measures of association with the female, that they increased their parental effort with increased access to the female during the mating period. The three measures in Fig. 1 differ in how well they reflect the beta male's share of matings. Measure (a) is the poorest predictor because it includes not only time the beta male is alone with the female, and thus able to mate with her undisturbed, but also time when both alpha and beta are with her, when there are frequent chases and matings are often interrupted (Davies 1985). Measure (b) is better, because it includes only exclusive mating access but it is expressed as a proportion of the total time available, some of which will include time when the female is alone and so when the competing male does not have access. Measure (c) gives the best predictor of share of matings because it is the share of exclusive access time only, that is to say the share of the time when matings can occur. Using partial correlation to eliminate the effects of the other two variables shows that measure (a) itself has no significant relationship with provisioning share while (c) remains highly significant (see legend to Fig. 1). Thus the measure that most strongly reflects mating

Animal Behaviour, 43, 5

732

..~

0

o I

I

I

I

I

I

o_

I ,

~

o

~,~

II ~

--~,.~

~NN ..o-~

o

o

,~, v

. o

c.o 6

c\

~V .o

o o

.,~

0

@

~o,

Q

@

9

I --

I

I

I

I

~-~,o

I

I

I

0

I

~0O-0_o

,-~

~

0~',-! .-0

o

o~p

I

I 0 co

I

.~ ~.=

~V

",u = ~.s.., o ~,-~ o

9

1 | v

~-~

I 0 '
I

I 0 r

I

I 0

alum o4aq ,~q spee~, elOUJ #o eOD~.u~a~ad

H V o

".n~ ~ ~ ~"~'.,u

.

Davies et al.." Paternity and parental effort

IO0

9

O0

9

9

9

733

9

8o o

~

40

0 0

g o

m (3_

~

000 I 0

9 I

9 I 20

I

I 40

I

1 60

I

I 80

I

I ]00

P e r c e n t a g e of exclusive meting access enjoyed by beta male

Figure2. A beta male's share of the paternity of a brood, assessed by DNA fingerprinting, in relation to his share of the exclusive mating access with the female (only cases of natural variation included here); Spearman rank correlation (corrected for ties), rs=0.685 , N=28, P<0.01. For the 20 cases where the beta male obtained some mating access, rs = 0-665, P<0-01.

share is the one most strongly correlated with provisioning share. Figure 2 shows that an increased share of mating access leads to an increased share of the paternity of the brood. Even considering just those cases where a beta male obtained mating access, the paternity share increased with mating share. Thus males do not simply have a 'feed versus do not feed' rule based on mating success, but vary their parental effort in relation to their probability of paternity. The fact that Fig. lc is the strongest of the correlations suggests that a male might monitor his paternity share by comparing his exclusive mating access with that of his rival, rather than by just assessing his own total amount of access (Fig. lb). The problem with analysis based on natural variation, of course, is that the relationship in Fig. lc may not be a causal one. For example, those beta males who are better competitors may both gain more mating access and be able to invest more in offspring care. In the next section, therefore, we use an experimental approach to test whether increased mating share led to increased parental effort. The experiments also provided another advantage.

Although on average there was no significant variation in the share of exclusive access gained by alpha and beta males at different stages of the mating period (Hatchwell & Davies 1992a), matings at different stages may vary in their chances of fertilizing eggs and we were interested to test whether males placed different values on copulations achieved at different times in relation to egg laying.

EXPERIMENTAL MANIPULATION MATING ACCESS

OF

Experimental Design Females began to solicit matings at around the time the nest was completed, usually from 2 to 9 days before the first egg was laid, and males continued to copulate at a high rate (0" 5 to 2.5 times per h) throughout egg laying, right up to the start of incubation (Hatchwell & Davies 1992a). Eggs were laid one per day, within the first 3 h after dawn, and the female usually began incubation on the day the last egg was laid. M o s t clutches were of three or four eggs (range two to five).

Animal Behaviour, 43, 5

734

i

Days before egg laying

-7 I

-5 I

-5

I

I

t

i

-I I

I

Egg laying i I

Incubation

I

le

2e

5e

9

9

9

I

I

I

I

I I

I

I

A

I

I

I

B I

I I

I I

C

c

c

.-t

I

I

Figure3. The stages of experimentalremovals of monogamous males and alpha males in polyandry and polygynandry. In this example, the female lays a clutch of three eggs (one egg per day) and begins incubation on the day the last egg is laid. The temporary removals, for 3-day periods, were at various stages indicated by A to C, Males removed at stage A (incubation) gained normal mating access. Males removedat stage B gained mating access prior to egg layingbut not on days that eggs were laid. Males removedat stage C had reduced mating access but gained at least some accessduring egg laying. The mating access of monogamous males (N= 15) and alpha males in polyandry and polygynandry (N = 28) was varied experimentally by removals of 3 days' duration at various stages during the mating period or incubation (Fig. 3). When monogamous males were removed, neighbouringmales came onto the territory and mated with the female. When alpha males were removed, resident beta males took over control of the female and were just as successful in keeping neighbours at bay as when the alpha male was in charge (Hatchwell & Davies 1992b). It ]night be thought that these experiments, which allowed males windows of access at different stages of the mating period, placed the birds in an unnatural situation. However, such variation also occurs naturally. In some cases of polyandry, beta males gained access only at particular stages and in polygynandry alpha males frequently left females unguarded to attend to other females in their mating system, thus allowing beta males free access at various stages of the mating cycle. Although the design of the experiment in Fig. 3 enabled us to vary mating access while giving all removed males the same 3-day period in captivity, three complications emerged in practice. (1) We could not predict in advance exactly when a female would begin laying so removals before laying were done in initial ignorance of their exact stage in the mating cycle. Sometimes we failed to catch a male at the first attempt and a removal planned for before laying turned out to be done during laying if the female had meanwhile begun her clutch.

(2) There was some variation in the time of day at which birds were caught and released. We caught most males early in the morning (0445 0630 British Summer Time (B.S.T.)) before the time of egg laying, so if the female laid her first egg later that morning we scored the male as having mating access 'before laying only' if he subsequently failed to gain matings before the onset of incubation. Thus, in the analysis below we define the period 'during laying' as from the time of laying of the first egg to the laying of the last egg, when incubation usually began. Birds caught at dawn were released after 3 days but a little later in the morning (0630-0830 B.S.T.), at a time by which most eggs due to be laid that day would already have been laid. The relevance of this will become clear later. The few birds caught at dusk (1900-2100 B.S.T.) were released at the same time of day 3 days later. (3) Although all the removed males settled back quickly on their territories when released, some experienced reduced mating access beyond the 3day exclusion period because the replacement male became dominant and continued to monopolize the female. We show later that a change in dominance status per se did not influence chick-feeding effort, only a change in mating access; nevertheless the increased period of reduced mating access did disrupt the neatness of the experimental design. Influence of Removals on Whether Males Feed

All males removed during incubation (controls), after the mating period had ended, gained normal mating access both before and during egg laying

Davies et al.: Paternity and parental effort

735

Table I. Influenceof stage of experimental removal on whether males fed the young No. removed males who fed young Stage removed (see Fig. 3)

Mating access to female

Incubation

(A)

Mating period

(B)

Mating period

(C)

Beforeand during laying Beforelaying only Beforeand during laying

Monogamous males

Alpha males

Total

7/7

4/4

I 1/11*

0/2

1/5

1/7'*

6/6

19/19

25/25

*Comparing totals in A and B, P < 0.0 I. **Comparing totals in B and C, P<0.001. (Fisher exact tests). and all fed the young (Table I). Thus the experience of the temporary removal itself did not disturb the male's normal chick-feeding behaviour. Males who were removed at stage C in Fig. 3, who thus had reduced mating access but who were present during at least some days of egg laying, likewise all fed the young. By contrast, of the seven males who were removed throughout the entire egg-laying period only one fed the young, even though all seven had enjoyed several days mating access prior to removal and even though all seven settled back on their territories (Table I). Now considering the responses of the beta males who gained access as a result of alpha male removal, 11 gained access both before and during laying and all helped to feed the young, while 13 gained access during laying only and all 13 likewise fed the young. These results suggest that some mating access during egg laying itself is necessary to cause chick feeding. Even a single day's access during egg laying was sufficient; eight of the removed males gained access only on the day the first egg was laid (i.e. removed at dusk on that day or at dawn the next day and gained no more matings before incubation) and three gained access only on the day before the last egg was laid (removed before laying of the first egg and released on the morning of the day before the laying of the last egg), yet all 11 fed the young. Likewise, of the beta males who gained increased mating access due to the alpha male's removal, six gained access only on the day before the last egg was laid yet all six helped to feed the young. Males were clearly interested in the nest contents and inspected the nest frequently during the mating period. When we released males back on their territories after the removals, one of the first things they

did was to visit the nest. In an experiment reported elsewhere, we have shown that males use the appearance of an egg in the nest as one of the cues to value their copulations; when a model egg was placed in the nest and males were removed within a day of laying of the first real egg, they fed the young even though absent throughout egg laying (Hatchwell & Davies 1992a). Of the seven males included in stage B in Table I, four were removed before the day of the first egg while three were removed at dawn on the day of the first egg, but before the egg was laid. Because these three were removed before laying occurred, they have been included in the category 'mating access before laying only'. The one male who fed the young was one of these three birds. It would be interesting to increase this sample of birds removed on the morning of the first egg by post-laying removals, to test whether the switch from not feeding to feeding occurred when the male realised that the female had laid. In some other species of birds, there is good evidence that males who fail to mate with a female may destroy her clutch or kill young nestlings and so induce the female to lay a replacement clutch, thereby hastening the day they have a chance to sire their own young (Crook & Shields 1985; Robertson & Stutchbury 1988; Koenig 1990). There is indirect evidence that beta male dunnocks who fail to mate with a female may also behave in this way (Davies 1986). However, the experimental removals provided no evidence for infanticide by males that were removed throughout the egg-laying period (Table II). For all measures, reproductive success from these nests was no different from nests where removals had occurred at other times and so where

736

Animal Behaviour, 43, 5

Table II. Influenceof experimental removals throughout the egg-layingperiod on measures of reproductive success

Removed male Measures of reproductive success % Nests where some eggs hatched successfully Mean % eggs per nest that hatched successfully(_+ 1 sE) % Nests where some young fledged Mean % eggs that produced fledged young

Had mating access during laying (N = 55)

Had no mating accessduring laying (N = 19)

Significance of difference

83.6

73.7

%2,=0.912; NS

64.4 + 4.8

63"9+ 9"9

z* = --0'318; NS

41.2 30.2 + 5.5

42'1 35"8_+10"7

%z~=0'005; NS z*= --0'467; NS

Sample sizesare larger than in Table I because they includenests that failed to produce young (wherechick feedingcould obviouslynot be scored) and some other removals from Hatchwell & Davies (1992b). *Mann-Whitney U-tests,two-tailed.

the removed male had consequently helped to provision the brood. Males who were removed throughout egg laying had nevertheless enjoyed mating access for up to a week prior to removal so removals of longer than 3 days would be needed to reduce mating access to zero. It is possible that there are two thresholds of access influencing male behaviour, with some access being sufficient to prevent infanticide but access during egg laying itself being necessary to cause chick feeding. Influence of Removals on Paternity

Experimental versus controls Table III summarizes the effects of the removals on paternity. The data include only cases where our observations showed that one male had monopolized all the mating access prior to removal, in other words monogamous males and those alpha males in polyandry and polygynandry who had been successful in keeping beta males at bay. Thus the only chance that another male had to fertilize eggs was after the removal. The control data (removals after clutch completion and no removals) show that our assessment of male monopolization was quite accurate. Of the cases where we had recorded one male as gaining 100% of the mating access, only two out of 100 chicks (one chick in each of two out of 34 broods) were sired by another male. One was sired by a neighbour but we were unable to test the other against potential sires because of insufficient sample for DNA analysis.

Compared with these controls, experimental removals (before and during egg laying) clearly increased the chance that another male would father the young (Table III; control versus experimental totals for number of young, Z21=35.81, P<0-001; for number of broods, %21=12.32, P<0.001). There were 13 broods where the removed male lost paternity. Eight cases involved removal of an alpha male from polyandry or polygynandry. Observation showed that in all cases the resident beta male quickly took over the female when the alpha male was removed; in some cases he was already mating with the female while we were taking the alpha male out of the mist net! The beta males then guarded the females and kept neighbours at bay just as successfully as the alpha males had done (Hatchwell & Davies 1992b). DNA fingerprinting confirmed that in all cases the beta male was the sire of the young lost by the alpha (total of 18 young from eight broods sired by the beta male). In six cases the beta male sired the whole brood and in two cases he sired one of two young with the removed alpha siring the other. In five removals of a monogamous male, the removed male lost paternity. In four of these cases a neighbouring male took over the female and the neighbour sired two out of three, one out of two, three out of four and two out of three of the young, with the removed resident siring the others in the brood. In the final case we did not observe any replacement male with the female, yet two of the three young were sired by another male with the remaining chick sired by the removed resident.

Davies et al.." Paternity and parental effort

737

Table Ill. Influenceof stage of removal of the male who was monopolizingexclusivemating access on the replacement male's successin fertilizingeggs, as assessed by DNA fingerprinting No. young (%) sired by replacement male Stage resident male removed

Observed*

Predicted

No. broods (%) where replacementmale sired young Observed

Predicted

Experimental removals

Before first egg laid Day first egg laid After day of first egg, but before clutch completion Total before/during laying (mating period)

20/26 (77) 8/16 (50) 0/32 (0) 28/74 (38)

29/37 (78) 12/22 (55) 5/45 (11) 46/104 (44)

9/10 (90) 4/5 (80) 0/12 (0) 13/27 (48)

10/10 (100) 5/5 (100) 5/12 (42) 20/27 (74)

0/18 0/88 0/106 (0)

1/5 1/29 2/34 (6)

0/5 0/29 0/34 (0)

Controls

Control removals, after clutch completed No removalt Total controls

1/ 12 1/88 2/100 (2)

The predicted values are calculated assuming that eggs are fertilized 24 h before they are laid, that replacement males take over the female and that there is 100% second male sperm precedence (see text). *Totals less than in predicted column because not all eggs hatched or not all chicks sampled (e.g. due to starvation). "['Includes 1988 data (Burke et al. 1989) and 1989 1990 data for cases where observations indicated that one male monopolized matings throughout (monogamy plus some cases of polyandry).

Variation with stage of removal Table III also shows that, within the experimental removals, there was significant variation in paternity loss with stage of removal, with earlier removals during the mating period increasing the chance that a replacement male gained paternity (considering the three stages of removal; for proportion of young sired by replacement, ~2z= 37.45, P < 0.001; for proportion of broods with replacement male paternity, X22=20.34, P<0.001). Although all removed males had enjoyed mating access for several days before removal, provided they were removed before the first egg was laid the replacement males sired most of the brood (Table III). Laboratory experiments with zebra finches, Taeniopygia guttata (Birkhead et al. 1988) have revealed last male sperm precedence but we cannot tell from our data whether this accounts for the replacement male advantage in the dunnocks because number of copulations by removed and replacement males will also have varied. Thus replacement males may have been more likely to fertilize eggs because of last male sperm precedence, because they gained more matings at a critical time, or a combination of these two. Whatever the

mechanism, it is particularly interesting to find strong replacement male paternity in dunnocks because of this species' extraordinary pre-copulatory display, where males stimulate females to eject sperm from previous matings prior to copulation (Davies 1983). Although the ejected sperm may not originate from the female's sperm storage tubules, but rather from the vagina and cloaca, the display presumably enhances the chances of a replacement male gaining paternity (Birkhead et al. 1991). In Table III we have calculated the predicted numbers of young sired by replacement males if there is 100% second male advantage and assuming eggs are fertilized 24 h before they are laid (Lake 1975). We have assumed that the resident male sires all the eggs fertilized before he is removed, which seems reasonable given the control data above. Removals were performed either at dusk or at dawn, simply because birds were easier to catch in the twilight. Experiments with mallards, Anas platyrhynehos, have shown that only inseminations of a female up to 1 h after egg laying have a chance of fertilizing the egg laid the following morning (Cheng et al. 1983). Although some males may have taken over in time to fertilize eggs due to be laid 24 h after the removal, many did not, so to be

Animal Behaviour, 43, 5

738

I00

9

9

O0

00

8o 0

E 0

0

60 9

0

0

9

4~ -~ 40 ~6

g 2O

1~

9

I

1 I I I I I I I I 0 20 40 60 80 t00 Predicted percentage of clutch fertilized by replacement male if I 0 0 % second male sperm precedence

Figure 4, Relationship for the 32 broods in Table III where a male was removed (27 experimental plus five controls) between predicted percentage of the clutch sired by a replacement male, assuming 100% replacement male sperm precedence (see text), and the observed percentage sired by the replacement male, as assessed by DNA fingerprinting. 9 removals of alpha males from polyandry and polygynandry, where the replacement male was the resident beta male; (3: removals of monogamous males, where the replacement was a neighbouring male. Spearman rank correlation (corrected for ties), r s = 0.814, P < 0"01. Excluding the predicted values of zero, r s = 0"756, N = 20, P < 0'01.

conservative we have assumed that replacements were unable to fertilize the ova ovulated on the morning of a dawn removal or on the morning following a dusk removal. This assumption may underestimate second male success slightly. Thus, for example, we assume that a male removed at dusk on day - 2 or at dawn on day - l (day before first egg) will fertilize the first egg of the clutch and lose paternity of the remainder; a male removed on the morning of the first egg will fertilize the first two eggs (the first egg, and the second egg, fertilized soon after the first egg is laid); a male removed at dawn on the day before the laying of the last egg will fertilize the entire clutch (because by the time the replacement male takes over the last egg will have already been fertilized). Table III and Fig. 4 show that there was no significant difference between the observed success of replacement males and these predictions based on the assumptions above (totals for experimental removals; no. of young column, Z21 = 0-488; no. of broods column, %21=2-80, both NS). However, a problem with the analysis is that in many broods not

all eggs hatched, and in some broods not all chicks were sampled for D N A (because of starvation in the nest and their removal by adults). Thus in some cases in Fig. 4 the observed proportions exceed the predicted values; e.g. the replacement male was predicted to fertilize three out of four eggs and was observed to father three out of three young, with one egg failing to hatch. There were only five broods where the replacement male had the chance to gain some paternity and where the entire clutch hatched and was fingerprinted. The predicted versus observed proportions of chicks sired by the replacement males were as follows: 3/3 versus 3/3; 2/4 versus 0/4; 1/4 versus 0/4; 1/4 versus 0/4 and 4/4 versus 3/4. These limited data suggest that second male advantage was often less than 100%, as also shown in laboratory experiments with zebra finches, where it is about 80% (Birkhead et al. 1988). Male dunnocks continue to copulate and to compete for matings throughout egg laying, right up to the laying of the last egg o f the clutch (Hatchwell & Davies 1992a), so it was interesting to confirm that

Davies et al.: Paternity and parental effort copulations performed after laying had begun could fertilize eggs. For the five cases where we removed a male at dawn on the day the first egg was laid, and have assumed that he would have fertilized the first two eggs of the clutch (see above), the predicted versus observed proportions of chicks sired by the replacement male were as follows: 2/4 versus 1/2; 3/5 versus 2/4; 3/5 versus 3/3, 2/4 versus 2/3 and 2/4 versus 0/4. Thus replacement males gained considerable success at this stage and overall their paternity was not significantly different from the predictions in Table III (second row of experimental removals). F o r removals performed later on in laying (day of second egg or later; third row of experimental removals in Table III), the replacement males obviously had less opportunity to fertilize eggs. If the clutch was just of three eggs, then the replacement was unlikely to be able to fertilize any of them. If the clutch was of four eggs, then he would have the chance to fertilize the last egg. The limited data in Table III suggest that replacement male success at these later stages may have been poorer than earlier on. Of the five broods where the replacement was predicted to fertilize eggs, the predicted versus observed paternity was as follows: 1/4 versus 0/4; 1/4 versus 0/3; t/4 versus 0/4; 1/4 versus 0/3 and 1/4 versus 0/3. Obviously the problems of incomplete sampling of the clutch for paternity and other variation arising under field conditions make this conclusion tentative.

Are male responses adaptive? We can use these paternity results to assess whether male responses to the removals were adaptive. Males removed before egg laying have no definite marker by which to value their copulations, though female appearance or behaviour on the day before the first egg is laid apparently conveys some information (Hatchwell & Davies 1992a). Given that females solicit matings several days before the laying of the first egg, and given the high degree of paternity gained by replacement males, it makes good sense that a male removed before egg laying, and absent throughout laying, refuses to feed the chicks despite his previous mating access: he has a high chance of not fathering any of the brood. By contrast, males present up to the start of egg laying are guaranteed some paternity provided they have monopolized matings up to then, so their provisioning of the brood makes good adaptive sense.

739

Table I l l suggests that males who gain access only late on in laying, after the day of the first egg, are unlikely to achieve much paternity. Males who gain access only on the last day before clutch completion have little or no chance of fertilizing eggs because the last egg will have been fertilized at dawn that day. Copulations throughout the last day are therefore worthless, yet males continue to copulate and compete for matings right up to the onset of incubation. A likely explanation is that males do not know that the following morning's egg is the last one (Hatchwetl & Davies 1992a). However, once the female begins incubation it would clearly pay males to backdate and devalue the previous day's copulations. They clearly do not do this: all nine males who gained mating access during laying only on the last day before clutch completion, when there were no more eggs to fertilize, nevertheless fed the young (see above). Thus the male's chick-feeding rule of 'feed the young if I gained mating access during egg laying', makes good sense in relation to loss of paternity to replacement males early on in the mating cycle but it has two drawbacks; first to undervalue matings achieved before laying, which might fertilize at least the first egg, and second to overvalue matings later on. The D N A fingerprinting results provided instances of both kinds of 'mistakes' by males. In three cases where a male was removed on the day before laying and later gained no further mating access (and so did not help to feed the brood) he nevertheless gained paternity (2/2, 1/3 and 1/4 chicks fathered). In four cases a male gained mating access only on the last day before incubation, and at a time when the last egg would already have been fertilized, and in all four cases the fingerprinting confirmed that he did not father any of the chicks (0/3, 0/3, 0/1, 0/2), yet in all cases he helped to feed them. Influence of Removals on Male Feeding Rate

We now turn from the male's decision of whether or not to feed the young, to the question of how hard he should work once he has decided to provision. Because males valued mating access only during egg laying, we have examined a male's chick-feeding effort in relation to his access on egglaying days. For example, a female who laid a clutch of four eggs has 3 days of mating access during laying (days of first, second and third eggs; incubation begins on the day of the fourth egg).

740

Animal Behaviour, 43, 5

Where males shared mating access, we have given equal weight to each day (as the males themselves appear to do). For example, a male who monopolized matings and who was removed at dawn on the day of the second egg would be scored as having 1 day's access, while his replacement would enjoy 2 days' access if there was a clutch of four eggs. Thus in this case the removed male's share of access would be 33 %. In eases where access was shared on a given day, the share was weighted accordingly.

Alpha versus beta male share of chick feeding in trios We first consider cases of polyandry and polygynandry where both alpha and beta males helped to provision the young. There were 17 cases where alpha males had been removed for 3-day periods at various stages of the mating cycle, thus varying experimentally their share of access during the egglaying period. Figure 5a shows that the alpha male's provisioning rate increased with his share of mating access. The beta male's provisioning rate also varied in the same way with his own mating access, being greater when alpha males were removed for more of the egg-laying period, thus allowing beta males an increased share of the access at this time (Fig. 5b). Female provisioning rate (in relation to that expected for a given brood size, see Hatchwell & Davies 1990) did not vary with alpha male share of access (rs = 0' 144, N = 17, NS). In Fig. 6 we have expressed these experimental results as the alpha male's share of male feeds to the brood in relation to his share of the mating access during laying. The result agrees with that found earlier from natural variation (Fig. lc), supporting the view that there is a causal link between share of mating access and share of chick feeding. When alpha males were released back on their territories following the removal, there was a brief period of chasing and fights between the two males. Sometimes the original dominance order was restored, with the removed bird retaining alpha status, but sometimes the dominance order reversed and the original alpha became beta (Hatchwell & Davies 1992b). Figure 6 shows that there was no effect of dominance status on chick feeding share; the relationship between provisioning share and mating share was the same whether the removed male maintained rank or lost rank. This provides nice evidence that it is mating access that influences chick feeding, not dominance rank per se. When

beta males were experimentallygiven a greater share of the mating access, they brought a greater proportion of the male feeds irrespective of their dominance rank later on at the time of chick feeding.

Monogamous male effort in pairs How do alpha and beta males come to this seemingly fair split of work in relation to their share of matings? We think some kind of bargaining must occur between the males, as suggested in Houston & Davies (1985), because our removals of monogamous males produced a different result. Although the monogamous removals led to increased mating access by neighbours, in almost all cases the neighbours went back to help their own females during chick feeding. Thus the removed monogamous male was the only male available to help the female. Provided he gained some access during egg laying his work rate in chick provisioning was unaffected by his share of mating access (Fig. 7). Access on just one egg-laying day caused the same provisioning rate as access throughout egg laying even though DNA fingerprinting showed that males did not have full paternity of the brood. In all cases the male brought around 50% of the feeds to the nest, as is normal for paired males (Hatchwell & Davies 1990). Female provisioning rate did not vary in relation to the monogamous male's share of access (rs = 0.255, N = 15, NS).

Why different resultsfor pairs and trios? Why does share of mating access influence alpha and beta male effort in trios but not that of monogamous males in pairs? One simple explanation could be a difference in knowledge. Alpha males could easily assess that they have lost paternity because when they return to their territory they find the beta male in residence and later see him feeding the chicks. By contrast because neighbour replacements return to their own territories, when a monogamous male is released he may simply find the female incubating and so has no way of assessing whether he has lost paternity during his absence. However, there are three reasons for dismissing this as the explanation. First, in some cases the monogamous male was returned before clutch completion, and so in time to find a neighbour with his female. Second, when an alpha male is removed during chick-feeding from a territory where both males are helping the female to feed the brood, the beta male immediately increases his provisioning

Davies et al.." Paternity andparental effort

741

a) J~

ca. e~

+2 +1

o

0

E m o

$

-I -2

|

-5

Q

I

-4

0

J~ 0

-5

I 0

I

I 20

I

I 40

I

I 60

I

I 80

I

I I00

I

1 60

I

I 80

I

I IOO

(b) +2

Z E o 2

+I

o

9

II

,

-I

X I

9

--5

I 0

I

I 20

I

I 40

Percentage of e x c l u s i v e access by alpha made on e g g - l a y i n g days

Figure 5. Provisioning rates of nestlings by alpha and beta males in polyandry and polygynandry, where both males helped to feed the brood, in relation to their experimentally varied share of the mating access during egg laying, caused by 3-day removals of alpha males (N= 17) at various stages of the mating cycle. Provisioning rates are observed rates per h in relation to those expected from average alpha and beta provisioning rates to a brood of a given size (Hatchwell & Davies 1990). (a) Alpha males; Spearman rank correlation, rs = 0.667, P < 0.01. (b) Beta males; r s = - 0.553, P < 0'05. Expressing a male's provisioning rate in relation to his own share of the mating access (arcsine transformed access data); for alpha males, Y= - 4.05 + 0.0642X, and for beta males Y= - 3.15 + 0.044X. ANCOVA shows no difference in slope (F~.30= 0.57) or elevation (F~,31= 0.05). Thus both males increased work rate in relation to mating share in the same way.

rate to that o f a m o n o g a m o u s male with full paternity (Hatchwell & Davies 1990). He could surely assess b o t h from the fact that matings were shared and chick-feeding was shared that he did not

gain full paternity. Third, in polygynandry alpha and beta males often share paternity o f two synchronously laying females a n d then each male helps to feed one o f the broods. The male's w o r k rate is not

Animal Behaviour, 43, 5

742

9

IO0

0

~:

0

8o

0

o.

g

60 0

E

~6 g

4O

0

9

o o

ft,

zo

I 0

I

I 20

I

i 40

I

I 60

I

I 80

I

I I O0

Percenfoge of exclusive occess by olphe mole on e g g - [ e y i n g doys

Figure 6. Relationship between an alpha male's experimentally varied share of mating access (caused by removals at different stages of the mating period) and his share of male feeds brought to the nestlings. Points at (0,0) on the graph indicate that the beta male gained all the mating access and did all the chick feeding. Points at (100, 100) indicate that the alpha male gained all the mating access and did all the chick feeding. Other points indicate that both matings and chickfeeding were shared between the two males. For all 23 points (including cases where one male provisioned and both males provisioned) Spearman rank correlation, r s = 0.869, P < 0.001. For the 17 cases where both males fed, r s = 0.702, P < 0.01. In some cases the removed male retained alpha status when released back on his territory ( 9 in other cases he lost status and became beta (9 ANCOVA showed that there was no difference between these two cases (arcsine transformed data, slope FI,19 = 0"01; elevation F~.zo= 0.34). related to mating share but is the same as that of a m o n o g a m o u s male with full paternity (Davies & Hatchwell, in press). We suggest the following explanation. In theory, we might expect a m o n o g a m o u s male to respond to paternity loss in the same way as he does to a reduction in brood size, namely by reducing his provisioning rate (Hatchwell & Davies 1990). However these two cases are not equivalent in practice because a male apparently cannot recognize his own young in a multiply sired brood (Burke et al. 1989). If a lone male decreased his effort in response to paternity loss, the whole brood would suffer, including those chicks sired by him. Certainly females increase their effort in response to decreased male effort, but it is not sufficient to compensate fully for the male's reduction, so nestling weight and survival is reduced as a result (Hatchwell & Davies 1990). By contrast, when there are two males on a territory a male who loses paternity does so to another male helping at the same nest. Thus any reduction

in parental effort by one male in relation to lost paternity is counteracted by an increase in effort by the other male, who has gained increased paternity. The similar relationships in Fig. 5a and b (see analysis in legend) suggests that any reduction in parental effort by one male is completely compensated for by an increase in effort by the other male. Thus, whereas the chicks sired by a m o n o g a m o u s male would suffer if he reduced his effort in relation to paternity loss, the chicks sired by one of two polyandrous males do not. Further experiments are needed to reveal the mechanism that gives rise to the compensation in effort by two males who share paternity and provision at the same nest. Males may vary their parental effort not only in relation to their paternity share but also in response to each other's parental effort (Houston & Davies 1985; Wright & Cuthill 1989), so it would be useful to manipulate both these factors independently. To complete this story we also need to quantify how male care influences parental survival as well

Davies et al.: Paternity and parental effort

743

{a) 6O el)

IL

Expected

4o

20

I P o

I 2o

I

t

I

40

I

I

I

I

I

80

60

I00

(b)

..c:

o

i

2

9

Expected

L

I

-Z 0

t

0

I

I

20

I

1

40

I

I

60

I

I

80

I

I

I00

Percentage of egg-laying days male hod access to female

Figure 7. Chick provisioning by monogamous males in relation to their experimentally varied mating access during egg laying, caused by 3-day removals at different stages of the mating period (N= 17). In all cases the removed male was the only male available to assist the female in chick feeding. (a) % Total feeds by male. The line is the expected 50% for normal cases of monogamy. (b) Observed male provisioning rate per h in relation to that expected from a monogamous male for a brood of a given size (from Hatchwell & Davies 1990). Provided the male gains some mating access during laying, there is clearly no influence of amount of access on chick feeding effort. as nestling survival. In principle there are circumstances when it may pay even a single male to reduce his effort in response to paternity loss because of the benefits of saved investment for future, more valuable broods. Previous studies have shown variable male responses to paternity loss through extra-pair copulations; in some cases males may reduce parental effort (Moller 1988) while in others they apparently do not (Morton 1987; Westneat 1988). Male responses may vary depending on the benefit of provisioning to the current brood and the cost this has to the male's future reproduction (Houston & Davies 1985). One factor that could influence parental effort is stage of the breeding season; at the start of the

season a male has further breeding opportunities in the near future, whereas at the end of the season he has to survive the long winter before another opportunity arises. However, we found no seasonal effect on our experimental results. F o r polyandrous males there was no influence of season on the deviations in chick-feeding rates from the fitted regressions in Fig. 5 (for alpha males, r s = - 0-125; for beta males, r s = 0.183; both NS). Likewise, for m o n o g a m o u s males the deviations from the expected lines in Fig. 7 were not related to season (for Fig. 7a, rs=0-139; for b, r s = -0"360; both NS). Most male dunnocks have only a few breeding attempts in their lives and their help with chick feeding has a marked effect on reproductive success (Davies & Hatchwell, in press). It is

Animal Behaviour, 43, 5

744

perhaps not surprising, therefore, that they place high value on each brood with lone males provisioning at a high rate despite paternity loss to save the lives of the chicks in the brood which may be their own.

GENERAL DISCUSSION How Good are Male Chick-feeding Rules? Our main conclusion is that males follow two simple rules. First, they use an indirect cue to guide their chick feeding in relation to paternity, namely 'feed the brood, provided I have mated with the female during the egg-laying period', with the start of the egg-laying period being marked by the appearance of the first egg in the nest and the end by a change in female behaviour when the last egg is laid (Hatchwell & Davies 1992a). Second, their parental effort is related to whether there is another male helping at the same nest. If there is, then the two males' share of provisioning is related to their share of mating access. If a male provisions alone, then he works at the same rate irrespective of his mating share. The DNA fingerprinting shows that these rules work well in promoting an individual's reproductive success because mating access is a good predictor of paternity. Furthermore, the male's demand for some mating access up to the time of egg laying itself is adaptive because loss of paternity to replacement males can severely reduce the success of inseminations early on in the mating period. However, the rules are not completely foolproof. A share of the matings does not guarantee a share of the paternity, especially late on in the mating period when most, if not all, of the eggs have been fertilized. Nevertheless, males value these late matings as much as those gained earlier on. This raises the question of whether there could be better rules. The best one of all, in the sense of guaranteeing that a male provisions only his own young, would be for a male to recognize his own chicks. Why don't dunnocks do this? There is evidence that some social insects and mammals can discriminate close kin from more distant kin, even when these are raised in a common environment such as the same nest (Holmes & Sherman 1982; Page et al. 1989). In these cases it is likely that discrimination is based on genetic odour labels, with individuals perhaps comparing the similarity of their own label with those of others to recognize

close relatives, a process Holmes & Sherman have called 'phenotype matching'. Birds do not have the array of odour cues available to insects and mammals, and phenotype matching based on visual or vocal cues seems unlikely, given that nestlings change so markedly in appearance and size during development. The failure of dunnocks to discriminate their own sired young is not surprising, given that they feed young cuckoos, Cuculus r and the young of other species introduced experimentally into their nests (Davies & Brooke 1989). So one answer to the problem of why there is no paternity marker is simply that dunnocks are unable to pass on a distinctive label to their offspring, and are forced to rely on cruder, indirect, cues to paternity instead. Another, more entertaining, possibility is that genes for paternity markers may not spread because males have lost out in an evolutionary conflict of interest with the female and the chicks (see also Beecher 1988). If a chick exhibited a paternity marker, it would be fed only by its father. If it hid its identity, it might be fed by two males and so gain more nourishment. The female, who is mother of the whole brood, would likewise benefit if two males cared for the young even in cases where only one was the father, so genes that suppressed paternity markers might be favoured. A model to explore conflicts of interest over paternity advertisement would be very worthwhile. Different Rules in Different Species? Our results make an interesting comparison with those of Koenig (1990) for the acorn woodpecker, Melanerpesformicivorus, where several males also shared females in a polygynandrous mating system. When Koenig removed subordinate males during the mating period, so they did not have the opportunity to sire any young, some of them nevertheless still helped to provision the brood. Their different chick-feeding response makes a nice contrast with the dunnocks because in the woodpeckers the males are close relatives, often brothers or father and sons. Thus kin selection may favour the provisioning of young sired by other males in the same mating system. In dunnocks, the males are not close relatives (Davies 1990) and so there is no benefit to be gained from feeding another male's offspring. Behavioural ecologists interested in function should, perhaps, ask more questions of the form 'why have different mechanisms been favoured in

Davies' et al.." Paternity and parental effort different species?' T h e subject is, in part, a c o m p a r a tive study o f b e h a v i o u r a l m e c h a n i s m s in relation to ecological circumstances.

ACKNOWLEDGMENTS W e t h a n k N.E.R.C. for funding the field work in C a m b r i d g e ; S.E.R.C. for funding the D N A fingerp r i n t i n g in Leicester; the H o m e Office a n d N a t u r e C o n s e r v a n c y Council f o r licences for the experimental work; C. D. Pigott, P. Orriss a n d N. Villis, together with the G a r d e n staff, for allowing us the freedom of the C a m b r i d g e University Botanic G a r d e n ; Alec Jeffreys for providing minisatellite probes; Tim Birkhead, Janis Dickinson, Alasdair H o u s t o n , Walter Koenig, Ian Owens a n d T h e l m a Rowell for helpful discussion a n d Mike Bruford for technical advice.

REFERENCES Beecher, M. D. 1988. Kin recognition in birds. Behav. Genet., 18, 465-482. Birkhead, T. R., Burke, T., Zann, R., Hunter, F. M. & Krupa, A. P. 1990. Extra-pair paternity and intraspecific brood parasitism in wild zebra finches Taeniopygia guttata, revealed by DNA fingerprinting. Behav. Ecol. Sociobiol., 27, 315 324. Birkhead, T. R., Hatchwell, B. J. & Davies, N. B. 1991. Sperm competition and the reproductive organs of the male and female dunnock (Prunella modularis). Ibis, 133, 306-311. Birkhead, T. R., Pellatt, J. E. & Hunter, F. M. 1988. Extra-pair copulation and sperm competition in the zebra finch. Nature, Lond., 334, 60-62. Burke, T. & Bruford, M. W. 1987. DNA fingerprinting in birds. Nature, Lond., 327, 149-152. Burke, T., Davies, N. B., Bruford, M. W. & Hatchwell, B. J. 1989. Parental care and mating behaviour o fpolyandrous dunnocks Prunella modularis related to paternity by DNA fingerprinting. Nature, Lond., 338, 249 251. Cheng, K. M., Burns, J. T. & McKinney, F. 1983. Forced copulation in captive mallards. III. Sperm competition. Auk, 100, 302-310. Crook, J. R. & Shields, W. M. 1985. Sexually selected infanticide by adult male barn swallows. Anim. Behav., 33, 754-761. Davies, N. B. 1983. Polyandry, cloaca-pecking and sperm competition in dunnocks. Nature, Lond., 302, 334-336. Davies, N. B. 1985. Cooperation and conflict among dunnocks Prunella modularis in a variable mating system. Anim. Behav., 33, 628 648. Davies, N. B. 1986. Reproductive success of dunnocks Prunella modularis in a variable mating system. I. Factors influencing provisioning rate, nestling weight and fledging success. J. Anim. Ecol., 55, 123-138. Davies, N. B. 1990. Dunnocks: cooperation and conflict among males and females in a variable mating system.

745

In: Cooperative Breeding in Birds (Ed. by P. B. Stacey & W. D. Koenig), pp. 457-485. Cambridge: Cambridge University Press. Davies, N. B. & Brooke, M. de L. 1989. An experimental study of co-evolution between the cuckoo Cuculus canorus and its hosts. II. Host egg markings, chick discrimination and general discussion. J. Anim. Ecol., 58, 225-236. Davies, N. B. & Hatchwell, B. J. In press. The value of male parental care and its influence on reproductive allocation by male and female dunnocks Prunella modularis. J. Anim. Ecol. Hatchwell, B. J. & Davies, N. B. 1990. Provisioning of nestlings by dunnocks, Prunella modularis, in pairs and trios: compensation reactions by males and females. Behav. Ecol. Sociobiol., 27, 199-209. Hatchwell, B. J. & Davies, N. B. 1992a. An experimental study of mating competition in monogamous and polyandrous dunnocks, Prunella modular&. I. Mateguarding and copulations. Anim. Behav., 43, 595-609. Hatchwell, B. J. & Davies, N. B. 1992b. An experimental study of mating competition in monogamous and polyandrous dunnocks, Prunella modularis. II. Influence of removal and replacement experiments on mating systems. Anim. Behav., 43, 611-622. Holmes, W. G. & Sherman, P. W. 1982. The ontogeny of kin recognition in two species of ground squirrels. Am. Zool., 22, 491 517. Houston, A. I. & Davies, N. B. 1985. Evolution of cooperation and life history in dunnocks. In: Behavioural Ecology: the Ecological Consequences of Adaptive Behaviour (Ed. by R. Sibly & R. H. Smith), pp. 471 487. British Ecological Society Symposium. Oxford: Blackwell Scientific Publications. Jeffreys, A. J., Wilson, V. & Thein, S. L. 1985. Hypervariable 'minisatellite' regions in human DNA. Nature, Lond., 314, 67-73. Koenig, W. D. 1990. Opportunity of parentage and nest destruction in polygynandrous acorn woodpeckers, Melanerpes f ormicivorus. Behav. Ecol., 1, 55-61. Lake, P. E. 1975. Gamete production and the fertile period with particular reference to domesticated birds. Symp. zool. Soc. Lond., 35, 225-244. Moiler, A. P. 1988. Paternity and parental care in the swallow, Hirundo rustica. Anim. Behav., 36, 996 1005. Morton, E. S. 1987. Variation in mate guarding intensity by male purple martins. Behaviour, 101, 211-224. Page, R. E., Robinson, G. E. & Fondrk, M. K. 1989. Genetic specialists, kin recognition and nepotism in honey-bee colonies. Nature, Lond., 338, 576-579. Robertson, R. J. & Stutchbury, B. J. 1988. Experimental evidence for sexually selected infanticide in tree swallows. Anim. Behav., 36, 749 753. Westneat, D. F. 1988. Male parental care and extrapair copulations in the indigo bunting. Auk, 105, 149-160. Wetton, J. H., Carter, R. E., Parkin, D. T. & Walters, D. 1987. Demographic study of a wild house sparrow population by DNA 'fingerprinting'. Nature, Lond., 327, 147-149. Wright, J. & Cuthill, I. 1989. Manipulation of sex differences in parental care. Behav. Ecol. Sociobiol., 25, 171-181.