Nordic Society Oikos

The Socioecology of Body Size Variation in the Primitively Eusocial Sweat Bee, ligatus ( : ) Author(s): Miriam H. Richards and Laurence Packer Source: Oikos, Vol. 77, No. 1 (Oct., 1996), pp. 68-76 Published by: Wiley on behalf of Nordic Society Oikos Stable URL: http://www.jstor.org/stable/3545586 Accessed: 14-02-2016 03:43 UTC

REFERENCES Linked references are available on JSTOR for this article: http://www.jstor.org/stable/3545586?seq=1&cid=pdf-reference#references_tab_contents

You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/ info/about/policies/terms.jsp

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].

Nordic Society Oikos and Wiley are collaborating with JSTOR to digitize, preserve and extend access to Oikos.

http://www.jstor.org

This content downloaded from 130.63.180.147 on Sun, 14 Feb 2016 03:43:15 UTC All use subject to JSTOR Terms and Conditions OlKOS 77: 68-76. Copenhagen1996

The socioecologyof bodysize variationin theprimitively eusocial sweatbee, Halictusligatus (Hymenoptera: Halictidae)

Miriam H. Richardsand LaurencePacker

Richards,M. H. and Packer,L. 1996. The socioecologyof body size variationin the primitivelyeusocial sweat bee, Halictus ligatus(Hymenoptera: Halictidae). - Oikos 77: 68-76.

Patternsof demographicand social variationexhibited by the sweat bee Halictus ligatus,reveal that the expressionof eusocial colony organizationis associated with local environmentalconditions, harsher conditions promoting more classical eusocial- ityand gentlerconditions promoting diminished . We examineda variety of factorsaffecting body size variationof H. ligatusat a nestingaggregation near Victoriain southernOntario, Canada duringthe summersof 1984, 1990, and 1991. Body size variationwas associated withgender and reproductivecaste as expected, and therewere significant colony effects on size as well. An unexpectedresult was the dramaticeffect of local environmentalconditions on body size. Changes in weather patternsfrom year to yearresulted in significantbody size variationamong all classes of adult sweatbees, includingqueens, workers, gynes, and males. In 1990 cool, rainy weatherled to the productionof brood withrelatively small body sizes,while in 1984 and 1991, relativelydry, warm weatherhad the opposite effect.Weather probably influencedbrood body size by affectingthe abilityof adult femalebees to gather provisionsfor the larvae,and so indirectlyaffecting food availabilityto brood. Since queens are produceda yearbefore their own workers,environmental factors affecting theirrelative body sizes are uncorrelated.Nevertheless, the degreeof eusocial colony organizationis stronglyaffected by variation in the relativesizes of queens and workers,and this illustratesthe extentto whichstochastic environmental variation structuresopportunities for behavioural interactions in primitivelyeusocial bees. The patternof inversevariation between temperature and body size in thisbee contradicts the generalpattern of increasedbody size at highertemperatures generally seen in insectsand otherectotherms, and indicatesthat social insectscould be an exception to this rule.

M. H. Richardsand L. Packer,Dept of Biology,York Univ., North York, ON, Canada M3J 1P3 (presentaddress of MHR: Dept of Biological Sciences, Brock Univ.,St. Catharines,ON, Canada L2S 3A1).

For ectothermicorganisms, two major environmental adult body sizes (Ray 1960, Atkinson 1994, Berrigan influenceson body size are food availabilityand devel- and Charnov 1994). In specieswith parental care, adult opmental temperature.In ,the more food an body size may also be influencedby factorsaffecting individualconsumes during the larval stages,the larger the ability of adults to provision the young. Many it is as an adult. This apparentlysimple relationship is insectsare mass provisioners,providing a larva withall complicatedby the effectsof temperature.At higher the food it will consume before developing into an temperatures,insects may develop faster,and so move adult. In such species,there exists the opportunityfor fromone larval stage to the next more quickly,spend provisionersto preciselycontrol the size of theiroff- less timeas feedingjuveniles, and end up withsmaller springin responseto ecological factorssuch as season

Accepted20 February1996 Copyright?) OIKOS 1996 ISSN 0030-1299 Printedin Ireland - all rightsreserved

68 OIKOS 77:1 (1996)

This content downloaded from 130.63.180.147 on Sun, 14 Feb 2016 03:43:15 UTC All use subject to JSTOR Terms and Conditions Table 1. The influenceof weatheron the demographyand social behaviourof VictoriaH. ligatusin 1984, 1990,and 1991 (table slightlymodified from Richards and Packer 1995).

Year

1984 1990 1991 Temperature warm cool hot Rainfall below average heavy moderate Nest initiationand synchronous synchronous asynchronous firstbrood emergence Nest and brood survival good poor good No. firstbrood per nest 6.65 5.83 >9.0 Firstbrood sex ratio 14.6 11.5 5.5 (% male) No. workersper nest 5.65 5.16 8.5 Second brood emergence protandrous not protandrous protandrous Ovipositionpatterns dominatedby queens large contributionby workers Colony organization more eusocial more eusocial less eusocial and food availability,and social factorssuch as the survival,produced few workers, and queens dominated population-widesex ratio (Wrenschand Ebbert 1993). oviposition.Under more favourableconditions due to In social insects,body size is also an importantcorollary unusuallywarm temperatures,colony and brood sur- of reproductivecaste (Breed 1976, Packer and Knerer vival were high, large numbersof workerswere pro- 1985, Richards and Packer 1994), so environmental duced, and workerslaid a large proportionof eggs in influenceson body size are inextricablyentwined with the reproductivebrood. Such effectsmay also underlie those affectingdemographic and social variation. continent-wide,north-south dines of increasingcolony A seriesof recentstudies points to the role of envi- size and decreasingeusociality in H. ligatus.In southern ronmentalvariation in organizingthe behaviour of areas with long nestingseasons, there are very large primitivelysocial sweat bees (Wcislo in press). In sev- numbersof workersper nest,making queen controlof eral species,there is markedgeographic variation in the workerbehaviour less effectivethan in northernareas expressionof eusociality,with populations in harsh withshort nesting seasons wherethere are many fewer climates (e.g. high altitudes)reverting to solitarybe- workersper nest. haviour (Sakagami and Munakata 1972, Packer 1990, In H. ligatus and other primitivelyeusocial sweat 1991, Richards 1994a, b, Eickwort et al. in press). bees,queens controlworker behaviour through physical Variationin social behaviourand colony organization aggressionso large numbersof workersmake queen is also observed in sweat bees with less extremebe- controlof workerbehaviour less effective(Kukuk and haviouralflexibility. One exampleis Halictus ligatus,a May 1991). However,adult body size must also be an primitivelyeusocial species that has been studied importantfactor governing the outcomeof social inter- throughoutits geographicalrange, and in which soli- actions. Halictus ligatus queens have been observed tary populations have never been observed. Halictus pummellingworkers into the sides of nest tunnelsor ligatusexhibits strong variation in the degreeof euso- forcingthem out the entranceto forage (Packer and cial colony organization:harsher environmental condi- Richards unpubl.), and the larger size of the queens tionspromote more classicallyeusocial behaviour,and clearly confersan advantage in such confrontations. gentlerconditions lead to diminishedeusociality, trends Variationin body size would thereforebe expectedto evidentat both local and continentalscales (Kirkton contributesubstantially to variationin the expression 1968,Michener and Bennett1977, Richards and Packer of eusocial behaviour.Halictus ligatusexhibits consid- 1995, Richards et al. 1995). By diminishedeusociality, erable geographic variation in body size (Kirkton we mean that workers'reproductive efforts shift from 1968), bees fromnorthern populations being generally assistingthe queen to raise the queen's reproductive smallerthan bees fromsouthern populations. The dine offspring,to raisingtheir own reproductiveoffspring, in body size is very likelyrelated to the north-south includinggynes. dines of increasingcolony size and decreasingeusocial- Our studies of a population of H. ligatus near ity,and likewise,could also mirrorlocal patternsof size Toronto in southernOntario (Table 1; Richards and variation. Packer 1995, Richardset al. 1995) indicatethat annual In this study,we investigatevariation in adult body variationin local weatherconditions directly influences size and its social and demographiccorrelates and the demographyof sweat bee colonies, and indirectly consequencesas part of our overall studyof environ- influencesthe natureof social interactionsamong adult mentallyorganized social behaviourof H. ligatus.We females.Under adverseconditions caused by excessive foundthat sweat bee body size was stronglyinfluenced rainfall,sweat bee colonies experiencedpoor brood by local environmentalconditions at Victoria, poor

OIKOS 77:1 (1996) 69

This content downloaded from 130.63.180.147 on Sun, 14 Feb 2016 03:43:15 UTC All use subject to JSTOR Terms and Conditions weatherleading to smallerbody size and good weather cation as described in Richards and Packer (1994). to larger sizes, a patternthat mirrorsthe continental Pupae and larvae thathad consumedall theirfood were north-southline of increasingbody size. We also transferredto small chambersin wax-linedPetri dishes found that the pattern of body size variation was and raised to adulthood in the laboratory.Occasion- complicatedby familial,behavioural and reproductive ally, we also attemptedto raise medium-sizedlarvae factorsthat determinethe nature of eusocial interac- that had consumedonly about half theirpollen mass. tions in thisbee. Invariably,these larvae ceased feedingand almostall of themdied, but in a fewcases, theysuccessfully pupated, becomingvery tiny pupae. Althoughnone of thesetiny pupae wereincluded in our size analyses,this shows the amount of food consumed Methods directrelationship between as a larva and finaladult body size. The lifecycle of H. ligatusnesting at this aggregation Adult bees collectedin the fieldwere placed in sepa- near Victoriain southernOntario is summarizedhere; ratemicrocentrifuge tubes on ice. Subsequentstorage of furtherdetails may be found in Packer (1986a, b), all bees was in separate microcentrifugetubes at Packer and Knerer (1986), and Richards and Packer -80'C. Assessmentof female caste was based on a (1995). Extensivenest excavationswere conducted at combination of body size (workers are invariably this aggregation from June to September of 1984 smallerthan the queens in theirown nest),observations (Packer 1986a, b, Packer and Knerer 1986), 1990 and of foraging activity by individually marked bees 1991 (Richardsand Packer 1995, Richardset al. 1995). (foundress-queensforage in spring,workers forage in Mated foundressesemerge from their overwintering summer,and gynesdo not forage),degree of mandibu- hibernaculato establishnests in late May or earlyJune. lar wear (an indicatorof diggingactivity, engaged in Most nestsare foundedby a singlefemale but thereare mainlyby foundressesand workers),whether or not a occasional multifoundressassociations. Foundresses femalehad mated (foundress-queensare always mated forage for several weeks to provisionthe firstbrood, but many workers are not), and degree of ovarian whichis composed mainlyof small femalesdestined to development(queens and some workershave highly become workers,and a fewmales. The workersforage developedovaries, but gynesnever do). In nestswithout and provisionthe second or reproductivebrood, which foundress-queens,replacement queens wereidentified as is composed of males and females technicallycalled worker-sizedfemales with a much higher degree of gynes.After emergence, the gynes mate, dig overwinter- ovarian developmentthan any of theirnestmates. ing hibernaculabeneath the natal nest, and enterdia- Head widths(HW) of adults and pupae were mea- pause preparatoryto becomingthe next generationof suredas the distanceacross the widestpart of the head foundressesthe followingspring. (includingthe compound eyes). Wing lengths were mea- Climate data were obtained fromthe Environment sured as the longestlinear distancefrom the point of Canada meteorologicalstation at Lester B. Pearson attachmentto the wing tip. Badly damaged wings(i.e. InternationalAirport, about 15 km fromthe fieldsite. those so severelyfrayed or torn that the wing tip was Details of summerdegree-day accumulation and precip- gone) or ones that were not fullyexpanded were not itationfor the years 1983-1984 and 1989-1991 may be measured. Head measurementswere made with a found in Richards and Packer (1995). Summariesof binocular microscopeaccurate to 0.05 mm and wing variationin the weatherand previouslypublished pat- measurementswith an ocular micrometeraccurate to ternsof demographicand social variationat Victoria 0.1 mm. In 1990 and 1991 foragingbees were caught, are shown in Table 1. marked,and theirhead widthsmeasured in the field Only single-foundressnests are includedin thisstudy. with the ocular micrometer.Microscope and ocular Depending on the time of excavation,nests contained micrometermeasurements agreed to within0.1 mm and queens, workers,newly emerged adult brood in their are consideredtogether. The proportionalsize differ- brood cells or in hibernacula(gynes only), pupae, or ence betweenqueens and workerswas calculatedas larvae. Sweat bees are mass provisioners,constructing a mass of pollen and nectarwhich is the sole source of (queen HW - workerHW)/queen HW food forthe developinglarva. A larva does not pupate until it has consumed its entirepollen mass, so the amountof food supplieddirectly influences adult body Statisticalanalyses size (but see Richards and Packer 1994). Undamaged pollen masses were collected in separate microcen- We used Model I (OrdinaryLeast Squares) regression trifugetubes and storedin ice. Gyne-producingpollen to examine the effectsof several variables on size. masses were identifiedby theirlarge size and peculiar Model II (Reduced Major Axis) regressionis oftenused saddle shape (Boomsma and Eickwort1993, Richards for morphometricdata, but Model I was justifiedfor and Packer 1994). Dry weightswere obtainedby desic- severalreasons (Sokal and Rohlf 1981): a) causal rather

70 OIKOS 77:1 (1996)

This content downloaded from 130.63.180.147 on Sun, 14 Feb 2016 03:43:15 UTC All use subject to JSTOR Terms and Conditions than associativerelationships were examined,b) mea- Queens foundin 1991 are fromthe same groupof adult surementerror was not expectedto scale withsize and femalesas gynes found in 1990 (see below). In 1984 was equal among all classes of individuals,and c) Type Packer found that small foundresses(HW < 2.6 mm) II regressionis not available for analysisof covariance were less likelyto establishnests successfullyand did (ANCOVA) or multipleregression. not surviveas long as large foundresses.However, in We investigatedthe effectsof a varietyof indepen- neither 1990 or 1991, did the average size of dentvariables, including size and numberof foragersin foundresseschange fromearly to mid-summer(Table a nest,year, and nestof origin,on head widthvariation 3A). Moreover, in 1991, foundressesthat failed to among queens,workers, gynes, and males. Workersize produceany brood wereactually somewhat larger than was the mean size of the workersin a colony and successfulfoundresses, although not significantlyso workernumber was the numberof live workersat the (Table 3B). We conclude that thereis no unequivocal timethe colonywas excavated.The independenteffects evidencethat foundress head widthaffects the probabil- of queen and worker size were analysed separately ity of nest failure. because the two variables were highlycorrelated (see Results). When more than one bee fromthe same nest was available (as for workers,gynes, and males), we Workersize used nestedANOVA, nestingthe variable nest within years. Statisticaltesting of the Mean Square (MS) for The largestworkers were produced in 1984 by large the grouped variable (in this case, year) in an unbal- queens underfavourable environmental conditions (Ta- anced nested ANOVA is difficult(Sokal and Rohlf bles 2 and 4). In 1990,small workerswere produced by 1981). An approximatesignificance test is obtainedby large queens under poor environmentalconditions. In comparingMSgroups/MSsubgroupsq or in thiscase, MSyear/ 1991, small workerswere produced by small queens MSnest- under favourableenvironmental conditions. In single- In order to estimatehow much of the variance in foundressnests, the workerswere invariablysmaller adult body size could be ascribedto annual or colony- than the queen that produced them,but the size of of-origin(nest) effects,we used nestedANOVA. How- workerswas stronglycorrelated with the size of the ever, the estimatedvalues should be interpretedwith queen (r = 0.49, n = 150,p < 0.0001). Year-to-yearvari- caution since theyare based on unbalanceddesigns. ation accounts for 10.7% of the variance in worker Whenevert-tests were used to compare the mean head width,while queen and nest effectstogether ac- sizes of two groups,F-tests for homogeneityof vari- count for 30.2% (nested ANOVA, n = 379). We used ances were also done. In all comparisons,the variances ANCOVA to partitionthe effectsof variationdue to were not significantlydifferent (p > 0.05). the queen-workersize correlation,further variation as-

Table2. Averagesizes of queens, workers, gynes, and males in Results 1984,1990, and 1991.Mean head widths were compared using the Student-Newman-Keulstest (c = 0.05) forannual com- The average sizes of foundresses(queens), workers, parisonswithin a sex or caste.Significant differences among yearsare indicatedby 'S' (small)or 'L' gynes,and males in nestsexcavated in 1984, 1990, and (large). 1991 are shownin Table 2. Note thatthe queens in this Sex or caste Year Head width Winglength sample were producedthe year beforethey were exca- mean+ sd (n) mean+ sd (n) vated, that is duringthe summersof 1983, 1989, and Queens1 1984L 2.92+ 0.20 (33) - 1990. At the simplestlevel (analysed with one-way 1990L 2.96 + 0.21 (33) - ANOVA, Table 2), thereare highlysignificant changes 1991S 2.74+0.19 (40) - in the averagesizes of adult bees fromyear to year.The Workers2 1984 L 2.58 + 0.11 (136) - 1990 S 2.53 + 0.17 (83) 5.60 + 0.29 (12) sizes of the larval provisionmasses that produce them 1991 S 2.48 +0.18 (159) 5.43 +0.28 (59) may also change fromyear to year. Gyne-producing Gynes3 1984 L 2.97 + 0.20 (5) - pollen masses were significantlylarger in 1991 (dry 1990 S 2.74 + 0.17 (46) 6.06 + 0.41 (13) weightmean =41.3 mg, sd = 2.8, n = 13) than in 1990 1991 L 2.90 + 0.18 (213) 6.31 + 0.37 (106) Males4 1984 S 2.17 + 0.13 (9) - (mean 34.9, sd= 7.8, n = 34; t = 2,271, df=45.0, p < 1990 L 2.29 + 0.15 (57) 5.76 + 0.39 (24) 0.05). Too few worker and male-producingpollen 1991 L 2.33 +0.16 (184) 5.88 +0.35 (89) masses were collectedfor statisticalcomparison. 'ANOVA: F= 13.26, df= 2, 103, p =0.0001; SNK: MSE= 0.040. 2ANOVA: F= 18.53,df = 2, 363,p = 0.0001;SNK: MSE = 0.023. Queen size 3ANOVA: F= 14.31,df = 2, 239,p = 0.0001;SNK: MSE= 0.033. The largestqueens were foundin 1984 and 1990,while 4ANOVA: F=4.89, df=2, 229, p=0.0083; SNK: MSE= thosefound in 1991 weresignificantly smaller (Table 2). 0.024.

OIKOS 77:1 (1996) 71

This content downloaded from 130.63.180.147 on Sun, 14 Feb 2016 03:43:15 UTC All use subject to JSTOR Terms and Conditions Table 3. Two methodsof comparingfoundress size and nestingsuccess in H. ligatus.A. Comparisonof head widthsof foundressesfound in activenests during the period of foragingby foundresses (weeks 5-7) and followingthe emergence of the firstworkers (weeks 11-14). Althoughthe mean size increasedin bothyears, the changes were not significant.B. Effectof foundresssize on probabilityof totalnest failure (failure to produceany brood at all) in 1991,based on a sampleof nests initiatedin earlyspring. Packer (1986a, b) definedfemales with HW < 2.6 mmas small;here we also use 2.8 mmas a cut-off point.Log-likelihood X2 (G) testscompare the frequencies of failedand successfulfoundresses. Year Earlynests Mid-summernests Comparisonof means(t) and mean+ sd (n) mean+ sd (n) variances(F) 1990 2.93+ 0.19 (22) 3.02+ 0.21 (6) t -2.02, df= 26, ns F= 1.12, df= 5,21, ns 1991 2.71 +0.18 (22) 2.79+0.24 (12) t=-1.07, df= 32, ns F= 1.82, df= 11,21,ns

Foundresssize Failed Successful Comparison (mm) HW<2.6 3 5 G=0.348, df= 1, ns HW>2.6 8 22

HW<2.8 6 15 G= 0.003, df=l, ns HW >2.8 5 12

Mean HW + sd 2.76 + 0.23 2.72 + 0.17 t = 0.667,df = 36,ns

cribableto beingraised in a commonnest, and environ- and workerswere small. Gynes exhibitedthe greatest mental variationascribable to year effects,and found amount of annual variation in size (25.2%/oof total thatall thesefactors contribute significantly to variance variance) and a similaramount of inter-nestvariation in workerhead width(Table 4). (27.2%, nested ANOVA, n = 264). Depending on Since the annual pattern of worker size variation whetherqueens or workersdetermine the sizes of larval differedfrom that observedfor queens (Table 2), the provisionmasses, gyne size variationcould resultfrom proportionalsize differencebetween them also changed variationin queen size, workersize, or the numberof from year to year (1984: 12.7%, 1990: 15.8%, 1991: foragerscontributing to a pollenmass (estimatedby the 11.4%). Size dimorphismwas significantlylower in 1991 number of workersat the time of nest excavation). than in either 1984 or 1990 (one-way ANOVA: F= These possibilitiesare examinedin Table 5. In general, 3.67, df= 2,18, p <0.05; this statisticis based on a nests with large queens or large workerstend to pro- sample in whichone queen-workerpair was randomly duce large gynes,but this effectis modifiedby year. drawn fromeach nest with two or more workersin orderto avoid biases caused by large nests).The great- est size differenceoccurred in 1990, the year with the Table5. ANCOVAtables for factors contributing to variation poorestweather conditions, and the smallestdifference in head widthof gynesfrom 1984, 1990, and 1991.Nested analysisof covariance models including the effects of queen or in 1991, the year withthe best weatherconditions. workerhead width are considered separately because these two variablesare highlycorrelated. ANCOVA 1: nestedmodel includingthe effect of queensize, year, and nesteffects not attributabletoqueen HW. ANCOVA 2: nestedmodel including Gynesize theeffect of mean worker size, number of workers per colony, year,and nesteffects not attributable to workerHW. The largestgynes were produced in 1984 when both queens and workerswere also large and weathercondi- Source df MS F p tionswere In favourable. 1991 weatherconditions were ANCOVA 1 (Y = gyneHW) 21 0.095 4.70 0.0001 also veryfavourable and gynesproduced that year were QueenHW 1 0.162 8.03 0.0068 almost as large as in 1984 despitethe fact that queens Year 2 0.288 14.26 0.0001 Nest(year)1 18 0.070 3.45 0.0003 Error 47 0.020 Table4. Analysisof covariancetable for factors contributing to variationin workerhead width in 1984,1990, and 1991. ANCOVA 2 (Y =gyne HW) 37 0.107 4.93 0.0001 Mean workerHW 1 0.512 23.63 0.0068 Source df MS F p No. workers 1 0.127 5.852 0.0168 Year 2 0.254 11.71 0.0001 Model (Y = workerHW) 56 0.057 3.71 0.0001 Nest (year)1 33 0.085 3.92 0.0001 Queen HW 1 1.168 75.84 0.0001 Error 143 0.085 Year 2 0.117 10.89 0.0001 Nest (year) 53 0.032 2.08 0.0003 l Nestedvariable (pun not intended). Error 207 0.015 2 The correlationbetween gyne head width and workernum- beris negative.

72 OIKOS 77:1 (1996)

This content downloaded from 130.63.180.147 on Sun, 14 Feb 2016 03:43:15 UTC All use subject to JSTOR Terms and Conditions Table6. ANCOVAtables for factors contributing to variation Ontario.Ecological factorsaffecting both body size and in head widthof malesfrom 1984, 1990, and 1991.Nested ANCOVAmodels including the effects of queen or worker head the degree of queen-workersize dimorphisminclude widthare considered separately because these two variables are temperature,rainfall, and food availability.The latter highlycorrelated. ANCOVA 1: nestedmodel including the also counts as a social factoraffecting body size, be- effectof queensize, year, and nesteffects not attributable to cause both caste and genderinfluence the total amount queenHW. ANCOVA2: nestedmodel including the effect of meanworker size, number of workers per colony, year, and nest of food with which any particularlarva will be pro- effectsnot attributable to workerHW. vided. Anothersocial factoris the tendencyfor size to be correlatedamong nestmates. These resultsare impor- Source df MS F p tant forseveral reasons. First,they illustrate that there ANCOVA 1 (Y = male HW) 25 0.038 2.12 0.0094 can be a large amount of environmentallybased varia- QueenHW 1 0.015 0.83 ns tion in body size, even in an insectwith parental care in Year 2 0.034 1.92 ns Nest(year)' 22 0.039 2.20 0.0087 whichjuveniles do not have to foragefor themselves. Error 59 0.018 Second, they illustratethe major role environmental factorsplay in structuringopportunities for social inter- ANCOVA2 (Y = maleHW) 44 0.043 2.20 0.0004 action in insects.Third, they contrast with results from Mean workerHW 1 0.131 6.76 0.0105 No. workers 1 0.030 6.03 ns diverseectotherms, including insects, showing that cold Year 2 0.117 6.03 0.0032 temperaturesresult in maturationlater and at a larger Nest(year)' 40 0.037 1.91 0.0036 size compared to developmentat highertemperatures. Error 125 0.019 We discuss theseissues below, emphasizingthe role of 1 Nestedvariable. the environmentin determiningphenotypic variation in size and the consequencesof thisvariation for conflicts Gyne size was negativelyassociated with the numberof betweenqueens and workersand among workersover live workersin a nest at the timeof excavation. reproductiveopportunities in the nest. Foundresses and gynes are membersof the same Annual variationin body size of H. ligatusnesting at reproductivecaste, so comparisonsof their sizes are Victoria mirroredthe patternsof demographicand instructive.In 1984, under favourableweather condi- social variationobserved among the bees at this site tions and when foundresseswere large,the gynespro- (Richards and Packer 1995, Richards et al. 1995). In duced were similarin size to the foundresses(t = 0.625, 1984,when both temperaturesand rainfallwere moder- df= 37, ns). In 1990, when weatherconditions were ate and average, bees experiencedrelatively high nest poor but foundresseswere again large,the gyneswere and brood survival,brood body sizes were generally considerablysmaller than the foundresses(t = 5.2357, large,and the degreeof queen-workersize dimorphism df= 77, p < 0.0001). These 1990 gynes became the was intermediate.In 1990, when excessiverainfall and foundressesof 1991. In the warm summerof 1991, the cool temperaturesresulted in poor nest and brood foundressesproduced gynes significantlylarger than survival,brood body size was generallysmall, the degree themselves(t = 5.08, df= 251, p < 0.0001). The size of of queen-workersize dimorphismwas high,and queens gynesdid not seem to be associatedwith overwintering dominated ovipositionof the reproductivebrood. In success: 1990 gynesdid not differin size fromthe 1991 1991,when unusually warm temperatures and moderate generationof foundresses(means: t = 0.078, df= 84, ns; rainfallcombined to produce good nest and brood variance:F= 1.15, df= 39,45, ns). survival,large brood body size, and low queen-worker size dimorphism,workers significantly increased their share of reproductivebrood oviposition.Clearly, local Male size environmentalvariation is an importantextrinsic orga- In males, annual effectsaccounted for only 7.4% of the nizer (Wcislo in press) of colony social interactions variationin head widthwith 20.7% originatingin differ- mediatedby changesin body size and demography. ences between nests (nested ANOVA, n = 250), an The strongeffects of local weathervariation on H. amount smaller than seen in any group of females ligatusbody size were not unexpected:the patternat (Table 2). In fact,the annual effectmay be illusory,a Victoriaparalleled the known north-southdine of in- statisticaleffect of the small 1984 sample (n = 9, but creasing body size across North America (Kirkton compare ANCOVAs 1 and 2, Table 6). Male size was 1968). In insectsgenerally, adult body size is a function not correlatedwith queen size or workernumber but of the quantityand quality of food consumedduring was significantlyassociated withworker size (Table 6). the larval stages. The dramaticeffect of food quantity on body size in H. ligatusis observed in half-grown larvae that cease feedingin the laboratoryand meta- Discussion morphoseinto tinypupae about halfthe size of normal pupae. However,the effect of food quantityand quality In thispaper we reportsignificant variation in body size is complicatedby the effectsof larval developmental in all fourclasses of adult Halictus ligatusin southern temperatures.At highertemperatures, insects generally

OIKOS 77:1 (1996) 73

This content downloaded from 130.63.180.147 on Sun, 14 Feb 2016 03:43:15 UTC All use subject to JSTOR Terms and Conditions develop faster(i.e. theychange from one larval stageto poor weatherconstrains foraging time, less pollen and the next more quickly)and spend less time as feeding nectar are gatheredand fewer,smaller bees are pro- juveniles,thus producingsmaller body sizes at higher duced. Apparentlythis is what happened duringthe temperatures(Ray 1960, Atkinson1994, Berriganand excessivelyrainy summer of 1990,when a generationof Charnov 1994). This is the opposite of the patterns large-sized foundressesproduced workers, and then observed in H. ligatus at Victoria or across North gynes,that were very small in size. The rain also caused America.Can these observationsbe reconciled? high rates of workerbrood mortality(Richards and We have no informationabout the physiological Packer 1995). The smallgynes of 1990 became thesmall efficienciesor metabolicrates of juvenileH. ligatus,but queens of 1991, but unusuallywarm and relativelydry highertemperatures apparently increase developmental weatherallowed themto producemany, comparatively rates:in 1991, unusuallywarm temperaturesled to the large-sizedworkers, and eventually,large-sized gynes as appearance of workerbrood pupae about 3.5 weeks well. A similarnegative effect of poor weatheron body earlierthan expected (Richards and Packer 1995). In the size has been observedin a populationof Victoria aggregation,brood develop undergroundat cinctipesnesting in a marginalenvironment in Nova depths of about 5-25 cm (Packer and Knerer 1986), Scotia (Packer et al. 1989). Interestingly,Kamm (1974) earlierbrood generallybeing closer to the surfacethan foundthat lower temperatures led workersof a labora- laterbrood. While beingunderground would undoubt- torypopulation of Lasioglossumzephyrum to construct edlyhave mitigatedvariation in temperature,this would larger brood cells with larger provision masses, but have been less so for the workerbrood than for the these bees would have been shelteredfrom severe reproductivebrood sinceworker brood cells werecloser weathersuch as rain. to the soil surface. So it seems that developmental The year-to-yearpattern of size variationin males temperatureper se did have the expectedeffect on H. was slightlydifferent from that observed in females,the ligatus,that is to increasedevelopmental rate. In most smallestmales being found in 1984 when the largest insects,the larvae are independentand mustforage for workersand gyneswere produced.The lower variance themselves,and would have less timeto do so if higher of male size may indicate as yet undiscoveredsize temperaturescause themto reachthe stage of metamor- constraintson males, especiallyif male body size is phosis more quickly. In mass provisionerssuch as positivelycorrelated with sperm production(Berrigan sweatbees, thelarvae are suppliedwith a predetermined and Locke 1991) or if it affects sustained flight amount of food beforethe egg is even laid. Since they efficiency.Alternatively, the sizes of males may have always consumethe entirefood mass beforepupation, more to do with sex allocation decisionsmade by the developmentaltemperatures do not affectthe amount of queen (Richardset al. 1995). Patternsof sex allocation food theyconsume. In fact,the amount of food con- are predictedto changein responseto the same environ- sumed is influencedby weather conditions affecting mental influencesthat affectworker productivity and foragingfemales. Thus insectsthat mass-provision their may reflectthe efforts of thequeen to bias thereproduc- brood, and possiblyothers that exhibitparental care, tive brood sex ratio towardmales in the period before may be exceptions to the general rule that higher she loses controlto the workers(Richards et al. 1995). temperatureslead to decreased body size in insects, At Victoria,H. ligatusfrom the same nest tendedto because juvenilesdo not have to findtheir own food. be similar in size. Foundresses constructthe pollen The influenceof weatheron H. ligatusbody size was masses that produce workers and first-broodmales, probablydetermined by directeffects on foragingand while workersforage for and probablyalso construct brood provisioningby adult females.The influencesof reproductivebrood provision masses. It is not clear summer temperaturesand weather are confounded, whetherthe queens or the workersare the finalarbiters sincein southernOntario, rainy weather is oftenassoci- of pollen ball size. This is importantbecause correla- ated with lower temperaturesand sunnyweather with tions between the sizes of foragersand brood could highertemperatures. Sweat bees do not forageduring result eitherfrom geneticcorrelation of parents and rainyweather and at Victoria,H. ligatuswere incapable offspringor simplyrepresent a scaling effectresulting of flyingwhen air temperaturesdropped below 14'C from foragersusing their own bodies as a basis for (Richards 1994a), but above this threshold,lower air measuringthe sizes of theprovision masses. It would be temperaturesprobably did not interferewith sweat bee difficultto distinguishbetween these two causes of foragingactivity. Female bees are able to exertprecise correlationexcept in a laboratoryexperiment in which controlover the amount of pollen and nectarin larval thesizes of provisionmasses could be artificiallymanip- provisionmasses, which show consistentdifferences in ulated (Thomas 1993). The fact that male body size size and shape,depending on whetherthey will produce correlateswith worker size but not queen size may workers,gynes or males (Boomsma and Eickwort1993, indicatethat workers make thedecisions about the sizes Richardsand Packer 1994). Since the pollen and nectar of reproductivebrood pollen masses, although it is for a single provisionmass are collected in one day likelythat queens controldecisions about the sex of the (Richards and Packer unpubl.), we suspectthat when brood (Richards et al. 1995). In any case, there is a

74 OIKOS 77:1 (1996)

This content downloaded from 130.63.180.147 on Sun, 14 Feb 2016 03:43:15 UTC All use subject to JSTOR Terms and Conditions strongtendency for large H. ligatusqueens to produce sioglossumzephyrum, large bees also tend to be domi- large workersthat in turn produce large gynes and nant to smallerbees, and large workersare more likely males (with fewer data, Packer and Knerer [1986] to lay eggs than small workers,apparently because the reached a differentconclusion, because theyfound no queens findsmall workers easier to control(Kukuk and correlationbetween forager size and brood cell volume, May 1991). All these observationssupport the con- which is correlatedwith brood head width [Kumar tention that the degree of size dimorphismbetween 1975]). queens and workersor subordinatesis an important Frank and Crespi (1989) have suggestedthat when determinantof the strengthof eusocial colony organi- there are more workersin a nest, they cooperatively zation (Breed 1976, Packer and Knerer 1985). producelarger gynes. If thiswere true, it would suggest Year-to-yearvariation in the degreeof queen-worker that largernumbers of workersin 1991 would explain size dimorphism is essentially stochastic, because largergyne body sizes that year. However, aftercor- queens and workersare produced in differentyears, rectingfor size similarityamong nestmates,we actually and environmentalconditions affecting their sizes are found an inverserelationship between worker number uncorrelated.Why don't queens simplymake workers and gynesize, and Packer and Knerer(1986) foundno that are sufficientlysmaller than themselvesto domi- correlationbetween worker number and the volume of nate easily?If workersize is associatedwith ergonomic gyne brood cells. Frank and Crespi's hypothesisde- efficiency,predation risk, longevity,or success as a pends on the additional assumptionthat reproductive replacementqueen, then there may be a minimum brood ovipositionis monopolized by queens, a situa- workersize below whichthere is a negativeimpact on tion which did not occur at Victoria in 1991. Many thenest's second brood productivity.Even in nestswith gyneswere the daughters,rather than the sisters,of the viable queens, workersmay still produce substantial workersthat produced them, so the workersin a large portions of the reproductivebrood when the rate at nestwould have been cooperatingto provisiona brood which larval provisions are collected outstrips the of daughters,sisters, and nieces. Furthermore,it ap- queen's egg-layingability (Richards and Packer 1995). pears that in some colonies with large numbers of It may be more of a disadvantagefor small queens to workers,they interferewith each other or that some make workersso small that they can easily be con- workersare more inclined to be "lazy" (Eshel and trolled,than to allow them a share of reproductive Motro 1988, Richards unpubl.). Finally,it is possible brood oviposition. that our estimateof workernumber is not accurate For temperatesweat bees, local weatherpatterns are sincewe used the numberof workersalive when a nest unpredictable,especially over the course of an entire was excavated,rather than the actual numberof work- summeror fromyear to year, but they have drastic ers provisioningparticular brood. effectson nestingsuccess, brood productivity,and body We had expectedto findthat in femalesweat bees, size, all of which impinge directlyon colony social specificallyfoundresses, large body size would be asso- organization.In the face of such unpredictability,Hal- ciated withenhanced survival or reproductivity(Honek ictus ligatus has maintained a high degree of be- 1993). We found no evidence for this prediction,al- haviouralflexibility in whichthe expressionof eusocial thoughthe selectiveeffects of body size could be more behaviouris continuouslyadjusted in responseto the subtle than detectablefrom our data. Similar results harshness of local environmentalconditions. Under have been found in studies of the blue orchard bee poor conditionsand at higherlatitudes, eusocial be- Osmia lignaria(Tepedino and Torchio 1989). Perhaps haviour is more stronglyexpressed than under fa- there is no selectiveadvantage of large body size in vourableconditions or at lower latitudes. females,except as it relates to social aspects of caste (overall body size mightbe determinedby inter-rather Acknowledgements- We thank Sandy Jackson,Diana than intraspecificcompetition [J. Seger pers. comm.]). Richards,Adonis and DimitriSkandalis, and especiallyBrian Since thereis broad overlapin the size distributionsof Feldmanfor able and cheerfulfield assistance. We also thank and David Berrigan,Art Woods,Jon Seger,Jay Evans, Vicky queens workers(Packer 1986a, b, Richards and Rowntree,and JoelShore for helpful and constructivecriti- Packer 1994), body size per se does not determinea cism.This study was supported by Canadian National Science female'sreproductive or social status.Relatively small and EngineeringResearch Council grants to bothauthors and femalesare more susceptibleto manipulationby large a SigmaXi grantto MHR. females(Kukuk and May 1991), but reproductiveop- portunitiesare also affectedby females'physiological conditionand the social milieu of the nest (Richards and Packer 1994). In short,large H. ligatusfemales References often,but not always,become foundresses,which often Atkinson,D. 1994. Temperatureand organismsize - a biolog- but not always,monopolize oviposition at the expense ical law for ectotherms?- Adv. Ecol. Res. 25: 1-58. Berrigan,D. and Locke, S. J. 1991. Body size and male of subordinatefoundresses (Packer 1986b) and workers reproductiveperformance in the fleshfly, Neobellieria bul- (Richards et al. 1995). In laboratorycolonies of La- lata. - J. Physiol.37: 575-581.

OIKOS 77:1 (1996) 75

This content downloaded from 130.63.180.147 on Sun, 14 Feb 2016 03:43:15 UTC All use subject to JSTOR Terms and Conditions - and Charnov,E. L. 1994. Reaction normsfor age and size turein sweatbees of the subgenusEvylaeus (Hymenoptera; at maturityin responseto temperature:a puzzle for life Halictidae): a phylogeneticapproach. - Behav. Ecol. So- historians.- Oikos 70: 474-478. ciobiol. 29: 153-160. Boomsma, J. J. and Eickwort.G. C. 1993. Colony structure, - and Knerer,G. 1985. Social evolutionand its correlatesin provisioningand sex allocation in the sweat bee Halictus bees of the subgenusEvylaeus (Hymenoptera; Halictidae). ligatus(Hymenoptera: Halictidae). - Biol. J. Linn. Soc. 48: - Behav. Ecol. Sociobiol. 17: 143-149. 355-377. - and Knerer,G. 1986. An analysisof variationin the nest Breed,M. D. 1976. The evolutionof social behaviorin primi- architectureof Halictus ligatusin Ontario. - InsectesSoc. tivelysocial bees: A multivariateanalysis. - Evolution30: 33: 190-204. 234-240. - , Jessome,V., Lockerbie,C. and Sampson, B. 1989. The Eickwort,G. C., Eickwort,J. M., Gordon, J. and Eickwort, phenology and social biology of four sweat bees in a M. A. In press.Solitary behavior in a high-altitudepopula- marginal environment:Cape Breton Island. - Can. J. tion of the social sweat bee (Hy- Zool. 67: 2871-2877. menoptera:Halictidae). - In: Crespi, B. and Choe, J. C. Ray, C. 1960. The applicationof Bergmann'sand Allen's rules (eds), Social competitionand cooperationin insectsand to the poikilotherms.- J. Morphol. 106: 85-108. arachnids: II. Evolution of Sociality. Princeton Univ. Richards,M. H. 1994a. Social responsesto changingenviron- Press, Princeton,NJ. ments: Reproductivebehaviour and reproductiveoptions Eshel, I. and Motro,U. 1988. The threebrothers' problem: kin in a primitivelyeusocial sweat bee. - PhD thesis,York selectionwith more than one potentialhelper. 1. The case Univ., North York, Ontario. of immediatehelp. - Am. Nat. 132: 550-566. - 1994b. Social evolutionin the genus Halictus: A phyloge- Frank, S. A. and Crespi, B. J. 1989. Synergismbetween neticapproach. - InsectesSoc. 41: 315-325. sib-rearingand sex ratio in Hymenoptera.- Behav. Ecol. - and Packer, L. 1994. Trophic aspects of caste determina- Sociobiol. 24: 155-162. tion in a primitivelyeusocial sweat bee. - Behav. Ecol. Honek, A. 1993. Intraspecificvariation in body size and Sociobiol. 34: 385-391. fecundityin insects: a general relationship.- Oikos 66: - and Packer, L. 1995. Annual variation in survival and 483-492. reproductionof the primitivelyeusocial sweatbee, Halictus Kamm, D. R. 1974. Effectsof temperature,day length,and ligatus (Hymenoptera:Halictidae). - Can. J. Zool. 73: numberof adults on the sizes of cells and offspringin a 933-941. primitivelysocial bee (Hymenoptera:Halictidae). - J. - , Packer, L. and Seger, J. 1995. Unexpectedpatterns of Kans. Entomol. Soc. 47: 8-18. parentageand relatednessin a primitivelyeusocial bee. - Kirkton,R. M. 1968. Biosystematicanalysis of variationof Nature 373: 239-241. Halictus (Halictus) ligatusSay (Hymenoptera,Halictidae). Sakagami, S. F. and Munakata, M. 1972. Distributionand - PhD thesis,Purdue Univ., Michigan. bionomicsof a transpalaearcticeusocial halictinebee, La- Kukuk, P. F. and May, B. 1991. Colony dynamics in a sioglossum(Evylaeus) calceatum,in northernJapan, with primitivelyeusocial halictinebee Lasioglossum(Dialictus) referenceto its solitarylife cycle at highaltitude. - J. Fac. zephyrum(Hymenoptera: Halictidae). - Insectes Soc. 38: Sci. Hokkaido Univ. Ser. VI Zool. 18: 411-439. 171-189. Sokal, R. R. and Rohlf, F. J. 1981. Biometry,2nd ed. - Kumar, S. 1975. Relationsamong bee size, cell size, and caste, Freeman,New York. in Lasioglossumzephyrum (Hymenoptera, Halictidae). - J. Tepedino,V. J. and Torchio,P. F. 1989. Influenceof nesthole Kans. Entomol. Soc. 48: 374-380. selectionon sex ratio and progenysize in Osmia lignaria Michener, C. D. and Bennett, F. D. 1977. Geographical propinqua(Hymenoptera: Megachilidae). - Ann. Entomol. variation in nestingbiology and social organizationof Soc. Am. 82: 355-360. Halictus ligatus.- Univ. Kans. Sci. Bull. 51: 233-260. Thomas, R. H. 1993. Ecology of body size in Drosophila Packer, L. 1986a. The social organisationof Halictus ligatus buzzatii: untanglingthe effectsof temperatureand nutri- (Hymenoptera;Halictidae) in southernOntario. - Can. J. tion. - Ecol. Entomol. 18: 84-90. Zool. 64: 2317-2324. Wcislo,W. T. In press.Behavioral environments of sweatbees - 1986b. Multiplefoundress associations in a temperatepop- (Halictinae)in relationto variabilityin social organization. ulationof Halictusligatus. - Can. J. Zool. 64: 2325-2332. - In: Crespi,B. and Choe, J. C. (eds), Social competition - 1990. Solitary and eusocial nests in a population of and cooperationin insectsand arachnids:II. Evolutionof Augochlorellastriata (Provancher) (Hymenoptera; Halicti- Sociality.Princeton Univ. Press,Princeton, NJ. dae) at the northernedge of its range. - Behav. Ecol. Wrensch, D. L. and Ebbert, M. A. 1993. Evolution and Sociobiol. 27: 339-344. diversityof sex ratio in insectsand mites.- Chapman and - 1991. The evolutionof social behavior and nest architec- Hall, New York.

76 OIKOS 77:1 (1996)

This content downloaded from 130.63.180.147 on Sun, 14 Feb 2016 03:43:15 UTC All use subject to JSTOR Terms and Conditions