MATE CHOICE IS NOT IMPORTANT FOR FEMALE REPRODUCTIVE SUCCESS IN THE COMMON (CARPODACUS ERYTHRINUS)

MATS BJ6RKLUND Departmentof Zoology,Uppsala University, Box 561, S-751 22 Uppsala,

ABSTRACT.--Istudied mate choice in relation to female reproductive successin the Common Rosefinch(Carpodacus erythrinus) over a four-year period. Approximately 50-99% of the vari- ance in female reproductive successwas due to nest predation during incubation. Because malesdid not defend nest sitesand femalesprobably face time constraints,I concludethat they paired unselectively.This is supportedby the lack of correlation between male phe- notypic characteristicsand pairing successand the random pattern in male pairing order. Received22 March 1989,accepted 12 July 1989.

A CENTRALissue in the understanding of avi- male mate choice. Models of female mate choice an mating systemsis female mate choice (Oring commonlyassume that femalesare free to choose 1982). Femalesare thought to choosein order among males (Bradbury and Gibson 1983), an to maximize fecundity (Halliday 1978, Searcy assumption that is doubtful if, for example, 1982, Kirkpatrick 1985, Andersson 1986). In males form dominance hierarchies (Beehler and species where males defend resources, that Foster 1988). Thus, it is also necessaryto ex- means that females primarily choosefeatures amine the freedom of femalesto chooseamong such as good feeding territories, safe nesting males. places,or males'prospective parental effort, or The Common Rosefinch(Carpodacus erythri- at least cues that indicate such material benefits nus)is a small (22 g), sexually dimorphic, car- (Nisbet 1973, 1977;Halliday 1983;Alatalo et al. dueline . The first-yearmales and females 1986; Yasukawa et al. 1987). If no immediate are indistinguishable. Cardueline in benefitsare obtainedfrom males(e.g. as in lek general maintain only a small territory around species;Vehrencamp and Bradbury 1984), the the nest, and males do not defend territories importance of male genetic quality should in- before nesting. In contrast to other , crease.Female mate choice may then be based they feed their young largely on seedsrather on charactersthat are indicativeof malequality, than solely on insects(Newton 1972, Stjernberg such as size or song rate (Zahavi 1977; Anders- 1979, Bozhko 1980). Individuals arrive from their son 1982, 1986; Kodric-Brown and Brown 1984). winter quartersto their Swedishbreeding areas Finally, if the variation among males in such in late May and begin to breed very soon after charactersis small, the availability of males is arrival. relatively high, and the time available for pair- The song of the Common Rosefinchis short ing is short,then nonrandomfemale mate choice (ca.1.0 s) and canbe classifiedinto discretesong- no longer pays, and a random mating pattern types(Bj•Srklund in press-a).A male with a par- in relation to male characteristicsis predicted ticular songtype does not change to another (Parker 1983, Hubbell and Johnson1987, Light- songtype over a season,and songtypesare in- body and Weatherhead 1987). variablewith regardto number, shape,and or- Female choice is best understood from the dering of syllables.Hence, males do not have female point of view. To maximize immediate song repertoireswhich might serve complex reproductivesuccess, females are likely to choose communication functions. Further, males do not malesthat increasetheir probability of success. have any visual displays,except that they select For example, if female reproductive successis prominent song perches in high trees. Males determinedto a large degreeby territory qual- feed their matesat a low rate during incubation ity, femalesare expectedto choosemales on the but not during early pairing stages(Stjernberg basis of their territories. Knowledge of the de- 1979, pers. obs.). terminants of female reproductive successis Autumn migration begins in late July and therefore crucial in the understanding of fe- ends in mid-August.This unusually early mi-

35 The Auk 107:35-44. January1990 36 MATSBJ6RKLUND [Auk,Vol. 107

Fig. 1. Examplesof the relationshipbetween male songperch (dots)and subsequentnest site (stars)in 1988.Stippled areasare arable land, crossedline is railway, and other lines are roads. gration is probably related to the fact that post- same morning), measured, and individually color- nuptial molt in the Common Rosefinch(un- banded.I measuredwing length (flattened),tail length, like other European passerines)occurs after tarsuslength (measuredas the distancebetween the migration rather than before (Stjernberg1979), extreme bending points at the intertarsal joint and shortening the breeding seasonconsiderably. the toes),bill length, width, and depth (measuredat I studied female mate choice in the Common the front of the nostrils).In 1988,plumage variability wasrecorded. Among males older than 1 yr, the throat, Rosefinch. In particular, I sought factors that breast,and rump were always bright red, while the determine female reproductivesuccess and ex- belly was either white or light red (never as intense amined whether a choice of mate had an effect. as the breast). The head, tail, back, coverts, and re- I also examined whether females were free to migesvaried from no red to light red to bright red. choose among males and, if so, if they chose Males were classifiedas having a high brightnessif among prospective mates or simply accepted at leasttwo of thesebody partswere bright red, and the first male they encountered. as having low brightnessotherwise. No femaleswere caught.They did not respondto playback,as malesdid; and to avoid desertions,I did METHODS not try captureat the nest.Males captured in one year did not respondto playbackthe next year, although The study was performed at R•ittvik, Central Swe- different tapeswere used. den (60ø52'N,15ø6'E) during May-July of 1985-1988 Males were observed, for 5-min periods with at in an areaof ca.1.5 km 2. The studyarea consists large- least 30-min intervals, between 0300 (sunrise) and ly of deciduousforest of differentheight (usually2- 1000.Only marked were followed. Numbers of 15 m) depending on the successionalstage (Fig. 1). songsand fights were recorded. Each male was ob- Pasturesand newly abandoned pastures together with served several times per day during the pre-pairing arable land constitute the remainder. In the aban- and pre-nestingperiods. Nestlings were banded and doned pasturesand the clear cuttings,there are scat- weighedon the sixthday after hatching,because open- teredbushes usually 1-2 m in height.The bushesand nestingpassetines' young can prematurelydesert the treesare mainlywillow (Salixspp.), birch (Betulaspp.), nest upon disturbance (Stjernberg 1979). Occasion- aspen(Populus trernula), alder (Alnusincana), and - ally, upon banding, the biggestchick tried to desert, cherry (Prunuspadus). Of specialimportance is the but it was forced to stay by putting the other chicks very rich abundanceof currant bushes(Ribes rubrum) on top of it. No further visits to the nest were made. in which manyof the nestswere placed.The study I observedfeeding frequencyfor 1 h at eachnest each area is a part of a larger area with arable land and morning (days 1-6), a total of 88 h over 2 yr (1987- gardens., usually single males,occurred 1988).In 1988, I followed 13 pairs for up to 30 min. in theseareas at very low densities. The distance between the pair members and their As malesarrived, they were caught(usually the activity were recorded every other minute. Move- January1990] MateChoice in theCommon Rosefinch 37

TABLE1. Eigenvector loadings and variance ex- (I) partitioned into episodesof selection(Arnold and plained for principal componentsextracted from a Wade 1984). These episodeswere variance in number correlationmatrix of body and bill characters. of eggs laid (W•), variance in number of eggs that hatched(W2), and variance in number of young that Principal component reached 6 days of age (W3), which was used as the Character 1 2 3 measure of fitness.

Body Tarsuslength 0.577 0.470 0.753 RESULTS Wing length 0.577 0.530 0.658 Tail length 0.577 -0.703 0.0063 Generalbreeding behavior.--On average, males % Variance 47.8 35.1 17.1 arrived 3-4 daysbefore females (1985: male me- Bill dian date of arrival was 22 May, female 26 May, Length 0.577 0.738 0.086 P = 0.015, n = 19, two-tailed Mann-Whitney Depth 0.577 -0.596 0.547 Width 0.577 -0.315 -0.833 U-test; 1986:male 26 May, female 30 May, P = % Variance 40.7 36.1 23.0 0.020,n = 31; 1987:male 27 May, female30 May, P = 0.027, n = 28; 1988: male 24 May, female 25 May, P = 0.012, n = 24). In general, arrival date accountsfor mostof the variance in pairing mentswere divided into short (< 20 m) and long (> 20 date (Table 2). The mean (+SE) number of days m), and the memberof the pair that initiated a move- ment was noted. We used a Sony Walkman Profes- betweenarrival and pairing (controllingfor the sional WM-D6C tape recorder with a SennheiserME- later arrival of females)was 1.6 + 0.25 days, n 80 microphoneto recordmales' singing upon arrival. = 19 in 1985, 1.8 + 0.22 days, n = 31 in 1986, Songswere analyzed by a UNISCAN sonograph. 1.3 + 0.19 days,n = 28 in 1987,and 1.1 + 0.20 To calculatebody and bill size and shape,the prin- days,n = 24 in 1988 (Fig. 2). Most maleswere cipal componentsanalysis (PCA) developedby Som- unpaired for only a short period. Becauseof the ers(1986) was used. Unlike ordinaryPCAs, it produces earlier arrival of males, there were nearly al- an isometricvector of size (first principal component ways more males present than females, and rather than an allometric) with all charactersthat de- newly arrived femalescould chooseamong sev- scribesize loaded equally without any shapecom- eral males (Fig 3). The pairing processis rapid ponent. The secondand third componentsdescribe shape in body and bill, respectively,after the effect and single females were not observed during of size has been removed (Table 1). The PCA is based the 4 yr of study. on 80 males.As a measureof pairing successof males, Nest building and egg laying begin on av- I used the residualsfrom the regressionof pairing erage 9.3 days (SD = 3.03, n = 41, all years date on the arrival date, which is the deviation (in combined)after pairing. Becausethere was lit- days) from the expectedvalue of an average male tle variation in time betweenpairing and laying arriving on a particular date. Hence, the term pairing of the first egg, earlier pairing resultsin earlier successimplies not how many females a male might egg laying. I observed no caseswhere males attract but the relative order in which he is paired, participatedin nest building. controlling for the averagedelay for malesthat is due I observed 19 males in 1985, 31 in 1986, 28 in to the later arrival of females. The residuals were back-correlated with the x-variable (arrival date) to 1987,and 27 in 1988,all of which becamepaired. ensurethat pairing successwas not related to arrival An important fact is that the number of first- date per se (Table 2). year males (recognizable by the femalelike The variance in female reproductive successwas plumage) was very low: only 2-4 advertising studiedby calculatingthe opportunityfor selection males were present each year (Bj6rklund in

TABLE2. Regressionsof arrival date (x) on pairing date (y) of Common Rosefinch males in 1985-1988.

Residuals vs. arrival

Year Model r F df P rs P 1985 y = 0.73x + 10.07 0.88 57.33 1, 17 <0.0001 -0.14 >0.50 1986 y = 0.86x + 6.23 0.87 87.03 1, 29 <0.0001 0.12 >0.50 1987 y = 0.77x + 8.77 0.85 68.69 1, 25 <0.0001 0.06 >0.50 1988 y = 0.55x + 12.83 0.73 24.45 1, 22 <0.0001 -0.05 >0.50 38 M•TSBJ•RKLUND [Auk,Vol. 107

7 1985 12' 1987

11'

10-

9.

8.

I I I •_ E 5' d I I I I Latency Z 4- 0 -- Od 03 3-

1986 I

I I I I , , Latency 0 •' Od 03

13-- 1988

12,

11

8-

7- , , Latency 6-

Fig. 2. Latency (days) between male arrival and 5- pairing dates in 1985 and 1986 (above), and 1987 and 1988 (right). 4- press-b).Hence, age of males was not likely to 3- be involved in female choice.Because the song 2- rate and general behavior differ greatly be- 1 tween unpairedand pairedmales, unpaired and advertisingmales were detectedeasily. Float- ing, nonadvertisingmales may have beenover- looked. In 1987, 1 male had 2 females and 2 , , , , , , Latency nests.When paired, malesgreatly reducedtheir singing rates,and some even stopped singing altogether (1985: unpaired males• + SE = 8.01 + 0.67 strophes/rain, paired males • = 4.16 + 0.67, F = 4.89, df = 1, 275, P < 0.001; 1987: 0.75, F = 11.60, df = 1, 190, P < 0.001, tested unpairedmales g = 7.94 + 0.52, paired males againstwithin-male variance;1986: unpaired œ= 4.38 + 0.65, F = 5.45, df = 1,200, P < 0.001; males• = 8.30 + 0.59, paired males• = 5.34 +_ 1988:unpaired males g = 7.78 + 0.27, paired January1990] MateChoice in theCommon Rosefinch 39

os os 10, 1985 10- 1986

20 25 i 5 20 25 1 5 May June May June

os os 10. 1987 10- 1988

20 25 1 5 20 25 1 5 -May June May June Fig. 3. Numbersof malesavailable during the seasonmeasured as operationalsex ratio (OS). OS = (No. males+ 1)/(No. femalesarriving + 1). males œ= 2.52 + 0.39, F = 150.03, df = 1, 138, The most important component of variance in P < 0.001).All males,including the bigamous female reproductivesuccess was the number of male, startedto sing again during egg laying eggshatching (Table 3). This variation was due and moved away from the nest during incu- to predation, and each year ca. 37% of the nests bation, presumably to attract another female. were destroyed(Bj•rklund in press-b).Nests Becausemany malesleft the studyarea during close(<20 m) to an edgewere muchmore likely incubation,the numberof femalesper malewas to fail than nests farther away (Bj•rklund in probably underestimated.The distancemoved press-b). It is incumbent on females to choose couldbe considerable:one male (in 1987)sang safebreeding sites.It is thereforeimportant to 3 km from its first nest. All males, however, analyzeif malesdefend such sites and if females regularlyvisited the nestingfemales and later choose males on the basis of safe nest sites. helped in feedingthe young. Common Rosefinch males do not defend ex- Malesguarded their matesintensely. Overall, clusive feeding territories (Newton 1972, 862min of observationswere gathered(œ + SD Stjernberg 1979), but song perches were de- = 63 min + 41.1, n = 13 pairs, range 20-146 fendedto somedegree (two malesrarely sang min/pair). Males were observedfarther than 20 in the sametree at the sametime). After pairing, m from their mates ca. 22.5% of the time (SD = two pairs might foragesimultaneously in the 16.4,range 0-56.3%, n = 13 males).Further, when samelarge tree. Song percheswere in higher males initiated long movements, females fol- than averagetrees, in solitary bushesin open lowed the males in 31.6% (24/76) of the cases; fields, on wires, and near forest edges. Nests but, when females initiated such a movement, were not necessarilynear these song perches their mates followed in 85.4% (74/87) of the (seeFig. 1 for 3 examples).In many casesthe cases(X 2 = 45.0, P < 0.001). There were no dif- pair moved away after pairing, often consid- ferencesin initiation of short (53.8% for males erable distances(Stjernberg 1979). They were vs. 46.2% for females, X2 = 1.31, P > 0.10) or usuallyseen in the area on the first few days long movements(46.6% for males vs. 53.4%for after pairingbut, with time, movedfarther away females, X2 = 0.37, P > 0.10). When males lost from the songperch (which might then be used contactwith their females,they startedto sing by another male). This happened in 4/19 (21%) (relative song frequency <20 m away: 28.7%, casesin 1985, 19/31 (62%) in males 1986, 16/28 >20 m away: 76.8%,P < 0.025, T = 77, Wilcoxon (57%) in 1987, and 16/27 (59%) casesin 1988. I test, two-tailed). suggest(see Table 4) that females chosenest Determinantsof femalereproductive success.-- siteswithout regard to the positionof malesong 40 MAtSBJ•RKLUND [Auk,Vol. 107

TABLE3. Femalereproductive success in 1985-1988.Sample sizes are in parentheses.W is fitness,V is variance, and I is opportunity for selection.

1985 (13) 1986 (t0) 1987 (13) 1988 (13) Number of eggslaid !• 5.08 4.90 5.00 5.08 Vw• 0.078 0.100 0.167 0.077 I• 0.0030 0.91% 0.0042 0.55% 0.0067 0.76% 0.0030 0.36% Number of eggshatched •2 3.85 3.00 3.00 3.64 Vw2 4.81 6.67 6.17 5.79 I2 0.324 98. t % 0.74 t 98.5% 0.685 78.6% 0.437 52.4% Number of 6-day-old young !•t3 3.77 2.90 2.46 3.00 Vw• 4.69 6.32 4.94 5.54 I3 0.330 0.752 0.871 0.834

perches. Thus, prospective nest sites were care probably does not influence the actual not defendedby unpaired males,and females number of fledged young (see above), it can were not likely to choosemales on the basisof influence the condition of the young. If so, a resources males held or behavior. positivecorrelation between male feeding rate In 1987and 1988,a considerableproportion and mean brood massis expectedwhen brood of variancein female reproductivesuccess was size and female feeding rate are held constant. due to a differencein the number of young Becausemale or female feeding rates,or mean reachingday 6. These losseswere due either to brood mass(male feedings.h-•.young -• 1987:œ unhatchedeggs or to chicksthat died at a very + SD = 0.18 + 0.079, n = 8, 1988: œ = 0.20 ___ early stage.The number of chicksat first check 0.034,n = 8, P > 0.10, U-test;female feedings. (nearly always day 0) did not differ from the h -•.young -• 1987:œ = 0.21 + 0.084, n = 8, 1988: number of young that reachedday 6. In one œ= 0.15 + 0.078, P > 0.10, U-test; mean brood case,a nest was destroyedat day 5. The causes weight 1987:œ + SD = 17.11 ___1.50, n = 8, 1988: and frequency of unhatched eggs (failure to œ= 15.66 ___1.33, n = 8, P > 0.10, U-test, tests developor not fertilized) were unknown. New- in all casestwo-tailed), did not differ among ly hatched, unfeathered chicks are vulnerable years,data were pooled.Male feedingsper hour to bad weather.In 1987,mean temperature dur- for each young was not correlatedwith mean ing hatching period was ca. 5øCbelow normal brood mass (r = 0.078, n = 16, P > 0.10), nor (data from local national weather station), so if was female feedingsper hour for each young femaleswere away during hatching,these loss- (r = 0.056, n = 16, P > 0.10), nor male and female es could be due to environmental causes. It is therefore important to analyze whether the amount of paternal care is predictable from a TABLE4. Distancebetween nesting site and the male's male'sbehavior before pairing and if the amount songperch. Figuresgiven are number (n) of nests of paternal care influences the quality of the found and number (n*) of nestsincluding disap- pearing pairs assumingthat these nestsare >200 young. m from the songperch. Common Rosefinches feed their nestlings largely on seeds.Mean feeding rate was œ+ SE Distance = 0.8 + 0.09 feedingsper hour (range 0.4-1.2, class(m) n % n* % n = 9 pairs)for females,and 0.6 + 0.10feedings <50 25 44.6 25 26.3 per hour (range 0.2-1.2) for males in 1987 (P = 50-100' t0 17.9 t0 10.5 0.05, sign test). In 1988, mean feeding rate was 100-150 8 14.3 8 8.4 t50-200 b 4 7.1 4 4.2 0.7 ___0.09 feedingsper hour (range 0.4-1.3, n >200 9 16.1 48 50.5 = 9 pairs) for females,and 1.0 + 0.07 feedings • Median distance. per hour for males(range 0.7-1.4, P = 0.10,Wil- bMedian distanceincluding the disappearingpairs assumingthat coxontest, two-tailed). Although male parental they nest >200 m from the songperch. January1990] MateChoice in theCommon Rosefinch 41

TABLE5. Product-momentcorrelations between pairing successand male characteristicsin 1985-1988.Num- bersin parenthesesfollowing eachyear are samplesizes; if different from year samplesize, n followsvalue.

Pairing success Character 1985 (19) 1986 (28) 1987 (31) 1988 (20) Body PC1 0.13 0.12 -0.02 -0.13 PC2 -0.11 -0.049 -0.21 -0.28 PC3 -0.11 -0.10 -0.048 0.33

Bill PC1 0.18 0.16 0.032 -0.15 PC2 - 0.012 - 0.094 0.32 0.31 PC3 0.17 -0.12 -0.042 -0.21 Songrate 0.11 -0.16 -0.39* 0.02 Mean frequency(kHz) -- 0.17 (26) -0.15 (22) 0.31 (17) Strophelength (s) -- -0.13 (26) 0.11 (22) 0.27

* P < 0.05. feeding rate combined (r = -0.14, n = 16, P > recordedonly in 1988,and only on malescaught 0.10). The partial correlation of male feeding that year (males were not recaptured from pre- rate and mean brood mass,holding female feed- vious years, see Methods). There was no rela- ing rate constant,was not significant(r = -0.14, tionship between pairing successand male n = 16,P > 0.10).Hence, the actualmale feeding plumage brightness (high: œ +_ SD = 0.65 +_ rate did not influence the condition of the 1.29, n = 5; low: œ = 0.87 +_ 2.33, n = 5, P > young. This doesnot necessarilymean that pa- 0.10,Mann-Whitney U-test). Thus,it is unlikely ternal care is unimportant,only that the vari- that females chosemates accordingto pheno- ation in feeding rate is. I concludethat females typic variation in plumage. are unlikely to exert mate choice on the basis If male dominance interferes with active fe- of fine-level prospectivepaternal care. male choice,then we would expecta correlation Femalechoice and male phenotypiccharacteris- between pairing successand dominance in tics.--Femalesprobably did not benefit mate- males. In the Common Rosefinch, size is a char- rially from matechoice. This suggests that choice acterdetermining dominance(BjSrklund 1989), basedon the genetic quality of malescan be of and correlationsbetween size and pairing suc- greaterimportance. For geneticquality to be an cessare expected.No such correlationswere important factorin mate choicerequires a vari- found (Table 5). It is likely that femalesare not ance in genetic quality of males.This variance constrainedby maledominance to pair with the must be correlatedwith somephenotypic trait first unpaired, advertisingmale they happen to that is easilyperceived by females,and females encounter upon arrival. Unconstrained and must have time to chooseamong males.There nonselective female choice would cause the was a significantvariation in songrate among variation in male pairing successto be random, males (1985: F = 5.56, df = 15, 190; 1986: F = and the distribution of number of daysbetween 4.05, df = 31,275; 1987: F = 2.33, df = 26, 200; arrival and pairing amongmales would be dis- 1988: F = 1.04, df = 21, 138, all years except tributed randomly. 1988 P < 0.001, ANOVA), but song rate corre- In all 4 yr, the number of daysbetween male lated with pairing successonly in 1987, where arrival and pairing (the later arrival of females there was a negative correlation between song controlled for) was not significantly different rate and pairing success(Table 5). In 1988,there from a Poisson distribution (Poisson was used was no significant between-male variation in because mean and variance in latency were songrate, probablybecause many malesdid not equal, see above, 1985:D = 0.16, P > 0.10, n = sing before pairing. No other morphologicalor 17; 1986: D = 0.15, P > 0.10, n = 31; 1987: D = behavioral characteristic that was measured cor- 0.18, P > 0.10, n = 28; 1988: D = 0.21, P > 0.10, relateswith pairing successin any year (Table n = 24) (Fig. 2). This alsoheld for female arrival 5). (1985: D = 0.069, P > 0.10, n = 17 days; 1986: Unfortunately,variation in plumagecolor was D = 0.13, P > 0.10, n = 13; 1987: D = 0.09, P > 42 MATSBJ6RKLUND [Auk,Vol. 107

0.10, n = 11; 1988: D = 0.21, P > 0.10, n = 10, er males, but their mates usually are not far Kolmogorov-Smirnofftest). Thus, the timing of away, and I never observed extrapair copula- pair formation in the Common Rosefinch can tions. Thus, a decreasedsong rate can reduce be regarded as a random process. the risk of extrapair copulations.The impor- tance of mate guarding may also be the cause DISCUSSION of the departure of pairs from the area soon after pair formation.Pairs may leaveareas where Females'reproductive success was largely de- the densityof malesis high to avoidharassment termined by choiceof nestsite, which was un- from other males. relatedto malebehavior. The probabilityof suc- The fact that the number of young or mean cess was independent of male physical brood masswas not correlatedwith male pa- characteristicsor behavior.Consequently, I ex- rental care doesnot necessarilymean that the pectedpairing to be randomwith regardto these contribution of the male is unimportant. Be- characteristics:exactly what I found. However, causeyoung are fed so seldom,growth is un- the number of females per male was far from likely to be determinedby the number of pa- random. If females chose the first male they rental feeding trips. Instead,absence of a male encountered,whether paired or unpaired,the during the nestlingperiod may affectthe time number of femalesper male would be normally femalescan brood the young,which in turn can (Poisson) distributed (Sutherland 1985). Then, affectthe survival of the young. This may ex- with a similarmean number of femalesper male plain female preferencefor unpaired males. (=1.0), we would expectto find approximately Similarly,female preference for unpairedmales equal proportions (1/3) of bachelors,monoga- mayexplain male movement away from the nest mouslypaired males,and polygynouslymated to seek subsequentmates (Alatalo et al. 1981). males.This was not the casein this population, Lightbody and Weatherhead(1987) argued where monogamy was the rule. Thus, females that female choicecan be regarded as neutral almostinvariably choseunpaired males. with regard to males or their territories if a The frequencyof polygyny can be underes- number of conditionsare met. Although these timated. Common Rosefinch males resume conditions were derived from the polygyny singingand moveaway from the nestsite around thresholdmodel for the evolutionof polygyny, the onsetof incubation;in somecases they move they apply to matechoice in predominantlymo- considerabledistances (Stjernberg 1979). Two nogamousspecies as well. Six conditions are other questions(why females prefer to mate necessary:(1) femalesdo not interact competi- with unpaired males and why males resume tively or cooperativelywhile settling, (2) food singing in anotherplace) are still open. supply in the territory is unimportant,(3) nest Female preference for unpaired males may sites are not limited, (4) variation in territory result because males reduce or discontinue their features is not correlated with variation in fe- singing after pairing. Consequently,newly ar- male reproductive success,(5) variation in male rived femalesmay fail to find the relatively si- "quality" is unrelated to female choice,and (6) lent, paired males. One reason males reduce or paternalcare is not necessaryfor female repro- stop singing when paired is mate guarding. ductive success. Malesthat continueto sing from their perches The CommonRosefinch feeds in a large area cannotsimultaneously guard their matesfrom around the nest, evidently up to several kilo- extrapaircopulations. Further, a singing, paired meters away (Stjernberg 1979). Furthermore, malemay attractnot only newly arrivedfemales nestsites are by no meanslimited, and variation but also other males (Alatalo et al. 1982), which in male quality is unrelated to female choice, may lead to extrapair copulations.Mate guard- thus satisfyingconditions 2, 3, and 5. Females ing by the male is certainly important in the occasionallyfight after pairing and, becauseter- Common Rose finch. Because males do not de- ritoriesare established only afternest building, fend exclusive territories, which may help to female interactionsare unlikely to influence reduce the frequency of extrapair copulation settling; this meets condition 1. attempts(M•ller 1987),paired malesshould stay The main influence on female reproductive near the female until after laying to guard successwas the probability of predation and against insemination by other males. Paired this may suggestthat territory quality is related CommonRosefinch females are courted by oth- to female choice. However, territories were es- January1990] MateChoice in theCommon Rosefinch 43 tablished only after nest building, and it is that the period to raise a brood and to gain highly likely that femaleschoose the nest site energy for the migration is much shorterthan rather than the male (Bjfirklundunpubl.). The most Holarctic migrating passetines. choice of nest site is also unrelated to male The main influence on female reproductive movementsbefore pairing, which satisfiescon- successin Common Rosefincheswas the prob- dition 4. Finally, I demonstratedthat paternal ability of predation during incubation. In fact, care was unrelated to the number of fledged given the circumstancesunder which broods young and that prospectivepaternal care was fail, I concludedthat predation was on females unpredictablefrom male prematingbehavior. rather than eggs(Bj•rklund unpubl.). The only I believe that female choice in the Common pattern found was that nests close to an edge Rosefinchis essentiallyneutral. At present,the between bush land and arable land were much frequencyof neutral femalechoice is unknown, more susceptiblethan nestsfarther away from but Lightbody and Weatherhead(1987) cite such an edge. There is thus considerablepres- studies where the postulated criteria are par- sure on females to choose safe nest sites. Gen- tially met. I suggestthat it may be more com- erally, femaleschoose nest sites that are covered mon than assumedpreviously, and that it cer- by leaves,which make it difficult for an air- tainly is worth further study. borne predator to locate (Bjfirklund unpubl.). One factor likely to strongly influence the Conversely,cats and mustelids regularly pa- degree of choice is time, becauseit greatly in- trolled theseedges after dark. Female Common fluencesthe costof searchingfor a mate (Alatalo Rosefinchesare known to stayon the nestwhen et al. 1988, Slagsvoid et al. 1988). Time con- approached(Stjernberg 1979, Bozhko 1980). This straintshave been suggestedas an explanation may be fatal when the predator is a cat. It is for a male Pied Flycatcher's(Ficedula hypoleuca) worth noticing that this population is lessthan ability to obtaintwo femalesalthough he gen- 25 yr old, and the predation pattern may be erallyprovides care for only onebrood (Alatalo different from the pattern elsewhere. Thus, fe- et al. 1981, 1986; Alatalo and Lundberg 1984; males choose nest sites according to one rule, Slagsvoidet al. 1988). Male Pied Flycatchers although there are other rules that may be at defend two or more territories (i.e. nest holes), leastequally important. This createsa high vari- generallyquite a distanceapart. Female choice ance in reproductivesuccess in the population. in the Pied Flycatcheris probablynot neutral I conclude that pair formation in the Com- because males defend a resource essential for mon Rosefinchis a random processwith regard reproduction,i.e. the nest hole. Subsequently, to male characteristics based on two facts. First, it has been shown that females chooseamong femalereproductive success is unrelatedto any available nest holes rather than among males resources males control, and females do not (Alatalo et al. 1986),although under certain cir- choose males on the basis of material benefits cumstancesfeatures of the males defending either controlledby the male(such as territory) theseholes may have some influence on female or correlatedwith male quality (feedingrates). choice(Slagsvoid et al. 1988). Second,because the number of arriving females The influence of time differs between these cannotbe predictedfrom one day to the next, two species.In the Pied Flycatcher,the clutch the number of available males from which to size decreases with time, and, under natural chooseis unpredictable.A hesistantfemale may conditions,the availabilityof suitablenest holes end up unpaired, becausethe time available for is limited (Alatalo et al. 1981, and others). Con- pair formation is limited. These facts,and sub- sequently,the costof searchingfor matesis re- sequent behavior, promote fast decisionsand a lated to an expectedbrood size decreasingwith low degreeof selectivity. time. In the Common Rosefinch, the situation is more drastic.The pairing period is short and ACKNOWLEDGMENTS unpredictable, and females may end up un- I thank Rauno Alatalo, JacobHSglund, Arne Lund- paired if they spend excessivetime choosing. berg,Bob Montgomerie, Anders Pape Moller, Torsten The breeding seasonis also shortenedbecause Stjernberg,and StaffanUlfstrand for valuablediscus- the Common Rosefincharrives at the breeding sionsand comments.I also thank Anders Sj81undfor groundca. 3 weekslater than the PiedFlycatch- excellentfieldwork in 1987 and 1988.This study was er (Alatalo et al. 1981), and autumn departure supportedby grantsfrom the RoyalAcademy of Sci- startsearly (late July to earlyAugust). This means ences(Hierta-Retzius, Ahlstrands, and Regnells). 44 MATSBJ6RKLUND [Auk,Vol. 107

LITERATURE CITED choice, and sexual selection. Am. Nat. 130: 91- 112. ALATALO, R. V., A. CARLSON& A. LUNDBERG. 1988. KIRKPATRICK,M. 1985. Evolution of female choice The searchcost in mate choiceof the Pied Fly- catcher. Anim. Behav. 36: 289-291. and male parental investmentin polygynous species:the demise of the "sexy son." Am. Nat. , , --, & S. ULFSTRAND. 1981. The 125: 788-810. conflictbetween male polygamyand femalemo- KODRIC-BROwN,A., & j. H. BROWN. 1984. Truth in nogamy:the caseof the Pied FlycatcherFicedula advertising:the kindsof traitsfavored by sexual hypoleuca.Am. Nat. 117: 738-753. selection. Am. Nat. 124: 309-323. , & A. LUNDBERG.1984. Polyterritorialpolyg- LIGHTBODY,J.P., & P. J.WEATHERHEAD. 1987. Polyg- yny in the Pied FlycatcherF&edula hypoleuca. Evi- yny in the Yellow-headed Blackbird: female dence for the deception hypothesis.Ann. Zool. Fennici 21: 217-228. choiceversus male competition.Anim. Behav.35: 1670-1684. ---, & M. BJ•RKLUND.1982. Can the song MOLLER, A.P. 1987. Intruders and defenders in avi- of male birds attract other males?An experiment an breedingterritories: the effectof spermcom- with the Pied FlycatcherFicedula hypoleuca. Bird Behav. 4: 42-45. petition. Oikos 48: 47-54. NEWTON, I. 1972. Finches. London, Collins. , --, & C. GLYNN. 1986. FemalePied Fly- NISBET,I. C. T. 1973. Court-shipfeeding, egg-size catcherschoose territory quality and not male characteristics. Nature 323: 152-153. and breedingsuccess in CommonTerns. Nature 241: 141-142. ANDERSSON,M. 1982. Sexual selection, natural se- 1977. Court-shipfeeding and clutch-sizein lection, and quality advertisement.Biol. J. Lin- naen Soc. 17: 375-393. Common Terns, Sternahirundo. Pp. 101-109 in Evolutionaryecology (B. Stonehouseand C. Per- 1986. Evolution of condition-dependentsex rins, Eds.). London, Macmillan. ornamentsand mating preferences:sexual selec- ORING,L.W. 1982. Avian mating systems.Pp. 1-92 tion basedon viability differences.Evolution 40: 804-816. inAvian biology (D. Farher,J. King, and K. Parkes, Eds.), vol. 6. New York, Academic Press. ARNOLD,S. J., & M. J. WADE. 1984. On the measure- ment of natural and sexual selection. Evolution PARKER,G. A. 1983. Mate quality and mating deci- sions. Pp. 141-166 in Mate choice (P. Bateson, 38: 709-719. Ed.). Cambridge,Cambridge Univ. Press. BEEHLER,B. M., & M. S. FOSTER. 1988. Hotshots, hot- SEARCY,W.A. 1982. The evolutionaryeffects of mate spots,and femalepreference in the organization selection.Annu. Rev. Ecol. Syst. 13: 57-85. of lek mating systems.Am. Nat. 131:203-219. SLAGSVOLD,T., J. T. LIFJELD,G. STENMARK,& T. BREIE- BJ•RKLUND,M. In press-a. Microgeographicvaria- HAGEN.1988. On the costofsearching fora mate tion in the songof the ScarletRosefinch Carpod- in female Pied FlycatchersFicedula hypoleuca. acuserythrinus. Ornis Scandinavica. Anim. Behav. 36: 433-442. --. In press-b.Is theabsence of first-yearScarlet SOMERS,K.M. 1986. Multivariate allometryand re- Rosefinch males from breeding grounds related to a male-biased sex ratio? Behaviour. movalof sizeand principal components analysis. Syst. Zool. 35: 259-268. 1989. Male contests in the Scarlet Rose finch STJERNBERG,T. 1979. Breedingbiology and popula- (Carpodacuserythrinus) in relation to asymmetries tion dynamicsof the ScarletRosefinch Carpodacus in resourceholding power and pairing status. erythrinus.Acta Zool. Fennica 157: 1-88. Behav. Ecol. Sociobiol. 25: 137-140. SUTHERLAND,W.J. 1985. Chancecan producea sex BRADBURY,J. W., & R. GIBSON. 1983. Leks and mate differencein variancein mating successand ex- choice. Pp. 109-138 in Mate choice (P. Bateson, plain Bateman'sdata. Anim. Behav.33: 1349-1352. Ed.). Cambridge,Cambridge Univ. Press. VEHRENCAMP,S. L., & J.W. BRADBURY.1984. Mating BOZHKO,S.I. 1980. Der Karmingimpel.NeueBrehm- systemsand ecology.Pp. 251-278 in Behavioural Bticherei 529. DDR Wittenberg Lutherstadt, A. ecology:an evolutionaryapproach (J. R. Krebs Ziemsen Verlag. and N. B. Davies, Eds.),2nd ed. Oxford, Blackwell HALLIDAY, T. R. 1978. Sexual selection and mate Sci. Publ. choice.Pp. 180-213 in Behaviouralecology, an YASUKAWA, K., R. L. KNIGHT, & S. KNIGHT SKAGEN. evolutionaryapproach (J. R. Krebsand N. B. Da- 1987. IS courtshipintensity a signalof malepa- vies, Eds.). Oxford, Blackwell Sci. Publ. rental care in Red-wingedBlackbirds (Agelaius 1983. The study of mate choice.Pp. 3-32 in phoeniceus)?Auk 104: 628-634. Mate choice(P. Bateson,Ed.). Cambridge,Cam- ZAHAVI,A. 1977. The costof honesty(further re- bridge Univ. Press. markson the handicapprinciple). J. Theor.Biol. HUBBELL,S. P., & L. K. JOHNSON. 1987. Environmen- 67: 603-605. tal variancein lifetime reproductivesuccess, mate