Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
The Heritability of External Morphology in Darwin's Ground Finches (Geospiza) on Isla Daphne Major, Galapagos Author(s): Peter T. Boag Source: Evolution, Vol. 37, No. 5 (Sep., 1983), pp. 877-894 Published by: Society for the Study of Evolution Stable URL: http://www.jstor.org/stable/2408404 Accessed: 06-03-2017 15:40 UTC JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://about.jstor.org/terms Society for the Study of Evolution is collaborating with JSTOR to digitize, preserve and extend access to Evolution This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms EVOLUTION INTERNATIONAL JOURNAL OF ORGANIC EVOLUTION PUBLISHED BY THE SOCIETY FOR THE STUDY OF EVOLUTION Vol. 37 September, 1983 No. 5 Evolution, 37(5), 1983, pp 877-894 THE HERITABILITY OF EXTERNAL MORPHOLOGY IN DARWIN'S GROUND FINCHES (GEOSPIZA) ON ISLA DAPHNE MAJOR, GALAPAGOS PETER T. BOAG1 Department of Biology, Trent University, Peterborough, Ontario K9J 7B8, Canada Received October 23, 1981. Revised December 20, 1982 bill length in mm) in terms of a genotypic value and an environmental deviation, i.e., P = G + E, or for the phenotypic variance of a population, Vp = VG + VE. The genotypic value of an individual can in turn be partitioned into a breeding value, a dominance deviation, and an interaction deviation, i.e., G = A + D + I, or VG = Ecologists use avian morphological measurements to develop and test evolu- tionary theories. The theories are usually based on genetic models, although little is known about the inheritance of such char- acters. In field studies it is commonly assumed that phenotypic variation closely reflects underlying genetic variation (Grant VA + VD + VI. Genetic resemblance be- et al., 1976), while theoreticians some- et al., 1976; Abbott et al., 1977; Boag and tween relatives is a function of breeding values; thus the variance of breeding values (the additive genetic variance or VA above), is an important property of a pop- Grant, 1981; Grant, 198 lb) have been ulation. times assume that heritabilities are equal to one (Long, 1974). P. R. Grant and his colleagues (Grant studying Darwin's ground finches (Geospiza) in the Galapagos, focusing on relationships between finch morphology and food supplies. Unlike Mendelian characters such as plumage polymorphisms (Mi- neau and Cooke, 1979), finch morphology involves metric characters and is studied using quantitative genetics (Falconer, 1981). Quantitative genetics describes the phenotypic value of an individual (e.g., its 1 Present address: Department of Biology, University, Kingston, Ontario, Canada K7L 3N6. Heritability in the narrow sense (h2) is the proportion of the phenotypic variance which is additive genetic (Falconer, 1981). The response of a trait to direct selection is a product of its heritability and the selection differential, i.e., R = h2S (Falconer, 1981). To argue that a metric trait is capable of evolutionary adaptation to a changing environment, one must demonstrate the existence of both additive genetic variation and appropriate selection pressures. Biochemical studies suggest that a large reservoir of genetic variation exists in most populations (Nevo, 1978; Clarke, Queen's 1979; Powell and Taylor, 1979). Attempts 877 This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms 878 PETER T. BOAG have been made to correlate electropho- responses in other characters (Falconer, retic or karyotypic variation with mor- 1981). In field studies, natural selection by phological variation (Soule et al., 1973; Patton et al., 1975; Rising and Shields, definition acts on the entire phenotype, and this complicates analyses of heritability and 1980), but the relationship remains ten- the response to selection (Lande, 1979). As tative (Kolata, 1974; Carson, 1976). Like- an indirect solution to this problem, linear wise, examples of selection are widespread combinations of characters based on prin- (O'Donald, 1973; Berry, 1977 p. 130-131; cipal components (PCs) have been used Dhondt et al., 1979; Boag and Grant, 1981; with Darwin's finches (Boag and Grant, Grant and Price, 198 1; Johnston and 1981), and thus it is important to dem- Fleischer, 1981), but there are few verte- onstrate that such combinations of char- brate traits of generally accepted ecologi- acters are themselves heritable. Multi- cal importance, with both a known quan- variate approaches to evolutionary titative genetic basis and a known selection problems sometimes assume a close correspondence between phenotypic (rp) and regime. Why have quantitative genetics not been genetic (rG) correlations; this is not always more widely used with natural popula- justified (Atchley and Rutledge, 1980), although rp and rG matrices may be similar tions of vertebrates? Certainly long term studies of individually marked animals have only recently become available, but in certain suites of characters (Leamy, also, quantitative genetics evolved as an et al., 1981). Such similarity simplifies the applied science, with control over matings and the environment individuals develop evolutionary interpretation of phenotypic in. The lack of experimental control com- is useful to partition phenotypic correla- plicates application of the methodology to tions into their genetic and environmental field studies, and the interpretation of field (rE) components (Falconer, 1981). 1977; Lande, 1979; Arnold, 1981; Atchley correlations, and hence when possible, it data. For instance, quantitative genetics Boag and Grant (1978) found high her- assumes that environmental deviations are itabilities for morphological traits in the independent, random normal deviates. If medium ground finch (Geospiza fortis) they depend on the breeding value of an during 1976 on Isla Daphne Major, Ga- individual, or if deviations among groups lapagos. Here I describe the growth and repeatabilities of the seven traits involved, re-examine evidence for assortative mat- of individuals or traits are structured, this assumption is violated. There has been concern about such effects in human stud- ing, and extend the preliminary heritabil- ies, particularly with respect to genotype- ity results using additional data from 1978. environment correlations (Feldman and I also examine covariation among the Lewontin, 1975). However, recent avian characters by calculating rp, rG, and rE matrices. These analyses confirm that PC 1 field studies (Perrins and Jones, 1974; Boag and Grant, 1978; Greenwood et al., 1979; Ojanen et al., 1979; Smith and Zach, 1979; van Noordwijk et al., 1980; Garnett, 1981) have produced high heritabilities, with little evidence for genotype-environment correlations (Smith and Dhondt, 1981; Dhondt, 1982; Lessells, 1982). Another problem is that external mor- is a highly heritable index of overall body size, which responded to strong selection on Daphne in 1977 (Boag and Grant, 1981). In addition I provide preliminary estimates of heritabilities in a second species of Darwin's finch, the cactus finch (G. scandens). phological measurements are usually cor- METHODS related. In a laboratory setting one can Most of the adult G. fortis and G. scandens have been measured and individually banded on Isla Daphne Major, Galapagos. In 1976 (N = 481) and 1978 (N = 384) estimate the heritability of a single character, subject it to direct artificial selection, and observe the pattern of correlated This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms FINCH nestlings were banded, and some later mist-netted and measured. By identifying marked parents at the nest, one can relate their measurements to those of their subsequently caught young. Additional details of the field work are described in Boag (1981). Seven morphological variables were ex- amined: weight in grams (WGT), and flattened wing chord (WNG), tarsus length from the nuchal notch to the lowest undivided scute (TRS), bill length from the HERITABILITIES 879 involved. Geospiza fortis and scandens reach asymptotic weight, wing length, and tarsus length by six weeks of age, while bill length, depth and width asymptote in G. fortis by eight weeks after hatching. Geospiza scandens bills grow more slowly, with bill length taking up to 12 weeks to asymptote (Boag, 1981). Geospiza fortis under eight weeks old were excluded from the analysis, and as a compromise between sample size and age, the scandens bill depth at the anterior edge of the nares analyses excluded young under nine weeks old. This group had heritabilities similar to separately analyzed eight and 12 week in a plane perpendicular to the commis- old scandens. sure (BDT), bill width at the widest ex- To minimize nonrandom associations between phenotypes and environments, Smith and Dhondt (1980), Ricklefs and Peters (1980), and Dhondt (1982) swapped clutches among parents. This has not yet been feasible on Daphne, but a less pow- interior edge of the nares to the tip (BLG), panse of the lower mandible (BWD), and bill length at a depth of 4 mm (LA4), measured by marking the upper bill at a total bill depth of 4 mm, and then measuring the distance from that plane to the tip, all measured in mm (Abbott et al., 1977; Boag, 198 1). LA4 is usually larger in smaller individuals. After a repeatability analysis, measurements were averaged for birds measured on more than one occa- sion. Corrections were applied to measurements made by P. R. Grant in 1975, based on adult birds measured by both of us. A reference group of 10 museum specimens was also used to maintain measurement uniformity during the study. Repeatabilities were estimated using variance components from analyses of variance (ANOVAs) (Sokal and Rohlf, 1969; Falconer, 1981). I used repeatabilities to assess measurement error by catch- ing 10 G. fortis at Academy Bay, Isla Santa Cruz, and measuring each bird four times in four hours, in random order and under typical field conditions. I predicted upper limits for heritability estimates (Falconer, 1981) by calculating repeatabilities of adult Daphne finches remeasured during the study. Such long term repeatabilities contain variation due to changes in measurement technique with time, as well as growth and abrasion of morphology. Calculation of heritabilities assumes that a character is the same structure, in the same ontogenetic stage, in all individuals erful, a posteriori check for obvious genotype-environment correlations is to test for consistent associations between the morphology of parents and their offspring under different developmental environments, such as different nesting habitats or clutch sizes. I tested each character in males, females, and offspring separately in four one-way ANOVAs: variation among three habitats (outer slope, inner slope, and plateau; see Boag [1981] for study areas), among dates of first egg laid (early, middle or late), among clutch sizes, and among numbers of young fledged. The 84 tests were carried out only on G. fortis because scandens numbers were too small when split into the required categories. Young Darwin's finches cannot be sexed without performing laparotomies, and thus all young were pooled, increasing the error of h2 estimates because of the variation in broods due to sexual dimorphism (the traits are 4-8% larger in males compared to females). Smith and Zach (1979) standardized for sex and year in their analysis of song sparrow (Melospiza melodia) morphology and Greenwood et al. (1979) standardized for annual variation in the heritability of great tit (Parus major) dispersal. One can also add or multiply by correction This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms 880 PETER T. BOAG factors (Perrins and Jones, 1974). Stan- group correlation matrix, which is empha- dardization may alter regression slopes, so sized in the Results. Despite strong selec- where possible a better procedure is to pool tion in 1977 (Boag and Grant, 198 1), fortis sums of squares and crossproducts from covariance matrices for the two years were the groups of data and calculate pooled homogeneous (Box's M-test, N = 275, regressions. I pooled the 1976 and 1978 G. P > .05). fortis data, but also discuss the results for rG and rE matrices were calculated using each year. These provide a replicated test standard procedures (Falconer, 198 1), with for heritabilities based on different parents the emphasis on the midparent-offspring and offspring, following a major pertur- data. In regression-based genetic correla- bation in 1977 (Boag and Grant, 1981). tions, there are two cross-covariances pos- Four methods were used to calculate sible for each pair of characters, and I used heritabilities: separate regressions of off- their arithmetic mean (Van Vleck and spring on male and female only and mid- Henderson, 1961): parent and full sib intraclass correlations (Becker, 1975; Falconer, 1981). The slope of the regression of mean offspring mea- rG = (COVxyl + COVYx1)/ 2 V(COVXXCOVY,). surements on the mean of their parents' measurements (the midparent value) di- Here COVx,,, is the covariance between rectly estimates heritability. The regressions of offspring means on their mother's ter y' in the offspring. Family size weighting was not used. I followed Cheverud or father's measurements alone estimate (1982) in calculating standard errors for rG single parent heritabilities when multi- and rE, although his Eq. 4 is somewhat misleading, and I thus referred also to plied by two. All regressions used a family character x in the midparent and charac- size weighting technique (Falconer, 1963). ANOVA was used to estimate variance Reeve (1955). components for intraclass correlations (So- from rp matrices using SPSS (Nie et al., Principal components were extracted kal and Rohlf, 1969); full sib heritabilities 1975). PCs summarize linear trends in a are twice the intraclass correlation, and in addition to additive genetic variance, contain variance due to nonadditive genetic and common environment effects. Only four pairs of G. scandens young and their parents were measured in 1976. In 1977 no fortis bred and no scandens correlated data set, and component scores typic qualities are most strongly inherited young survived to be measured (Boag, multivariate structure with that seen in the 1981). The 1978 scandens were initially phenotypic correlations. analyzed in the same fashion as 1978 fortis; because scandens were not subject to the strong directional selection experi- calculated for parents and offspring can be used to suggest which overall pheno(Cheverud, 1981). SPSS was also used to extract PCs fromfortis rG and rE matrices (Atchley et al., 1981), to compare their Log10 scales were used in the correlation analyses because allometry produces curvilinear relationships between many of enced byfortis in 1977 (Boag and Grant, these variables (see also Ricklefs and 1981), I added the four 1976 pairs to the Travis, 1979). Other analyses showed that data, and found that heritabilities re- log10 transformation had little effect on mained the same. heritabilities, phenotypic or genetic cor- Phenotypic correlations of the seven log10 relations, or PC analyses (Kempthorne, transformed variables were generated using all adults and offspring involved in the formed covariance matrices were also sim- 1960 p. 12-23). PC analyses of logl0-trans- univariate heritability calculations, with ilar to correlation based analyses. Given separate matrices for the two years of G. fortis data. The 1976 and 1978fortis data that many of the quantities involved are dimensionless and that no character had were also used to produce a pooled within a coefficient of variation (CV) over 11% This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms FINCH HERITABILITIES 881 (Table 4), the robustness of these calcu- TABLE 1. Repeatabilities of Geospiza measure- lations to differences of scale is not sur- ments. The Santa Cruz birds are 10 G. fortis each prising (Lande, 1979). RESULTS measured four times on Isla Santa Cruz. The other data are forfinches captured between 1975 and 1978 on Isla Daphne Major and measured two to four times Repeatability of Measurements each. The All group contains males, females and birds Mechanical measurement error is negligible in the seven characters, as repeat- P. R. Grant in 1975. N is the number of different abilities within a single day are high (Ta- of measurements made. of uncertain sex, with some measurements made by individuals measured, followed by the total number ble 1). Over longer time intervals, weight changes and seasonal feather and beak growth or abrasion do alter repeatabili- ties. Table 1 shows that weight, wing, and LA4 have lower long term repeatabilities than tarsus and bill length, depth and width. G. scandens repeatabilities tend to be lower than in G. fortis, and females of both species have lower weight repeatabilities than males. Phenotypes and Environments Seven of the 84 ANOVAs across environmental categories were significant in 1976, although there were no strong de- viations of parent and offspring means in the same direction in a given environment (Boag and Grant, 1978). In 1978, males living on the inner slopes were largest, significantly so for tarsus length (Duncan Multiple Range tests, P < .05). Females showed a similar, but not significant pattern. Inner slope young were significantly larger than other young except for bill length and width. Inner slope young were also older, but when age was controlled for as a covariate, the area effect remained significant. There were no significant differences in 1978 among early, middle, or late season nesters. 1978 males with small clutches had a significantly longer LA4, while females with four egg clutches were significantly larger than those with smaller or larger clutches in weight, bill depth, and LA4. Young from four egg clutches were also Santa Character Cruz Males Females All G. fortis WGT 1.00 .75 .51 WNG .98 .60 .79 TRS .96 .92 .72 BLG .99 .95 .98 BDT 1.00 .93 .96 BWD 1.00 .99 .98 LA4 .88 .59 .48 .73 .75 .91 .96 .94 .98 .59 .97 .84 .82 .77 N 10, 40 42, 90 9, 19 118, 246 G. scandens WGT .84 WNG .58 .25 .62 .55 .62 TRS .97 .96 BLG .94 .76 BDT .94 .89 BWD .98 .95 .94 .80 .87 .95 .61 LA4 .67 .44 .85 .77 .76 N 14, 28 10, 20 36, 74 larger in weight, bill length, bill depth, and bill width. Young showed no significant variation across numbers of young fledged, Thus G. fortis on the inner slopes in 1978 were larger than in other areas. Large adults survived best in 1977 (Boag and Grant, 1981), and birds breeding on the inner slopes in 1976 survived especially well (Boag, 1981). If inner slope territories were also better for rearing young in 1978, inner slope fortis should have shown increased breeding success. But inner slope larger than young from larger or smaller clutches, with significant differences for bill nests were no more successful than nests in other areas (Boag, 1981). Alternatively, depth, width, and LA4. There was no sigif there was a strong territory quality efnificant variation in male size across numfect, inner slope young might have reached bers of young fledged, while females fledga larger size relative to their parents than ing four or more young were significantly young raised elsewhere. Again, there was This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms 882 PETER T. BOAG TABLE 2. Product moment correlations between G. fortis and G. scandens mates on Daphne in 1976, 1977, and 1978. All pairs shown fledged one or more young. No G. fortis bred in 1977. Significance levels in this and subsequent tables: *, P < .05; **, P < .01; * P < .001. G Character WGT -.06 BDT 40 1976 -.10 .19 .13 .28 23 .02 .09 no obvious tendency for this to be the case. .24 .49* -.32 .16 11 .41 -.16 .30 -.17 16 1978 -.39 -.15 -.42 -.04 -.32 1977 -.47 -.17 -.15 .41** -.10 scandens .24 -.15 .35* LA4 1978 -.26 .30 BWD G -.15 .38* BLG N 1976 .14 WNG TRS fortis -.24 -.11 20 new averages for birds remeasured since Thus large parents did produce large young 1976. The revised data support the origi- in the inner slope area, but this does not nal conclusion of significant assortative seem to have been augmented by a cor- mating for some characters in 1976fortis, relation between large parents and supe- but a comparison of correlations in suc- rior breeding resources. cessful and unsuccessful 1976 pairs (Sokal and Rohlf, 1969) showed no significant Correlations Between Mates difference between the two groups. I also Boag and Grant (1978) showed that the used canonical correlation to simulta- morphology of 1976 G. fortis mates was neously examine all the characters in correlated, especially in successful pairs. groups of male and female mates; again Table 2 summarizes correlations between only the results for the successful 1976for- mates fledging at least one young. The 1976 tis sample approached significance (R2 = fortis values have been recalculated, using .62, X249= 58.01, P = .18; P values for TABLE 3. Pooled 1976 and 1978 G. fortis heritabilities, followed by averages for separate 1976 and 1978 estimates. Presented as h2 + SE, based on family size weighted regressions of offspring means on their midparents, male parents, andfemale parents, and intraclass correlations amongfull sibs. N is the number of family groups followed by the total number of offspring involved. Significance levels are those of the corresponding regression or analysis of variance. Character Midparent Male Female Sib Pooled WGT .91 + .09*** .95 ? .22*** .81 ?.17*** .78 ? .19*** WNG .84 ? .14*** .82 ? .24*** .61 ?.17*** .42 ? .20* TRS .71 .10*** .61 ? .19** .61 ? .13*** .32 ? .20* BLG .65 + .15*** .75 ? .27** .92 ? .17*** 1.11 ? .17*** BDT .79 ? .09*** 1.03 ? .19*** .87 + .15*** 1.12 ? .17*** BWD .90 + .10*** 1.02 ? .22*** .96 ? .17*** .99 ? .18*** LA4 .35 ? .10** .55 ? .74 N 39, 82 46, .18** .37 .82 85 57, ? .14** .74 111 .64 ? .20*** .77 46, 129 1976 .75 N 22, 26 N 17, 56 34, .82 44 .77 32, 37 .87 18, 38 1978 .72 12, .77 41 .71 25, 74 .71 28, 91 This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms FINCH other samples were ?. 90). This suggests HERITABILITIES I11.0- * 0 that the assortative mating in even this group is relatively weak overall. Table 2 shows that this correlation is not present at all in another year (1978) or species 1978 E y 0.74x+2.50 r= there were no significant differences between 1976 and 1978 except for tarsus in the male-offspring regression (t-test of 1976 versus 1978 tarsus regression slopes, t = 2.20, df. = 80, P < .05). The absence of 0 9 m0 0 - 1976 Geospiza fortis Heritabilities lar, and when compared prior to pooling, 0 63% 0 heritabilities in 1976 and 1978 were simi- 0 -C 2~~~~~~~ 10.0- (scandens). Table 3 shows that average G. fortis 883 O / O 04 y= 0.82x+ 1.27 a 9.0 .0 58 0 cn 0' / oo 0~~~~ 8.0- ' '0 8.0 0 9.0 10.0 11.0 Midparent Bill Depth (mm) inflated female-offspring heritabilities FIG. 1. Regressions of G. fortis mean offspring suggests that maternal effects are mini- bill depth on midparent bill depth in 1976 (open cir- mal. More surprisingly, average full sib estimates are also only slightly larger than cles) and 1978 (solid circles). See also Table 3. The slopes of the lines are not significantly different, but analysis of covariance showed that the 1978 line has other estimates, suggesting that common a significantly higher Y-intercept. Crosses indicate environment and nonadditive genetic ef- bivariate means. fects are also not very important in this population. Taken individually or pooled, the two years' data suggest that about 76% of the phenotypic variation in external morphology of G. fortis is heritable. relatively larger young were produced in 1978. Table 4 compares means for the two Despite the similarity of the two years' years, showing that the main effect is a results, the small differences may be of depression of offspring size in 1976. The biological interest. For instance, 1976 single parent regression heritabilities were small size of 1976 young appeared per- larger than in 1978, as expected given the and remeasured the next year showed no weak assortative mating observed in 1976 (Falconer, 1981). Full sib heritabilities tively smaller young in 1976 presumably were also larger in 1976, again expected if interfamily variation was increased by density dependent environmental effects, as discussed below. Environmental Effects Analysis of covariance showed that while midparent-offspring regression slopes were similar in the two years, yintercepts were greater in 1978, signifi- cantly so for tarsus (P < .05), bill depth (P < .01), and LA4 (P < .01). Figure 1 illustrates this for bill depth; the 1978 data are shifted to the right because 1978 parents were larger than 1976 parents due to manent, as 10 young measured in 1976 evidence of "catch-up growth." The relaresulted from poor conditions for growth that year, due to high breeding densities, later reduced by the 1977 drought on Daphne (Boag and Grant, 1981). Thus in January 1976 thefortis population size was 1144, with territories (x7 ? SE) of 203.6 + 18.9 'm2 (N = 19), while in January 1978 the population size was 181, with territo- ries of 477.8 ? 55.0 m2 (N = 9). It is also worth noting that the homogeneity of these regression slopes in the face of the different environments of 1976 and 1978 argues against the presence of genotype-environment interactions (Falconer, 198 1). natural selection (Boag and Grant, 1981), Uncertain Paternity and the points are shifted upwards both because of the larger parents and because 1978 data using males seen to rear off- Regressions originally calculated for the This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms 884 PETER T. BOAG TABLE 4. Means (x) and coefficients of variation (CV = 100 x SD/x) for G. fortis parents and offspring used in 1976 and 1978 midparent-offspring h2 analyses (Table 3). Adult statistics are based on equal numbers of males andfemales making up midparents, and thus the CVs are based on the total parental variance, not the variance of midparent values. PC1 scores are discussed in the text, and are based on the rp PCA in Table 8. The PC scores are standardized, and thus variances are given here instead of CVs, to indicate how the various subsamples depart from the overall variance of 1.0. 1976 Adults Character WGT 15.86 WNG TRS x 8.05 68.09 1978 Offspring CV 15.05 3.48 x Adults CV 10.15 66.15 3.79 x 16.75 69.09 Offspring CV x CV 9.68 16.48 9.21 3.33 67.95 2.85 3.57 18.75 3.42 18.58 3.53 19.08 3.80 19.09 BLG 10.74 6.64 10.35 8.13 11.04 5.55 10.83 6.31 BDT 9.56 9.29 8.96 9.12 9.90 8.62 9.88 6.79 BWD LA4 8.74 3.56 PC1 .03 N 6.66 9.25 .97 8.45 3.68 -.69 44 7.72 8.49 1.31 26 spring gave low heritabilities (average h2 for the seven characters in the 22 original 1978 male-offspring families was .35). 8.95 3.46 .57 6.94 8.89 6.13 9.96 3.41 6.66 1.14 34 .39 .78 58 same parents were pooled. To justify the pooling, eight G. fortis females were selected which had produced multiple broods High heritabilities were obtained in 1978 with the same male, and for which I had using other methods, suggesting either that recaptured and measured at least two off- strong maternal or nest environment effects were present, or that 1978 paternities spring from two or more broods. ANOVA were uncertain. There were behavioral and breeding data to support the last alternative. 1976 G. fortis pairs each raised only one clutch, whereas 1978 fortis females produced three clutches (Boag, 1981). In 24% of the 1978 renestings, females left a nents, showing that the variance between territory for a new male, sometimes before her brood had fledged, leaving the first male to rear those young. In several cases the female left her first mate and laid in the nest of a new male in less than a week, suggesting that some eggs could have been fertilized by the original male. Thus I used the relatively small sample of faithful pairs and their broods to produce the 1978 maleoffspring heritabilities in Table 3, despite large standard errors. In other analyses uncertain paternity is less important and a conservative revision of parentage was used to add several families to the known was used to calculate variance compobroods was not significant for any char- acter (mean component for the seven char- acters was 6.7%), with the relative sizes of the among females components (44.4%) and within broods components (48.9%) determining the heritabilities of the traits. Young in successive broods did of course have highly significantly different ages; the absence of significant between brood morphological variance components confirms that excluding young under eight weeks old minimized differences in morphology due to age. Geospiza scandens Heritabilities Heritabilities were calculated in a similar fashion for the combined 1976 and 1978 G. scandens sample discussed in the Methods. Table 5 shows several differ- 1978 faithful pairs, based on nest chron- ences from the G. fortis data. Less than ologies and behavioral observations. half of the phenotypic variance in scan- Differences Between Broods Because 1978 pairs produced several clutches, fully grown young having the dens morphology is heritable on average, and the rankings of the average heritabilities for the seven characters do not match those of fortis (Spearman correlation, r, = This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms FINCH HERITABILITIES 885 TABLE 5. G. scandens heritabilities, combining 1978 data with four 1976 families. Presented as h2 ? SE. N is the number offamily groups followed by the total number of offspring. Character WGT .58 WNG .12 Midparent Male Female Sib + .39 .37 ? .32 .14 + .58 .35 ? .39 ? .26 .11 ? .42 .19 ? .42 .35 ? .39 TRS .92 ? .23*** 1.26 ? .36*** .94 + .39* .88 ? .35** BLG .32 BDT .14 BWD LA4 ? .34 .60 It N ?.24 + 1.29 .32 ? .02 .42 .20** 29 ? .44 .62 .43 16, ?.30*** ? .37 ?.41 -.18 .22*** .59 20, .04 -.01 35 -.23; here and subsequently, significance ? ? .39 .45 .56 ? .39 ?.70 .56 ? .39 ? -.17 .33 ?.34 .18 20, .30 34 .06 + .38 .44 15, 38 1980; Grant, 198 la) and warrants further study. levels of r, are omitted because of lack of independence). Geospiza scandens females also switched mates in 1978, but herita- Correlations Between Characters bilities and nest chronologies suggested that Phenotypic, genetic, and environmental few of the scandens families involved correlation matrices for G. fortis are shown young of uncertain paternity. Some of the in Table 6. The genetic correlations for reduction in the heritabilities of scandens fortis are high, so that the environmental reflects lower repeatabilities in this species; correlations are largely "residuals" (Lea- note particularly that scandens females my, 1977). Separate analyses of the 1976 display low repeatabilities and low herita- and 1978 fortis data sets did, however, bilities. The only consistently high scandens heritability was for tarsus length, show good overall matching of matrix elements for all three types of correlations which had an intermediate value infortis. (rank correlations between 1976 and 1978 Nestlings It would simplify field work and in- matrix elements: rp, r, = .92; rG, r, = .66; rE, r, = .52). Table 6 shows there is also good correspondence between the ele- crease sample sizes in studies such as this ments of the rp and rG matrices (r, = .8 if offspring could be measured before they but no such similarity between the rE and fledged. Unfortunately these methodolog- either the rp (r, = .13) or rG (r, = -.12) ical advantages are offset by theoretical matrices. difficulties in interpreting such data (Fal- Phenotypic correlations were generally coner, 1981). Nevertheless, some workers lower in G. scandens (Tables 6 and 7; av- have examined resemblance between fully erage absolute value of rp infortis was .56 grown parents and their growing offspring and in scandens .39). Geospiza scandens had lower heritabilities, with larger stan- (Brooke, 1977; Grant, 198 la). When I regressed the mean weight of 1978 fortis dard errors, and this is reflected in the un- young at banding (7 days old) on the weight reliable genetic and environmental corre- of their midparents, I found a slope of lations shown in Table 7. They are .61 + .28 (P < .05). Such a regression in- presented largely to illustrate the sampling corporates a type of cross-covariance, and errors possible when rG is calculated from with a proper estimate for "7 day weight" small samples using low and imprecise genetic variance could give a genetic co- heritability estimates. While in fortis ge- variance between "7 day" and "adult" netic correlations are high compared to weights (Falconer, 1981). Parent-off- environmental correlations (average ab- spring resemblance during ontogeny has not received wide attention (Vandenburg solute values, rG = .75; rE = .36), the reverse is true in scandens (rG = .45; rE = .86) (Falconer, 1981). and Falkner, 1965; Atchley and Rutledge, This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms 886 PETER T. BOAG TABLE 6. Phenotypic, genetic and environmental correlation matrices for G. fortis. rp based on 1976 and 1978 pooled within group correlations, N = 275. All correlations have been multiplied by 102 and decimal points omitted. Diagonal of rp matrix contains phenotypic standard deviations (log10) multipled by 102. rp values > 12 are significant. Upper parts of rG and rE matrices contain standard errors. Diagonal of rG matrix contains h2 values based on log1o scale. WGT WNG TRS BLG BDT BWD LA4 rp WGT 4.03 WNG +64 TRS +60 BLG BDT 1.45 +56 1.59 +69 +60 +52 2.79 +76 +64 +54 +81 BWD +75 +60 LA4 -34 -27 +52 -20 3.64 +78 -12 +88 2.94 -46 -41 3.57 rG WGT .85 .10 .13 .10 .07 .07 .26 .11 .28 .17 .32 + 95 + 95 +71 .67 .07 .08 .34 +87 +87 +75 +90 .82 .04 .23 WNG +88 .89 .15 .12 .10 TRS +89 +68 .65 .18 .15 BLG BDT BWD +94 +78 +61 +89 +93 .90 LA4 -55 -30 -43 -48 -79 -70 .24 .39 WGT .56 rE 3.11 WNG -98 TRS -26 BLG BDT BWD LA4 1.06 1.02 1.03 .80 1.83 2.51 +21 .49 .76 -13 -68 .46 .64 .44 +19 -75 -02 +57 .73 .54 -97 +26 +47 +60 .75 -59 -07 -37 +15 2.69 3.68 .66 .79 .38 +02 +28 -04 +04 I extracted principal components from ilar results from separate 1976 and 1978 the G. fortis rG and rE matrices for com- analyses, suggests that environmental cor- parison with the phenotypic PCA discussed below. Both matrices were ill-con- are large and positive, like the phenotypic relations among BLG, BDT, and BWD ditioned for PCA, a not uncommon result and genetic correlations among these (Leamy, 1977; Arnold, 1981). BMDP characters, while there are strong negative (Dixon and Brown, 1979) did not accept correlations between WGT or WNG and either matrix, but SPSS (Nie et al., 1975) did, as it calculates simple principal components without inverting the correlation matrix. These admittedly strained analy- most other characters. This hints at dif- ses are shown in Table 8; the genetic correlations display a multivariate structure similar to the phenotypic correlations, while the environmental matrix has a different structure. This was confirmed by cluster analyses of the correlation matrices (BMDP1M, with average linkage, Dixon ferent pathways of environmental covariation in these two groups of characters. WGT and WNG are likely to be influenced by the condition of a bird (Table 1), reflecting recent, largely external influences. Environmental covariation among beak characters may more closely reflect internal effects, active primarily during periods of growth. and Brown, 1979), an alternative to using Heritabilities of Principal Components PCA on ill-conditioned matrices (Leamy, Table 8 shows unrotated PCs extracted from rp matrices in Tables 6 and 7. In all cases PCs 5 through 7 were not significant 1977). The G. fortis rE matrix, backed by sim- This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms FINCH HERITABILITIES 887 TABLE 7. rp, rG and rE matrices for G. scandens, as in Table 6, with N = 92. rp values > 20 are significant. WGT WNG TRS BLG BDT BWD LA4 rp WGT 3.55 WNG +66 TRS +35 BLG BDT +36 +56 +44 BWD LA4 1.40 +43 +04 1.52 +58 +11 2.04 +56 +47 +36 +51 +38 +39 +10 -40 +38 2.19 +75 -32 1.96 -14 3.44 rG WGT .58 WNG +57 .44 .33 .7 9 .99 .59 .41 .25 .50 1.53 3.74 1.27 .64 TRS +99 +61 .94 .53 1.26 .52 .25 BLG -13 -48 -10 .32 3.23 1.02 .44 BDT +30 -82 +77 -90 .09 1.25 1.89 BWD +13 -33 -01 -27 +29 .35 .55 LA4 +10 -20 -66 +60 -102 +17 .83 rE WGT .41 9.00 .61 .43 .52 1.00 WNG +79 1.21 .41 .34 .41 .94 TRS -235 +31 2.81 2.82 5.82 8.78 BLG +114 +100 +83 .38 .44 .72 BDT +60 +82 +106 +64 .30 .66 BWD +70 +87 +195 +72 +91 1.08 LA4 -11 +53 +174 +22 -13 -67 (Ricklefs and Travis, 1980). Heritabilities were calculated for both species using the same methods as before, applied to PC scores. Table 8 shows that PC1, representing overall body size (Boag and Grant, 1981), was most strongly inherited in G. fortis, while PC2, representing bill point- edness, and PC3, representing a contrast between bill and non-bill characters, were most strongly inherited in scandens. Response to Selection The most important implication of high heritabilities is the prediction that selec- tion of G. fortis parents should result in a corresponding shift in the phenotypes of their offspring. Geospiza fortis underwent strong directional natural selection on Daphne between 1976 and 1978 (Boag and Grant, 1981), and thus the data in Table 4 can be used to determine whether the 1978 offspring display the predicted re- sponse. To test this, I standardized the 4). The 1978 parents displayed an average increase in size of .44 SD due to selection, while their offspring increased by .31 SD relative to the 1976 midparents. The already standardized PC1 means show a similar result; 1978 midparents were .54 SD larger, while their offspring showed a response of .36 SD. In both cases the increase in the size of 1978 young relative to the increase seen in their selected parents is close to that predicted by heritabilities of about 76% (Falconer, 1981). With larger sample sizes, these data could in principle be used to estimate realized heritabilities (h2r) (Falconer, 1981), although a multivariate generalization of the formula given at the beginning of this paper would be required to cope with the complexities of natural selection acting on correlated characters (i.e., A = GP-1S as opposed to R = h2S; see Lande, 1979). DISCUSSION 1978 midparent and offspring means with These data support the conclusions of respect to the means and standard devia- Boag and Grant (1978), namely that external morphology is strongly heritable in tions (SD) of the 1976 midparents (Table This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms 888 PETER T. BOAG TABLE 8. Unrotated PC loadings for G. fortis rp, natural populations of Darwin's finches. rG and rE matrices, and for G. scandens rp matrix. Geospiza fortis has particularly high her- EIG are eigenvalues, while % V is the percent of the itabilities for weight and bill dimensions, total phenotypic variance explained by each PC. Heritabilities of each PC, together with their standard errors and significance levels, are given below the fortis and scandens rp PCAs. The h2 estimates are based on PC scores used in the midparent-offspring regressions of Tables 3 and 5. and for overall body size as represented by PC1. Results for G. scandens are less complete, but suggest that heritabilities are lower in this species, and that in contrast to fortis, the highest heritabilities are for tarsus length and shape PCs 2 and 3. Character PC1 PC2 PC3 PC4 WGT .87 WNG TRS .71 BLG .04 .78 .85 BDT corresponds in part to high repeatabilities. .55 Garnett (1976) reported repeatabilities -.59 -.32 .29 -.04 higher than my long term values, but his .22 -.05 .26 -.09 for much growth, molt, abrasion or change in measurement technique to take place. -.09 .91 -.05 -.25 .31 .93 BWD -.02 .12 .21 The heritabilities of G. fortis are high compared to other birds (Table 9) and this G. fortis, rp -.06 data were collected over too short a period LA4 -.45 .88 .11 .01 EIG 4.48 .93 .62 .42 Smith and Zach's (1979) study is more % 63.9 13.3 8.8 6.0 comparable, and their repeatabilities were V h2 .75*** .36** .47** .32** SE .12 .10 .15 .10 G. fortis, rG WGT .98 .13 -.10 WNG .89 .37 .21 -.18 -.56 -.02 TRS .81 BLG .19 .95 BDT .98 BWD .19 .17 -.17 .94 .19 .01 .06 -.15 lection (Falconer, 1981). If female choice -.66 .75 .07 .06 .84 .44 .16 % 80.1 12.0 6.3 2.2 G. fortis, rE WGT .28 -1.00 -.02 .36 WNG -1.06 .38 -.04 -.05 .01 .59 .01 .75 .28 .19 BDT .82 BWD LA4 EIG .00 -.33 for other reasons. The mean dimensions of young produced in different years may differ (Table 4, Fig. 1), because of changes -.20 -.17 .94 -.05 1.08 .82 % V 44.8 27.2 15.4 11.6 G. scandens, rp WNG .83 TRS .59 BLG .65 BDT .82 BWD LA4 EIG % V .23 .37 -.36 .24 .15 -.09 -.51 .59 -.31 .44 .44 -.02 -.34 .11 -.04 .79 -.15 -.48 .10 -.10 .90 -.06 .30 3.35 1.65 .69 47.9 23.6 9.9 .45 6.4 h2 .20 1.07*** .64** -.09 SE .39 .20 .22 1981). This was not observed, suggesting .01 .57 .77 of an excess of males (Boag and Grant, -.03 .05 WGT mation of many new pairs in the presence .81 .80 1.90 mate correlations, one would predict higher correlations in 1978 due to the for- that the 1976 pairs were weakly correlated .27 3.14 Correlations between mates affect heritability estimates and the response to se- had been responsible for the 1976 G. fortis 5.61 BLG plain the lower scandens heritabilities. .22 EIG TRS are lower in scandens, particularly scandens females, but do not completely ex- -.19 .20 LA4 V lower. Repeatabilities for some characters .30 in nestling food availability, and because selection alters parental phenotypes. Perhaps after several years of regular recruitment to the breeding population, pairs composed of a male and female reared in the same year may gradually accumulate, inducing a correlation into the overall breeding population. Van Noordwijk et al. (unpubl.) reported no correlations between great tit mates, while Smith and Zach (1979) noted only one significant, negative, correlation between song sparrow mates. The male-offspring heritabilities in 1978 fortis were complicated by uncertain pa- This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms FINCH HERITABILITIES 889 TABLE 9. Heritabilities of avian external morphology. Data for G. fortis are pooled midparent values from Tables 3 and 8, while G. scandens data are midparent values from Tables 5 and 8. Species WGT WNG TRS BLG BDT BWD PC 1 Reference Puffinus puffinus .661 .92 .761 .92 Brooke, 1977 Ficedula hypoleuca - - .5 7 - .5 7 Alatalo et al., unpubl. F. albicollis - - .59 - .59 Alatalo et al., unpubl. Parus major - - .52 - .52 Alatalo et al., unpubl. P. major .59 - .5 2 - .56 Van Noordwijk et al., 1980, unpubl. P. major - .76 .76 Garnett, 1981 P. major - - .62 - .62 Dhondt, 1982 Melospiza melodia .04 .14 .32 .33 .51 .50 - .31 Smith and Zach, 1979 Geospizafortis .91 .84 .71 .65 .79 .90 .75 .79 Present study G. scandens .58 .12 .92 .32 .14 .60 .20 .41 Present study G. conirostris 951 -.681 .87 .481 .661 741 - .87 Grant, 1981a Branta canadensis .20 - .15 .53 - .29 Lessells, 1982 1 Character unlikely to have been fully grown in offspring measured, a ternity. Mis-directed parental care may ance implied by high heritabilities is main- lead to selection for behavior to ensure pa- tained. Grant and Price (1981) discuss this ternity (Fujioka and Yamagishi, 198 1), but problem in detail, and here I review only evaluation of paternity in natural popu- those arguments specific to Daphne. lations is difficult without elaborate ex- The level of variation in a population is periments (Bray et al., 1975) or genetic markers (Mineau and Cooke, 1979). I tried, largely a function of mutation, genetic drift, migration, and selection (Falconer, 1981). Lande (1975) suggests that muta- with little success, to identify the paternity of questionable broods by examining the residuals of heritability regressions. How- maintain substantial amounts of polygenic ever, in studies with larger sample sizes, variation, but it is difficult to see why they outlier tests (Sokal and Rohlf, 1969) on male-offspring regressions may help as- should be more effective in Daphnefortis sess paternity. My results suggest that body size and tion and recombination may be able to compared to closely related species, i.e., scandens. Minimum annual population sizes for Daphne finches are available for bill dimensions ought to be especially re- seven years (Boag, 1981), and suggest that sponsive to natural selection in G. fortis. This is important, because considerable the effective population size (Ne) of fortis is no more than 460, while Ne for scan- effort has been devoted to studying the dens may be about 230. Skewed sex ratios evolutionary and functional relationships in years such as 1978 (Boag and Grant, between variation in bill or body size, and the feeding ecology, adaptive radiation and 120' and 80, respectively (Falconer, 1981). 1981) could reduce Ne to values as low as species recognition of Darwin's finches Thus genetic drift, if anything, should re- (Lack, 1947; Bowman, 1961; Grant et al., duce variation in both species, albeit more 1976; Abbott et al., 1977; Boag and Grant, so in scandens. 1981; Grant, 198lb; Ratcliffe, 1981). It has There is introgression between Daphne also been suggested that intraspecific vari- G. fortis and small ground finch (G. fuli- ation itself may be maintained by natural selection (Grant et al., 1976; Smith and Zach, 1979). As with biochemical varia- ginosa) immigrants at an annual rate of tion, the major problem is to explain how immigrants from Isla Santa Cruz (Boag, the large amount of additive genetic vari- 1981). Thus migration explains in part how 1. 1% to 3.5%; Daphnefortis probably also mate with morphologically distinct fortis This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms 890 PETER T. BOAG large Daphnefortis heritabilities might be as Santa Cruz, spatially varying selection maintained (Grant and Price, 1981). How- pressures may actually promote variation ever the effect of selection regime on ge- in fortis (Grant et al., 1976). Unfortu- netic variability is a matter for more de- nately, present theoretical models do not bate. There is little doubt that stabilizing adequately describe situations such as selection reduces variation, but the effect these. On Daphne, thefortis population is of different types of directional selection probably not often at genetic equilibrium, on variation remains uncertain (Grant and population dynamics are unpredictable, and selection pressures may change in direction and intensity from season to season and from year to year, with strong density and frequency dependent effects likely to be the rule rather than the exception (Clarke, 197 9). Another aspect of the uncertain rela- Price, 1981). Current data suggest that Daphne G. fortis are subject to temporally fluctuating directional selection on morphology, while scandens experience stabilizing selection (Grant et al., 1976; Boag, 1981; Boag and Grant, 1981; Grant and Price, 1981). The selection regimes of the species reflect their ecologies; Daphnefortis are opportunistic generalists, experiencing large population fluctuations and phenotypic shifts as rainfall alters seed abundances and sizes (Boag and Grant, 1981). When food is short, intraspecific competition among fortis appears intense, and diets expand, while intraspecific correlations between morphology and diet (Boag and Grant, tionship between selection and heritability is the paradox that in quantitative genetics, small heritabilities are associated with traits closely related to fitness (Falconer, 1981). The assumption is that fitness or "major fitness components" are continually subjected to unidirectional selection, eliminating additive genetic vari- ance. In nature it is usually difficult to find or measure a "major fitness component." 1981; Grant, 198 lb) imply that selection is Even characters such as clutch size in birds likely to be frequency dependent. Geospiza scandens experience less violent population changes because their specialized occupy intermediate optima and display significant amounts of heritable variation (van Noordwijk et al., 1980). This is because selection can reduce VA in total fitness, while at the same time pressures on individual components (e.g., clutch size) feeding and breeding ecology (Lack, 1947; Boag, 1981) buffers them against changes in food availability. In periods of food shortage scandens diets contract and interspecific competition with fortis predominates (Boag, 1981); there is no evidence for intraspecific correlations between morphology and diet in scandens. Theory predicts that fluctuating directional selection is unlikely to maintain genetic variation unless selection pressures vary spatially and not just temporally (Grant and Price, 1981). Despite this prediction, Mackay (1980) found that temporally varying directional selection maintained higher heritabilities in laboratory populations of Drosophila than did spatially varying selection. Thus the selection regimes currently seen on Daphne would reduce additive genetic variance in scandens, and might at least slow the loss of variation infortis. On larger islands, such cancel as negative genetic correlations are established between characters coupled to fitness by different mechanisms (Falconer, 1981). As quantitative genetic theory is integrated into evolutionary ecology, genetic, especially negative genetic, correlations will be the focus of much interest. Given the prevalence of pleiotropy, the evolution of almost any character is constrained or assisted by selection on a host of other characters; natural selection is always for "fitness," so that all observed phenotypic responses are in some sense correlated responses. This will apply to morphology (Lande, 1979; Atchley et al., 1981), reproductive characteristics (Lande, 1982), food preferences (Arnold, 1981), or even the complex behavioral "trade-offs" of forag- This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms FINCH HERITABILITIES 891 ing theory (McCleery, 1978). Before much imize complications found in natural progress can be made in such areas, it will populations, such as genotype-environ- be necessary to adapt quantitative genetic ment correlations, assortative mating, and theory to situations governed chiefly by natural selection. Such an approach is ca- natural selection operating on correlated pable of precise results, but lest carefully characters (Lande, 1979). One serious controlled experiments become ends in methodological problem is the notorious themselves, a constant effort must be made unreliability of genetic correlations; in some to test their generality and realism in field cases, reasonable estimates of rG may re- situations (Levins, 1968). I suggest that quire 1,000+ parent-offspring pairs (Van neither uncontrolled studies of natural Vleck and Henderson, 1961). A presum- populations nor elegant studies of labo- ably worse problem holds true for rE, al- ratory organisms alone will in the long run though less attention has been paid to its answer the question of how polygenically sampling variance (Cheverud, 1982). determined phenotypes evolve in response Another point other workers should to changing environments. Given this consider is that despite the strong genetic premise and the pressure for more rigor- influence, the environment did signifi- cantly affect finch morphology on Daphne, ous evolutionary explanations of phenotypic patterns (Gould and Lewontin, 1979), primarily by reducing the mean size of off- it remains surprising that there have been spring in a poor year. Nutritional condi- no parallel collections of quantitative ge- tions can alter the growth rates of nest- netic data from natural and controlled lings (Bryant, 1978), but no attention has laboratory populations of any single avian been paid to whether such effects produce species, using characters of ecological im- smaller adult birds; i.e., there is little in- portance and which historically have been formation on canalization or "norms of re- assigned high adaptive values. action" for avian morphological characters. Such information is important SUMMARY because selection acts on phenotypes. Even The repeatabilities of seven external if heritabilities within a single population morphological characters were high for and year are high, nutritional effects on Geospiza fortis and G. scandens on Isla the mean or variance of body size among Daphne Major, Galapagos, between 1975 fledglings could cloud the evolutionary and 1978. Geospizafortis mates displayed significance of the high, sometimes selec- positive assortative mating with respect to tive, first year mortality seen in many birds morphology in 1976, but not 1978, while (Johnston and Fleischer, 1981; van scandens showed no consistent assortative Noordwijk et al., unpubl.). Also, most of mating in any year. Heritabilities infortis the populations in Table 9 are sedentary, were high in both years, averaging .76 based on three types of parent-offspring largely because sedentary or strongly phil- opatric species are chosen for long term regression and on full sib intraclass cor- population studies. In other species, local relations. Heritabilities in scandens were breeding populations may be composed of lower, averaging .46. 1978 heritabilities birds reared in a variety of other, some- were complicated by uncertain paternity in some fortis families. Genetic correla- times distant populations. If environmental effects produce locally unique devia- tions were also high among the seven tions in offspring phenotypes, such mixed characters in fortis, and had a multivari- breeding populations would display in- ate structure like that seen in the phenotypic correlations. Geospiza fortis had a creased environmental variances and decreased heritabilities. Individual philosophies differ (Kemp- thorne, 1977 p. 3), but quantitative geneticists have traditionally sought to min- high heritability for PC1, representing overall body size (h2 = .75), while scandens had high heritabilities for PC2, representing bill "pointedness" (1.07), and This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms 892 PETER T. BOAG PC3, representing a ratio between bill and non-bill characters (.64). As predicted by the heritabilities, fortis offspring showed a strong phenotypic response to natural selection of their parents in 1977. Geospiza fortis morphological variation may be influenced by introgression with G. fuliginosa and frequency dependent, tempo- rally fluctuating directional selection, while G. scandens morphology may reflect an evolutionary history predominated by stabilizing selection. of floristic diversity and interspecific competition. Ecol. Monogr. 47:151-184. ARNOLD, S. J. 1981. Behavioral variation in natural populations. I. Phenotypic, genetic, and environmental correlations between chemorecep- tive responses to prey in the garter snake, Thamnophis elegans. Evolution 35:489-509. ATCHLEY, W. R., ANDJ. J. RUTLEDGE. 1980. Genetic components of size and shape. I. Dynamics of components of phenotypic variability and covariability during ontogeny in the laboratory rat. Evolution 34:1161-1173. ATCHLEY, W. R., J. J. RUTLEDGE, AND D. E. CowLEY. 1981. Genetic components of size and shape. II. Multivariate covariance patterns in the rat and mouse skull. Evolution 35:1037-1055. BECKER, W. A. 1975. A Manual of Quantitative ACKNOWLEDGMENTS The data were collected while I was a graduate student in the Department of Bi- Genetics. Students Book Corp., Pullman. BERRY, R. J. 1977. Inheritance and Natural History. Collins, London. BOAG, P. T. 1981. Morphological variation in the ology at McGill University, with financial Darwin's finches (Geospizinae) of Daphne Major support from a National Science and Engineering Council of Canada (NSERC) Island, Galapagos. Ph.D. thesis, McGill Univ., scholarship, an NSERC operating grant to P. R. Grant, and a travel grant from the Frank M. Chapman Fund of the American Museum of Natural History. While the paper was written, I was sup- Montreal. BOAG, P. T., AND P. R. GRANT. 1978. Heritability of external morphology in Darwin's finches. Nature 274:793-794. 1981. Intense natural selection on a pop- ulation of Darwin's finches (Geospizinae) in the Galapagos. Science 2 14:82-85. ported by grants from the Trent Univer- BOWMAN, R. I. 1961. Morphological differentiation and adaptation in the Galapagos finches. sity research committee. I thank the Charles Darwin Research Station (CDRS) BRAY, 0. E., J. J. KENNELLY, AND J. L. GUARINO. and the Ecuadorean Government for their assistance in the Galapagos, particularly CDRS directors C. MacFarland and H. Hoeck, and national park intendente M. Cifuentes. P. R. Grant, B. R. Grant, L. M. Ratcliffe, E. Green, D. Nakashima, Univ. Calif. Publ. Zool. 58:1-326. 1975. Fertility of eggs produced on territories of vasectomized red-winged blackbirds. Wilson Bull. 87:187-195. BROOKE, M. DE L. 1977. The breeding biology of the Manx Shearwater. Ph.D. thesis, Univ. Oxford, Oxford. BRYANT, D. M. 1978. Environmental influences and B. Tompkins helped with fieldwork. on growth and survival of nestling house martins Delichon urbica. Ibis 120:271-283. P. Kuo, T. Buelow, and P. Northrup CARSON, H. L. 1976. Genetic differences between helped with computing at Trent. I thank R. Alatalo, L. Gustafsson and A. Lund- CHEVERUD, J. M. 1981. Variation in highly and berg, as well as A. J. van Noordwijk, for permission to cite unpublished work. Comments by P. R. Grant, K. Sittman, G. Bell, L. M. Ratcliffe, T. D. Price, R. Lande, J. Hegmann, M. G. Bulmer, A. J. van Noordwijk, and D. S. Falconer and his colleagues greatly improved the manuscript. LITERATURE CITED ABBOTT,I.,L. K. ABBOTT, ANDP. R. GRANT. 1977. Comparative ecology of Galapagos ground finches (Geospiza Gould): evaluation of the importance newly formed species. BioScience 26:700-701. lowly heritable morphological traits among social groups of rhesus macaques (Macaca mulatta) on Cayo Santiago. Evolution 35:75-83. 1982. Phenotypic, genetic, and environ- mental morphological integration in the cranium. Evolution 36:499-516. CLARKE, B. C. 1979. The evolution of genetic diversity. Proc. Roy. Soc. London B 205:453-474. DHONDT, A. A. 1982. Heritability of blue tit tarsus length from normal and cross-fostered broods. Evolution 36:418-419. DHONDT, A. A., R. EYCKERMAN, AND J. HUBLE. 1979. Will great tits become little tits? Biol. J. Linn. Soc. 11:289-294. DIXON, W. J., AND M. B. BROWN (eds.). 1979. BMDP-79. Biomedical Computer Programs P-Series. Univ. Calif. Press, Berkeley. This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms FINCH FALCONER, D. S. 1981. Introduction to Quantitative Genetics, 2nd ed. Longman, London. . 1963. Quantitative inheritance, p. 193-216. HERITABILITIES 893 quences of family size in the Canada Goose Branta canadensis. Ph.D. thesis, Univ. Oxford, Oxford. In W. J. Burdette (ed.), Methodology in Mam- LEVINS, R. 1968. Evolution in Changing Environ- malian Genetics. Holden-Day Inc., San Francis- ments. Princeton Univ. Press, Princeton. LONG, T. 1974. The role of heritability in adjust- co. FELDMAN, M. W., AND R. C. LEWONTIN. 1975. The heritability hang-up. Science 190:1163-1168. ing phenotypic variability. Amer. Natur. 108:140142. FuJIoKA, M., AND S. YAMAGISHI. 1981. Extra- MACKAY, T. F. C. 1980. Genetic variance, fitness, marital and pair copulations in the cattle egret. and homeostasis in varying environments: an ex- Auk 98:134-144. perimental check on the theory. Evolution 34: GARNETT, M. C. 1976. Some aspects of body size 1219-1222. in the great tit. Ph.D. thesis, Univ. Oxford, Ox- MCCLEERY, R. H. 1978. Optimal behavioural se- ford. quences and decision making, p. 377-410. In J. R. Krebs and N. B. Davies (eds.), Behavioural Ecology. Blackwell Scientific, Oxford. MINEAU, P., AND F. COOKE. 1979. Rape in the lesser snow goose. Behaviour 70:280-291. NEVO, E. 1978. Genetic variation in natural populations: patterns and theory. Theoret. Pop. Biol. . 1981. Body size, its heritability and influ- ence on juvenile survival among great tits Parus major. Ibis 123:31-41. GOULD, S. J., AND R. C. LEWONTIN. 1979. The spandrels of San Marco and the Panglossian par- adigm: a critique of the adaptationist programme. Proc. Roy. Soc. London B 205:581-598. GRANT, P. R. 1981a. Patterns of growth in Darwin's finches. Proc. Roy. Soc. London B 212: 403-432. 1981b. Speciation and the adaptive radiation of Darwin's finches. Amer. Sci. 69:653-663. GRANT, P. R., B. R. GRANT, J. N. M. SMITH, I. J. ABBOTT, AND L. K. ABBOTT. 1976. Darwin's finches: population variation and natural selection. Proc. Nat. Acad. Sci. USA 73:257-261. GRANT, P. R., ANDT. D. PRICE. 1981. Population variation in continuously varying traits as an eco- 13:121-177. NIE, N. H., C. H. HULL, J. G. JENKINS, K. STEINBRENNER, AND D. H. BENT. 1975. SPSS: Statistical Package for the Social Sciences. McGrawHill, N.Y. VAN NOORDWIJK, A. J., J. H. VAN BALEN, AND W. SCHARLOO. 1980. Heritability of ecologically important traits in the great tit, Parus major. Ardea 68:193-203. O'DONALD, P. 1973. A further analysis of Bum- pus' data: the intensity of natural selection. Evolution 27:398-404. logical genetics problem. Amer. Zool. 2 1:795-81 1. OJANEN, M., M. ORELL, AND R. A. VAISANEN. GREENWOOD, P. J., P. H. HARVEY, AND C. M. PER- 1979. Role of heredity in egg size variation in the great tit Parus major and the pied flycatcher Ficedula hypoleuca. Ornis Scand. 10:22-28. PATTON, J. L., S. V. YANG, AND P. MYERS. 1975. Genetic and morphologic divergence among in- RINS. 1979. The role of dispersal in the great tit (Parus major): the causes, consequences, and heritability of natal dispersal. J. Anim. Ecol. 48: 123-142. JOHNSTON, R. F., AND R. C. FLEISCHER. 1981. troduced rat populations (Rattus rattus) of the Overwinter mortality and sexual size dimorphism Galapagos archipelago, Ecuador. Syst. Zool. 24: in the house sparrow. Auk 98:503-511. 296-310. KEMPTHORNE, 0. (ed.). 1960. Biometrical Genetics. Pergamon Press, London. . 1977. Introduction, p. 3-18. In E. Pollak, 0. Kempthorne, and T. B. Bailey, Jr. (eds.), Proc. of the Internat. Conf. on Quantitative Genetics (1976). Iowa State Univ. Press, Ames. KOLATA, G. B. 1974. Population genetics: reevaluation of genetic variation. Science 184:452-454. LACK, D. 1947. Darwin's finches. Cambridge Univ. Press, Cambridge. LANDE, R. 1975. The maintenance of genetic variability by mutation in a polygenic character with linked loci. Genet. Res. 26:22 1-235. 1979. Quantitative genetic analysis of multivariate evolution, applied to brain: body size allometry. Evolution 33:402-416. 1982. A quantitative genetic theory of life history evolution. Ecology 63:607-615. PERRINS, C. M., AND P. J. JONES. 1974. The in- heritance of clutch-size in the great tit Parus major L. Condor 76:225-229. POWELL, J. R., AND C. E. TAYLOR. 1979. Genetic variation in ecologically diverse environments. Amer. Sci. 67:590-596. RATCLIFFE, L. M. 1981. Species recognition in Darwin's ground finches (Geospiza, Gould). Ph.D. thesis, McGill Univ., Montreal. REEVE, E. C. R. 1955. The variance of the genetic correlation coefficient. Biometrics 11:357-374. RICKLEFS, R. E., AND S. PETERS. 1981. Parental components of variance in growth rate and body size of nestling European starlings (Sturnus vul- garis) in eastern Pennsylvania. Auk 98:39-48. RICKLEFS, R. E., AND J. TRAVIS. 1980. A mor- phological approach to the study of avian community organization. Auk 97:321-338. LEAMY, L. 1977. Genetic and environmental cor- RISING, D., AND G. F. SHIELDS. 1980. Chromo- relations of morphometric traits in randombred somal and morphological correlates in two New World sparrows (Emberizidae). Evolution 34:654- house mice. Evolution 31:357-369. LESSELLS, C. M. 1982. Some causes and conse- 662. This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms 894 PETER T. BOAG SMITH, J. N. M., AND A. A. DHONDT. 1980. Experimental confirmation of heritable morphological variation in a natural population of song sparrows. Evolution 34:1155-1158. SMITH, J. N. M., AND R. ZACH. 1979. Heritability of some morphological characters in a song sparrow population. Evolution 33:460-467. SOKAL, R. R., AND F. J. ROHLF. 1969. Biometry. netic-phenetic variation correlation. Nature 242: 191-193. VANDENBURG, S. G., AND F. FALKNER. 1965. Hereditary factors in human growth. Hum. Biol. 37:357-365. VAN VLECK, L. D., AND C. R. HENDERSON. 1961. Empirical sampling estimates of genetic correlations. Biometrics 17:359-371. W. H. Freeman Co., San Francisco. SOULE, M., S. Y. YANG, M. G. W. WEILER, AND Corresponding Editor: S. J. Arnold G. C. GORMAN. 1973. Island lizards: the ge- This content downloaded from 132.174.254.72 on Mon, 06 Mar 2017 15:40:15 UTC All use subject to http://about.jstor.org/terms