PREDATION AND COLONIALITY IN CLIFF (PETR 0 CHELID ON PYRRHONO TA )

GERALD S. WILKINSON 1 AND GREGORYM. ENGLISH-LOEB2 •Departmentof Biology,University of Californiaat SanDiego, La Jolla, California 92093 USA, and 2Departmentof Biology, San Diego State University, San Diego, California 92182 USA

ABSTRACT.--Thehypothesis that the time requiredto detectan approachingpredator variesinversely with colonysize was tested.Ninety-five attacks by LoggerheadShrikes (Laniusludovicianus) and AmericanKestrels (Falco sparverius) were simulatedat six Cliff (Petrochelidon pyrrhonota) colonies. Colony size, which ranged from 18 to 320nests, did notexplain a significantportion of thevariance in thetime it tookto detectapproaching predatormodels, even when effectsdue to wind velocity,breeding stage, nest exposure, andair temperaturewere removed. Only air temperatureproved to be a significantpredictor of detectiontime. To determinewhether or not large coloniesmight reducethe relative amount of predationmore effectivelythan smallcolonies, we conductedperiodic predator censusesaround each colony and observedactual predator attacks. Relative predator density increasedonly five-fold, while colonysize increased by a factorof 20. Further,we foundno relationshipbetween attack rate and colonysize. Consequently, we suggestthat larger col- oniesmay dilute the effectof localpredators and therebysuffer less nestling predation on a per nest basis.Received 24 April 1981, accepted1 September1981.

MOST must detect and subsequently attackrate of eachpredator may rise. If nest- evade predatorsif they are to survive. Birds lings are being preyedupon, then the magni- that live in groupsmay have an advantageover tude of the dilution effectwill be greatestwhen solitary individuals, because they can ex- the colonyis highly synchronous.This idea is changeinformation about a predator'slocation quite similar to the predatorswamping or sa- and thereby reduce the risk of predation on tiation hypothesis(Lloyd and Dybas 1966, othergroup members. Although the evolution- Clark and Robertson 1979), which states that ary advantagesof suchalarm calling behavior groupssatiate predators by synchronizingvul- are still in dispute(see Harvey and Greenwood nerabilityin time or space,thus decreasing the 1978,for review), many workersbelieve that probabilitythat any given individualwill be one major evolutionaryadvantage to being in preyedupon. Accordingly,we assessthe va- a largegroup is earlydetection of approaching lidity of the dilution hypothesis by using predators.In this paper we describefield ex- predatorcensuses, direct observationsof at- perimentson the (Petrochelidon tempted predation on swallows,and breeding pyrrhonota)that testthe effectof colonysize on synchronyestimates at Cliff Swallowcolonies the time it takesto detectavian predators,and of different sizes. that demonstratethe ability of Cliff Swallows to discrminate predator from nonpredator METHODS models. In addition, we evaluate an alternative Study subject.---CliffSwallows were chosen as advantageto largecolony size, that of diluting subjectsof this investigationbecause they live in an individual's chancesof being captured(Ber- coloniesof differentsizes, emit high-pitched,single- tram 1978). note alarm calls (Samuel1971a), and show a stereo- The dilution hypothesisstates that any in- typeddisplay once an approachingpredator is de- dividual's probabilityof being capturedby an tected.They may build from 15 to severalthousand attackingpredator decreases as the numberof of their gourd-shapedmud nestsbeneath a bridge, overhangingcliff, or building eave (Mayhew 1958, his neighborsincreases, assuming all have an Samuel1971b). Consequently, only avian predators equal chance of being captured. Obviously, cangain access to theirnests easily. American Kes- this dilution effect will always occur unless trels(Falco sparverius) often attack Cliff Swallowcol- predationincreases with colonysize in com- onies(pers. obs.), as theydo BankSwallow ( pensation.This may occur in two ways.Either riparia) colonies(Freer 1973, Windsor and Emlen the number of predatorsmay increase,or the 1975).They attemptto grab perchedadults (Bonnot

459 The Auk 99: 459-467.July 1982 460 WILKINSONAND ENGLISH-LoEB [Auk, Vol. 99

Fig. 1. The four models:American Kestrel, Loggerhead Shrike, Cliff Swallow,and MourningDove.

1921,this study)or flying swallowsor to reachwith- colonyhad ceasedgiving a mobbingdisplay and had in nests for nestlings (this study). Loggerhead resumed normal activity. Shrikes(Lanius ludovicianus) capture most Cliff Swal- We presentedbetween 2 and 12 models(• = 8.4, lowsby pulling nestlingsout of nestswith their bills SE = 0.5) every9 rain (• = 8.9, SE = 0.4) at eachcol- (this study). ony. We beganeach of the 28 daysof presentations Field experiment.--Wesimulated attacks of avian at 0840(SE = 15 min) PacificDaylight SavingsTime. predatorsat six coloniesin San Diego County, Cal- Any trials in which the modelstopped or otherwise ifornia between 6 April and 9 June1978 to determine was towed unusuallywere omittedfrom all analyses. whetheror not largecolonies could detect approach- Excludingthose problem trials, the mean time, 7.0 ing predatormodels faster than smallercolonies. We s (SE= 0.1, n = 161), required to tow the models toweda stutfedpredator, either a LoggerheadShrike down the guidelines produced a flight speedof 4.3 or American Kestrel, or a stuffed nonpredator con- m/s, which is about one-half our estimates of either trol, either a Cliff Swallow or Mourning Dove (Ze- a shrike or kestrelapproaching a landing. naidamacroura), at the colonies(Fig. 1). The models Using our recordings,written observations,and emergedfrom a smallblind, traveledalong 30 m of the photographs,we measuredthe responsetime horizontalguide lines, which were orientedperpen- and the intensity of the display for each trial. We dicular to the colony stTucture,and then entered usedthe elapsedtime betweenwhen the modelleft another small blind beneath the center of the colony the blind and when the first alarm call was given as (Fig. 2). We designedthe apparatusto mimic an ac- the responsetime. We scoreddisplay intensity on a tual attack; kestrels and shrikes, however, sometimes six-point scale:zero equals no detectableresponse approacheda colonyfrom a greaterdistance or more laterallyto the colonyface. Due to differencesin to- pographyand the nestingstTuctures, the geometry of the apparatuswas similar but unique at eachcol- ony (Table 1). We timed and tape-recordedthe responseusing a stopwatch,Nagra IV-D tape recorderat 7¾2ips, and a SennheiserMKH 815T directional microphone and took a photographof the colonyat the end of the trial from within a blind adjacentto the starting pole. Written descriptionsof the swallows'response were made after each trial, based on a mutual deci- sionby the observers.We randomizedthe order in whichwe presentedthe models(except when testing for habituation and sensitization, in which case we presented the same model for many consecutive Fig. 2. Avian predatorattack simulation appa- trials). The next model was not presenteduntil the ratus. JULY1982] Predationon Cliff Swallows 461

TABLE1. Summary of colony characteristics.

Deviation from Deviation Distance Number of Direction perpen- from below (-) or Exposure active nests model dicular to horizontal above (+) (percentage Name of Nesting flown structure approach nests nests colony 1978 1979 structure from (degrees)(degrees) b (m) facing out) EC 320 674 Bridge W 0 0 -2.5 90-95 DM 147 208 Bridge W 35 0 -5.0 85-90 SF 96 100a Bridge E 19 -5 -3.5 0 SP 92 100a Building N 0 +8 -1.5 50 SV 76 92 Bridge E 8 -2 +1.0 0 SM 18 35 Bridge E 0 -2 -1.0 50 Theseare estimatesbased on a winter inventoryof the nestingsite. Negativevalues refer to downwardslope, positive values to upwardslope. and five indicates that the response was indistin- hatchingwas usedas a measureof colonysynchrony guishablefrom that given to real predatorsduring (Emlen and Demong 1975). an actual attack. We based intermediate intensity Predator censuses.--We periodically censused scoreson the proportion of birds in the colony that American Kestrelsand LoggerheadShrikes in a 1- gave alarm calls and entered the mobbing flock. km circlearound each colony during the 1979breed- Alarm callsalways preceded the formationof a mob- ing season. Five observationstations were estab- bing flock. In an actualattack, once an alarm callwas lished within each censusarea, one placed at the given, most birds vacatedtheir nestsand formed a colonyand the othersdistributed a mean distanceof coordinatedflock that wheeledat the predatorwhile 0.6 km (SE = 0.1) aroundthe colony.At eachstation individuals continually gave alarm calls. We com- an observerspent 5-15 min scanningthe surround- paredmedian mobbing response intensities for each ing vegetationfor perching kestrelsor shrikes. Be- model to determine whether or not swallows could causeboth AmericanKestrels (Balgooyen 1976) and discriminate between models. LoggerheadShrikes (Craig 1978)typically sit on el- We measured exposure of the nest, air tempera- evated perches, they are readily observedover long ture, wind velocity, and breedingstage, in addition distances.At most of the colonies,the surrounding to colonysize, for eachtrial, becauseeach may have terrain was primarily level and covered with low causalrelationships with responsetime. If the first vegetation.Both predatorspecies nested within the swallowto give the alarmtends to be sitting in the census area of at least four of the colonies. The cen- nestopening, then thosecolonies with greaternest susesincluded some of the nesting period of the exposureshould have lower detection times. We ar- predatorsand much of the first clutch of the swal- bitrarily divided nest exposureinto three equal cat- lows. egories•more than two-thirds of the nestsexposed, In addition to the experiments, in 1978 we con- one to two-thirds exposed, or less than one-third ducted 20 h of observations within a blind at the two exposed.If birdsin flight tend to give the first alarm, largestcolonies after hatchingto estimatepredation then detection may be more rapid when insectsare rates. We observedcolony DM, with 147 nests, for most available, such as periods of warm morning 490 min and colony EC, with 320 nests, for 795 min. temperaturesand low wind speeds(Johnson 1969). Statisticalanalyses.--We have used nonparametric If predatorsconcentrate their activitiesduring one statisticswhen the assumptionsof normality or ho- part of the breedingcycle, such as during the nest- mogeneityof varianceswere obviouslyviolated. To ling period, then the swallows may become more ascertainthe relative importanceof colony size and vigilantand detectpredators faster during that stage. the other four factorsin determiningresponse time, Breedingstage was divided into two categories,be- a hierarchical analysisof covariancewas performed. fore and afterhatching. If criesof nestlingscould be Becauseof the inherent lag time before responseto heard from at least half of the nests or at least half any stimulus, we expecteda skewed distribution to- of the returningadults carried food ratherthan nest ward short responsetimes to the simulated attacks. material,we scoredthe breedingstage as after hatch- Therefore, we transformedthe dependent variable, ing. We determinedthe synchronyof the breeding responsetime, logarithmically.The log transforma- stageat two coloniesby periodicallyexamining most tion removed heterogeneity among variances be- of the nestsin the colonyusing a smallflashlight and tween colonies (F•,i• = 4.94, P > 0.05 for the log mirror and estimatingthe meandate of hatchingfor transformed data; F6.•2= 19.68, P < 0.01 for the un- each nest. The standard deviation of the date of transformed data, F-max test, Sokal and Rohlf 1969). 462 WILKINSONAND ENGLISH-LoEB [Auk,Vol. 99

TABLE2. Summary of analysisof covariancefor the determinants of responsetime.

Squared multiple Corre- lation coef- ficient Source (R2) df F-ratio P

Main effects 0.205 6 3.48 <0.005 Colony size 0.009 I 0.91 >0.25 Breeding stage 0.016 I 1.63 >0.10 (3) (2) Nest exposure 0.040 2 2.04 >0.10 100 150 200 250 300 Temperature 0.135 I 13.74 <0.001 COLONY SIZE (NESTS) Wind velocity 0.005 1 0.57 >0.50 Interactions 0.195 13 1.70 >0.05 Fig. 3. Responsetime to the predator models as Explained 0.400 19 2.39 <0.005 a function of colony size. The crossline showsthe Residual 0.600 68 mean, the narrowline representsthe range,and the wide line equalsone standarderror aboveand below the mean. N equalssample size, and the numberof from nonpredator models for us to measure no responsesat each colonyis indicatedin paren- detection time. If we are to extrapolatefrom theses. our results to the natural situation, however, we must show that the Cliff Swallows recog- We used the regressionmethod (Nie et al. 1975)and nize the predator models as potential preda- includedonly two-way interactionterms. The effects tors. In addition, all trials must be statistically of colony size were evaluated independently, and independent to justify an analysis of covari- then the factoror covariatethat explainedthe great- ance. The following evidence supports these est amountof the remainingvariance was enteredin assumptions. subsequent steps. Interaction terms were evaluated We believe that all trials, whether consecu- after the main effects.All significancelevels for this tive or not, can be treated as independent rep- and other analysesare for two-tailed tests. licates for three reasons. First, we found no

RESULTS correlation (Spearmanrank) between display intensity and the interval between consecutive Predatordetection.-•Contrary to expectation, trials (rs = 0.034, P > 0.05, n = 133). We omit- our results indicate that there is no trend be- ted cases in which there was no response tween responsetime and colony size (Fig. 3). from this correlation, because they corre- Furthermore, the analysisof covarianceshows spondedto infintely long responsetimes. Sec- that colony size does not explain a significant ond, the order of presentationdid not signif- portion of the variancewhen entered first (Ta- icantlyaffect response time or display intensity ble 2) or when entered after the other four main (Table 3). Third, neither habituation nor sen- effects (F•,86= 0.52, P = 0.47), nor do nest ex- sitization occurred if the same model was posure, breeding stage, or wind velocity. Air towed repeatedly. We towed the kestrel on two temperature, however, explains a highly sig- series of consecutive trials (of 9 and 6 trials nificant portion of the variance.Thus, if there each) on 19 and 21 May at colony EC. The dis- is a correlationbetween swallow activity and play intensities for all 15 trials were scoredas temperature, this result implies that flying four. The cumulativeresponse times were not swallowsare the first to detectan approaching significantlydifferent from a hypotheticalcu- predator. In supportof this reasoning,we ob- mulative distribution in which the response served flying birds giving initial alarm callson time was held constant(P > 0.20 for each sam- numerous occasions. These results indicate ple, Kolmogorov-Smirnov one-sample test, that for this range of colonysizes, large colo- Siegel 1956), even though the median response nies detect predatorsno more quickly than do times were significantlydifferent on the two smaller colonies. days (P < 0.02, Mann-Whitney U-test). The swallows need not discrminatepredator The swallows did not discriminate (Mann- JULY1982] Predationon Cliff Swallows 463

TABLE3. Effect of order of presentationon mobbing intensity and responsetime.

Model sequence Mobbing Responsetime Intensity score a pb Time (s)a n pb Predator after predator versus 3.5 _+ 0.2 42 0.152 4.35+ 0.35 40 0.301 Predator after swallow or dove 3.1 + 0.2 43 4.81 + O.36 41

1.3 _+ 0.4 21 4.45 + 0.69 8 Swallow after predator versus 0.587 0.875 Swallow after swallow or dove 1.4 + 0.4 15 4.56 + 0.97 8

2.0 + 0.4 20 4.65 + 0.69 14 Dove after predator versus 0.277 0.935 Dove after swallow or dove 2.6 + 0.4 13 4.16 + 0.46 11

Values are means -+ SE. Mann-Whitney U-test.

Whitney U-tests, Sokal and Rohlf 1969) be- should not increaselinearly with colony size tween the two predator models or the two non- for the dilution effect to occur. Although there predator models in their mobbing responses. is a significant correlationbetween the esti- For the kestrel and shrike models P = 0.35 mated colony size in 1979 and the number of (n = 64 and 26), and for the dove and swallow predators per census (r.• = 0.597, P • 0.01, models P = 0.10 (n = 34 and 37). They did, n = 27, Spearman rank correlation), the rela- however, discriminate between predator (kes- tionship appears asymptotic, not linear (Fig. trel and shrike trials combined) and dove 4). That is, large colonieshave proportionately models (P • 0.001, n = 90 and 34) and pred- fewer predatorsin their vicinity than do small- ator and swallow models (P • 0.001, n = 90 er colonies. and 37). The responseswere directed only at Even though the number of predatorsmay the models, becausewe observed no response not increase in proportion to colony size, the if the apparatuswas operatedwithout a model. dilution effect could still be negated if each Median responsesfor the models are listed in predator attackedmore frequently. At the two Table 3. Although the predatormodels elicited colonies where we made observations, the a significantly greater responsethan the non- mean attack rate per hour was very similar predatormodels, the averageresponse was no- (0.70 at DM and 0.83 at EC). At colonyDM we ticeablyless than that given to an actualprcd- witnessed five predator attacksand at colony ator(which, by definition,was always a fiv•e). EC we observed 11 attacks. The daily rate at Our observations of the responses of Cliff DM was not significantly different from the Swallowsto birds approachingthe colonyalso rate at EC (P >0.40, n = 7 and 13, Mann~ suggestedan ability to discriminate between Whitney U-test). predators anti nonpredators.Like Bank Swal- Breeding synchronyenhances the dilution of lows (Hoogland and Sherman1976), Cliff Swal- predation. If synchronizedbreeding were de- lows gave alarm calls when they encountered termined by other factors, however, such as potential predators--Loggerhead Shrikes, the climate of a region, then dilution might be American Kestrels, Roadrunners (Geococcyx a result and not a causeof synchronousbreed- californianus),Cooper's Hawks (Accipitercoop- ing. To determine the extent that regional ef- erii), Red-shouldered Hawks (Buteo lineatus), fects have versus local colony effects,we com- Red-tailed Hawks (B. jamaicensis),and Great pared within-colony synchrony to between- Blue Herons (Ardea herodias). They ignored colony synchrony. Higher within- than be- other birds of similar size and shape,however, tween-colony synchronyimplies that birds are that would or couldnot capturethem--Swain- synchronizing their breeding in responseto son's Hawks (B. swainsoni),White-tailed Kites some aspectof the local colony area. The two (Elanusleucurus), Forster's Terns (Sternafor- colonies for which we determined hatching steri), Mourning Doves, Killdeer (Charadrius dates, SM and SV, were asynchronouswith vociferus),and Starlings(Sturnus vulgaris). respectto each other. That is, the colonymeans Predatordilution.--As stated previously, the of the mean dates of hatching for each nest number of potential predators near a colony were significantly different (P • 0.01, n = 54 464 W•LK•NSONAND E•GL•SH-LoEs [Auk, Vol. 99 and 12). But both colonieshad a high degree 32 of within-colony synchrony.The standardde- 6 viation of the hatch dates was 5.15 for SV and 6.50 for SM, which is within the range (3-6.5) of synchrony estimates for Bank Swallows (Emlenand Demong1975). Both Bank and Cliff swallows (Emlen 1952, 1954; Myres 1957; Em- len and Demong1975; Hoogland and Sherman 1976)are consideredto have high within-col- ony synchrony.We can concludefrom these data that within-colonysynchrony is greater than between-colony synchrony. Thus, Cliff 100 200 300 400 500 600 Swallowsmay synchronizenesting to reduce colony S•ZE predation via dilution. Alternative explana- Fig. 4. Density of American Kestrelsand Log- tions for breeding synchronyand colonyfor- gerheadShrikes in a 1-km radius circlearound each mation are discussed below. colonyas a function of colonysize. Rangeis given by a narrow line, one standard error by the wide DISCUSSION line, and mean by the crossline. N equalsthe num- ber of censusesat each colony. Much of the work on predator detection by individuals in groups has been done with flocksof birds. Althoughcolonies are different are needed to explaingroup sizesthat exceed from flocksin that they are not mobile aggre- that limit. Becausewe found no early detection gations,individuals in both flocksand colonies effect, the Cliff Swallow colonies we studied are still faced with the problem of detecting may have exceededthis limit. Hooglandand predators. Several workers (Pulliam 1973, Sherman (1976) showed that initiation of mob- Treisman 1975, Caraco 1979a, Lazarus 1979) bing of a stuffedweasel placed in a BankSwal- have argued from theoreticalbases that the low nest is more rapid for larger coloniesof time required to spot an approachingpredator swallows.Their experimentwas flawed, how- should decreasewith increasing flock size. ever, becausethey may have been measuring This theoreticalcorrelation has been supported the time elapsed between a disturbance in a number of laboratoryand field studiesof (causedby the experimenterplacing the weasel flocks of 50 or fewer birds (Lazarus 1972, Di- in the burrow) and the first mobbing response mondand Lazarus1974, Powell 1974, Siegfried by swallowsthat had subsequentlyreturned to and Underhill 1975, Kenward 1978) and con- the nest site. This measur• may or may not tradicted in one (Lazarus 1979). Caraco (1979a), correlatewith the time required to detect an Hoogland (1981), and Lazarus(1979) suggest, approachingpredator. however, that predatordetection time should Althoughwe haveshown that large colonies not continue to decrease with increases in do not detect predatormodels faster than do group size. They predictthat individualswill small coloniesof Cliff Swallows,large colonies spend more time searchingfor food or inter- would still benefit if they detectedpredators at actingwith othermembers of the grouprather greaterdistances than did smallcolonies. If our than continuingto scanfor predators.Several apparatushad been set up within the zone of studies (Lazarus 1972, Powell 1974, Abramson detection,that region in which predatorsare 1979, Caraco 1979b, Hoogland 1979, Bertram alwaysdetected by the smallestcolony, then 1980)have documenteda decreasein scanning we would not havedetected any distanceeffect rate and an increasein feeding rate per indi- if one existed. We observed that neither kes- vidual with increasinggroup sizes. Siegfried trels nor shrikes were mobbed, however, while and Underhill (1975)and Hoogland and Sher- perchingin full view of colonyEC, our largest man (1976)have found higher levels of aggres- colony, before attacking.Unfortunately, we sion in larger goups.These data suggestthat could not estimate the detection distance for the time needed to detect an approaching approachingpredators. The zone of detection predatormay becomeasymptotic as groupsize couldbe determinedby varyingthe guide-line increasesand, therefore,that other hypotheses length at eachcolony. 1982] Predationon Cliff Swallows 465

Our results indicate that nesting Cliff Swal- theirterritory sizes to the density?f theirprin- lows discriminatedbetween the predator and cipal prey (Maher 1970, Anderson nonpredator models. Two aspectsof their re- 1976), which may reduce the dilution effect. sponses need discussion, however. First, the Snapp (1976) reported that the fraction of both responses to the predator models were typi- eggsand young lost to predatorswas constant cally similar but lessintense than the responses for Barn Swallows ( rustica) in 47 col- to actual predators. Part of this difference may onies containing from 1 to 30 nests. Unfortu- be due to the blind at the end of the apparatus, nately, she did not report the losseson a per- which concealed the model after it had been nest basis. Furthermore, if predators are not towed. On thosetrials during which a predator territorial, coloniesmay suffer higher preda- model stopped half-way, the response was tion than solitary nesting birds (Burger 1974). more intenseand prolonged.It is not clearwhy Although our data are few, predator attackrate the swallows apparently become neither ha- appearsto be independent of colony size. In bituatednor sensitizedto the predatormodels addition, the predator censusesindicate that nor why they will ignore a kestrel perched at predator densities increasedat most five-fold, 100 m but will mob an approachingor depart- while swallow densitiesincreased by a factor ing kestrel (pers. obs.). The swallows may be of 20. Since the American Kestrel and Logger- mini.mizing energyexpenditure or maximizing head Shrike are territorial (Balgooyan 1976, feeding or nest-building time (see Withers Craig 1978, respectively),it is likely that dilu- 1977). If so, it is to be expectedthat the swal- tion of these predatorsdoes occur. To test this lows will exhibit the responseonly when there idea more carefully, however, more extensive is the genuine danger of an attacking kestrel information is needed on predator encounter rather than waste time or energy on a kestrel and attack rates. that may not be intending to attack. From this study we cannotconclude that the Second, on some occasions the swallows did dilution of predatorsis the only factorthat cur- give alarm calls and the display to Cliff Swal- rently favors Cliff Swallow coloniality, even low and Mourning Dove models. The re- though the data supportthis simple hypothesis. sponsesto the swallow and dove models were There are other alternativehypotheses that also not false alarms causedby some other stimu- explain how predation rates are decreasedin lus, becausethe display was directed at the group-living birds. Predation could be higher models. One explanation for the responsesto on peripheral nests, which would favor colon- the dove modelsis that the swallowsgave the ial nesting (Hamilton 1971);predators may be alarm before they had indentified the ap- deterredby mobbing colonymembers (Hoog- proachingshape. The responsesdirected at the land and Sherman1976); colonies may be cryp- swallowmodel may be due to colonymembers tic and found lessoften by randomly searching recognizing the approaching model as a for- predators than dispersed nests (Paloheimo eign swallow. On several occasions we ob- 1971, Vine 1971, Treisman 1975); or the re- served colony membersrespond to approach- sponsedisplays of the swallowsmay confuse ing Cliff Swallows by attacking them and or distractattacking predators (Humphries and giving alarm calls. Eggs are commonlyfound Driver 1967, Neill and Cullen 1974). In addi- beneath Cliff Swallow nests (Emlen 1954, this tion, group-living may benefit by in- study), and Emlen (1954) suggestedthat this creasing foraging efficiency, or groups may may result from foreign swallowsrolling them form as a result of a shortageof suitablebreed- out while attemptingto usurp a nest. ing areas(Lack 1968,Alexander 1974). Whether The evidence in favor of the predator dilu- or not any of these alternatives are plausible tion hypothesisfor other colonialnesting birds explanations for colony formation and main- is inconclusive. For example, Nisbet (1975) tenancemust await further study. They could showedthat the number of tern chickspreyed be testedby monitoringthe dispersionof prey, upon was constant, although the number of manipulatingprey dispersionand colonysize, susceptiblechicks fluctuated during the nest- and following markedbirds from the colonyto ing season. Becausepredation did not vary the feeding areas.Only carefullydesigned field with prey availability, this supportsa dilution experiments will permit the exclusion of one effect. Some avian predators, though, adjust or more of these hypotheses.The extensive 466 WILKINSONANO ENGLISI-•-LoEs [Auk,Vol. 99 variation in Cliff Swallow colonysize suggests vigilance in lifeß Brain, Behav. Evol. 9: that no single hypothesis will adequately ex- 60-79. plain group size in this species. EM•EN, J. T. 1952. Socialbehavior in nesting Cliff Swallowsß Condor 54: 177-199. ß 1954. Territory, nest building and pair for- ACKNOWLEDGMENTS mation in the Cliff Swallowß Auk 71: 16-35. We thank Michael Beecher,Jack Bradbury, Boyd EMLEN,S. T., & N.J. DEMONG.1975. Adaptive sig- Collier, William Hazen, John Hoogland, Raleigh nificanceof synchronizedbreeding in a colonial Robertson,Sandra Vehrencamp,James Whitehead, : a new hypothesis.Science 188: 1029-1031. and an anonymous reviewer for comments, which FREER,V. M. 1973. Sparrow Hawk predation on improved the manuscript. Lora Dickerhoff, Gary Bank Swallows. Wilson Bull. 85: 231-233ß McCracken, and Laurie Terwilliger supplied valu- HAMILTON,W. D. 1971. Geometry for the selfish able assistancein the field during portions of the herd. J. Theor. Biol. 31: 295-311. study. Travel and computercosts were generously HARVEY,P. H., & P. J. GREENWOODß1978. Antipre- defrayed by the Departmentof Biology of the Uni- dator defense strategies:some evolutionary versity of California at San Diego. Equipment was problemsßPp. 129-151 in Behavioralecology. An purchased on NSF grant BNS76-04400to JackBrad- evolutionary approach (J. R. Krebs and N. B. bury. We are indebted to William J. Hamilton, III Davies, Eds.). Oxford, Blackwellsß who initially encouragedus when we began this HOOGLAND,J. L. 1979. The effectof colonysize on study. individual alertnessof prairie dogs (Sciuridae: Cynomysspp.). Anim. Behav. 27: 394-407. In pressß The evolution of coloniality in LITERATURE CITED white-tailed and black-tailed prairie dogs ABRAMSON,M. 1979. Vigilance as a factor influenc- (Sciuridae: Cynomysleucurus and C. ludovici- ing flock formation among curlews, Numenius anus)ßEcology 62: 252-272ß arguata. Ibis 121: 213-216. , & P. W. SaERM•N. 1976. Advantagesand ALEXANDER, g. D. 1974. The evolution of social be- disadvantagesof Bank Swallow (Ripariariparia) havior. Ann. Rev. Ecol.Syst. 5: 325-383. coloniality. Ecol.Monogr. 46: 33-58. ANDERSON,M. 1976.Predation and kleptoparasitism HUMPHRIES, D. A., & P.M. DRIVER. 1967. Erratic by skuasin a shetlandsea-bird colonyß Ibis 118: display as a device against predators. Science 208-217. 156: 1767-1768. BALGOOYEN,T. G. 1976. Behaviorand ecologyof the JoaNsoN, C. G. 1969. Migration and dispersalof in- American Kestrel (Falco sparverius,L.) in the sectsby flightß London, Methuen. Sierra Nevada of California. Univ. Calif. Publ. KENWARD, R. E. 1978. Hawks and doves: attack suc- Zool. 103: 1-85. cessand selectionin Goshawkat woodpigeons. BERTRAM,B.C. R. 1978. Living in groups:predators J. Anim. Ecol. 47: 449-460. and prey. Pp. 64-96 in Behavioralecology. An LACK,D. 1968. Ecologicaladaptations for breeding evolutionary approach (J. R. Krebs and N. B. in birdsß London, Methuen. Davies, Eds). Oxford, Blackwells. LAZARUS, J. 1972. Natural selection and the func- ß 1980. Vigilance and group size in ostriches. tions of flockingin birds: a reply to Mutton. Ibis Anim. Behav. 28: 278-286. 114: 556-558. BONNOT,P. 1921. SparrowHawk capturesswallow. ß 1979. The early warning functionof flocking Condor 23: 136. in birds: an experimentalstudy with captive BURGER,J. 1974. Breeding biology and ecologyof Quelea. Anim. Behav. 27: 855•66. the Brown-hoodedGull in ArgentinaßAuk 91: LLOYD,M., & H. S. DYBAS. 1966. The periodical 601-613. cicadaproblemß I. Populationecology. Evolution CARACO,T. 1979a. Time budgetingand group size: 20: 133-149ß a theory. Ecology60: 611-617. MAHER,W. J. 1970. The PomefineJaeger as a brown 1979b. Time budgeting and group size: a lemming predator in northern Alaska. Wilson test of theory. Ecology60: 618-627. Bull. 82: 130-157. CLARX,K. L., & R. J. ROBERTSON.1979. Spatial and MAYHEW,W. W. 1958. The biology of the Cliff Swal- temporal multi-speciesnesting aggregationsin low in California. Condor 60: 6-37. birds as anti-parasiteand anti-predator defen- MYRES,M. T. 1957. Clutch size and laying dates in ses. Behav. Ecol. Sociobiol. 5: 539-571. Cliff Swallow coloniesß Condor 59: 311-316. CRAIG,R. B. 1978. An analysisof the predatorybe- NEILL, S. R. ST. J., & J. M. CULLEN.1974. Experi- havior of the LoggerheadShrikeß Auk 95: 221- mentson whetherschooling by their prey affects 234. the hunting behavior of cephalopodsand fish DIMOND, S., •: J. L•ZARUS. 1974. The problem of predators. J. Zool. London 172: 549-569. JULY1982] Predationon Cliff Swallows 467

NIE, N.H., C. H. HULL, J. G. JENKINS, K. STEIN- S•EGFR•ED,W. R., & L. G. UNDERHILL. 1975. Flock- BRENNER,&: D. H. BENT. 1975. Statistical pack- ing as an anti-predator strategy in doves. Anim. age for the socialsciences. New York, McGraw- Behav. 23: 504-508. Hill. SNAPP,B. D. 1976. Colonial breeding in the Barn NISBET,I. C. T. 1975. Selectiveeffects of predation Swallow (Hirundo rustica) and its adaptive sig- on a tern colony. Condor 77: 221-226. nificance. Condor 78: 471-480. PALOHE•MO,J. E. 1971. A stochastictheory of search: SOKAL,R. R., & F. J. ROHLF. 1969. Blumetry. San implicationsfor predator-prey situations.Math. Francisco, W. H. Freeman. Biosciences 12: 105-132. TREISMAN, M. 1975. Predation and the evolution of POWELL,G. V. N. 1974. Experimentalanalysis of the gregariousness.I. Models for concealmentand socialvalue of flockingby Starlings(Sturnus vul- evasion. Anim. Behav. 23: 779-800. garis) in relation to predation and foraging. VINE, I. 1971. Risk of visual detection and pursuit Anim. Behav. 22: 501-505. by a predator and the selective advantage of PULLIAM,H. R. 1973. On the advantagesof flocking. flocking behavior. J. Theor. Biol. 30: 405-422. J. Theor. Biol. 38: 419-422. W•NDSOR,D., & S. T. EMLEN. 1975. Predator-prey SAMUEL,D. E. 1971a. Vocal repertoiresof sympatric interactions of adult and prefledgling Bank Barn and Cliff swallows. Auk 88: 839-855. Swallows and American Kestrels. Condor 77: 1971b. The breeding biology of Barn and 359-361. Cliff swallowsin West Virginia. Wilson Bull. 83: WITHERS,P. C. 1977. Energeticaspects of reproduc- 284-301. tion by the Cliff Swallow. Auk 94: 718-725. SInGEL,S. 1956. Nonparametric statisticsfor the be- havioral sciences. New York, McGraw-Hill.