Evohaion, 43(6),1989, pp. 1273-1284

INBREEDING IN DARWIN'S MEDIUM GROUND ( FORTIS)

H, LISLE GIBBS'" Museum ofZoology and Department ofBiology, The University ofMichigan, Ann Arbor, MI48109-1079

AND PETER R. GRANT Department ofBiology, Princeton University, Princeton, NJ 08544-1008

Abstract. - We studied the frequency and causes of inbreeding and its effect on reproductive success in a population of Darwin's Medium Ground Finches iGeospiza fortis) on Isla Daphne Major, Galapagos, during four breeding seasons (1981, 1983, 1984, and 1987). Pedigree analysis showed that levels of inbreeding were low but comparable with those observed in other . For pairs with at least half of their grandparents known, approximately 20% of all pairings were between detectably related birds. The frequency ofpairings between closely related birds (coefficient of kinship [] ~ 0.250) among all pairs was 0.6%. We detected no effect of inbreeding on repro­ ductive success, although sample sizes were small. The observed reproductive output of related pairs was not significantly different from the output ofunrelated pairs, and there was no correlation between a pair's kinship coefficient and an estimate of the potential magnitude of inbreeding depression. Comparisons with a study ofGreat Tits (Parus major) by van Noordwijk and Scharloo (1981) suggest that, even if present, the fitness costs of inbreeding in this population of G. fortis would be low. Observed levels of inbreeding in each breeding episode were accurately predicted by simulations of random mating in which relatedness had no influence on pairing between in­ dividuals. This result suggests that levels ofinbreeding in this population are determined more by demographic factors than by behavioral avoidance of mating with kin.

Received June I, 1988. Accepted March 24,1989

It has often been argued that inbreeding direct estimates of inbreeding and the has important evolutionary consequences mechanisms responsible can only be ob­ in vertebrate populations (Mayr, 1963; tained from long-term studies of marked Wright, 1969). For example, levels of in­ individuals (e.g., studies on the Great Tit breeding can influence the overall amount [Parus major]: van Noordwijk and Scharloo of genetic variation present within popu­ [1981]; van Noordwijk et at [1985]; van lations and the degree ofkinship among in­ Tienderen and van Noordwijk [1988]). dividuals, while the negative effects of in­ However, the following key questions about breeding on fitness may play an important inbreeding remain largely unanswered. 1) role in the of a wide variety of How frequently does inbreeding occur? 2) behaviors, including patterns ofmate choice What is the effect, if any, of inbreeding on (Bateson, 1978) and sex-specific patterns of reproductive success? 3) What is the relative dispersal within species (Greenwood and importance of factors such as dispersal, Harvey, 1982). population structure, and behavioral avoid­ Inbreeding is generally defined as mating ance ofmating with kin in determining levels between related individuals, with kinship of inbreeding? judged from some arbitrarily defined ref­ Population-level estimates ofinbreeding, erence population or pedigree (see Shields inferred from pedigrees, have been calcu­ [1982] and Ralls et at [1986] for recent dis­ lated for several species (Bulmer, 1973; cussions). Despite its potential importance, Greenwood et at, 1978; van Noordwijk and inbreeding has been rarely studied in nat­ Scharloo, 1981; Payne et at, 1985; Ralls et ural populations of birds, mainly because at, 1986; Rowley et at, 1986; Craig and Jamieson, 1988). Most studies, however, have only estimated levels of inbreeding I Present address: Department of Biology, Queen's University, Kingston, ON K7L 3N6, Canada. from matings between close relatives (par­ 2 To whom all correspondence should be addressed. ents and offspring or siblings breeding with 1273 1274 H. L. GIBBS AND P. R. GRANT each other). As Templeton (1987) suggests, ulation size, degree of population subdivi­ and van Noordwijk and Scharloo's (1981) sion, and variance in reproductive success study ofinbreeding in Great Tits (Parus ma­ among individuals. jor) demonstrates, matings between more The purpose ofour study was to examine distant relatives can make an important the level, causes, and fitness consequences contribution to observed levels of inbreed­ of inbreeding in a population of Darwin's ing and should be incorporated into esti­ Medium Ground Finches iGeospiza fortis) mates of the overall frequency of inbreed­ on Isla Daphne Major, Galapagos. This ing. population has been intensively studied Studies that examine the fitness conse­ since 1976 (see Boag and Grant, 1984; Price, quences ofinbreeding in wild birds are even 1985; Gibbs and Grant, 1987a), with the rarer. Inbreeding in domesticated or captive result that practically all adults in the pop­ birds generally has a negative effect on com­ ulation have been individually marked since ponents of fitness such as egg hatchability 1978. Thus, detailed pedigrees and life-his­ and production (Shoffner, 1948; Lerner, tory information are available for many in­ 1954; Sittmann et aI., 1966; but see Daniell dividuals. The specific goals of the study and Murray [1986]). However, the effects of were to I) measure the level of inbreeding, inbreeding on reproductive success in wild using a pedigree analysis, 2) determine birds vary. Hatching success is negatively whether there was an effect of pair-relat­ correlated with pair relatedness in Great Tits edness on reproductive success, and 3) iden­ (van Noordwijk and Scharloo, 1981; see also tify the factors responsible for the observed Greenwood et al. [1978]). In contrast, highly level of inbreeding in the population. inbred pairs ofSplendid Fairy Wrens (Mal­ urus splendens) show no decrease in success compared with less inbred birds (Payne et MATERIALS AND METHODS aI., 1985; Rowley et aI., 1986). Clearly, more Populations studies are needed to determine whether in­ The population of Darwin's Medium breeding usually is associated with a fitness Ground Finches (Geospizafortis) has been cost. studied intensively on Isla Daphne Major, Moreover, the relative importance ofbe­ Galapagos, for the first 6-9 months ofeach havioral avoidance and population struc­ year from 1976 until the present (Boag and ture in generating observed levels of in­ Grant, 1984; Price, 1985; Gibbs and Grant, breeding in wild bird populations is largely 1987a). Isla Daphne Major (henceforth, unknown. There is indirect evidence, based "Daphne") is a small island «40 ha) sit­ on dispersal patterns, that individuals in uated in the middle of the Galapagos ar­ some species avoid mating with close kin chipelago. A description ofits geological and (Koenig and Pitelka, 1979). Ralls et ai. floristic characteristics is given in Boag and (1986) suggest that the low percentage of Grant (1984). Over 95% of all G. fortis in­ matings observed between close kin in a dividuals in the population have been band­ number ofpopulations implies that behav­ ed with a metal band and a unique com­ ioral avoidance of mating with close rela­ bination ofthree colored plastic bands since tives is common among birds. These stud­ 1978. Due to the demography of the pop­ ies, however, did not attempt to compare ulation, the 1981 breeding season was the observed levels ofinbreeding with those ex­ first in which matings between related birds pected ifbirds were mating at random with could be detected. Owing to annual varia­ respect to relatedness. This comparison has tion in rainfall, birds do not breed in every only been attempted for a single population year. Here, we present data on inbreeding of Great Tits (van Noordwijk et aI., 1985; in four years (1981, 1983, 1984, and 1987) van Tienderen and van Noordwijk, 1988), when enough reproduction occurred for us and the results demonstrated that pairing to identify breeding pairs reliably and to was random with respect to relatedness. This obtain information on their reproductive finding suggests that the level ofinbreeding success. in this population is determined mainly by Finches breed in response to rain, which aspects ofpopulation structure such as pop- can fall any time during December-July. INBREEDING IN FINCHES 1275

During each of the four breeding seasons are based on calculations involving effective reported here, all nests were marked, and population size (see Crow and Kimura, the parents were identified. Nest contents 1970): 1) the demographic data needed to were regularly checked to determine clutch estimate N; are extensive and difficult to size, hatching, and fledging success. We are obtain for natural populations; 2) changes confident that more than 95% of all nests in our knowledge of ancestry in the popu­ were found, because repeated systematic lation, which strongly influence detected searches ofall potential nest sites were con­ levels of inbreeding, can be incorporated ducted and because fledglings without bands into our estimates of expected levels of in­ were rarely seen. Young birds that were alive breeding; and 3) with simulations, we were in January ofthe year following the season able to generate a distribution of expected of their birth were recorded and identified values ofkinship under random mating and, during repeated and standardized visual hence, to obtain a plausible estimate of the censuses and other observations (Boag and variance of the distribution. Grant, 1984; Gibbs and Grant, 1987b). We compared observed and random pat­ terns of pairing during the first episode of Inbreeding Analyses breeding in each of four years: 1981, 1983, Inbreeding is defined here as any mating 1984, and 1987. We did the same for the between detectably related individuals (see middle of the 1983 season (April), when van Noordwijk and Scharloo, 1981). The large numbers of young birds, born earlier degree of relatedness between members of in the year, began breeding (see Gibbs et aI., a pair is described by their coefficient of 1984; Gibbs and Grant, 1987a). We con­ kinship (cP), which is the probability that an sidered only the first episode of mating in allele chosen randomly from either individ­ most years, because the formation of most ual would be identical by descent (Malecot, new pairs takes place at the beginning of 1948). The coefficient of kinship of a pair each breeding season. For each breeding ep­ is equal, by definition, to the coefficient of isode, we used the following procedure to inbreeding of their offspring. Kinship coef­ generate a frequency distribution of ex­ ficients for all pairs in each breeding season pected levels ofkinship ifpairing were ran­ were calculated from pedigrees with the SAS dom with respect to relatedness. Supplemental Library Program INBREED (Reinhardt, 1979). We assumed that pairs 1) To determine which birds could po­ with no detected common ancestors had a tentially pair at the start of a year, we kinship coefficient of zero. Clearly, the ob­ identified all reproductively active fe­ served level of kinship is a minimum esti­ males and males on the island from cen­ mate, because common ancestors for many sus data. During 1981-1984, breeding fe­ birds were not known. The depth of the males were identified as all females that pedigree for the population varied consid­ laid eggs within the first 30 days of the erably through time; in 1981, the minimum breeding season. In 1987, the sex ratio detectable level of kinship was cP = 0.125. was slightly female-biased; hence, all po­ By 1987, this value had decreased to 0.008. tentially reproductive females did not breed immediately. In this year, female Expected Levels ofInbreeding breeders were identified as birds that had To determine whether the frequency of reproduced in previous years. In all years, inbreeding is greater or less than expected males were identified as birds that ac­ by chance, we followed the empirically based tively defended territories. approach ofvan Noordwijk et al. (1985) and 2) We then excluded from this group van Tiederen and van Noordwijk (1988) and those males and females that had bred used computer simulations of pairing be­ with the same mate in the previous breed­ tween individuals to determine whether re­ ing season, on the assumption that these latedness influenced mate choice and, hence, individuals would not be available for re­ levels ofinbreeding. This approach has three mating. advantages over the less direct ways of es­ 3) We then paired each member ofthe timating expected levels of inbreeding that limiting sex (females in all years except 1276 H. L. GIBBS AND P. R. GRANT

1987) with a single randomly chosen that bred in the same year, the median dis­ member of the opposite sex and calcu­ persal distance was greater for females than lated cP for each randomly formed pair. for males (females: median distance = 263 The mean level of inbreeding for the en­ m, range = 37-634 m, N= 104; males: me­ tire round of simulated mating was cal­ dian distance = 190 m; range = 18-525 m, culated as the mean cP value for all pairs N = 61; Mann Whitney U test, P < 0.05). (randomly formed pairs and long-term This difference between the sexes violates pairs combined). Both randomly and pre­ the "equal chance" assumption, because the viously formed pairs were combined in observed distances suggest that females can order to make this estimated value di­ potentially disperse over the entire island rectly comparable with the observed val­ while males are more restricted in their ue (see step 5). movements. It may be unimportant, how­ 4) This procedure was then repeated ever, since females (the wide-ranging sex) 1,000 times to generate an expected dis­ were also the "choosing" sex in all years tribution of random pairings for each except 1987, due to the male-biased sex ra­ breeding episode. tio on the island during 1981-1984 (Price, 5) Observed levels of inbreeding were 1984; Gibbs and Grant, unpubI.). The sec­ calculated as the mean cP value for first ond assumption, that pairing occurs simul­ pairings of all breeding females (1981­ taneously, is generally upheld in our pop­ 1984) or breeding males (1987) within a ulation; pairing at the start of a breeding given reproductive episode. season occurs rapidly, usually over a 1-2­ week period (Price, 1984). Our simulation ofrandom mating for this population makes several assumptions that Success ofRelated Pairs are somewhat unrealistic about the pool of We evaluated the effect of inbreeding on potential mates and their pairing behavior. eight related measures ofreproductive suc­ First, we assumed that patterns ofdispersal cess at different stages ofthe life cycle: clutch do not influence the probability ofeach un­ size, number ofyoung that hatched, number mated bird pairing with each unmated ofyoung that fledged, number ofyoung that member of the opposite sex. Van Tiederen survived to the following year, percentage and van Noordwijk (1988) discuss how the ofeggs that hatched, percentage ofeggs that dispersal patterns of individuals can influ­ fledged, percentage of hatched young that ence levels of inbreeding in a natural pop­ fledged,and percentage offledged young that ulation. Second, we assumed that pairing survived. We compared the observed suc­ occurs simultaneously among all unpaired cess ofrelated pairs with an estimate oftheir birds during a breeding episode, i.e., that expected performance had they been unre­ the population is not subdivided as a result lated. This estimate was calculated from a ofdifferent individuals breeding at different sample of pairs from the same season that times. were not known to be related; pairs were These assumptions, however, may not matched for factors known to have a prox­ deviate very strongly from the observed be­ imate effect on reproductive success, such havior of the birds. First, the small size of as pair age (Gibbs et aI., 1984) and seasonal the island (maximum N-5 diameter = 743 changes in reproductive success (Boag and m; maximum E-W diameter = 560 m), and Grant, 1984; Price, 1985; Gibbs and Grant, the movement ofjuvenile and nonbreeding 1987a). Greenwood et al. (1978) and Row­ birds over much of the island (Gibbs and ley et al. (1986) adopted a similar approach Grant, unpubI.), suggest that mate choice is in their studies ofinbreeding in other birds. probably not constrained by dispersal op­ Some undetected closely related pairs were portunities. Natal dispersal probably occurs probably included in the nonrelated group, over the entire island; however, judging from particularly in the early years of the study, very limited data, females disperse farther when the pedigree for the population was than males, as has been found in some other shallow. To reduce these pairs to a mini­ passerine species (Greenwood and Harvey, mum, we used only pairs from age-month 1982). For example, for birds born in 1983 classes that had the greatest number of INBREEDING IN FINCHES 1277

TABLE I. Occurrence ofdetected inbreeding in G. fortis on Isla Daphne Major from 1981 to 1987.

Detected close inbreeding Detected inbreeding (tI> ". 0.125) Number Number Number Mean Pair type of pairs of pairs % of pairs % inbreeding (tI» Unique pairs (male and female identified) 583 27 4.6 8 1.4 0.00394 Parents of male and female known 345 26 7.5 7 2.0 0.00592 Parents and at least half of the grandparents known 70 17 24.3 0 0 0.0120 All grandparents known 10 2 20.0 0 0 0.0110 known ancestors and were represented by Scharloo (1981), observed levels ofinbreed­ at least 15 pairs. Approximately halfofthe ing are strongly influenced by the extent of related pairs bred more than once within a our knowledge ofthe ancestry ofindividuals year or in different breeding seasons. Since in a population. The positive effect ofped­ the performances of these birds during re­ igree depth on detected inbreeding is clear peated breeding attempts were not indepen­ in these data. Detected matings between re­ dent, we averaged the observed measures of lated birds increased from 4.6% for all pairs success from two or more attempts, did the to more than 20% for pairs for which at least same for the expected measures ofsuccess, halfofthe grandparents were known (Table and thereby obtained single values for the 1). The comparable estimates of mean pair expected and observed performances ofeach relatedness are 0.004 and 0.012, respec­ related pair over the study. We were not tively. Detected levels of close inbreeding able to examine the effect of the level of (4) ~ 0.125), often cited as a measure of inbreeding among individual females on population levels of inbreeding (cf. Green­ possible maternal traits such as clutch size wood et aI., 1978), ranged from 0% to 2% (cf. van Noordwijk and Scharloo, 1981; but for all pairs in different categories. Thus, on see Gibbs [1988]) because only three inbred average, for birds with at least half of their females reproduced during our study (see grandparents known, a minimum of20% of below). their matings are between related birds, re­ sulting in a minimum mean 4> value of0.01 for all pairs. RESULTS Matings Between Related Birds Effect ofRelatedness on Pairing Table 1 shows observed levels ofinbreed­ To see whether pairing was nonrandom ing (percentage matings between related in­ with respect to relatedness, we compared dividuals) for different subsets ofall unique observed levels of inbreeding with those pairings between different males and differ­ generated from simulated episodes of pair­ ent females. In the period 1981-1987,4.6% ing in which mating between birds was ran­ (27 of 583) ofall pairings were between re­ dom with respect to relatedness. Figure 1 lated individuals, with coefficients of kin­ compares the distributions of mean relat­ ship ranging from 0.008 to 0.250. The ob­ edness per episode of random mating (for served distribution of 4> among inbred 1,000 episodes ofpairing) with the observed matings was: 0.008 (N = 1 pair), 0.016 (N value for five separate breeding periods. = 3), 0.031 (N = 3),0.047 (N = 1),0.063 There is a clear effect of pedigree depth on (N = 10),0.094 (N = 1),0.125 (N = 5), and both the continuity and symmetry of the 0.250 (N = 3). Matings between distant kin random distributions. For 1981, the distri­ (4) < 0.125) make an important numerical bution is strongly skewed to the right, and contribution to the overall levels ofinbreed­ there are frequent gaps along the abcissa. By ing in this population; more than 70% ofall 1987 the center ofthe distribution had shift­ inbred matings are between distant kin, and ed to the right, reflecting the increase in de­ these matings contribute 40.1 % to the tected levels ofinbreeding, although the dis­ summed values of 4> for all pairings. tribution itselfwas still skewed to the right. As emphasized by van Noordwijk and The lack oflow relatedness values in certain 1278 H. L. GIBBS AND P. R. GRANT

50

1983L 40 Obs 1981 1983E 1 30 Obs Obs ~ 1 20

"0• 10 c ::J 0 II: 0 '0 0 0.5 1.0 1.5 C 50 II o 1984 1981 CD 40 Q. Obs Obs 30 1 1 20

10

0 0 0.5 1.0 1.5 0 0.5 1.0 1.5

Mean Relatedness Per Round 01 Random Mating (x 10 2)

FIG. I. Distributions of mean inbreeding per round ofpairing for simulations ofrandom mating by G. fortis with respect to relatedness. Five breeding episodes are shown, which simulate pairing during the first round of breeding in 1981, 1983 (1983E), 1984, and 1987 and for an episode ofpairing in the middle ofthe 1983 breeding season (1983L; see Materials and Methods). Observed levels ofinbreeding are indicated by arrows above each distribution. periods (e.g., 1983E) as compared to others inbreeding appear to be extremely close to (e.g., 1981 and 1983L) is due to the varying those expected ifpairing between individual persistence of small numbers of long-term birds were random with respect to relat­ inbred pairs from one year to the next. For edness. each breeding episode, however, the ob­ The use of population-wide estimates of served level ofinbreeding (mean relatedness pairing (i.e., mean relatedness per round of offirst pairings by each member ofthe lim­ pairing for all birds) could mask subtle ef­ iting sex in each episode; see Materials and fects ofrelatedness on pairing. For example, Methods) is very close to the middle value of each of the random distributions. TABLE 2. Expected versus observed levels ofinbreed­ Table 2 gives a quantitative description ing in G. fortis from 1981 to 1987. All values for '" of how well random mating predicts ob­ have been multiplied by 100. Only the first breeding served inbreeding in each breeding episode. episode is used for 1981, 1984, and 1987 (see text). For each episode, the difference between the Observed median value of cP from the random distri­ number of cP Percentage of pairs per bution and the observed value of is ex­ simulations breeding tremely small: the maximum difference is Median with values episode from random less than (number of 0.021, or less than 7% ofthe observed value Year distribution Observed observed related pairs) of cPo The fact that, in each case, almost 1981 0.176 0.164 49.0 76 (I) exactly half of the random simulations 1983 (range: 48-56%) have values less than the Early 0.536 0.554 55.4 79 (3) observed value also demonstrates the close Late 0.335 0.314 47.9 185 (7) correspondence between the expected ran­ 1984 0.432 0.440 51.7 164 (8) dom values and the observed values of cPo Thus, in all years, the observed values of 1987 0.508 0.510 52.6 93 (4) INBREEDING IN FINCHES 1279

TABLE 3. Expected and observed frequencies of related matings among newly formed pairs in each year. Expected values were calculated from simulations for each breeding episode and only include pairings with t/> values greater than zero.

Relatedness

< 0.125 >: 0.125 Expected Observed Expected Observed Year N % N % N % N % 1981 0 0 0 0 92 100 100 1983 Early 34 21 I 50 127 79 I 50 Late 376 60 3 50 247 40 3 50 1984 679 68 5 83 320 32 17 1987 214 75 9 90 70 25 10 ifavoidance ofpairing with close kin is off­ pie sizes are small, making statistical tests set by preferential pairing with more distant inappropriate, but in no case is there strong kin (see Bateson, 1978), then the mean kin­ evidence that close kin are being avoided ship coefficient, averaged across all pairs, while more distant kin are preferred (Table would be similar to one calculated if pairing 3). were random. To see whether such a subtle pattern ofmate choice occurred, for the sim­ Reproductive Success ofRelated Pairs ulations for each breeding episode we de­ To determine whether inbreeding affects termined all possible pairings between re­ reproductive success, we compared the ob­ lated individuals and calculated the numbers served success of related pairs with esti­ of pairs in two categories of relatedness, mates ofexpected success based on the per­ those with kinship coefficients of at least formance of pairs not known to be related 0.125 and those with coefficients less than (Table 4). A paired comparison ofobserved 0.125 but greater than zero. We then com­ with expected values for all related pairs pared these numbers of pairs with similar (mean cJ> = 0.085; N = 27) shows no signif­ values for observed newly formed pairings icant differences for eight components of between related birds in each period. Sam- reproductive success.

TABLE 4. Expected and observed measures of reproductive success for related pairs of G./or/is. P values are from a paired t test. MSD is the smallest difference between expected and observed that can be detected for a given sample using a two-tailed t test with a 0.05 level ofsignificance (Zar, 1984). The percentage ofyoung that fledged was measured only for matings from which at least one egg hatched; the percentage of young that survived was measured only for matings from which at least one offspring fledged.

Types of pairs Measure of success N Expected Observed P MSD All related Clutch size 27 3.49 3.50 0.06 0.95 0.30 ( > 0) Number ofeggs hatched 27 2.01 1.65 1.64 O.ll 0.45 Number of young fledged 27 1.67 1.40 1.28 0.21 0.43 Number of young survived 16 0.69 0.52 1.09 0.29 0.34 Percentage of eggs hatched 27 54 45 1.74 0.09 II Percentage of young fledged 21 79 85 0.90 0.38 13 Percentage of eggs fledged 27 45 38 1.38 0.18 II Percentage of young survived 14 48 33 1.58 0.14 21 Closely related Clutch size 8 3.57 3.34 0.69 0.51 0.77 ( ~ 0.125) Number ofeggs hatched 8 1.92 1.68 0.52 0.62 1.12 Number of young fledged 8 1.58 1.52 0.18 0.90 1.07 Number of young survived 6 0.77 0.72 0.12 0.91 0.99 Percentage of eggs hatched 8 51 44 0.77 0.46 22 Percentage of young fledged 6 79 75 0.40 0.71 25 Percentage ofeggs fledged 8 41 36 0.51 0.62 22 Percentage of young survived 5 51 39 1.11 0.33 30 1280 H. L. GIBBS AND P. R. GRANT

• Eggs Hatched 1.0 Young Fledged "0 0 •ell "0 • 0.5 8 ii:..... 8 "0 11 11 "0 ~• - u 0 8 U• :I: •Do • ell )( c W ~ -0.5 I 0 "0 > • ci •~ Z -1.0 it -'2• g 1.0 c .5! "0 •ell "0 ">• • 0.5 • ii: 8 c ..... 8 c "0 11 9 6 ~• 6 • u 0 + ::E• <} :I:• ell c ~ -0.5 0 > ~ -1.0

.008-.049 .063-.094 .125-.250

Kinship Coefficient Cia .. FIG. 2. The relationship between average deviations (observed success - expected success) for different measures of reproductive success and the coefficient of kinship for related pairs of G. fortis. The measures of success were: number ofeggs hatched and number ofyoung fledged (above), and percentage ofeggs hatched and percentage of hatched young that fledged (below). Values are means ± SE. Sample sizes (numbers of pairs) are given above each point.

To determine whether significant differ­ depression is, in fact, greatest for the least ences in reproductive success occur only related class, not for the most related one. among closely related (cJ> 2: 0.125) pairs (see Rank correlations between the differences Greenwood et aI., 1978; Ralls et aI., 1986), in each of the eight measures of reproduc­ we repeated the above analyses using only tive success in Table 4 and cJ> ranged between pairs with a cJ> value of0.125 or greater (Ta­ -0.29 and 0.05. None ofthe differences was ble 4). As before, there was no detected dif­ significant, and the signs ofthe correlations ference between observed and expected val­ varied, indicating that there was no consis­ ues ofreproductive success for related pairs, tent relationship between degree ofinbreed­ although sample sizes are small (N = 5-8) ing and success. and the values for the minimum detectable significant difference (MSD) show that the power of the statistical tests is low. Inbred Offspring If inbreeding depression occurred, the Inbred offspring are both viable and fer­ magnitude ofthis depression should covary tile. A total of eight inbred offspring, pro­ positively with pair relatedness; we would duced from matings between related parents therefore expect a significant correlation be­ in early 1983, were recruited into the breed­ tween cJ> and the difference between expected ing population from late 1983 to 1987. They and observed values of reproductive suc­ consisted of five males, including four cess. There is no such relationship for four brothers from the same clutch and three fe­ measures of reproductive success (Fig. 2). males, each from different nests. All the The upper part of the figure shows that the brothers had coefficients of inbreeding (F) INBREEDING IN FINCHES 1281 of0.125, while the fifth male had an Fvalue vides a less complete but more widely avail­ of0.063; two ofthe females had F= 0.125, able comparative measure of the levels of while the other had an F = 0.063. All these inbreeding in populations. Ralls et al. (1986) birds had at least one nest, and their clutch summarize the available estimates for na­ sizes ranged from two to five eggs. One of tive birds that do not breed cooperatively, the male siblings had a total of six nests, using mostly unpublished data. The per­ while two ofthe inbred females (F = 0.125 centage for five mainland and one island and 0.063) had seven nests each. Six ofthe population offour species ranges from 0.4% eight birds fledged at least one young. for the Pied Flycatcher(Ficedula hypoleuca) to 3.0% for the Vlieland Great Tit popula­ DISCUSSION tion. The equivalent figure for G. fortis on Daphne is 0.6%. Thus, inbreeding among There are still relatively few studies of G. fortis on Daphne appears to be as fre­ inbreeding in birds based on pedigree anal­ quent as in some other species with similar ysis (for review see Ralls et al. [1986]). social behavior but somewhat rarer than in Nonetheless, the work done so far suggests the Great Tit. that observed levels ofinbreeding are com­ The consistently low percentages ofclosely parable among species with similar social inbred matings in these populations sug­ behavior. As shown in this study, levels of gested to Ralls et al. (1986) that close in­ inbreeding appear to be determined more breeding may be avoided behaviorally. This by population structure than by behavioral appears unlikely to us, since pairing occurs avoidance ofmating with relatives (see also at random with respect to relatedness in two van Noordwijk et al. [1985] and van Tien­ species, Darwin's Medium Ground Finches deren and van Noordwijk [1988]). In con­ and Great Tits, which span the range (0.6%­ trast, the effects ofinbreeding on reproduc­ 3.0%) in terms of percentages of closely tive success differ considerably between inbred pairs. More likely, the similarity in populations. levels ofinbreeding between studies is a re­ sult ofsimilarity in the effective population Levels ofInbreeding sizes of the birds under study (van Tien­ To compare the frequency of inbreeding deren and van Noordwijk, 1988). We em­ in different populations, it is necessary to phasize, as Ralls et al. (1986) acknowledge, take into account the differences in pedigree that inbreeding avoidance can only be di­ depth. This can be done in two related ways: rectly demonstrated by comparing the ob­ I) by estimating a mean c/> for a sample of served pattern ofmating with a random ex­ birds with a certain minimum number of pectation. known ancestors and 2) by estimating the percentage of all matings made up of pair­ ings between particular classes of relatives. EJ.rectofInbreeding To our knowledge, van Noordwijk and The major difference between this study Scharloo (1981) presented the only previous and the detailed work on Great Tits (Green­ estimate of inbreeding based on the first wood et aI., 1978; van Noordwijk and method. They estimated the mean value of Scharloo, 1981) is that we could detect no c/> to be 0.036 for the Vlieland population effect ofinbreeding on reproductive success, of Great Tits, based on pairs for which at whereas van Noordwijk and Scharloo (1981) least grandparents were known. The equiv­ detected a strong and significant effect, alent value for G. fortis is 0.0 II (Table I), mainly on hatching success. Our result is suggesting that inbreeding may be some­ not without precedent. Rowley et al, (1986) what lower in the finches. However, the es­ recently reported that reproductive success timate for G. fortis is much less reliable, did not differ between very closely related because it is based on a much smaller sam­ (c/> ~ 0.250) and less closely related (c/> < ple size: all grandparents were known for 0.250) pairs ofSplendid Fairy Wrens (Ma­ 33% of all Great Tit pairs, but for only 2% lurus splendens) (see also Payne et al. [1985], ofall Darwin's pairs. James et al. [1987], and Payne et al. [1987]). The percentage of all matings between Clearly, our sample sizes of related pairs very closely related birds (c/> ~ 0.250) pro- offinches were small, particularly the num- 1282 H. L. GIBBS AND P. R. GRANT ber of closely related pairs (¢ ~ 0.125), in ing males and 20.3-49.1% of breeding fe­ which possible costs of inbreeding are ex­ males (depending on area) were presumably pected to be highest. Ouranalyses ofsuccess nonrelated immigrants (van Tienderen and thus have a conservative bias because some van Noordwijk, 1988). In contrast, the pairs of birds that we assumed to be unre­ Daphne population ofG.fortis is much more lated may in fact share recent common closed, with a maximum of 0-3% immi­ ancestors. Therefore, more data are needed grants per breeding season. Although some to confirm the conclusion that costs of in­ hybridization also occurs (see Gibbs and breeding are small or nonexistent in these Grant, 1987a), it is infrequent (approxi­ birds. Nevertheless, based on the argument mately 2% per breeding season). Severe presented below, we suggest that, even if population bottlenecks on Daphne can also additional data were to result in the detec­ occur, with the number ofbreeding females tion ofa cost, its magnitude would probably ranging from more than 200 birds to 28 still be substantially lower than the cost of individuals in one year (Price et aI., 1984). inbreeding observed in Great Tits. These factors probably combine to reduce If, as in Great Tits, the negative effects of the long-term effective population size ofG. inbreeding act mainly to reduce hatching fortis to a level much lower than that ofthe success and if the magnitude of this effect Vlieland Great Tit population. Thus, dele­ can be estimated as the difference in success terious alleles may have been purged from between related and nonrelated pairs, then the finch population in the past (see Tem­ we can compare the potential cost of in­ pleton, 1987), accounting for the difference breeding in G. fortis with the demonstrated in the effects of inbreeding on reproductive cost in Great Tits. Restricting the analysis success between the two species. This hy­ to closely related pairs (¢ ~ 0.125), the re­ pothesis could be indirectly tested by com­ duction in the number of eggs hatched is paring current levels ofheterozygosity in the approximately 1.25 for pairs of Great Tits G. fortis and Great Tit populations using and 0.24 for pairs of G. fortis. Part of the genetic markers (cf. Yang and Patton, 1981). difference can be explained by the larger The prediction is that the G. fortis popula­ mean clutch size of Great Tits, which is tion should have lower levels of variation about twice that ofG.fortis (van Noordwijk than the Great Tit population (Nei et al., et al., 1981). Nevertheless, on a percentage 1975). basis, the loss in eggs hatched is 18% in Great Tits and only 7% in G. fortis (values Avoidance ofInbreeding? for Great Tits were estimated from fig. 4 There is behavioral evidence that close and table 5 in van Noordwijk and Scharloo inbreeding is avoided in some cooperatively [1981]; those for G. fortis were based on the breeding birds (e.g., Koenig and Pitelka, data in Table 4). The loss for the finches 1979), even though the supposed reason for appears to be substantially lower than the such behavior (a cost to inbreeding) has not loss for Great Tits; however, the small non­ been demonstrated. Nonetheless, the close significant increase in recruitment for inbred correspondence between observed levels of Great Tit young relative to outbred young inbreeding and those predicted under a may reduce this difference slightly (van model ofrandom pairing with regard to re­ Noordwijk and Scharloo, 1981). latedness strongly suggests that population Even though present levels of inbreeding structure is the primary determinant ofob­ are similar, the difference between Great Tits served levels of inbreeding in the G. fortis and G. fortis in the effect of inbreeding on population on Isla Daphne Major. Similar­ reproductive success may be due to histor­ ly, van Noordwijk et aI. (1985) and van ical differences in population size and struc­ Tienderen and van Noordwijk (1988) found ture that have led to reduced levels of het­ little difference between the mean kinship erozygosity and lower numbers of coefficients ofobserved and randomly gen­ deleterious alleles in the G. fortis popula­ erated pairs of Great Tits. These studies of tion. The Great Tit population on Vlieland the Great Tit and our study of G. fortis are appears to be relatively open and outbred: the only ones that have generated expected over the study period, 8.5-30.8% ofbreed- levels of inbreeding under models of ran- INBREEDING IN FINCHES 1283 dom pairing and then compared the results Biology, University of Michigan, and a with observations from a wild population. Hinsdale-Walker Scholarship from the Mu­ The lack ofdiscrimination against mating seum of Zoology, University of Michigan. with relatives does not appear to be related to the presence or absence of a cost to in­ LITERATURE CITED breeding. Both Great Tits, which show a BATESON, P. P. G. 1978. Sexual imprinting and op­ cost, and G. fortis. for which there is no timal outbreeding. Nature 273:659-660. evidence of a cost, mate at random with BEECHER, M. D. 1982. Signature systems and kin relatives. It is likely that factors other than recognition. Amer. Zool. 23:477-490. kinship have a stronger effect on mate choice BOAG, P. T., AND P. R. GRANT. 1984. Darwin's in both species. In G. fortis. these factors Finches (Geospiza) on Isla Daphne Major, Gala­ pagos: Breeding and feeding ecology in a climati­ include the size and age of the birds and cally variable environment. Ecol. Monogr. 54:463­ territory size (Price, 1984). In G. conirostris 489. on Isla Genovesa, age-related experience and BULMER, M. G. 1973. Inbreeding in the Great Tit. behavior of males is of overriding impor­ Heredity 30:313-325. CRAIG, J. L., AND I. JAMIESON. 1988. Incestuous mat­ tance in attracting a mate (Grant and Grant, ing in a communal bird: A family affair. Amer. 1987). Another possibility is that kin rec­ Natur. 131:58-70. ognition is poorly developed in all these CROW, J. F., AND M. KIMURA. 1970. An Introduction species. In birds, kin recognition appears to to Population Genetic Theory. Harper and Row, develop through social association, rather N.Y. DANIELL, A., AND N. D. MURRAY. 1986. Effects of than through direct assessment of genetic inbreeding in the Budgerigar (Melopsittacus undu­ relatedness (Beecher, 1982; Payne et al., latus: Aves: Psittacidae). Zoo BioI. 5:233-238. 1988). Thus, while parents or nestmates GIBBS, H. L. 1988. Heritability and selection on clutch could recognize each other, the opportunity size in Darwin's Medium Ground Finches (Geo­ spiza fortis). Evolution 42:750-762. to develop recognition of more distant rel­ GIBBS, H. L., AND P. R. GRANT. 1987a. Ecological atives or siblings born at different times may consequences of an exceptionally strong EI Nino be absent (Grant, 1984). event on Darwin's Finches. Ecology 68: 1735-1746. To conclude, while levels of inbreeding --. 1987b. Adult survivorship in Darwin's are comparable to those found in other pas­ Ground Finch populations in a variable environ­ ment. J. Anim. Ecol. 56:797-813. serines, there was no detectable effect ofin­ GIBBS, H. L., P. R. GRANT, AND J. WEILAND. 1984. breeding on reproductive success, or of re­ Breeding of Darwin's Finches at an exceptionally latedness on mate choice in G. fortis. early age in an EI Nino year. Auk 101:872-874. Population structure seems to be a more GRANT, B. R. 1984. The significance ofsong variation in a population of Darwin's Finches. Behavior 89: important determinant oflevels ofinbreed­ 90-116. ing in these birds than does behavioral GRANT, B. R., AND P. R. GRANT. 1987. Mate choice avoidance of mating with kin. in Darwin's Finches. BioI. J. Linn. Soc. 32:247­ 270. ACKNOWLEDGMENTS GREENWOOD, P. J., AND P. H. HARVEY. 1982. The natal and breeding dispersal of birds. Ann. Rev. We thank P. Boag, M. Bulmer, P. Myers, Ecol. Syst. 13:1-2 I. B. Grant, R. Payne, T. Price, P. van Tien­ GREENWOOD, P. J., P. H. HARVEY, AND eM. PERRINS. deren, and especially P. Smouse for com­ 1978. Inbreeding and dispersal in the Great Tit. ments and discussion; G. Keyes, S. Latta, Nature 271:52-54. JAMES, P. C, L. Q. OLIPHANT, AND I. G. WARKENTIN. D. McCullough, and J. Weiland, for help 1987. Close inbreeding in the Merlin (Falco co­ with the field work; R. Bush and J. Tworek lumbariusi. Wilson Bull. 99:718-719. for assistance with the analysis; and T. Price KOENIG, W. D.. AND F. A. PITELKA. 1979. Related­ for the use ofhis data. This work was carried ness and inbreeding avoidance in the communally out with the permission of the Direccion nesting Acorn Woodpecker. Science 206: 1103-1105. LERNER, I. M. 1954. Genetic Homeostasis. Wiley, General de Desarrollo, Quito, and with the N.Y. help of the Servicio Parque Nacional Ga­ MALECOT, G. 1948. Les Mathernatiques de l'Here­ lapagos, and the Charles Darwin Research dite, Masson, Paris, France. Station. It was supported by NSF grants MAYR, E. 1963. Species and Evolution. Har­ vard Univ. Press, Cambridge, MA. to P.R.G. H.L.G. was supported by an NEI,M., T. MARUYAMA,ANDR.CHAKRABORTY. 1975. NSERC (Canada) Postgraduate Scholar­ The bottleneck effect and genetic variability in pop­ ship, Block Grants from the Department of ulations. Evolution 29: 1-10. 1284 H. L. GIBBS AND P. R. GRANT

PAYNE, R. B., L. L. PAYNE, AND S. M. DoEHLERT. SHOFFNER, R. N. 1948. The reaction of fowl to in­ 1987. Song, mate choice, and the question of kin breeding. Poultry Sci. 27:448-452. recognition in a migratory song bird. Anim. Behav, SITTMANN, K., H. ABPLANALP, AND R. A. FRASER. 35:35-47. 1966. Inbreeding depression in Japanese Quail. PAYNE, R. B., L. L. PAYNE, AND I. ROWLEY. 1985. Genetics 54:371-379. Splendid wren Malurus splendens response to cuc­ TEMPLETON, A. J. 1987. Inferences on natural pop­ koos: An experimental test of social organization ulation structure from genetic studies on captive in a communal bird. Behaviour 94:108-127. mammalian populations, pp. 257-272. In B. D. --. 1988. Kinship and nest defense in co-oper­ Chepko-Slade and Z. T. Halpin (eds.), Mammalian ative birds: Splendid Fairy Wrens (Malurus splen­ Dispersal Patterns. Univ. Chicago Press, Chicago, dens). Anim. Behav. 36:939-941. IL. PRICE, T. D. 1984. Sexual selection and body size, VAN NooRDWUK, A. J., AND W. ScHARLOO. 1981. In­ , and territory variables in a population of breeding in an island population of the Great Tit. Darwin's Finches. Evolution 38:327-341. Evolution 35:674-688. ---. 1985. Reproductive responses to varying food VAN NooRDWUK, A. L, J. H. VAN BALEN, AND W. supply in a population ofDarwin's Finches: Clutch SCHARLOO. 1981. Genetic and environmental size, growth rates, and hatching synchrony. Oeco­ components ofvariation in clutch size ofthe Great logica 66:411-416. Tit (Parus major). Neth. J. Zool. 31:342-373. PRICE, T. D., P. R. GRANT, AND P. T. BoAG. 1984. VAN NooRDWUK, A. J., P. H. VAN TIENDEREN, G. DE Genetic changes in the morphological differentia­ ]ONG, AND J. H. VAN BALEN. 1985. Geneological tion ofDarwin's Ground Finches, pp. 49-66. In K. evidence for random mating in a natural population Wohrmann and V. Loescheke (eds.), Population of the Great Tit (Parus major L.). Naturwissen­ Biology and Evolution. Springer-Verlag, Berlin, W. shaften 72:104-106. Ger. VAN TIENDEREN, P. H., AND A. J. VAN NooRDWUK. RALLS, K., P. H. HARVEY, AND A. M. LYLES. 1986. 1988. Dispersal, kinship and inbreeding in an is­ Inbreeding in natural populations of birds and land population of the Great Tit. J. Evol, Biol. I: mammals, pp. 35-56. In M. Soule (ed.), Conser­ 117-137. vation Biology: The Science ofScarcity and Diver­ WRIGHT, S. 1969. Evolution and the Genetics of Pop­ sity. Sinauer, Sunderland, MA. ulations, Vol. 2. The Theory ofGene Frequencies. REINHARDT, P. S. (ed.). 1979. SAS Supplemental Univ. Chicago Press, Chicago, IL. User's Guide. SAS Inst., Inc., Cary, NC. YANG, S. Y., AND J. L. PATTON. 1981. Genic vari­ ROWLEY, I., E. RUSSELL, AND M. BROOKER. 1986. ability and differentiation in Galapagos finches. Auk Inbreeding: Benefits outweigh costs. Anim. Behav. 98:230-242. 34:939-941. ZAR, J. H. 1984. Biostatistical Analysis, 2nd Ed. SHIELDS, W. M. 1982. Philopatry, Inbreeding, and Prentice-Hall, Englewood Cliffs, NJ. the Evolution of Sex. State Univ. New York Press, Albany, NY. Corresponding Editor: G. de Jong