Geographicalvariationinbloodparasitesinferalpigeons:therole ofvectors



Prevalenceandintensityofbloodparasites areknowntovaryinspacewithinasame species,yet the causes underlying such variation are poorly known. Theoretically, blood parasites variation can be attributedto differences to exposure to parasite vectorsand/or todifferencesinhostsusceptibility. Here,weshowthat prevalenceof HaemoproteuscolumbaeinferalpigeonsColumbalivia variedamong fivenear-by populations(range15%–100%),paralleled byvariation intheabundance ofitsmain vector,thelousefliesPseudolynchiacanariensis.Geographic variation inintensity of blood parasites didnotcovarywithabundance ofvectors.Withinpopulations,older individuals hadahigherprobabilityofbeingparasitized than youngerones,whereas younger birds, wheninfected, sufferedhigher intensities. Furthermore,wefound no evidence of sex-related differences neither in prevalence nor in intensity of blood parasite infections. To demonstratethat geographical variation in prevalence was actually duetodifferencesinvectorexposure, weconducted twoexperiments based ontranslocationofunparasitizedpigeons from avector-free area to an area where both theparasite and vector wereabundant.With thefirstexperiment, wedemon- strated thatunparasitized pigeonswerenotresistant totheparasitebecausewhen transmissionwas possible pigeons became parasitized in a fewmonths. With the secondexperiment, inwhichhalfofthepigeons wereprevented from contacts with the vector, weruled out the posibility that pigeons weconsidered asunparasitized would have suffered from latent infections. Therefore, both observational and experimentalevidencesupports theview thatvectorabundance isthemajorfactor influencingthespatial variation inprevalence ofH.columbaeinpigeons.

D.Sol ()andR. Jovani,Dept deBiologiaAnimal (Vertebrats), Univ.deBarcelona,Av.Diagonal645,E-08028Barcelona,Spain.–J.Torres,Deptde Parasitologia,Univ.deBarcelona,Av.Diagonal643,E-08028Barcelona,Spain.

Parasitescannegativelyaffectdifferentfitness compo- nents of their hosts and, hence, they may potentially exert importantecological and evolutionary pressures on hosts (reviewedinMøller 1997).The role ofpara- sites in sexual selection has received special attention eversinceHamiltonandZuk(1982) suggestedthe hypothesis that secondary sexual characters have evolved to signal heritable resistance to parasites in matechoicecontext.Subsequent interspecifictestsusing blood parasite prevalence haveprovided both support andconflictingevidence forthehypothesis.Thesedis- crepanciesmightinpart beduetothelackofattention that hasbeenpaidtotheparasites’biology(Endlerand

Lyles 1989, Weatherhead and Bennett 1991, 1992). TestsoftheHamiltonandZuk(1982)hypothesis using blood parasites assume that one can reliably estimate the parasite prevalence ofeach host species,and that failure to detect the parasite in some hosts reflects resistance andnotdifferential exposuretotheparasite. Yet, the informationconcerning within speciesvaria- tioninbloodparasites iscurrently toolimitedtoensure that theseassumptions arefulfilled(Merila¨ etal.1995). Clearly, westillneedto learn much more about para- sites beforebeingabletoconductrobusttestsofhy- pothesesconcerningtheecological andevolutionary consequences ofhost-parasiterelationships.

Although prevalenceandintensity ofblood parasites are knowntovarybothspatiallyandtemporally (Weatherheadand Bennett 1991,1992,Bennett et al.

1995, Merila¨etal.1995, Boschetal.1997), veryfew studieshavequantifiedtheextentof thisvariationand, consequently, itscausesarenotwellunderstood.Inthis paper, weconsider thegeographic variation inhaema- tozoan parasites inferalpigeons Columbalivia,Gmel. Haematozoan parasites are obligate parasites, which useblood-sucking Diptera asvectors(Valkiunas 1996). Forsuchvector-transmittedmicroparasites,variation in prevalence canariseboth from factors intrinsic to the hosts (e.g. genotype resistance, behaviour or state of health)andfromextrinsicfactors, suchasdifferencesin exposure to vectors (van Riper et al. 1986,Atkinson and van Riper 1991, Holmstad and Skorping 1998, Wiehnetal.1999). Parasitemia canalsobeageorsex depended (e.g. Weatherhead and Bennett 1992, Mc- Curdy et al. 1998), so these factors need to be con- trolled toavoid biasintheresults.

The specificaims of our study weretherefore two- fold.We firstanalysedthegeographicvariation in prevalenceandintensityof bloodparasitesamongfive nearby populationsofferalpigeons. Becauseconsider- ablevariation inprevalencewasfound, wethenstudied theinfluence ofintrinsicandextrinsicfactorsin deter- miningthespatialvariation. Toinvestigatewhether geographicaldifferencesinprevalencewere duetothe degreeofvectorexposure, rather than duetoresistance toparasites, weperformed aseriesofexperimentsbased ontranslocationofpigeons from avector-free area to an area where vectors were abundant. We reasoned that, ifalowprevalenceinapopulationwasduetothe scarcityof vectors,prevalenceshouldincreasewhen exposingindividuals tothevector.

Vertebratehost,haemoatozoaparasiteand louseflyvector

Feral pigeonsstemfromdomesticated rockdovesthat, after intense artificial selection, escaped or were re- leased and established themselves in rural and urban areas (Johnstonand Janiga 1995).Pigeons are seden- tary and undertakeforaging, flight,breeding, and roostinginflocks oraggregationsofflocks. Theyare granivoresandfeedinthestreetsofcities orinfields surroundingcitiesandvillages.Although urban pigeons habituatetopeople,rural pigeonsareastimidasother wildspeciesofbirds. Asaresult ofartificial selection, feral pigeonsareabletoreproducealmostcontinuously although in practice most individuals stop breeding during thecoldermonths. Malepigeonsdefendasmall nest-siteterritory andparticipate innestbuilding, incu- bation andfeedingthenestlings.Nestsareusuallybuilt inholesandtheyareoftenre-usedinsubsequent breed- ingattemps.

All haemoparasites observed in the blood samples wereidentifiedasHaemoproteuscolumbae(Kruse;Api- complexa:Haematozoa)according tothedescription of the macro and microgametocytes reported in Bennett andPeirce(1990).Protozoa belongingtothegenus Haemoproteusareprimarily parasites ofbirdsandhave theirsexualphasesin insectsotherthanmosquitoes (reviewedinSchmidt and Roberts 1989,Atkinson and van Riper 1991). Haemoproteus columbae, the type species of the genus, is a cosmopolitan parasite of pigeons. After 21–32dprepatentperiod (Ahmed and Mohammed 1977, cited in Atkinson and van Riper

1991),theasexualphase(merozoites) entererithrocytes to become pigmented gametocytes in the circulating blood.Thepathogenesis inpigeonsseems tobeslight, although atpresent itisnot wellknown (seeAtkinson and vanRiper 1991and referencestherein).

Hippoboscidflies (Diptera, Hippoboscidae)are,at present, theonlyknown vectorofH.columbaeinferal pigeons (Bennett and Peirce 1990and references therein). Bennett andPeirce(1990)suggestthat cerato- pogonid flies,a group of vectors of Haemoproteus in other pigeonspecies,couldalsobeasuitable vectorof H. columbae.However, ceratopogonidfliesare associ- atedwithforestorhumidzonesrather thanurban areas (Kettle1982),andhence,itseemsunlikelythattheycan actasvectorsinourH.columbae-pigeonsystem.Inthe secondexperiment (seebelow),wereport anadditional argument tosupport thisidea.Samplesofflies were captured andidentifiedaspigeonflies Pseudolynchia canariensisaccording to Theodor and Oldroyd (1964). Thepigeonfly isawinged,blood-sucking parasite commoninpigeonsthroughout mostofthewarm, temperate regions of the world. Louse fliesuse bird nests for reproducing and, perhaps for this reason, oftenshowaseasonalpatternof abundance witha maximum just after the breeding season (but seeFig.


Fig. 1. Seasonal variation in louse fly prevalence in adult pigeons from Barcelona (empty bars) and Moia` (solid bars). The absence of bar in spring and summer indicates zero prevalence. Sample sizeswere,from spring to winter, 34,29,

40 and 18 pigeons for Barcelona, and 23, 30, 45 and 47 pigeons forMoia`,respectively.

Table 1. Characteristicsofthefivepigeonpopulationsstudied.

PopulationnameNumber ofCapturedPeriod ofcapturesHabitattype

captures / juveniles / adults
Agramunt / 1 / 8 / 9 / July / rural
Barcelona / 12 / 38 / 52 / March–July / urban
Granollers / 2 / 26 / 32 / April–June / urban
Moia` / 2 / 2 / 62 / February–March / rural
OlesadeMontserrat / 1 / 17 / 11 / July / rural


We sampled free-ranging pigeons from Februaryto July1998infiveareas ofCatalonia,NE Spain. These areas are representative of the most typical habitats used by pigeons, i.e. urban and rural habitats. Al- though thedifferent areas wererelativelyclosetoeach other(minimumandmaximumdistancesbetween two populationswere,respectively,25and50km),we are confidentthatpigeonsfromeach areaformtruepopu- lations(seeSolandSenar1995).Theperiod ofcapture coincided with theperiod inwhichadult pigeons tend to besexuallyactive.Oneto 12captures werecarried outineachpopulationbetween10.00and16.00husing Yunik traps. Data on the month of capture, sample effortandtypeofhabitat inthefive populationsare reported in Table 1.In Barcelona, pigeons werecap- tured at twosites;thezoological garden and thecam- pusof theUniv.of Barcelona.Becauseenoughdatato provide statistically reliableresults wereonlyavailable forthreepopulations(see Table1),withinpopulation analyseswillberestricted tothesepopulations.

Aftercapture, pigeonswereenclosedinacagebefore being examined,whichgenerallyoccurredimmediately (delayof severalhoursin examiningpigeonsonlyoc- curredintheAgramunt sample).Foreachpigeon,we recorded age,sexand intensity ofblood parasites, and number oflouseflies.Agewasassessedbyexamining iriscolour, cerethickness and unmolted juvenilefeath- ers, which allow to distinguish juveniles(6months old)and adults (Kautz 1985).Sexofadults wasdeter- minedbyexaminingthewidth oftheisquion, whichis atraditionaltechnique adopted bypigeonkeepersthat hasbeenshowntobeareliablewaytosexpigeons(Sol unpubl.); assignmentofsomebirdsofuncertain sexwas confirmed byobservationofsexualbehaviour.

For each pigeon, one or two blood smears were obtained fromadropofbloodsqueezedfromaclipped toenail onto aglassslide(Davidar and Morton1993). Each slidewasrapidly air dried, fixedwith methanol for10minandstainedwithGiemsastain.Amagnifica- tionof×1000wasusedtoidentifyandquantify blood parasites. Bloodslideswereexaminedfor10minbefore declarednegative.Forpositiveslides, thenumberof parasited cells wascounted in100microscopic fieldsby moving randomly over the slide in areas where the

bloodcells formedamonolayer (Bennettetal.1995). Intensityis expressedasthenumberofparasitized erythrocytes per100randomly selectedfields.Inafew cases, erythrocytes were disposed in multilayers, and thesesmearswerenot usedintheanalyses.

Pigeonflyabundance wasestimated bycounting the number offliesduring10-minsearchforallpartsofthe body. Because pigeons were enclosed in cages before being examined, itispossible that pigeon fliesmoved among thehosts,andhence,comparison offly loads betweenagesandsexes shouldbetakenwithcaution. It should benoted, however,that suchaggregations were not much different from roosting orforaging aggrega- tions ofpigeons.

Parasitism is known to vary among seasons and years,whichcouldhavebiasedsomeof ourresults.To evaluate thepotentially moreimportantsourceofbias, i.e.theseasonal variation inpigeonflies,weuseddata gathered fromspring1997towinter1998inMoia` and Barcelona populations.Toanalyseiftheobservedgeo- graphicpatterns holdbetweenyears,wecompared data gatheredinthespringsof1997 and1998 fromthe Barcelona populationand in the winters of 1998and

1999from theMoia` population(seeResults).


Because wefound significant differences inprevalence ofH.columbaebetweenpopulations(see Results),we furtherinvestigatedthereasonsforsuch differencesby means ofaseriesofexperiments. The firstexperiment wasdesignedtotestifunparasitizedpigeonswereresis- tant or susceptible to theparasite. In March 1998,we moved 48of the 52unparasitizedadult pigeons cap- turedinMoia` (whereprevalenceofbloodparasites was lowandlouseflies werescarce)toBarcelona (where prevalence ofblood parasites washigh and louseflies were relatively abundant). Pigeons were captured as partofacontrolprogramme conducted bytheCity Council. Individuals were divided at random in four groupsof12birds,andeachofthegroupswasenclosed in a 100×60×60 cm outdooraviary for 5 months. Throughoutthestudy, pigeonshad freeaccesstofood (amixtureofseedsspecialforpigeons),waterandgrit. To facilitate vector transmission, we attracted local

free-rangingpigeonstotheexperimentalcages using artificialfeeders.Presenceofblood parasites andabun- dance oflouse fliesofpigeons weredetermined every two months. At the end of the study, pigeons were released.

Becausepigeonswere foundtobesusceptibletoH. columbaeinfection,wedesignedasecondexperimentto ruleout thepossibilitythat theparasite wasinlatency and emerged asaconsequence ofadebilitationofthe resistance against parasites (perhaps due to social stress, diet deficiencies or captivity). The experiment was alsousefulinrulingoutthepossibilitythatH. columbaewastransmittedbyceratopogonidflies. The protocol wasthesameusedinthefirstexperiment,with the only difference that in this case two of the four outdoorcageswerecoveredwithanettokeeppigeon flies out. However, the holes of the net were large enough to permit the entrance of ceratopogonidflies. Wepredicted that ifH.columbaewasnotinlatencyin thepigeons and pigeonflieswerethesolevector capa- bleoftransmittingtheparasite, onlypigeons from the experimental groups would become parasited after a certain time. However, ifany of these statements did not hold, pigeons from both the experimental and control groups would become parasitized. The experi- mentwascarriedoutwithsix unparasitizedpigeons captured inMoia` inFebruary 1999,andwitheight additional pigeons captured in March 1999. Pigeons from each of these groups were randomly divided amongthecontrolandexperimentalcages. Amongthe pigeonscaptured inMarch, onewasalreadyparasitized with H. columbae;this individual wasenclosed inthe experimental cageafter ensuring that itwasnot carry- inganypigeonflies.Attheendofthestudy,allpigeons werereleased.


Testsofindependencewere performedwiththeFisher exacttestforsmallsample sizes(n30)and withthe x2-test in the other situations. Continuous variables were analysedwithparametric testswhentheymetthe criteria ofnormality. For variables that did not meet thatcriteriaandcouldnotbe normalisedby standard transformations, alternative non-parametric testswere used(seebelow).


Ofatotal of257pigeons examined, 206wereinfected with H. columbae.Because sexor age of hosts could affectparasite prevalenceand intensity ofparasitation, we analysed sex and age-related differences in para- sitemiawithin populationsbefore exploring geographi- cal differences. In the only two populations where samplesize ofmalesandfemaleswerelargeenough (Barcelona and Granollers),there wasnoevidencefor sex-related differences in prevalence, intensity, or in vector abundance (Table 2).Wetherefore pooled data onboth sexesforfurther analyses.

Whenage-classeswerecompared withinpopulations, prevalenceofH.columbaeinjuvenileswasconsistently lower than that in adults, whilethe parasite intensity wassignificantly higher (Table 3).Abundanceof flies didnotdifferbetweenage-classes(Table3).Becauseof theage-relateddifferences,thefollowing analyseswere carried out separately forjuvenilesand adults.

Geographicalvariationinbloodparasitesand vectors

Prevalence ofH. columbaeshowed considerable varia- tion among populations(Table 3).For adults, preva- lencewas100%infourofthefivepopulations,whileit was extremely low (14.8%) for the Moia` population (x2=121.17, p=0.0001). A similar figure emerged when comparing prevalence in juveniles, although in Moia` thesamplesize(N=2)wastoolowtobeconclu- sive.Regarding intensityofH.columbae,nodifferences werefound betweenpopulations(Table 3).

Differences between populations in abundance of louseflies werestatistically significant,andasimilar pattern ofabundance wasfound inboth juvenilesand adults(Pearsonproduct-momentcorrelation between mean number of fliesin juveniles vsadults: r=0.94, N=5, p=0.017). The population with low parasite prevalencewas preciselytheonewherevectorswere scarcer.

The low parasite prevalence and vector abundance foundinMoia` cannot beanartifact associated withthe method of sampling, because the same method was usedforthefivepopulations.Neither canitbeaneffect

Table 2. Prevalence (%ofbirds infected) and intensity (mean number ofredblood cellsinfectedamong parasitized birds) of Haemoproteuscolumbae,andlousefly abundance (meannumber offliesperhost)infemaleandmalepigeonsfromtwonear-by populations.(Samplesize:Barcelona: females=13,males=14;Granollers: females=12,males=17).



IntensityLouse flies

females / males / p / females / males / p / females / males / p
Barcelona / 100 / 100 / n.s.a / 1.32(0.79) / 1.46(0.34) / n.s.b / 1.29(0.34) / 2.00(0.66) / n.s.b
Granollers / 100 / 100 / n.s.a / 3.45(1.55) / 2.37(0.64) / n.s.b / 0.69(0.29) / 0.39(0.18) / n.s.b

Note: ax2;bt-test.

Table3. Prevalence(%ofbirdsinfected)andintensity(meannumber ofredblood cellsinfectedamong parasitized birds)ofH.

columbae,andpigeonflyabundance (meannumber offliesperhost)injuvenileandadultpigeonsfromfivenear-bypopulations.


PrevalenceIntensity (SE)

Louse flies(SE)

Agramunt / 37.5 / 100.0 / 0.009d / – / 2.24(1.26) / – / 0.38(0.38) / 0.11(0.11) / n.s.b
Barcelona / 85.3 / 100.0 / 0.028a / 8.21(1.75) / 1.47(0.43) / 0.001c / 1.43(0.31) / 1.49(0.24) / n.s.b
Granollers86.9100.00.088a 7.71(1.93)2.82(0.73)0.025c0.35(0.09)0.52(0.16) n.s.c
Moia` / 0.0 / 14.8 / – / – / 1.55(1.18) / – / – / 0.02(0.02) / –
Olesade / 93.8 / 100.0 / n.s.d / 9.51(2.80) / 1.97(0.65) / 0.004c / 0.64(0.24) / 0.40(0.16) / n.s.b


Note: ax2 test; bt-test; ct-test with separate variance estimates, dFisher exact test, eF-test, fKruskall-Wallistest. n.s. non significant. Thedash denotes insufficientsample size.

ofseason(Moia`was theearliestpopulationsampled), becauseinMoia` vectorswerescarceallyear,while in Barcelonatheywerealwaysabundant (Fig.1).Inter- population differences in louse flies loads and H. columbae prevalencealsoholdbetweenyears.Forex- ample,thecapture carriedoutinMoia` betweenFebru- ary and March 1999gaveonly three out of15adults withH.columbae,andnoindividualwithpigeonflies,a figure very similar to that given in Table 3. In Barcelona, ontheother hand, thesameprevalences of both H. columbaeand pigeonfliesweremeasured ina sampleof34 adultpigeonscaptured in1997,andina sampleof50other adults captured in1998atthesame siteand season (100%prevalence ofH. columbaeand

50%prevalence ofpigeonfliesinboth samples).


The aim ofthe firstexperiment wasto test ifpigeons from Moia` (where prevalence of blood parasites was low) were susceptibleorresistanttoH.columbae. We therefore translocatedpigeons(N=48)fromMoia` toa sitewhereboth theparasite and thevectorwereabun- dant. After two months of exposure to the vector in outdoor aviaries,prevalenceofH.columbaeincreased from14.8%to53.9%, andtwomonthslaternearlyall pigeons(95%) wereparasitized (Fig.2,above);the remainingunparasitized pigeonsbecameparasitized during thenextmonth. Louseflyabundance paralleled theincreaseinparasite prevalence(Fig.2,below).This resultdemonstrates thatpigeonsfromMoia` werenot resistant but susceptible totheparasite.

Inthesecondexperiment, weenclosedhalfofatotal of14pigeonsintwonormal outdoorcages(experimen- talgroup)andtheotherhalfintwosimilarcages but coveredwitha netthatpreventedanycontactwiththe pigeonflyvector(control group). Aftertwomonths, all individuals from the experimental groups were para- sitedbutnoneof theindividualsin thecontrolgroup. Pigeons from the control groups werenot parasitized

evenafter fourmonths. Louseflieswereonlyfound in pigeonsfromthe experimentalgroups.Thisexperiment demonstratesthat theparasite wasnotinlatencyinthe host, but was acquired via vector transmission.In addition, theresult supports theideathat ceratopogo- nid fliesdo not act as vectors of H. columbaein the study area (seeMethods).


Although itiswellestablished that infections byblood parasites canvaryinspace(Weatherheadand Bennett

1991, 1992, Bennett et al. 1995, Merila¨ et al. 1995, Boschetal.1997),theextent and causesofthisvaria- tionarepoorly known. Here,weshowthat geographic variation inprevalenceofH.columbaeinferalpigeons

Fig. 2. Progression oftheprevalence ofH. columbae(above) and abundance ofpigeon flies(below)inpigeons from Moia` exposedtothevector inexperimental cages(seeMethods).

canbehighandthatsuchvariation canbeattributedto geographic variation in abundance of vectors. These results are in accordance with the increasingly recog- nisedviewthat vectors playacentral roleinparasite- host relationships (Anderson and May 1982, Ewald

1983,Piersma 1997,Poulin 1998).

Our results are in line with previous findings that older individuals haveahigher probabilitytobepara- sitized by haematozoanparasites than younger ones, whileyoungerbirds,wheninfected,sufferhigherinten- sities (e.g.Seutin1994, Merila¨etal.1995). Ahigher prevalenceinadultsmightbetheresultofalongertime ofexposure to theparasites. However, parasite preva- lenceinjuvenileswasalsohigh(upto94%)inmostof thestudied populations,whichsuggeststhat infections generallyoccuredatanearlyage. Thelowerparasite intensity inadults, ontheother hand, istobeexpected if older birds acquire a certain degree of immunity against parasites (Merila¨ et al. 1995). Alternatively, adultswithhighintensityofparasitescouldbe under- represented inthepopulationduetotheirhigherriskof mortality; thislasthypothesis seemshoweverlessprob- ableinour casebecausecurrent evidenceindicate that H.columbae rarelyproducesmortality inpigeons(see Atkinson and vanRiper 1991).

We foundnoevidenceforsex-relateddifferencesin prevalenceorintensityofbloodparasites (seeMcCurdy etal.1998). Becauseinoursystemparasiteprevalence largelydepends onexposuretovectors(seebelow),the absenceofsex-relateddifferencesinprevalencesuggests that the sexeswere equally exposed to vectors. This agreeswiththefactthat, inpigeons,both sexesusethe habitat and participate in the breeding duties in a similarfashion(Johnston andJaniga1995).Lackof differencesinlousefly abundance betweensexesalso points tothisconclusion, although wecannot ruleout thepossibilitythat thiswasaresultofare-distribution offliesamong pigeons when they weremaintainedin thecagesbeforebeingexamined(butseeMethods). On the other hand, the similarity inintensity between fe- malesand malescanmean that both sexesareequally susceptibletotheparasite, whichinturn canmeanthat the supposed immunosuppressiveeffectofcertain sex- ualhormones (Folstad andKarter 1992) areoflittle importance inthecaseofhaemoproteus parasitesin pigeons.

Amajor findingofthisstudywasthat prevalenceof H.columbaecanvarydramatically among pigeonpop- ulations, evenwithinarelativelyrestrictedgeographical region. Parasite prevalence ranged from 15%to 100% foradults, andfrom0to94%forjuveniles.Incontrast, the intensity ofthe parasite within each age-classwas verysimilar among populations.Geographic variation inprevalence can theoretically beattributedto differ- ences inexposuretoparasitevectorsand/or hostvul- nerability toparasites. Inthisstudy, observationaland experimental evidenceindicate that itistheabundance

ofvectorthat playsamajor roleincausingdifferences inprevalence. In the populationfrom Moia`,inwhich parasite prevalencewaslow,louseflieswereveryscarce ascompared withtheother populations.Yet,thisrela- tionshipdoesnotimply causalitybecauseotherfactors, suchasgeneticpredisposition, stateofhealthorbe- haviour, could alsovaryamong populationsinasimi- larway.Wetherefore conducted experiments todiscern betweenthecompetinghypotheses. Thefirstexperiment demonstrated thatpigeonswerenotresistant butsus- ceptibletotheparasite, suggestingthat among popula- tion differences in parasitemia were largely due to among populationdifferences in exposure to vectors. The second experiment provided further evidence for this, demonstratingthat the parasites were not in la- tencyinthe hosts but they wereacquired viavectors. Thus, a second major finding of our study is that spatial variation in prevalence of H. columbae was linked totheabundance ofvectors.

Theoretical models ofhost-parasiteinteractions pre- dict that, for a parasite to persist and spread in a population,itisnotthevectordensitybutthevectorto hostratiowhatmustexceedacriticallevel(Begonetal.

1996, Hudson and Dobson 1997). According to the results,thedensityofvectors,andnotthehostdensity, is the key factor in our parasite-host system. This assertion isalso supportedbythe observationthat in Moia`,wherethevectorswerescarceand prevalenceof H. columbaewaslow,most pigeons rest and breed in the same building and often concentrateto forage in thesamegrainfactory.Thissuggeststhatin Moia`the effectivehostdensitywasatleastashighasthat inthe other populations.On the other hand, our data also show that low densities of pigeon flies are able to maintainahighprevalenceof bloodparasitesin the population(e.g. Agramuntpopulation),a result that has already been reported in other studies (e.g. van Riper etal.1986).Inour case,hightransmissionrates can have been favoured both for the mobility of the louseflyandtheopportunitiesfortransmissionthat the hostoffers(e.g. flockingbehaviour andcontinuous breeding).

Determining thecausesofvariation invectorabun- dance between populationswas not the scope of this study but, nevertheless, some hypotheses can be ad- vanced.Theseincludevariations inlocalclimaticcondi- tions(Senaretal.1994, Valkiunas1996), habitat characteristics (Figuerola 1999,Tella et al. 1999)and anti-parasitevariation in responses of the host (Hart

1997).With data from onlyfivepopulations,littlecan besaidinsupport ofoneoranother hypothesis rather than tonote that vectors tended tobemore abundant in urbanareasthanin ruralareas.Ingeneral,vector abundance isassumed toattain thehighestdensitiesin forestedareas(e.g.Merila¨ etal.1995,Tellaetal.1999). Although this can betrue for certain familiesofvec- tors, thefactthat pigeonflieswereespeciallyabundant