Resource Allocation, Brood Production and Cannibalism During Colony Founding in the Fire Ant , Solenopsis Invicta Author(S): Walter R

Resource Allocation, Brood Production and Cannibalism during Colony Founding in the Fire Ant , Solenopsis invicta Author(s): Walter R. Tschinkel Source: Behavioral Ecology and Sociobiology, Vol. 33, No. 4 (1993), pp. 209-223 Published by: Springer Stable URL: http://www.jstor.org/stable/4600872 Accessed: 03-11-2015 17:37 UTC

REFERENCES Linked references are available on JSTOR for this article: http://www.jstor.org/stable/4600872?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].

Springer is collaborating with JSTOR to digitize, preserve and extend access to Behavioral Ecology and Sociobiology.

http://www.jstor.org

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions Behav Ecol Sociobiol (1993) 33:209-223 Behavioral Ecology and Sociobiology ? Springer-Verlag1993

Resourceallocation, brood productionand cannibalismduring colony foundingin the fire ant, Solenopsisinvicta

Walter R. Tschinkel Department of Biological Science, Florida State University, Tallahassee,FL 32306-3050,USA

Received February 12, 1993/Acceptedafter revision July 3, 1993

Summary.The colony foundingcharacteristics of newly Key words: Pleometrosis - Formicidae - Worker size - mated fire ant queens from monogyne colonies were Cannibalism - Haplometrosis - Colony development. studied in the field and in the laboratoryunder haplo- and pleometroticconditions. Initial queen weight (live) was not correlatedwith subsequentprogeny production. Duringfounding, queens lost a mean of 54% of theirlean Introduction weight, 73% of their fat weight and 67% of their energy content. The percentageof fat decreasedfrom 44% to Ant queensfound new coloniesin one of two ways:either 33%. Queens lost weight or energy in relation to the accompaniedby workers(dependent founding) or unac- amount of progeny they produced(Figs. 1, 2). The effi- companied (independent founding) (Holldobler and ciencyof the conversionof queento progenyincreased as Wilson 1977). Among the majority of independently more progenywere produced,leading to a declinein the foundingspecies, founding is claustral,meaning that the unit cost of progeny (Fig. 3). The more minims a queen queenseals herselfin the foundingchamber and rearsthe produced,the lower the mean weightof theseminims and first brood of workersby drawingon nutrientstores in the fasterthey developed(Fig. 4). In a fieldexperiment on her body (Holldoblerand Wilson 1990). The need for pleometroticfounding, total brood increasedwith queen such largenutrient stores has led to higherlevels of fat in number,peaked betweenfour and seven queens and de- species with claustrallyfounding queens than in those clined with 10 queens (Fig. 5). Brood developedfaster at foundingdependently (Keller and Passera1989). Most of the sunny, warmersite, but total productionand queen this materialis sequesteredbetween adult eclosion and survivalwas higher at the shady site. As queen density the nuptial flight (Kellerand Passera1989). increased,production per queen decreasedas a negative In some claustral ant species, queens may improve exponentialin which the exponent estimatedsensitivity founding success through pleometrosis (co-operative of brood productionto queen-crowdingand the constant colony founding) (Holldobler and Wilson 1977). The estimatedthe productionby solo queens (Fig. 9). These main features of colony founding in Solenopsisinvicta effects of queen number were confirmedin laboratory have been describedby Markinet al. (1972),including the experiments.The decreaseof productionper queen was existenceof pleometrosis.Tschinkel and Howard(1983) small and not always detectable during the egg-laying showed positive relationshipsbetween queen settlement phase, but brood attritionwas always strong duringthe densityand pleometrosis,and betweenpleometrosis and larval period and increasedwith queen number(Figs. 8, foundingsuccess. Pleometrosis may lead to lower queen 10).While airbornefactors may have contributedto this mortality and greaterworker production(Waloff 1957; inhibition,most of the brood reductionwas due to other Bartz and Holldobler 1982; Tschinkel and Howard causes, probably cannibalism.For a given number of 1983).Larger numbers of workersin the firstbrood result minims, increased queen number increased the mean in an advantagein the ensuingcompetition among incip- weight of these minims,an effectthat resultedboth from ient colonies (Bartz and Holldobler 1982; Pollock and a lower minim productionper queen and from cannibal- Rissing 1989;Tschinkel 1992a,b). Central to this compe- ism of dead queens by survivors (Fig. 11). Cannibal tition is brood raiding,the reciprocalstealing of brood queens lost much less weight to producea given number fromneighboring incipient nests by workers(Pollock and of minims than unfed control queens,and these minims Rissing 1989; Tschinkel1992a). Victory usually goes to were heavier(Fig. 12). the nest with more workers,emphasizing the importance of pleometrosisand worker production.Workers soon execute supernumeraryqueens (Tschinkeland Howard

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 210

1983),but how they choose the single,surviving queen is tive competition and regulation.Embryonation of eggs unknown. as a regulatorymechanism in foundingcolonies is unex- In most ants, workers produced by the founding plored. queen(s)are distinctlysmaller than any others in the life Althoughthe potentialfor regulationwithin the claus- cycle and are called minims or nanitics (Holldoblerand tral chamberexists, little is known about processesand Wilson 1990).Vander Meer (1986) suggested that minims relationshipsduring this period. In this paper I relate are a behaviorallyand biochemicallydistinct caste. Cer- queenweight, queen number, weight loss, minimnumber, tainly,the role of minimsin brood raidingsuggests some minim size and development rate during the claustral distinctiveminim behavior,but the degreeto which this period,clarifying patterns of the allocation of queen re- is limited to minims is not yet clear (Tschinkel1992a). sourcesunder haplo- and pleometroticconditions. I also The small size of minims may be an adaptivefeature of show that cannibalismof larvaeand queensstretches the fire-antlife history: Porterand Tschinkel(1986) showed resourceswithin the sealedfounding chamber and affects that for a given weight of workers,minims rearedmore allocation patterns. brood than did minor workers of S. invicta.For many species of ants, rapid colony growthis an importantele- ment in survivaland competition(Vargo 1988). Materialsand methods Pollock and Rissing (1989) have pointed out that, whetherpleometrotic or haplometrotic,the sealedfound- Generalmethods. Queens were aspiratedfrom the ground and under ing chamberis a closed system with fixed resources.For litter on the afternoonsof large mating flights in Tallahassee,Flor- the denizens of the founding chamber,the question is ida, USA. All colonies in this area are monogyne. For colony found- how to allocatethese fixedresources in orderbest to meet ing, queens were placed singly or in groups,depending on the exper- the needs of survival,worker production and the post- iment, into one of two types of founding nests. The first consisted of small test tubes half filled with water retained by a cotton plug. claustralbrood raidingcompetition (Tschinkel 1992a, b). After addition of the queen(s),these were retained with a second Natural selection should optimize the trade-offbetween cotton plug in the mouth of the tube. The second type consisted of worker number and worker size (Porter and Tschinkel blocks of dental plaster 15 x 15 x 3 cm with 25 flat-bottomedholes 1986). Higher worker numbers favor success in brood 1.5 cm in diameter in a 5 x 5 array. Queens were placed into the raiding (Tschinkel 1992a) and colony growth rate holes in numberscalled for in the experimentaldesign. For labora- (Tschinkeland Howard 1983), while larger worker size tory use, the sides of the plasterblocks were waxed to reducedrying, and the plates were covered, top and bottom, with glass and held at favors worker longevity (Porter and Tschinkel 1985a), 30?C. Plates were moistened as needed. For field use, the plaster stress resistance and defensive value. In addition, the founding plates were unwaxed. queen must balance allocation of resourcesfor worker Data consisted of counts and weights of the contents of found- productionagainst those for maintenanceof herselfand ing nests. For several analyses brood were lumped into "young her brood, and perhapsalso for a reservefor post-claus- brood" consisting of eggs and first-instarlarvae, and "olderbrood" tral needs. consisting of the remaining larval instars, pupae and adult minim workers. These allocations are, of course,adjusted over evolu- Fat contents were determined by dry weight loss after ether tionarytime, but there is an indicationthat some adjust- extraction. Energy contents were calculated using 39.33 J/mg fat ment may be possible within each founding chamber. and 18.87J/mg of lean weight (Peakin 1972). Wood and Tschinkel(1981) showed that the small size of S. invictaminims was at least partlytrophogenic in orig- Field experiment.Queens were collected on the afternoon after the in, becausewhen newly mated queens were adoptedinto mating flight of 10 May 1982, and placed in the founding plates in groups of 1, 2, 4, 7, or 10, arrangedin Latin squares.The plates were queenlessnests, the size of workersin the firstbrood was covered with a fine-mesh brass screen, wired tightly and buried at largerthan minims.Generally, similar conclusions were the test sites. The sunny site contained scatteredweedy vegetation, reached by Goetsch (1937) and Passera(1977) working and received full sun most of the day. The shady site was under with Pheidole pallidula,and by Petersen-Braun(1977) sweetgum trees less than 100 m from the sunny site. At each site, with Monomoriumpharaonis. These findings suggest that three founding plates, each containing five of each group size, were the outcome of larval developmentwithin the claustral buried at 5 cm depth and three at 15 cm (six plates per site). This entire experimentwas repeateda second time afterthe mating flight chambermay not be fixed, and may be modifiedby the of 21 June 1982. In the first replicate, the plates at the sunny site rearingenvironment. The number of minims produced were terminatedafter 30 days and those in the shady site after 40 by foundingqueens has been shown to be important,but days. For the second replicate, all plates were excavated after 35 variationof their size may also affectsuch traits as their days. longevityand resistanceto stressand dehydration(Hood The experimentaldesign thus consisted of the independentvari- and Tschinkel1990). ables: queen number(5 levels);site (2 levels);depth (2 levels);founding plate (a blocking factor, 3 levels); replicate(a blocking factor, 2 During the claustralphase, developing larvae of most levels). There were 300 nests in each replicate, 60 at each queen ant species are nourishedat least partlythrough trophic density. Each replicate used 1440 queens. eggs laid for the sole purposeof being eaten (Holldobler and Wilson 1990; Baroni-Urbani1991). Such trophic Stainingeggs to detect embryonation.Five-day-old eggs werecollect- eggs are often non-embryonated.Voss (1985) refined a ed and stained with the Feulgen stain method of Schmuckand Metz methodfor detectingembryonation in eggs. Both insemi- (1931) as modified by Voss (1985). Because DNA is stained pink, this method allows non-developingtrophic eggs to be differentiated nated and virgin queensin maturecolonies can lay non- from nucleated, embryonated eggs. Eggs were classified into un- embryonatedeggs, especiallyin polygynecolonies (Vargo stained, non-embryonatedeggs (class 0) and classes 1-3 containing and Ross 1989),where they may play a role in reproduc- increasinglydeveloped embryos.

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 211

Table 1. Initial weights (mg) and fat content of queens from the mating flights of 29 May, 3 June and 19 June 1991

Measures Mating flight

29 May 3 June 19 June F ratios

x SD x SD x SD df Ratio r

Live weight 16.3a 1.22 14.9b 1.23 15.1b 1.13 3, 100 42.9 22% Dry weight 7.58a 0.69 7.92a 0.82 8.53b 0.62 3, 50 22.9 23% % Dry weight 47.0a 2.65 52.5b 3.41 55.6c 1.57 3, 50 34.0 26% Lean weight 4.24a 0.38 4.51b 0.40 4.65b 0.36 3, 50 15.1 17% Fat weight 3.35a 0.43 3.41a 0.52 3.88b 0.36 3, 50 22.4 23% % Fat 44.0a,b 3.01 42.8a 3.94 45.5 b 2.24 3, 50 9.31 11 %

The F-ratios and dfs referto comparisons among flights by ANOVA. All P values were less than 0.005 after application of the Bonferroni correction for table-wide significance.Means with the same superscriptindicate pairs of values not significantlydifferent (ANOVA)

Data analysis before and after foundingwere determinedby comparison betweendata sets. Queenslost a mean of 8.55 mg live Data were analyzed using Minitab (Ryan et al. 1982) by multiple weight (about 54%)during founding. Loss of dry weight regression using dummy variables, or by analysis of variance averaged64% (5.14mg) revealingthat the percentwater (ANOVA) and the Newman-Keuls Test. Occasionally data were transformedto stabilize the variance. A few outliers were deleted of queens averaged about 12% higher after founding after analysis of residuals. The field experiment was analyzed by (59%) than before (47%).Loss of fat weight was about ANOVA using BMDP8V (Dixon et al. 1985). 2.59 mg, and was proportionallyhigher (73%) than the loss of lean weight (2.54mg; 57%).This was reflectedin a decline of queen percent fat from about 44% before Results foundingto about 33% after.When these declineswere expressedin energeticterms (joules),queens lost 67% of Haplometroticfounding their energycontent duringfounding. Again, there were significantdifferences among the Initial queen weight, andfat content. Queens were collect- threeflights. Queens from each subsequentflight lost less ed after 3 mating flightsin 1991 (29 May, 3 June,and 19 live weight(ANOVA, F2210 = 10.1; P<0.001; r2 = 9%), June).From each flight 50 queenswere used to determine but this orderwas reversedfor dry, lean and fat weights, initial weights and fat content (Table 1). reflectingthe higher water content of queens from the Over all three flights, queen live weights averaged first flight. 15.4 mg (SD = 1.19), dry weights 8.01 mg (SD = 0.71), Together,these data show that colony founding re- lean weights 4.47 mg (SD = 0.38), and fat weights sults in very large losses of weight and energy content, 3.55 mg (SD = 0.44). Dry matter made up 51.7% that fat contributesdisproportionately to these losses, (SD = 2.54) of the live weight, and fat made up 44.1% and that queensfrom differentmating flightscan be sig- (SD = 3.06) of the dry weight. Similarlive weights were nificantlydifferent in many or most of these measures. reported by Fletcher and Blum (1983), Porter and Tschinkel(1986) and Markinet al. (1972)and are charac- Brood production in relation to queen characteristics. Dry teristicof the monogyne form of the fire ant. and lean weightsof groupedminims and groupedbrood For most measures,there were significantdifferences were then determinedfor the same 300 post-founding betweenmating flights (Table 1). For all measuresexcept nests from the above experiment.Minims and pupae percentfat, the 29 May flight was significantlydifferent werecounted. The parametersand the total r2values are fromthe 19 Juneflight, with the 3 Juneflight more similar given in Table 2. In the discussionbelow most r2 values to the earlierflight in some measuresand the later flight excludethe effectof matingflight and thus differfrom the in others. Dry, lean and fat weights increasedwith each total r2 in Table 2. mating flight.The 29 May queens containedthe highest While it might be expected that heavierqueens pro- proportionof water, resultingin the highestlive weight. duced more minims,this was not the case (Table2; re- The percentfat varied only little, although significantly. gression 1). The numberof minimswas not significantly relatedto the initial queenweight (slope not significantly Queen weight loss during founding After completion, in differentfrom zero, 0.4% of total variation).While total the laboratory, of colony founding by another 100 progenyweight increased significantly with increasedini- queens from each mating flight, the queens and their tial queen weight (Table2, regression4), this trend was progeny were killed by freezing. Live, dry and lean weak and accountedfor only 2.5% of the variation. weights of the queens were determined.Because it is not On the other hand, the number of minims increased possible to take the initial dry weight of a queen that significantlywith increasedweight lost by the queendur- producedminims, only differencesin mean queenweights ing founding (Table 2, regression2; Fig. 1). The regres-

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 212

Table 2. Coefficientsfor regressionsof brood production variables,in the form y = ax + b

Regression Variables y-intercept Slope Total r2 number y x MF b SEM a SEM

1 No. min. Q live wt. All - - 1 -2.36 9.90 - - 9.4% 2, 3 4.60 9.00 - -- 2 No. min. Q wt. loss All - - 4.11 0.36 1 -23.9 3.43 - -- 44.8% 2, 3 -15.8 3.15 - -- 3 Wt. progeny Q wt. loss All - - 0.31 0.028 1 -0.87 0.27 - -- 41% 2, 3 -0.45 0.24 - -- 4 Wt. progeny Q live wt. All - 0.15 0.040 1 -0.48 0.66 - -- 7.5% 2, 3 -0.12 0.60 - - MF=mating5 flight;Wt. 1=29 progeny May, 2=3 June,Final 3=19 Q June.dry wt. Q=queen, All others- in the regression- after -0.56 application of0.079 the Bonferroni correc- min. = minim, AS % = arcsine square root transformedproportion. 1, 2 tion for3.57 table-wide 0.22 significance.- A single slope-- difference22%became All listed coefficients are significantlydifferent from zero or from3 non-significant3.83 0.24after this correction- -- 6 Tot. min. dry wt. No. min All 0.12 0.022 0.069 0.0012 96% 7 AS% min. of No. min. All - - 12.2 1.25 total progeny wt. 1 13.5 1.30 - -- 84% 40' 2, 3 sion16.4 equation 1.39showed -that queensthat -- producedno minims lost up to 3.8 mg (29 May) and 5.8 8 AS% min. of Tot. prog.R2= wt. 45% All - - 12.2 1.25 mg (3, 19 June). total progeny wt. 1 Thereafter,22.2 2.87every 1 mg- of live weight-- lost resulted39% in a 2, 3 mean27.9 of 4.115.74 additional - minims.Queen -- weight loss ex- + 9 30-+ wt/min *+ CNo. min. 1 plained0.089 27%0.0016 of the variationin minim-- number. * *a + 46 + 0 * Total- - 10-4 .I*++ +.I 000 1, 3 weight of progeny-6.57 x similarly0.8 x 10-4 increased with +0 + 00 + + 0 + *o 2 queen0.11 weight 0.0020 loss (Table-0.00122, regression 0.00013). About50% 31% of en ++ + + 0 + 40 0 00 3 the 0.094variation 0.0019 - - E + ** + 0 o o in progeny weight was accounted for by ._ m-o oo0 *- 20 + + a o queenweight loss. Queenslost an averageof 1.5 or 2.8 mg live weightin orderto producethe firstoffspring individ- + o* a a a a O + a ual. Thereafter,they produced 0.31 mg progeny (dry Z + + + a 0+ + 4+0 + weight)for every 1 mg live weight lost. Because queens 0+ + 10- + 0 + a a averagedabout 50% dry weightinitially, this meant that * a ao 0.6 mg dry a progeny were producedfor every additional + (dry) 1 mg lost. * oab + a a * + o0 0 0 In energeticterms, queens lost 90 J without progeny + +0 o aa.ao +O-* production(intercept = -31.2). Over 70% of this can be 4 6 8 10 12 14 accounted for as maintenancecost, using 0.029 J/lean Queen weight loss mg/h (Tschinkel,unpublished data). Mean lean weight (mg, live) declined57% over 30 days at 30?C.Thereafter, each ad- Fig. 1. Number of minims produced in relation to the live weight ditional joule lost resulted in 0.34 J of progeny lost by the queen during founding. Data are from three different (r2 = 50%) (Fig. 2). mating flights (29 May, squares; 3 June, circles; 19 June, crosses). Efficiency(fraction of queen weight-lossconverted to The more minims a queen produced the more weight she lost brood) of the conversionof queen to progenyincreased (r2 = 45%). The ranges of brood production and weight loss are with the total weight of brood produced(Fig. 3), rising high. Queens producing no brood nevertheless lost a substantial from 4% amount of weight or 8% to about 35% (70% if convertedto a

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 213

100 0.4 + R2= 50% R2= 83% 80~~~~~~~~~~~~ + + + + 0.3 + 0 + 4. ++0 >1 ~~~~~~~~+++0z g 60+ +t.0**~ + 0 2 3 4 2) P+0 + 00 0 - 0.1 ~ ~~~~~ + +++ 60 0 0 ~~+ *040o + 0 +3 C 0 0 +~~~ 0.1 'aLb,1 75 ~~~+ 0+ ++ 0 2 * 0 1+ El~~~~~~~E - - >. 0 ++I C

0.01 ElI

Fi.). Effiienc was esiaesmgbodprmgqenwih

0 1 2 3 4 Queen energy loss Total weight brood (mg) (joule) Fig. 2. Total energy in progeny in relation to the energy lost by the Fig. 3. Efficiencyof brood production in relation to the weight of queen during founding. Data are for three differentmating flights brood produced. Data are from three differentmating flights (as (as Fig. 1). The higher the energy invested in progeny, the more Fig. 1). Efficiencywas estimated as mg brood per mg queen weight energy the queen lost (r2 = 50%). The ranges of energy loss and lost. The cost of brood per mg decreased (efficiencyincreased) as brood productionare quite high. Queens producing no brood nev- more brood was produced (r2 = 83%) ertheless lost a substantial amount of energy, possibly as maintenance cost

dry-weight basis). Variation in progeny weight accounted Efficiency of conversion of queen to progeny, either in for 74%/ of the variation in efficiency while mating flight weight or energetic terms, was also significantly higher in contributed an additional 9%/. Thus, queens that pro- queens from the later flights, adding 7% and 25% to the duced more progeny did so at a lower unit cost, largely regressions of these variables against the amount of because of the large fixed cost that did not result in progeny (in mg or J). brood. This fixed cost is probably maintenance respira- As expected, total progeny weight was negatively re- tion. lated to the queen's final weight (Table 2, regression 5). When the basis was energy content, the efficiency of For every additional 1 mg (dry) retained by the queen, conversion of queen to progeny was much lower. Because total progeny weight was less by 0.56 mg (dry). Final the initial joule content of queens rearing progeny cannot queen weight accounted for 18% of the variation while be known, this conversion factor was calculated from the mating flight added 4% more (22% explained). weight loss during founding, the mean energy content (J/mg) of queens before and after founding, and the First generation brood as a cohort. Inspection of brood progeny produced (J). Although efficiency increased sig- status and numbers suggested that the total brood in nificantly with increased weight loss (slope = 0.013; experimental nests was a cohort - that is, larvae from the t = 5.08, P <0.001), this factor accounted for only 2.2% first batch of eggs developed together while few further of the total variation, while mating flight explained 25% larvae were produced until the brood consisted of a sub- (see below). stantial fraction of minims. This means that the differ- All measures of brood production were significantly ences in counts of total brood represent largely differ- different for at least one mating flight, and the inclusion ences in the size of this first cohort, not differences in the of mating flight in regressions added considerably to the development rate. The temporal pattern of egg embry- explained variation (Table 2). While initial queen weight onation confirmed this. Five days after nest initiation, was not related to the number of minims produced, over 40% of the eggs were in stages 2 and 3 of develop- queens from the second and third flights produced signif- ment, while only about 4% were in stage 1, indicating icantly more minims, adding 9%/ to the explained vari- that most of the embryonated eggs were among the first ance. Similarly, this added 5%/to the explained variance laid. The remaining 56% were non-embryonated trophic of regression 4 (initial queen weight and total progeny eggs laid later. These patterns are in general agreement weight) and 17% to the regression 2 (number of minims with those described by Markin et al. (1972) and Voss and vs. the queen weight loss). For the regression of progeny Blum (1987). afgainst queen weight loss, the second and third mating flight:produced significantly more progeny than the first Progeny characteris-tics.As minim production increased, (t = 5.34; P <0.001) adding 9%/ to the explained varia- the individual weight and development time of minims tion. decreased. When minim production was near zero, mni-

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 214

the square root of the number of adult minims (best fit) (Table 2, regression 7). As the number of adult minims 0.12 *+ increased from near zero to 30, the proportion of progeny +0 weight present as minims rose from 6-8% (depending on mating fight) to 79-83% (Fig. 4B). Minim number ac- E 0.10 * 5 + counted for 83% of total variation in percent minims, ._ 0 * + + 0+ with mating flight adding only another 1%. When total progeny weight was used as the x patterns were U * z id0+ + variable, X +0+0 X a O a similar, although the explained variation dropped to CL + 0.08 + +S31 + 39% (Table 2, regression 8). + 0+~~~~~~~~~ In overview, queen live weight failed to predict production during founding, but queens of higher fertility 0.06 lost more weight and produced more minims, which developed faster and weighed less than minims of queens with lower fertility.

A ,00.04 , 0 10 20 30 40 Pleometrotic founding 80 R2= 84% Field experiment. Queen number, site and depth explained 11% of the variation in total brood. Queen num- 70 0 0 ber had a highly significant effect on the total number of + + +d . brood produced (exclusive of eggs) (ANOVA, data trans- + 60- + formed to ln(y+ 1), F4,16 = 19.33; P<0.001) (Fig. 5). At + DI0D S all sites and depths, total production climbed from single 50 * 0R o = 40] * ? + * + ? ++1*O queens to a maximum at about four to seven queens, and then declined as queen number increased to ten. The de- .2c + a~~~~~D 1 |,L~~~~ X ~30i40 o0 020 10 0 o 40 cline was more rapid and the peak sharper for nests at the ECU~~~~ 20 no0 a 0 sunny site, resulting in a significant site-by-queen number interaction (F4,64 = 3.87; P<0.01; 9.4% of explained 0 40 00 + lb~ variance). Total production was also 20% higher at the shady site in both replicates (F1,16= 7.06; P<0.05; 8.2% 20- of explained variance). A replicate-by-depth-by-queen interaction accounted for another 11% of the explained B 10 a variance (F464 = 4.70; P <0.005). Overall, queen number 0 10 20 30 40 explained about half of the variation in total production, with site and depth making smaller contributions, either No. minims directly or in interaction with queen number. Fig. 4. A As minim production increased, the individual weight of Brood development rate was estimated as the fraction minims decreased.B Minims developedfaster when higher numbers of brood which were adult minim workers (arcsin square were produced. Development rate was estimated as the fraction of root transformed). Brood development rate was not sig- the brood which were minims aftera fixed elapsed time (r2 = 84%). Data are for 3 differentmating flights (as Fig. 1) nificantly affected by queen number but was much higher at the sunny site (site: F1360= 300; P< <0.001) and at shallower depth (depth: F1 360 = 45.5; P < <0.001). In the sun, brood was 57% (SE = 2.0) minims at 5 cm and ims averaged between 0.09 and 0.11 mg dry weight (de- 46% (SE = 2.5) at 15 cm, while in the shade, brood was pending on mating flight). In groups of 30 minims, mean 24% (SE = 1.6) minims at 5 cm and 11% (SE = 1.4) at weight dropped to about 0.07 mg, a decrease of 22-36% 15 cm. Site accounted for 61% of the explained variation (Fig. 4A; Table 2, regression 9). The number of minims in fraction minims, while depth accounted for 9.2%. A accounted for about 38% of the total variation in their significafit replicate and replicate-site interaction resulted weight, the higher minim weight of the third flight adding from allowing the shade treatment in replicate 1 to run 10 11% more (total 50% explained). Similar though some- days longer than the sun treatment, a source of artifact what weaker patterns emerged when total progeny not repeated in replicate 2. When this run-length effect is weight was used as the x variable. eliminated, the results of the two replicates were similar. The decline in individual minim weight was correlated The probable source of these development rate differ- with the decline in their development time. Because ences was soil temperature, which varied dramatically queens begin egg-laying on the same day and all nests of according to site and depth. The mean afternoon temper- a replicate were later terminated on the same day, devel- atures (1200-1500 hours) at the sunny site were 39?C at opment rate was estimated by the proportion of the total 5 cm and 33?C at 15 cm. Under sunny, dry conditions, progeny weight which was adult minims. The proportion the temperature at 5 cm occasionally reached 45?C. was arcsin square root transformed and regressed against Mean afternoon temperatures at the shady site were 26?C

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 215

0 5cm o 15cm 20 20

10 0 0

Fig. 5. The relationship of total brood to X 0 I I I TX10II queen number, site and nest depth (filled o circles, 5 cm; open circles, 15 cm) in the field experiment.Queen and site had the largest effects,with production reaching a maximum between 4 and 7 queens at the 2- | 4 .~ ,' , . o0 l . . , , . . shady site o 2 4 7 1 0 0 2 4 7 10 Sunny Site Shady Site Queen Number

10 10 0 5 cm 0 cm1 15

6 number was maximum for 4-queen Z 2~~ 2 ~~tnests at the shallow, sunny site, l Y z ~~~~~~~~~~~~~~~~~~thoughthis was not the site with o0 * , W | W 0?- I . , I , . highest brood production 1 2 4 7 140 1 2 4 7 1e0 Sunny Site Shady Site Queen Number

and 25?Cat 5 and 15 cm, respectively.Morning tempera- significantly(depth: F1,16 = 14.9; P<0.005; 13% of ex- tures averagedabout 29?Cat the sunny site and 25?Cat plained variance; site: F1,16= 12.8; P<0.005; 11% of the shady site, with little differenceby depth. Soil mois- explained variation),with shallower and sunny treat- ture was also different,perhaps as a resultof the temper- mentshaving higher numbers of minims.Modest interac- ature differences.At the sunny site, the soil contained tions (5 and 4% of explainedvariance) of queen number 2.2% and 3.2% water at 5 and 15 cm, respectively.At with site and depth were associatedwith the very sharp shady sites these values were 3.5 and 3.8%.While it may peak at four queens found in the sunny, shallow treat- not have a direct effect on developmentrate, drier soil ment. Small but significantreplicate by site or depth in- heatsmore rapidly,amplifying the temperaturedifference teractionswere also found. Overall,while the sunny site between sites. had lower total brood production,the higher develop- From the point of view of post-claustralcolony suc- ment rate resulted in higher adult minim production cess, especially success in brood-raiding (Tschinkel within a fixed time, and a more sharplydefined optimal 1992a)the early production of a large number of adult queen number.Because early productionof large num- minims is of great importance.When the minims open bers of adult minims is important to colony founding the nest, brood not in the form of adult minims is of no success,the observedpatterns may indeed be important value in winning brood raids. Queen number had the for success. largest effect on the number of minims (F4,64 = 17.1; Queennumber had no significanteffect on queenmor- P <0.01), accountingfor 25% of the explainedvariation. tality (F464 = 0.83; n.s.) indicating that queens died of As with total brood, minims reach a maximumat inter- intrinsic or physical causes rather than the effects of mediatequeen numbers. Replicate had a significanteffect crowding. Of variation in percent live queens (arcsin (F1l16= 22.98; P < 0.002; 20% of explainedvariance) be- square root transformed)6% was explained by site cause shade treatmentsran 10 days longer than sun in (F1 16 = 38.6; P< <0.001): survivalwas 81% (SE = 3.2) replicate1. Replicate2 avoided this artifactand is shown in the shade and 48% (SE = 3.0) in the sun. Depth had in Fig. 6. Site and depth affectedthe numberof minims no significanteffect. It should be noted that after most

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 216

12 12 0 5 cm

m 0 Y10 0c 15 cm a) 0) 8 8

26 Fig. 7. Brood per queen in relation la to queen number, site and nest 2 depth (filledcircles, 5 cm; open circles, 15 cm) in the field experiment. 2 -2 Brood declined o 24 o12 4 7 1 per queen sharply with queen number.Depth also had 0 . . . . 0 . . . . ~~~~~~~~~~~~~~~~~~~aminor effect 1 2 4 7 1 0 1 2 4 7 1 0 Sunny Site Shady Site Queen Number minims have eclosed, they begin executingsupernumer- 800 ary queens,but these experimentswere terminatedprior 7 d to that stage. The existenceof a maximumfor brood or minimpro- 600 duction at intermediatequeen numbers suggested that 0 each queen'scontribution must have declinedrapidly as the number of queens in the group increased.This was confirmedby ANOVA of the brood per queen. Brood 0) 400 per queen showed a steady decline as queen numberin- c+/ 00m creased(F4,64 = 26.7; P <0.001) (Fig. 7) and queen number accountedfor 70% of the explainedvariation. Depth and replicate added another 5% each, with shallower Z 2002 nests being slightlymore productiveper queen.Analysis based on individual brood stages showed very similar patternsfor all of these. Becausesite and depth had little or no importance,it appears that this decline in repro- 0 2 4 6 8 10 12 ductive efficiencyis the result of interactionsamong the 60 - ants in the foundingchamber. This phenomenonis treat- ed in greaterdetail below. To succeed,a queen must not only survive,but pro- 0 50 duce brood as well. Successfulnests were minimallyde- E 4 fined as those with > 0 brood and live queens.Note that laE4 14 d 0 + there can be more than one successfulqueen in a nest. 0 Both queens and nests enjoyed higher success in shade 24d than sun. Nest success was maximal at intermediate 0N 30 queen numbers but queen success was independentof 0 queen number. 0 20

Effect of queennumber in laboratoryexperiments. Eight foundingplates were stockedwith queensat 1, 2, 4, 7 and 10 queens per group arrangedin a Latin square.Queens were collected on 24 May 1983. Nests in all eight plates werefirst censused after 7 days. Four of these plates were terminatedafter 14 days and their contentscensused and weighed.The other four were similarlyterminated, cen- Fig. 8A,B. Brood counts in the laboratory experiment. A Young sused and at 24 brood were a simple multiple of the number of queens, but B older weighed days. Census data were trans- brood showed a maximum at intermediate queen number. Nests formed to ln(y+ 1). Both census and weight data were were censused on days 7, 14 and 24. Bars indicate SE analyzedby ANOVA. Census of young brood (eggs and first instar larvae) below), though egg crop was significantly lower at 14 and showed that their numberincreased in direct relationto 24 days than at 7 days. queen numberat all three census times (Fig. 8A). Queen On the other hand, the older brood production curve numberthus seemed to have little effecton egg produc- had an optimal character with respect to queen number tion per queen(but see "Factorsreducing queen fertility" (Fig. 8B), much like that in the field experiment (Figs. 5

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 217

Table 3. Parametersof the regressionof In (brood/queen)against (- Q - D) for several experiments;brood/queen = y

Expt. In y Day no. K Slope Intercept r2 Dominant (condition) brood stage s SE In N1 SE N,

Field (L+P+M)/Q 30-40 4 0.14 0.019 pupae, minims (shade) (shade) 2.48 0.091 11.94 31 % 30-40 4 0.25 0.013 pupae, minims (sun) Lab-83 (E + L1)/Q 7 8 0 - 4.22 0.089 68.0 eggs 14 8 0 - 2.85 0.087 17.3 51% larvae 24 8 -0.046 0.013 3.98 0.095 54.0 pupae, minims (L+P+M)/Q 7 8 0.14 0.016 0.15 0.18 1.16 eggs 14 8 0.23 0.019 2.76 0.21 15.8 71% larvae 24 8 pupae, minims Embryo eggs/Q 5 8 0.012 0.0036 4.63 0.040 103 8.4 % eggs Poly minims/Q 30 8 0.09 0.010 2.66 0.11 14.3 29% minims, pupae

The slopes estimated the sensitivity of brood production to queen M = minims. Field = field experiment; lab-83 = laboratory experi- number,Q. The intercept,In N,, estimated the brood production of ment on effect of queen number; embryo=embryonation experi- solo queens (N1). D = K -(K + 1)/Q where K = a fitted constant (see ment; poly= effect of queen number on minim weight experiment text). E = eggs; LI = first instar larvae; L = older larvae; P = pupae; and 6), with peak numberincreasing greatly from 7 to 14 whereN is the mean numberof progeny,Q is the number days, then decliningsomewhat to 24 days. As in the field of queens,N1 the mean numberproduced by solo queens, experiment,ANOVA showed that this peak was based and s is a factorestimating sensitivity to crowding.When on a decline in the brood productionper queen. At 14 s = 0, crowdinghas no effectand N is simplya multiple days, when the oldest brood began to pupate, there was of the solo queencontribution, N1. As s becomesincreas- a significant decline in mature larvae per queen ingly positive, brood production declines ever more (F485 = 13.6; P<0.001) and pupae per queen rapidly with queen number.When s is negative,brood (F485 = 4.09; P <0.01) as queennumber increased. At 24 productionis stimulatedby crowding(not observedin S. days a substantial fraction of pupae had eclosed, and invicta).In some cases, the decrementaleffect of each ad- there were highly significantdeclines of productionper ditional queen became smaller,so that the effectof each queen of young larvae, mature larvae, pupae and adult queen,-Q, had to be decrementedby D, whereD is itself minims (F474 = 18.4-30.6;all P<0.001). These patterns a functionof Q: were similarwhen the total weight of brood was the de- D = K-(K+i)/Q pendent measure. Increased queen number caused a highly significant decline in the weight of brood per and K is a fittedconstant imparting increased non-linear- queen (ANOVA, F41,63 = 58.7; P< <0.000;1 r2 = 59 %). ity as K increases.The addition of 1 to the K in the Thus, as developmentof the pleometroticcolony pro- quotient means that when Q = 1, the exponentequals 0 ceeded, the brood contributionper queen declinedcon- and N = N1. When K = 0, the per-queen effect changes tinually over time, and this decline was more rapid for least with queen number. As Q becomes large D ap- largerfounding groups. proachesK. The lower productionper queen was reflectedin sig- The form nificantlyhigher final queen weights as queen number (ln N - ln Q) = ln N1 + s(-Q-D) increased(ANOVA, F4169 = 16.7;P<0.001; r2 = 29%), i.e., lower weight loss duringfounding. Whether this was yields a straight line with slope s and intercept ln N1 the resultof lower materialdonation rates or of cannibal- when (ln N-in Q)is plotted against(-Q-D). Comparison ism of brood or dead queens is consideredbelow. of linear regressionsof this form showed that intrinsic progeny production and sensitivityto crowding varied Fitting the production-queennumber curve. The decline of among differentstages, flights and conditions (Table3; productionper queen with increasedqueen number gen- Fig. 9). (The transformation[In (N + 1)-in Q], which in- erally resultedin a concave plot and was fitted reason- cluded groups producingno progeny,deviated strongly ably well by a negative exponential function. Queens from normality and was fitted less well by this regres- from different flights or under different conditions sion.) seemedto generatedifferent parameters. I reasonedthat Appliedto the fieldexperiment, the regressionshowed the total progenyproduced by an associationshould be that production of older brood by queens in the shade a multipleof what queensproduce by themselves,multi- was significantlyless sensitiveto crowdingthan those in plied by a crowdingsensitivity factor: the sun (s = 0.14 vs. 0.25; t1,300 = 4.61; P<0.001) (Table 3; Fig. 9A). There were no significantdifferences in the N = QN, es(-Q-D) intercepts(t-test), showing that solo queens in all treat-

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 218 - 3 1.0

0.9 I

0 0.8 2 0 - 0 -0.7

0 0) o 0.6- 0 0L 2 -- ~~~~~~Sun 0.5

0.4 1 2 4 7 A 0 , No. Queens -15 -10 -5 0 Fig. 10. Attrition of progeny in relation to queen number between -Q-D, K= 4 day 7 and 14 in the laboratoryexperiment. Data based on counts of all individuals present, with 55% of the eggs present on day 7 assumed to be embryonated.Attrition increasedwith queen num- 7Q-4Q 2Q 1 Q 6- 10tQ ber, probably through cannibalism.Bars indicate SE

Eggs + Larvae-I 7 d O. 4 _ - significantlywith queen number, but the effects were C 0r------?s ~~~d ~ ~ ~~~~~24 small. 4=D 2 - 14 d This was not true of older brood (larvae+pupae+ > -2- in 10 144d minims)which showeda sensitivityto queencrowding all experimentsat all colony developmentstages, with - Larvae-1 + -14 slopes ranging from 0.14 to 0.40 (Table 3). For older 0 pupae +24 2 mminims d brood in Fig. 9B the slope increasedsignificantly from 7 d 0.14 to 0.23 (t = 4.59; P<0.001) from 7 days to 14 days when mostly larvae were present, but did not increase -2 from 14 days to 24 days (the pupalperiod) indicating that most of the attritionwas probablyof larvae,not pupae. Once larvaepupated, crowding had no additionalnega- tive effects. -20 -1 5 -1 0 -5 0 Becauseolder brood werereduced by crowding,while eggs were not, it followed that brood must have disap- -Q-D, K=8 peared as they developed, and that the disappearance Fig. 9A,B. Brood production per queen plotted as the naturallog of was proportionallyhigher in larger queen groups. The brood/queen against -Q-D, a variable derived from queen number numberof progenypresent at 7 days that were still alive and k = a fitted constant definingD (see text for details).This trans- at 14 days declinedby 44% for single queen nests. This formation produces linear plots and allows the slopes and intercepts to be used as estimates of sensitivity to crowding and solo attritionincreased steadily with increasedqueen number, queen production, respectively.A Data from the field experiment, reaching88% attritionin 10-queennests (Fig. 10) (one- showing that brood production at the sunny site was more sensitive way ANOVA; F491 = 6.40; P<0.001). This increasein to queen crowding. B Data from the laboratory experiment,show- attritionrate, and the higherfinal weightof pleometrotic ing little effect of queen crowding on young brood. This increased queens are consistent with cannibalismof the vanished by day 7 and stabilized after day 14. Young brood declined and larvae. older brood increased from days 7 to 14 The performanceof solo queens (NI) differedamong experimentsand changedover time (Table3; Figs. 8 and ments producedabout 12 older progeny.Overall, varia- 9). In the laboratoryexperiment, solo eggs dipped to a tion in (-Q-D) and site explained18% of the variationin minimumat 14 days, while solo older brood rose from (In N-In Q). 0.81/queen at 7 days to 8.94/queenat 14 and 24 days. As brood developed,they increasedin sensitivityto This is consistentwith the cohortnature of the firstbrood queen crowding.In both the laboratoryand embryona- noted above and by Markinet al. (1972).Solo production tion experiments,the slope for early brood was zero or of eggs was also much higher than older brood, in part quite low (0.02-0.08;Fig. 9B),indicating that the number becauseonly about half the eggs were embryonated,but of eggs in groups of queenswas almost a simplemultiple also becausethere was 44%/larval attritioneven in solo of the egg productionof solo queens.This was true even queen nests (Fig. 10). at 14 and 24 days when older brood predominated(Table 3; Fig. 8). In another experiment(see "Factorsaffecting Effect of queen number on adult minim weight. Queens queen fertility"below), however, egg productiondeclined collectedafter the matingflights of 21 May, 3 Julyand 10

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 219

0.13 Because these simple analyses did not segregatethe effectof the numberof minims (see above) a more com- 0.12 plete analysisof covariancewas run. Queennumber was used as the treatmentfactor. Many of the dead queens E 0.11- appearedto have been partiallyeaten, so the numberof pieces (head, thorax, gaster) eaten and the number of minimswere used as covariates. t)j 0.10 ququeens O Z E The ANCOVA showed that all three factorshad significant effects on adult minim weight (queen number, 0.09 F2,92 = 22.7, P< 1 x10-5; number minims, F192 = 8.09, P < 0.006;number eaten, F1,92 = 7.53,P < 0.01).Each ad- 0.08 f4 queen ditional queen increased mean minim weight by 1 queen 0.0012mg, each additionalminim decreased it by 8 x 10-

A 0.07 mg, and each piece eaten by survivingqueens increased it 0 2 4 6 8 10 12 by 0.0005mg. Queen numberaccounted for 74% of the No. of Live Queens explainedvariance in minim weight, numberof minims

0.13- for 13% and numberof queen pieces eaten for 12%. A total of 37% of the variancewas explained. A one-way ANOVA of only those nests without 0.12 queen mortalityat 14 days confirmedthe positive effect E of queen numberitself on weight per minim, which increasedfrom 0.080 mg for singlequeens to 0.083for four- 0.11. queen groups and 0.10 for ten queens (F243 = 5.74; P<0.02; r2 = 22%). 0E Regressionshowed that queennumber, total weightof adult minimsand numberdead at 14 days all had significant effects on final queen weight (F351 = 15.7; 0.09 P<0.0001; r2 = 57%). Each additionalqueen increased the final queen weight by 0.14 mg and each dead queen by 0.14 mg. Each additionalminim decreasedthe weight B 0.08 0 1 2 3 4 5 6 per minim by 0.000084mg. About 69% of the explained No. of Dead Queens variation was accounted for by queen number,4% by dead queens,and 27% by numberof minims. Fig. 11. A Weight per minim in relation to the number of queens How mightthese factorsoperate to affectadult minim alive at the end of the claustral period. Colonies were started with and queen weight? As shown above for single queens, 1, 4 or 10 queens, as indicated. In the 4- and 10-queencolonies, the lower the number of live queens, the greater was minim weight. minim weight declined as the number produced in- B Weight per minim increased in relation to the number of dead creased.As queens founded in larger groups, they pro- queens. Data from same colonies as Fig. l1A. Bars indicate SE duced less brood per queen, resultingin heavier adult minims and less queen weight-loss. Perhaps more resources per larva or cannibalismof some larvae caused July 1990 were allowed to found colonies singly or in minim weight to increasewith queen numbereven when groups of four or ten. When most of the brood had no queens died to provide additional food. When co- eclosed,the colonies were killed for counting,drying and foundressesdid die, they provided additional nourish- weighing. ment, contributing to lower weight loss of surviving As before,groups of four queensproduced significant- queens and higherweight per minim. ly more minims than did single queens (X = 30.6 and 18.4, respectively),but in this experimentgroups of ten Does queen cannibalism contribute to brood production? were highest (X = 50.3). Newly mated queens from the mating flight of 26 June A one-wayANOVA showedthat the mean weightper 1991were set up singlyin 300 nest tubes.To 100 of these, adult minim increasedsignificantly with queen number a piece of killed queenwas added at 3-dayintervals (pre- (F2,187 = 30.3; P <0.001) suggesting that with more vious piece removed)(cannibal group). To another 100,a queens more materialwas invested in each minim. Fur- piece of tenebrionidbeetle larva of equivalentsize was ther analysis showed a more complex phenomenon. added(carnivore group), while the last 100 servedas con- ANOVA of mean minim weight against the numberof trols and were only opened and re-closedat 3-day inter- queens still alive at the approximateend of the larval vals. When most of the brood had eclosed to adult min- feedingperiod (14 days),showed that withinthe four-and ims, the nests were killed by freezingfor weighing and ten-queengroups, adult minimweight was higherin nests counting. withfewer live queens (F996 =3.4; P <0.005) (Fig. 1llA). Treatmentand number of adult minims had signifi- Conversely,minim weight rose 23% as the numberof cant effects on the weight per minim (Fig. 12A) dead queensat 14 days increasedfrom 0 to 5 (regression; (F9147 = 61; P < <0.001; r2 = 54%). Minims produced r2-18%) (Fig.1B). by cannibalqueens averaged 0.022 mg (24%-30%)heav-

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 220

0.160.16 R2= 54% more (both significantly higher than controls t = 4.91, (No. minims: 39% P<0.001, t = 15.1, P<0.001, respectively). Queen * Treatment:16%) 0.14 weight loss was 0.54 mg per mg of minims and did not differ by treatment (i.e., slopes were not significantly dif- + E + ferent). Weight of minims accounted for 26% of total - 0.12- * . variation of final queen dry weight, number eaten 11%

+ 03 and treatment for 35% more (70% explained). The number of pieces offered to carnivore and canni- QO.tO- + +o0 . +g: bal queens was, of course, the same. However, queens ate + @" o4 + * + + o + parts of a significantly larger number of beetle larval = = *S0.08|+ ++1. 0 9e~~1+0 a +3 n pieces (X 3.82) than of queen pieces (X 2.78) g I = o oD~~~10 a (tgg 2.87; P<0.01). Nevertheless, queen pieces were significantly more effective in raising final queen weight 0.06 00 and weight per minim, suggesting that queen pieces were either fed on more heavily than larval pieces, or were much more effective in achieving the observed results, or A 0.04 0 1 0 20 30 40 50 were partially eaten without detection. 8 R2_ 70% Factors reducing queenfertility. How and at what stage does the reduction of contribution per queen in

6 pleometrotic associations take place? It was shown above that queen number had only very minor effects on

3. $-+ .0 + egg-laying. However, the decline of larvae/queen with increased queen number could have been the result of a axX + 0. * .0 decline in the proportion of nucleated eggs (i.e., higher fraction of trophic eggs). Queens were placed as above m- 4 0 crTO*c + . + into founding plates in groups of 1, 2, 4, 7 and 10, and were allowed to lay and incubate eggs until the first eggs + +0+o300o+ 0 + + + were almost to hatch 5 The U_~ ~ 0 0 a1300 ready (about days). eggs were + stained (Voss 1985) to detect embryonation and the data 2- +i 000o6o0 +0 n 1 3=|3OBO analyzed by ANOVA after transformation to ln(y + 1). As noted above, the number of eggs increased almost, B 0~~~~~~~ but not quite, as a simple multiple of queen number (Fig. 8). Of greater interest is the finding that the number 0 10 20 30 40 50 of embryonated eggs per queen declined significantly with increased queen number (F494 = 2.64; P <0.05; No. minims r2 = 33?%),but so did the total number of eggs per queen Fig. 12. A Weight per minim in relation to feeding and the number (F494 = 3.55; P<0.01; r2 = 34%). As a result, there was of minims produced by solo queens. Queens allowed to feed on no significant effect of queen number on the fraction of queen pieces (cannibals, full circles) produced heavier minims than eggs embryonated, about 55% (F494 = 0.185; n.s.). As those fed beetle larvae (carnivores,crosses). Both groups of minims group size increased, individual queens laid somewhat were heavier than unfed controls (open squares).(Total r2 = 54%; for number of minims, 39%; treatment, 16%.) B Queen weight at fewer eggs, on average, or fewer survived for 6 days, but the end of the claustral period was much higher for cannibal than a constant fraction of these were embryonated. Reduc- carnivore queens, and both were higher than unfed controls tion in reproductive output thus began with egg-laying in (r2= 70%) this experiment but did not involve embryonation. The experiments in Figs. 8 and 9 found no such effect of queen number on egg production, suggesting that these small ier than those from control queens, while those from car- effects may sometimes be undetectable. nivore queens were 0.01 mg (11-14%) heavier. Both were It seemed possible that the negative effect of grouping significantly greater than the controls (t-test: t = 7.25, on reproductive output per queen was partly mediated P <0.001 and t = 4.05, P <0.001, respectively). by an airborne factor such as a pheromone. This was As previously, weight per minim declined significantly tested by two experiments - one in which the chambers with number of minims. Each additional minim caused a were divided by fine screening so that queens shared the ,drop of 0.00061 mg in mean weight. Number of minims airspace but had no physical contact, and the second in accounted for 39% of the total variance, while treatment which extracts of queens were applied to filter papers accounted for 16% (total 54%). which were used to line the chambers of solo queens. In addition to producing heavier minims, queens While the first experiment indicated that airborne factors given access to food lost less weight than controls might account for about a quarter of the reduction in (Fig. 12B). For equal minim production, carnivore output per queen, the second experiment failed to con- queens weighed 0.72 mg more than did controls, while firm a role for an extractable, airborne factor. Most of the cannibal queens weighed a whopping 2.48 mg (76%) decline in output per queen must result from factors re-

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 221 quiring queen contact. Whereas contact pheromones haps on physical factors. Whetherqueen behavior and might play a role in the inhibition of queen fertilityin physiology integrates all this complex information is founding associations, it seems unnecessaryto invoke presentlyunknown. pheromonesto explain the decline in brood output per Queen "condition"varied significantly among mating queen.The observedbrood attritionsuggested cannibal- flights, and these differenceswere detectable in queen ism. Brood cannibalismis common in ants (Holldobler weight, fat content, and in the production of brood. and Wilson 1990). Queenslost more than half theirweight during founding. Energyloss was almost 70%,and queenswere 11% lean- er after founding than before. This level of weight and Discussion energy loss is in line with other ant species that found colonies claustrally and independently (Keller and In Solenopsisinvicta, as in the majorityof ants, colony Passera 1989). foundingtakes place in the isolation of a sealedchamber. Predictinga queen'sminim production from her phys- Founding queens may join others but the founding ical characteristicswould be of interest.However, queen chamber never contains workers until the first brood live weight had no predictivevalue for queen fertility. ecloses.Whether alone or in groups,once the queensare This brings into question the assumption of Nonacs sealed in the foundingchamber, their total materialand (1992)that a queen'sweight is predictiveof her reproduc- labor resourcesare fixed.Their success,individually and tive competitiveness.The mean dry weights of queens as a group, may depend on how they allocate their re- from the differentmating flights correlatedweakly with sources to worker production, maintenanceand post- the differencesin mean production,but these relation- claustral reserves.In spite of its apparent simplicity,a ships are of no use in the predictionof individualfertility. numberof physicaland social influencesaffect the nature It is clear that brood productionis greatly affectedby of the firstbrood and the post-foundingqueen and there- variablesnot measuredin this study, other than weight. fore the success of the queens and associations. Within a single cohort of queens, brood production In this study, the fate of foundingnests was followed varied from none to 32 minims, and this was accompa- until most minim workers had eclosed. Although pro- nied by a weight loss of 4-12 mg, and an energy loss of duction of minimsmeans that the queens have been suc- 110-280 J. cessful in founding up to that point, becoming the re- Choiceof foundingnest site carriesa trade-off:queens productrixof a mature colony requiresthat the queen producedmore brood and survivedbetter at the cooler, surviveseveral ensuing phases: execution of supernumer- shady site, but they producedminims much earlierat the ary queens by minim workers (Tschinkeland Howard sunny site, even though their total productionwas lower 1983); brood raiding and queen usurpation in which there.It seemslikely that earlyadult minimproduction is neighboringincipient colonies attemptto steal each oth- more important than total brood, because pre-adult ers brood, and queens of losing nests attempt to usurp brood convey no advantage during brood raiding those of successfulones (Tschinkel1992a, b); earlycolony (Tschinkel1992a). The lower total productionat higher growth and territorial competition in which small temperatures may result from proportionally higher colonies can be overrunby largerones (unpublishedob- maintenancecosts at warmersites. servations).Clearly, "success" during the phase which is Why are some queens so much more productivethan the subjectof this paperis still farfrom assuring a queen's others? Consideringthe importanceof high minim pro- ultimatereproductive success. Factors contributing to an duction to the success of incipient colonies (Tschinkel individualqueen's ultimate success are of great interest, 1992a),it is difficultto arguethat low productionis adap- but were not the subjectof this paper. tive. Severalpossibilities can be suggested.Perhaps un- Groups of co-founding queens are more successful productivequeens could be productiveunder other con- duringthe claustraland incipientperiods than are single ditions(and vice versa).This hypothesiswould, of course, queens (Tschinkeland Howard 1983; Tschinkel 1992a). be more than a little difficultto test. Perhapsunder some In nature,pleometrosis was shown to be drivenby local conditions,low productivitycould be advantageous,by queen density and microtopography (Tschinkel and reservingmore stored nutrientsin the queen and allow- Howard 1983). Queens were more likely to join when ing longerpost-claustral periods without feeding. Queens local density was higher, leading to a highly clumped oftenfound nests in bareareas before much plant growth distributionof queensin nest chambers.Before the claus- has occurred, and it is my impression that incipient tral chamberis sealed, there is a great deal of joining, colonies do succumbto starvationsometimes. leaving and mutual inspection (pers. obs.), raising the Anotherpossibility is that low productionby an indi- possibilitythat queens are making active choices among vidual is advantageouswhen foundingin groups, again potential co-foundresseson the basis of accumulatedin- because more body reservesremain at workereclosion, formation.Some of the potentialeffects of those choices allowingthe individuala longerperiod in which to com- have been describedin this paper, but whetherthese ef- pete for reproductivedominance. Perhaps much of the fectsare translatedinto greaterultimate success is known variationsimply representsthe spreadof norms of reac- only in the case of workernumber (Tschinkel 1992a). The tion in populationsof newly mated queens. benefitsof joining an associationdepend upon the num- The size of minims was not fixed, but decreasedin ber of queens already in the group, the overall density nests with more minims.Although Wood and Tschinkel (and thereforefuture competition)in the area and per- (1981)showed that the size of minimscould be increased

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 222 by rearingthem in worker-containingnests, the present and effectswould have to be much lower,because fecun- data show that influenceswithin the foundingchamber dity inhibitionwas not always detectable.On the other prior to workereclosion also affectminim size. The na- hand, polygyny results in larger workersduring found- ture of these influences is similar to factors affecting ing, but smallerones in maturecolonies. workersize in colonies throughouttheir lives. Porterand Non-embryonated,trophic eggs are laid by both in- Tschinkel(1985b) showed that when a fixed numberof seminated and uninseminated queens in polygyne workerswas allowedto rearlarger numbers of larvae,the colonies. Vargo and Ross (1989) showed that polygyny resultingpupae were smaller.They suggestedthat as the reduced the embryonationrate of inseminatedqueens, ratio of care-giversto larvae decreased,the final size of but not uninseminatedones, whose embryonationrates larvae declinedbecause each receivedless care, on aver- weremuch lower to begin with. They suggestedthat mu- age. In the foundingnest, the more larvae a queen pro- tual inhibitionmay be involvedin this reduction.Ovipo- duces, the less care she can lavish on each, and the sition rate did not affectembryonation rate. I did not test smallerthe resultingworker. Tschinkel (1988) has applied for embryonationby individualqueens, but found that these observationsto the explanationof the ontogeny of the overallembryonation rate was not affectedby queen workerpolymorphism in S. invicta.Mean workerweight number.Thus, whereasboth embryonationrate and egg increasessmoothly over 6 ordersof magnitudeof colony laying rate are elementsof fertilityreduction in polygyne size, includingthe founding chamber,suggesting that a colonies, only egg-laying rate reduction operates in single mechanism,perhaps resemblingmass action, de- foundingassociations, and weakly at that. terminesworker size from the foundingchamber to the Finally, does founding chamber size increase with largestcolony. numberof queensin a pleometroticassociation? If so, it Weightsand census of queens and brood suggestthat would suggest that space/queenmay be an important the claustralperiod in pleometroticnests may not be as factor,because crowdingcould affectqueen interaction, peaceful as suggestedby Rissing and Pollock (1987) in efficiency of brood care and disease. Plaster casts of Veromessorpergandei. In this species, queens coexisted founding chambers showed some increase in chamber peacefully until the end of the claustral period, when size with queen number,but certainlyless than propor- fightingbroke out. In S. invicta,grouping does not have tional. The chambervolume per queen in all experimen- large effectson egg-layingby individualqueens, but the tal nests was much largerthan that in all natural,queen- largerthe groupthe higherthe larvalattrition rate during dug nests. On the other hand, nests in soil with their the claustralperiod and the fewerthe survivingbrood per verticaltunnel allowed more effectivedisposal or storage queen. Becausequeens in such groups end the claustral of dead queens, dead brood and waste. In field excava- period with higher body weights, it seems likely that tions, dead queens were often found immediatelyunder brood are cannibalized.The question of who eats whose the soil surfacein the driestpart of the nest, with the live brood is an interestingone, and opens the possibilitythat ants in the lower extreme.What role these factorsplay in queens compete even within the claustralnest through foundingsuccess meritsfuture research. differentiallarval cannibalism. Groups of ten queenstyp- ically producedonly 10-25% as much brood per queen Acknowledgements.I am grateful to Tracey Andreae, Kristen as did solo queens,suggesting a greatdeal of brood attri- Moegling and StephanieClark for competentand cheerfultechnical help, to Duane Meeterfor statisticaladvice and Don McInnesfor a tion beforethe pupal stage. critical reading of the manuscript. This research was carried out Cannibalismof queens was also shown to be an im- under NSF grants No. PCM 8022077,PCM 8309198,BSR 8502969 portant factor contributinggreatly to the post-claustral and BSR 8920710. This is paper no. 23 of the Fire Ant Research weightof survivingqueens, and to minimweight. In most Team. experiments,queen mortality rate was not related to queen number,so it seemslikely that queensdie of "nor- mal"causes and are not killed by chamber-mates.Those References that die, however,are often cannibalizedand appear to contribute to the improved weight and perhaps post- Baroni-UrbaniC (1991) Indiscriminateoophagy by ant larvae: an explanation for brood serial organization.Insectes Soc 38:229- claustralcondition of the cannibals.Higher weight per 239 minimmay have positive effectson minimlongevity and Bartz SH, Holldobler B (1982) Colony founding in Myrmecocystus resistanceto stress, perhaps helping the colony survive mimicusWheeler and the evolution of foundress associations. the incipientperiod. Behav Ecol Sociobiol 10:137-147 There are interestingparallels between polygyny in Dixon WJ (1985) BMDP statistical softwaremanual. University of foundingcolonies (pleometrosis)and in maturecolonies. CaliforniaPress, Berkeley In Fletcher DJC, Blum MS (1983)The inhibitorypheromone of queen mature polygyne colonies, as in one of the two lireants: effectsof disinhibitionon dealation and oviposition by pleometroticexperiments in this study, the fecundityper virgin queens. J Comp Physiol 153:467-475 queen was inverselyrelated to the numberof functional Goetsch W (1937) Die Entstehung der Soldaten im Ameisenstaat. queens in the colony (Vargoand Fletcher 1989).Vargo Naturwissenschaften25:803-808 (1992)showed that the queensin polygynecolonies mu- Holldobler B, EO Wilson (1977) The number of queens: an impor- tually inhibit each others' fecundity by means of a tant trait in ant evolution. Naturwissenschaften64:8-15 pheromone.It seems H6lldobler B, Wilson EO (1990) The ants. Belknap Press of Har- possible that a similarmechanism, vard University, Cambridge perhaps even the same pheromone, operates in Hood GW, TschinkelWR (1990) Dessication resistancein arboreal pleometrotic associations. However, pheromone levels and terrestrialants. Physiol Entomol 15:23-35

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions 223

Keller L, Passera L (1989). Size and fat content of gynes in relation Tschinkel WR (1992a) Brood raiding in the fire ant, Solenopsis to the mode of colony founding in ants (Hymenoptera:Formi- invicta: field and laboratory studies. Ann Entom Soc Am cidae). Oecologia 80:236-240 85:638-696 Markin GP, Collins HL, Dillier JH (1972) Colony founding by TschinkelWR (1992b)Brood raiding and the population dynamics queens of the red imported fire ant, Solenopsisinvicta. Ann En- of founding and incipient colonies of the fire ant, Solenopsis tomol Soc Am 65:1053-1058 invicta.Ecol Entomol 17:179-188 Nonacs P (1992) Queen condition and alate density affect pleomet- TschinkelWR, Howard DF (1983) Colony founding by pleometro- rosis in the ant Lasius pallitarsis.Insectes Soc 39:3-13 sis in the fire ant, Solenopsis invicta. Behav Ecol Sociobiol Passera L (1977) Production des Soldats dans les societes sortant 12:103-113 hibernation chez la fourmi Pheidole pallidula. Insectes Soc Vargo EL (1988) Effects of pleometrosis and colony size on the 24:136-146 production of sexuals in monogyne colonies of the fire ant So- Peakin G (1972) Aspects of productivityin Tetramoriumcaespitum lenopsis invicta. In: TragerJC (ed) Advances in Myrmecology. L. Ekol Pol 20:55-63 E.J. Brill, New York Petersen-BraunM (1977) Untersuchungenzur sozialen Organisa- Vargo EL (1992) Mutual pheromonal inhibition among queens in tion der Pharaoameise, Monomoriumpharaonis. Insectes Soc polygyne colonies of the fire ant Solenopsisinvicta. Behav Ecol 24:303-318 Sociobiol 31:205-210 Pollock GB, Rissing SW (1989) Interspecificbrood raiding,territo- Vargo EL, Fletcher DJC (1989) On the relationshipbetween queen riality and slavery in ants. Am Nat 133:61-70 number and fecundity in polygyne colonies of the fire ant So- Porter SD, Tschinkel WR (1985a) Fire ant polymorphism:the er- lenopsisinvicta. Physiol Entomol 14:223-232 gonomics of brood production. Behav Ecol Sociobiol 16:323- Vargo EL, Ross KG (1989) Differential viability of eggs laid by 336 queens in polygyne colonies of the fire ant, Solenopsisinvicta. J Porter SD, Tschinkel WR (1985b) Fire ant polymorphism (Hy- Insect Physiol 35:587-593 menoptera:Formicidae): factors affectingworker size. Ann En- VanderMeer RK (1986) Chemical taxonomy as a tool for separat- tomol Soc Am 78:381-386 ing Solenopsisspp. In: Lofgren CS, VanderMeer RK (eds) Fire Porter SD, TschinkelWR (1986) Adaptive value of nanitic workers ants and leaf-cutter ants: biology and managment. Westview, in incipient fire ant colonies. Ann Entomol Soc Am 79:723-726 Boulder, pp. 316-326 Rissing SW, Pollock GB (1987) Queen aggression,pleometrotic ad- Voss SH (1985) Rapid, simple DNA staining for fire ant eggs. J vantage and brood raiding in the ant Veromessorpagandei. An- Entomol Sci 20:47-49 im Behav 35:975-981 Voss SH, Blum MS (1987) Trophic and embryonated egg produc- Ryan TA, Joiner BL, Ryan BF (1982) Minitab referencemanual. tion in founding colonies of the fire ant Solenopsisinvicta (Hy- Duxburg Press, Boston menoptera: Formicidae).Sociobiol 13:271-278 SchmuckLM, Metz CB (1931)A method for study of chromosomes WaloffN (1957)The effectof the numberof queens of the ant Lasius in entire insect eggs. Science 74:600-601 flavus (Fab.)(Hym., Formicidae) on their survivaland on the rate Toom PM, Cupp E, Johnson CP, Griffin1 (1976)Utilization of body of development of the first brood. Insectes Soc 4:391-408 reservesfor minim brood developmentby queens of the import- Wood LA, TschinkelWR (1981)Quantification and modificationof ed fire ant, Solenopsisinvicta. J Insect Physiol 22:217-220 worker size variation in the fire ant Solenopsisinvicta. Insectes Tschinkel WR (1988) Colony growth and the ontogeny of worker Soc 28:117-128 polymorphism in the fire ant, Solenopsis invicta. Behav Ecol Sociobiol 22:103-115

This content downloaded from 128.186.14.5 on Tue, 03 Nov 2015 17:37:58 UTC All use subject to JSTOR Terms and Conditions