The Condor88:499-504 0 TheCooper Ornithological Society 1986

GEOGRAPHICAL ECOLOGY OF CLUTCH SIZE VARIATION IN NORTH AMERICAN WOODPECKERS

WALTER D. KOENIG HastingsReservation and Museum of VertebrateZoology, Universityof California, Berkeley, CA 94720

Abstract. Clutch sizes of North American woodpeckersexhibit strong positive latitudinal gra- dients. There is no overall effect of either estimated first egg date or proximity to the coast on clutch size. However, mean clutch size varies inversely with an index of winter productivity and estimates of breeding densities of all woodpecker speciescombined; clutch size is uncorrelated with summer productivity. These results support Ashmole’s hypothesisthat geographictrends in fecundity are determined, at least in part, by seasonalityof resources. Key words: Ashmoles’ hypothesis;clutch size; geographicvariation: Picidae; woodpeckers.

INTRODUCTION Specieswithin the family Picidae are widely Since Lack (1947, 1948) first focusedattention distributed and often common throughout on clutch size as a major topic of ornithological North America. Furthermore, despite a fair interest, geographicalvariation has served an degree of ecological diversity within the fam- important role in testing theories that seek to ily, all speciesare altricial, cavity nesting, and explain the evolution of clutch size (e.g., Cody (except for the Acorn Woodpecker Melanerpes 1966). Here I examine clutch size in North formicivorus) monogamous, and thus the im- American woodpeckers(Picidae) with the aim portanceof theselife-history featuresas sources of both describing the geographic patterns of of clutch size variation are minimized. variation and of relating those patterns to hy- MATERIALS AND METHODS potheses concerning the evolution of clutch size. I discusselsewhere clutch size variation Data on clutch size were obtained from col- within this taxon with respect to morphology lectionsin North American museumsand from and diet (Koenig, unpubl.); thus, here I focus an ongoing study of the Acorn Woodpecker at primarily on geographic and climatic corre- Hastings Reservation, Monterey Co., Califor- lations. nia (Koenig and Mumme, in press). In both A primary goal of the analyses performed cases,I used only clutches which were known here investigating the correlation between or thought to be complete based on informa- clutch size and climate is to examine Ash- tion given in the collector’s original data cards. mole’s hypothesis(Ashmole 196 1, 1963). This For each clutch I determined latitude, longi- hypothesis suggeststhat clutch size is deter- tude, estimated date of clutch initiation, and mined by resource abundance during the location in relation to a coastline (whether the breeding season relative to breeding popula- clutch was collected within 1” of a coast or tion density. Ashmole further proposed that inland). Date of clutch initiation was estimated breeding population density was determined as the collection date, minus the number of by winter resource abundance; hence, clutch eggs in the clutch, minus a correction factor size should be positively correlated with sea- based on an 1 l- to 14-day incubation period sonality of resources(i.e., the ratio of summer (Jackson 1977) and the estimated age of the to winter resources) rather than absolute re- clutch as determined by the stageof incubation sourceabundance during any one season.Ash- listed by the collector. For further details on mole’s hypothesishas been shownto be a pow- this estimator, see Koenig (1984a). erful predictor of geographicpatterns of clutch Data were handled in two ways depending size variation both in a seriesofpasserine com- on the goals of the analysis. First, individual munities (Ricklefs 1980) and in the Northern clutcheswere used as independent data points Flicker Colaptes aurutus (Koenig 1984a), and for analysesin which it was desirable that in- has also been implicated as being important traspecific or intrageneric variation be pre- in a variety of other taxa (Grant and Grant served intact. Second, clutches of the same 1980, Cockburn et al. 1983; see also Boyce speciescollected within the same 1” block of 1979). latitude and longitude (latilong block) were combined. Means of clutch sizescollected from all specieswithin a particular latilong block ’ ’ Received 14 February 1986. Final acceptance12 June were averaged to obtain a mean for all wood- 1986. pecker species occurring within that latilong 14991 500 WALTER D. KOENIG

TABLE 1. Mean and latitudinal correlations of clutch size in North American woodpecker speciesand genera.’

Latitudinal range Correlation with Species of sample Mean i- SD latitude (I,) n clutches

Colaptesauratus 26-54 (28”) 6.20 + 1.56 0.58*** 468 Melanerpescarolinus 25-43 (18”) 4.31 f 0.76 0.14 61 M. aurifrons 14-34 (2ooj 4.65 + 0.80 0.29* M. uropygialis 24-38 (14”) 3.75 * 0.93 -0.02 2; M. erythrocephalus 29-44 (150) 4.82 * 0.80 0.10 71 M. formicivorus 32-39 (7”) 5.06 * 1.06 0.14* 224 M. lewis 33-46 (13”) 5.88 + 1.19 0.32* 51 All Melanerpes 14-46 (32q 4.81 t 1.39 0.24*** 531 Sphyrapicusvarius 38-55 (17”) 4.93 f 1.02 0.42** 46 S. ruber 34-46 (12”) 4.69 f 0.72 0.22 42 S. nuchalis 36-45 (9“) 4.77 * 1.09 - 17 S. thyroideus 34-46 (12”) 5.67 * 0.85 0.12 49 All Sphyrapicus 34-55 (21 g 5.08 * 0.98 0.03 154 Dryocopuspileatus 26-55 (29”) 3.80 & 0.77 0.24* 92 Picoidesvillosus 28-62 (34”) 3.93 i 0.72 0.26*** 159 P. pubescens 28-61(33”) 4.81 f 0.78 0.27*** 205 P. scalaris 25-34 (9”) 3.81 + 0.76 -0.25 58 P. nuttallii 33-40 (79 4.34 + 0.77 -0.05 88 P. stricklandi 32 (0”) 3.50 t- 0.84 6 P. borealis 28-36 (8”) 3.68 ? 0.69 0.16 P. albolarvatus 33-46 (139 4.35 i 0.73 0.09 ;; P. arcticus 43-55 (12”) 3.89 rt 0.85 0.32 27 P. tridactylus 41-61 (20”) 3.87 f 0.64 15 All Picoides 25-62 (377 4.28 f 0.85 0.1F 681 All species 14-62 (484 4.93 * 1.45 0.22*** 1,926

= Correlations with latitude given only when the sample sire IS >25 clutches. *P < 0.05, **P < 0.01. ** P < 0.001. block. Mean latilong clutch size data were then American breeding bird survey program ad- used in analysesinvolving climate and popu- ministered by the U. S. Fish and Wildlife Ser- lation density, as these latter variables could vice (Bystrak 198 1). Mean abundance values not be independently calculated for every were calculated for each specieswithin each clutch, but could be reasonably determined for latilong block from over 18,000 surveys con- geographicareas on the order of latilong blocks. ducted over several thousand different routes. For each latilong block, I added the follow- I assumedthat woodpeckers constitute a rea- ing two climatic variables as determined from sonably cohesive “guild” (Root 1967), and the nearest station listed in Thornthwaite As- thus, that the major competitors of wood- sociates (1964): mean “summer” actual peckers are likely to be confamilial individu- evapotranspiration (AE), defined as the com- als, including conspecifics. Accordingly, I bined AE values for the three consecutive combined the abundances of all woodpecker months yielding the highest total, and mean speciesin analyses involving breeding densi- “winter” AE, defined as the combined AE val- ties. Further details can be found in Koenig ues for the three consecutive months yielding (1984a). the lowest total. Actual evapotranspiration is Taxonomy follows the AOU (1983, 1985). defined as precipitation minus runoff and per- In all, a total of 1,926 complete clutches from colation (Sellers 1965) and is directly related 21 species of North American woodpeckers to primary productivity in terrestrial environ- was analyzed. The Northern Flicker was ex- ments (Rosenzweig 1968). When combined cluded from some of the family-wide analyses into “summer” and “winter” values, AE data becauseI have analyzed it in detail elsewhere hasa theoretically straightforward relationship (Koenig 1984a, b), and becausethe strong lat- to seasonality of resourcesand is thus appro- itudinal trends and relative abundance of this priate for testing Ashmole’s hypothesis (Rick- taxon might otherwise unduly influence the lefs 1980; Koenig 1984a; but see below and results of analysesperformed here. Hussell 1985). Indeed, AE values have been previously shown to be a good predictor of RESULTS geographicvariation in clutch size in both pas- serine communities (Ricklefs 1980) and in the THE LATITUDINAL GRADIENT Northern Flicker (Koenig 1984a). Mean clutch sizes( f SD) for the 2 1 speciesof Data on the comparative abundanceof pop- woodpeckersconsidered are given in Table 1. ulations was determined from the North Also listed in Table 1 are the latitudinal ranges CLUTCH SIZE VARIATION IN WOODPECKERS 501

TABLE 2. Analysis of covariance of clutch size for in- four species(100%) in the first category yielded dividual speciesand for generaof North American wood- significantcorrelations with latitude, while only peckersexcluding Colaptesauratus. two of six (33%) in the second category and

Regres- one of four (25%) in the third category did so sion (x2 = 6.7, df = 2, P < 0.05). These results Variable slope F-Value df indicate that the interspecific differencesin the Species correlation between clutch size and latitude Covariates listed in Table 1 are largely an artifact: the Latitude 0.026 35.3*** 1, 1173 likelihood that a significant correlation occurs Est. first egg is a function of the sample size and the lati- date 0.000 0.9 1,1173 tudinal range over which samples were ob- Main effects - tained. ~. Species 22.9*** 19,1173 Proximity to Table 1 also lists mean clutch size for each the coast” - 0.1 1,1173 genusunder consideration. Clutch sizes of all -Explained five genera correlate positively with latitude, (R* = 0.286) - 12.8*** 39, 1173 and only that for sapsuckers(Sphyrupicus) is Genera not significant. (The sample size and latitu- Covariates dinal range for Sphyrupicusis large enough to Latitude 0.023 23.8*** 1 suggestthat this may reflect a real generic dif- Main effects ference.) Using all individual clutches com- Genus - 57.8*** 3 bined (either including or excluding C. auru- Explained tus), there is also a highly significantlatitudinal (R* = 0.120) - 49.3*** 4 gradient. Analyses of covariance similarly in- aWhether the clutchwas collected inland or within I” of the coast. dicate the importance of latitude as a predictor of clutch size within the family, even when C. aurutus is excluded (Table 2). over which clutches for each taxon were ob- The data in Table 2 also indicate that, at tained and correlation coefficients for clutch least on a familywide basis, neither estimated size with latitude for each taxon for which I first egg date nor proximity to the coast sig- was able to find 25 or more complete clutches. nificantly influence clutch size. In both the Of the 18 speciesin this category, correlation analyses in Table 2 the covariates (including coefficients were positive for 15 species (sig- latitude) were controlled prior to consideration nificantly so for 8 species spread throughout of the main factors. Thus, there remain highly all 5 genera), and negative for only 3 species significant interspecific and intergeneric dif- (none significantly so). ferencesin clutch size after controlling for lat- Examination of Table 1 suggeststhat the itude. significance of the latitudinal correlation for individual speciesdepends in part on the range ASHMOLE’S HYPOTHESIS over which clutcheswere collected and on the Ashmole’s hypothesispredicts that geographic samplesize. The first of theserelationships was variation in clutch size is correlated with sea- confirmed by a Mann-Whitney U-test com- sonality of resources(Ashmole 196 1, Ricklefs paring the latitudinal range for each species 1980). Prior analyses (Ricklefs 1980, Koenig according to whether the latitudinal correla- 1984a) have tested for such a correlation by tion with clutch size was significantat the 0.05 comparing clutch size with the ratio of summer level or not; a highly significant difference was to winter actual evapotranspiration. As point- found (z-value = 5.04, P < 0.00 1). The second ed out by Hussell (19 8 5) however, a stricter was tested by dividing speciesinto three cat- interpretation of Ashmole’s hypothesis sug- egories based on the number of clutches ob- geststhat the appropriate test is between clutch tained (N > 100, 51 to 100, and 25 to 50). All size and (S - IF)/’ IV, where S = summer and

TABLE 3. Spearmanrank correlationsbetween variables usingmean values for each latilong block. n = 297 latilongs except for correlations involving breeding density, for which n = 262 latilongs.

SummerAE- Breedingdensity Latitude SummerAE Winter AE winter AE all picids

Mean clutch size all picids 0.37*** -0.06 -0.32*** 0.00 -0.17** Latitude - -0.07 -0.76*** 0.06 -0.24*** Summer AE - - -0.05 0.97*** 0.32*** Winter AE - - - -0.22*** 0.22*** Summer AE-winter AE - - - - 0.31*** 502 WALTER D. KOENIG

TABLE 4. Regressionsof mean clutch size for all wood- tude controlling for ln[ w] yields an F-value of peckerscombined by latilongs on actual evapotranspira- 10.6, df = 1, 295, P -c 0.01). tion and breeding density. Variables are log-transformed. DISCUSSION There is a strong positive latitudinal gradient Actual evapotranspiration in clutch size among North American wood- Summer AE -0.019 0.6 1,294 peckers (Table 1). Fifteen of 18 species(83%) Winter AE -0.044 31.1*** 1,294 with adequate data and four of five genera Explained (80%) yield positive correlationsbetween clutch (RZ = 0.096) - 15.5*** 2,294 size and latitude, significantly so for eight Summer AE and density speciesand four genera. Only three specieshad Summer AE -0.014 0.3 1,259 negative correlations between clutch size and Density -0.112 5.0* 1,259 latitude; two of thesewere very weak and none Explained was significant. Furthermore, interspecific dif- (R* = 0.020) - 2.9 2,259 ferencesin the significanceof the observed lat-~ itudinal trends appear to be largely a function of sample size and the latitudinal range over W = winter AE. In Table 3 I present correla- which clutches were obtained. Thus, North tions between the mean clutch size of all picids, American woodpeckers join the ranks of the determined for each latilong block as described variety of other birds and mammals for which above, with latitude, S, IV, and S - IV, and strong latitudinal gradients in clutch size have the estimated breeding density of all wood- been described (e.g., Lack 1947, Lord 1960, peckers.Using thesedata there is again a strong Cody 1966). positive latitudinal gradient in clutch size. I have shown elsewhere(Koenig 1984a) that There is no correlation between variation in there is no correlation in C. aurutusbetween mean clutch size and either S or S - IV, while clutch size and either first eggdate or proximity clutch size correlates inversely with both W to the coast when controlling for latitude. The and breeding density. The correlation between identical pattern holds within the family as a S and S - W is close to unity, indicating that whole excluding C. aurutus (Table 2). Clutch Hussell’s (198 5) correction is unnecessaryfor size may still decreasewith date at a particular this data set. I will thus use only S, rather than locality; however, the latitudinal gradient S - IV, in the following analyses. masks whatever date or coastal effects may be The top half of Table 4 presentsthe results present on a finer geographicscale (see Moore of a multiple regressionof ln[mean clutch size] and Koenig 1986). by latilongs on ln[S] and ln[W’J As expected Table 2 also indicates that there are signif- based on the results in Table 3, W correlates icant interspecific and intergeneric differences inversely with mean clutch size, while there is in clutch size within the family even after con- no relationshipbetween clutch sizeand S. Thus, trolling for latitudinal differences.At least some variation in mean clutch size correlates with of these differences are related to differences S/W due to the inverse correlation between in foraging ecology and feeding behavior, with clutch size and IV. speciesmore dependent on wood-boring beetle On a proximate level, Ashmole’s hypothesis larvae tending to have smaller clutches than predicts that geographicpatterns in clutch size those dependent on other resources (Koenig, are determined by the ratio of resourcesavail- unpubl.). In addition, migratory speciestend able for reproduction to breeding population to have larger clutcheswhile specieswith great- density. This prediction is tested in the lower er sexual dimorphism in culmen length tend half of Table 4 with the latilong data using S to have smaller clutches.Nonetheless, a major and the estimates of population density for all geographicgradient within the family remains woodpeckers combined derived from the even after interspecific variation is removed. breeding bird surveys.The resultsindicate that Mean clutch size of North American wood- clutch size again is uncorrelated with S. How- peckers varies inversely with winter AE (IV) ever, mean clutch size decreasessignificantly and with the breeding density of all picids but with increasing density. does not correlate with summer AE (S) or with Although there is a strong correlation be- S - IV. These resultsare further supported by tween winter AE and latitude (Tables 3 and 4), multiple regressions(Table 4). Insofar as AE there remains a highly significant latitudinal values represent seasonal availability of re- gradient even after controlling for winter AE sources,these resultsthus support all four pre- (using the combined mean clutch size data, a dictions of Ashmole’s hypothesis(including the multiple regression of In[clutch size] on lati- proposition that winter resourceslimit breed- CLUTCH SIZE VARIATION IN WOODPECKERS 503 ing population size)delineated by Moller (1984, population, this finding eliminates the con- see also Hussell 1985): (1) clutch size is in- founding effects of latitudinal correlates such versely correlated with winter resources(Table as varying daylength. Nonetheless, even this 3), (2) clutch size is positively correlated with analysisfails to go beyond simple correlations, resourcesper breeding pair (lower half of Table and it obviously fails to directly address the 4) (3) clutch size is positively correlated with issue of geographic variation in clutch size. the ratio of summer to winter resources(upper Thus, experimental manipulation and consid- half of Table 4) and (4) clutch size is uncor- eration of time-activity budgets, as suggested related with summer resources(Table 3). by Hussell (1985), are desirable to test alter- These results, along with the community- native hypothesesof clutch size variation. wide analysisof Ricklefs (1980) and my earlier Seasonality of resourcesis unlikely to solve analysisof C. aurutus(Koenig 1984a), all lend the problem of geographicvariation in clutch support to the role of seasonalityof resources size (see also Moller 1984). However, its rel- in determining geographic variation in clutch ative successin a spectrum of analytical levels size. There are, however, several reasons to suggeststhat it influences clutch size in a va- continue investigation of the factors influenc- riety of taxa. Ashmole’s hypothesis clearly de- ing geographic trends in fecundity. First, sea- serves careful consideration in all studies of sonality of resourcescan explain only a small geographicvariation in fecundity. fraction (2.0 to 9.6% based on the R2 values in Table 4) of the total variance in clutch size, ACKNOWLEDGMENTS and there remains a strong, highly significant Data usedin theseanalyses were collectedfrom the Amer- latitudinal gradient even after controlling for ican Museum of Natural History, California Academy of winter AE. Similar results were found for C. Sciences,Chicago Academy of Sciences,Delaware Mu- seum of Natural History, Florida State Museum, Milwau- aurutusalone (Koenig 1984a). Thus, at best, kee County Public Museum, Moore Laboratory of Zool- Ashmole’s hypothesiswould appear to be able ogy, Museum of Vertebrate Zoology, National Museum to explain only a fraction of the geographic ofNatural History, San BernardinoCounty Museum, San- variation in clutch size within North American ta Barbara Museum of Natural History, and Western Foundation of Vertebrate Zoology; I thank those who aid- woodpeckers. ed and abetted my data collecting forays to these collec- Second,the analysesperformed here, as well tions. JanetConley, Nancy Joste,Ron Mumme, Pam Wil- as the earlier studies of Ricklefs (1980) and liams, and especially Maria Pereyra helped with various Koenig (1984a), also involve several short- aspectsof data collecting,coding, and analysis.D. Bystrak comings. The tests performed are only indi- provided the breeding bird censusdata, and Occidental College provided accessto computer facilities. Numerous rect, some of the data (in particular, the helpful comments were provided by Jon Greenlaw and breeding bird surveys) are only rough approx- Dennis Paulson.Financial assistancecame from the Frank imations, and no alternative hypothesesfor the M. Chapman fund of the American Museum of Natural observed geographic patterns have been rig- History. Field work on the Acorn Woodpecker at Hastings Reservation has been supported by NSF grant orously tested (Koenig 1984a). In addition, as BSR8410809. pointed out by Hussell (1985) even the ful- fillment of the above predictions of Ashmole’s hypothesis cannot be considered definitive in- LITERATURE CITED asmuchas many environmental and biological AMERICANORNITHOLOGISTS UNION.’ 1983. Check-list of variables correlate with latitude, and therefore North American Birds, 6th ed. American Omithol- with clutch size as well. I agree with Hussell ogists’ Union, Washington, DC. AMERICANORNITHOLOGISTS UNION.’ 1985. Thirtv-fifth (1985) that evidence beyond the correlations supplement to the American Ornithologists’ Union detected here and in prior studieswill be nec- check-list of North American Birds. Auk 102:680- essary before firm conclusionsconcerning the 686. evolution of geographicvariation in clutch size ASHMOLE,N. P. 1961. The biology of certain terns. Ph.D. diss. Oxford Univ., Oxford, England. can be drawn. ASHMOLE.N. P. 1963. The reaulation of numbers of One approach that goesbeyond the analyses tropical oceanic birds. Ibis 153:458-473. presentedhere is the investigation of intersea- BOYCE,M. S. 1979. Seasonalityand patterns of natural sonal variation in clutch size as a function of selection for life histories. Am. Nat. 114:569-583. population density and food resources. Such BYSTRAK, D. 1981. The North American breeding bird survey. Stud. Avian Biol. 6~34-4 1. an analysis for the Acorn Woodpecker indi- COCKBURN, A., A. K. LEE, AND R. W. MARTIN. 1983. catesthat yearly variation in clutch size within Macrogeographicvariation in litter size in Antechinus a population correlates significantly with the (Marsupialia, Dasyuridae). Evolution 37:86-95. resourcesavailable for breeding per individual CODY,M. L. 1966. A generaltheory of clutch size. Evo- lution 20: 174-184. as indicated by the number and energetic con- GRANT, P. R., AND B. R. GRANT. 1980. The breeding tent of stored acorns per bird (Koenig and and feedingcharacteristics of Darwin’s fincheson Isla Mumme, in press).Because it involves a single Genovesa, Galapagos.Ecol. Monogr. 50:38l-410. 504 WALTER D. KOENIG

HUSSELL,D.J.T. 1985. Clutch size, daylength, and sea- rameters of Swallows Hirundo rustica and House sonality of resources:comments on Ashmole’s hy- Martins Delichon urbica. Omis. Stand. 15:43-54. pothesis. Auk 102:632-634. MOORE,W. S., ANDW. D. KOENIG. 1986. Comparative JACKSON,J. A. 1977. How to determine the status of a reproductive successof Yellow-shafted, Red-shafted woodpecker nest. Living Bird 15:205-221. and hybrid flickers across a hybrid zone. Auk 103: KOENIG,W. D. 1984a. Geographic variation in clutch 42-51. size in the Northern Flicker (Colaptes auralus): sup- RICKLEFS,R. E. 1980. Geographic variation in clutch port for Ashmole’s hypothesis.Auk 101:698-706. sizeamong passerine birds: Ashmole ’s hypothesis.Auk KOENIG,W. D. 1984b. Clutch size of the Gilded Flicker. 97:38-49. Condor 86:89-90. ROOT,R. B. 1967. The niche exploitation pattern of the KOENIG,W. D., ANDR. L. MUMME. In press. Population Blue-gray Gnatcatcher. Ecol. Monogr. 37:3 17-350. ecology of the cooperatively breeding Acorn Wood- ROSENZWEIG,M. L. 1968. Net primary productivity of pecker. Princeton Univ. Press, NJ. terrestrialcommunities: predictions from climatolog- LACK,D. 1947. The significanceof clutch size. I, II. Ibis ical data. Am. Nat. 102:67-74. 89:302-352. SELLERS,W. D. 1965. Physical climatology. Univ. Chi- LACK,D. 1948. The significanceof clutch size. III. Ibis cago Press, IL. 90:25-45. - THORNTHWAITEASSOCIATES. 1964. Average climatic LORD,R. D., JR. 1960. Litter size and latitude in North water balance data of the continents, parts V, VI, and American mammals. Am. Midl. Nat. 64:488-499. VII. Lab. Climatol. Publ. Climatol. 17:l-6 15. MBLLER,A. P. 1984. Geographictrends in breeding pa-

Also commonly known as “The laughing Jackass” this brrd IS well known for Its unrque ‘cry’ The common kookaburra and tts rarer cousin. the blue wlnged kookaburra can be found throughout coastal Australia Australian, unique and within reach! Yes, the Kookaburra and over 300 bird species unrque to Australia await you on a Monarch Blrdlng Tour. Monarch Blrdlng Tours take you,throughout Australia In search of some of the most diverse avlfauna to be found anywhere In the world It’s a brrdlng tour you’ll never forget

AUSTRALIANBIRDING TOURS