INTERSPECIFIC AGGRESSIVE BEHAVIOR OF THE POLYANDROUS NORTHERN (JACANA $PINOSA)

MARTIN L. STEPHENS Allee Laboratoryof AnimalBehavior, The Universityof Chicago,940 East57th Street, Chicago,Illinois 60637 USA

ABSTRACT.--Adultsof the polyandrous Northern Jacana(Jacana spinosa) defend their off- spring against avian territorial intruders that are potential offspring predators. I investigated this defensive behavior in CostaRica during 1980 and 1981. Jacanasattacked 15 species;most attackswere against Purple Gallinules (Porphyrulamartinica), which are known predators of jacana eggs and offspring. Three factors influenced the probability of attacks:(1) spatial proximity of intruders to jacanaoffspring, (2) speciesidentity of intruders,and (3) stageof breeding (jacanaswere most responsive to intruders when young offspring were present). Attacks prevented intruders from reaching jacanaoffspring, and limited evidence suggests that attacksalso reduced the density of intruders near jacana nests. The female's role in offspring defense was substantial. Females participated in the defense of eggs and small young at levels comparableto thoseof males.This participationincluded: (1) joining ongoing attackswhen their mates solicited aid, (2) continuing attackswhile their mates led offspring away from intruders, and (3) launching attacks in the absence of their mates, especially during incubation. Received8 November1982; resubmitted 19 September•1983, accepted 12 January 1984.

NORTHERNJacanas (Jacana spinosa) have one RodriguezNational Wildlife Refuge,formerly known of the rarest breeding systemsknown among as Hacienda Palo Verde, in Guanacaste Province, Costa : resource-defensepolyandry (Jenni 1974, Rica.The study site is part of a marshnear the Temp- Emlen and Oring 1977, Graul et al. 1977). Fe- isque River. The marsh is an expanseof emergent vegetationinterrupted by open water and floating male Northern Jacanas defend territories that vegetation. Scattered"Palo Verde" trees (Parkinsonia include the smaller territories of their one to sp.) give the marsh a savannah-like appearance. four mates.Females are nearly 70%heavier than Hundreds of jacanasarrive for breeding at the marsh males and do virtually no incubating, brood- when it is inundated during the rainy season(May- ing, or escortingof the precocialyoung (Jenni November; Slud 1980). The study site was covered and Collier 1972, Jenni and Betts 1978). Fe- by floating vegetation(primarily Nymphaea,Eichornia, malesdo aid malesin defending offspringfrom and Neptunia)and borderedby emergents(primarily predators,however: males and females both at- Typha, Panicurn,Paspalum, and Eleocharis).Small tack birds of other species that are potential patchesof emergentsdotted the zone of floating vegetation. predatorsof jacanaeggs and chicks. Jenni and I conducted observations from towers 3-6 m tall Betts(1978) describethe behaviorsthat jacanas that were within the territories of focal . I employ in this interspecificaggression, docu- draped a blind over the tower before beginning sys- ment the time that malesand femalesspend in tematic observations.My approach to and ascent of attacking offspringpredators, and comment on the tower sometimes caused nonfocal birds to flee, other aspectsof this behavior. In this paper, I including speciesthat jacanasattack. In order to re- confirm and extend the observations of Jenni duce any effectsthis might have had on the data, I and Bettsand place emphasison: (1) the prox- waited in the blind at least 5 min, and often at least imate causes of interspecific attacks, (2) the 20 min, before beginning to sample. consequencesof such attacks,and (3) the role During 1981, observationperiods were 2 h long. that femalesplay in this form of parental care. During each period I simultaneouslyobserved one female and one of her mates, regardlessof the num- ber of mates the female had. I observed the same

METHODS mate during the entire period over which a particu- lar female was observed,except when events such as I studied Northern Jacanasfrom June to November clutch loss made it profitable for me to initiate ob- 1980 and May to December 1981 in the Dr. Rafael servations of a different mate and discontinue obser-

508 The Auk 101: 508-518. July 1984 July1984] JacanaInterspecific Aggression 509 vations of the first. The focal male and female are causeof the gradual transition from downy to juve- referred to as the "focal pair," even though some nal plumage. focal females were polyandrous. Nonparametricstatistical tests employed below are During 1980,observation periods were 80 min long. describedby Siegel (1956). When meansare strongly During each sample ! simultaneouslyobserved one affected by outliers, medians are reported. An as- female and one of her mates for 60 min; then, after sumption of many of the statisticaltests that follow a 5-min pause,! observedthe same female but a dif- is that attacks are temporally independent of one ferent mate for the remaining 15 min. Males of mo- another. ! determined that this was indeed the case nogamous females were observed for the entire pe- by plotting interattackintervals as a log survivorship riod. The same male was observedduring all 60-min function for each speciesthat was attacked and by segmentsof the period spent observing a particular comparing this distribution to the one that would be female, and these are the males in the male/female expected(the negative exponential) if attackswere comparisonsreported below. to occur at random with respect to one another (see Three types of sampling were conductedconcur- Slater 1974). Sample size was adequate for this com- rently during each observationperiod. Throughout parisonfor only the mostfrequently attackedspecies a period ! recorded the details of all attacks by the (Purple Gallinule, Porphyrula martinica). The ob- focal pair, including time and position of attack, served and expected distributions were not signifi- identity and behavior of attacker, speciesof victim, cantly different (Kolmogorov-Smirnovtest, n = 35, and distance moved by victim during or within 5 s P > 0.20), indicating that attacksagainst this species after attack (= Aggression Sampling). Successiveat- did not tend to occur in bouts. tacks against an individual were scored as separate attacks if the attacker engaged in nonaggressivebe- havior for at least 30 s between attacks. Second, ! RESULTS recorded the behavior and position of focal pairs and their offspring every 5 min (1980) or 10 min (1981) Behaviorduring attacks.--Jacanasattack other by scan sampling (Altmann 1974) (= Behavior and birds with a combination of primarily two be- PositionSampling). Finally, ! plotted the positionsof haviors describedby Jenni and Betts (1978). I all avian intruderson the focalfemale's territory every labeled thesebehaviors "swoops" and "threats." 20 min (1980) or 30 min (1981) by scan sampling (= In the "swoop," a jacanaflies toward an intrud- Species Scan Sampling). Positions were plotted on er and strikes it with its feet, simultaneously vegetation mapsof territories. SpeciesScan Sampling uttering a sharp call. Jacanassometimes feign was momentarily interrupted to record detailsof any ongoing attacks,whereas Behavior and Position Sam- the strike by thrusting the feet toward the in- pling was completedbefore recording attack data. truder while flying past it or by veering away Observationperiods began at 0600, 0750, 0940, 1230, from the intruder just before reaching it (pers. 1350, and 1440 during 1980 and at 0600, 0900, 1200, obs.). and 1500 during 1981. Observation periods were ro- In the "threat," a iacanacrouches in front of tated randomly among male/female pairs such that an intruder and displays its carpal spurs, si- approximately equal numbers of sampleswere ac- multaneously uttering a long, shrill vocaliza- cumulatedat each time of day for each focal pair. A tion that lenni et al. (1975) labeled the "scream." total of 16 pairs was observed,but observationswere The exposed yellow spurs are highlighted concentratedon 12 pairs. Of the latter, only one in- against the maroon forewings. lacanas were dividual was observedin both yearsof the study.All but two focal animals were captured in mist-nets and never seen striking birds of other specieswith color-banded for individual recognition. their spurs.See lenni and Betts(1978) for a full Components of the above methodology, such as description of threats. observation-periodlength, varied somewhatduring On rare occasions, other attack behaviors such early 1980,but thesechanges did not biasthe results. as charges or pecks were employed. Pecks oc- To examine temporal patterns, ! divided the repro- curred as an intruder turned to flee from an ductive period of male jacanasinto four stages:Pre- attacking iacana. True fights were also rare. incubation (the period before or between nesting at- Fights,which were seenin attacksupon Purple tempts, when no eggs or offspring are present), In- Gallinules only, consistedof both participants cubation (the first day of egg-laying to the last full iumpinõ up simultaneouslyand appearing to day of incubation), Downy (day 1, the day of hatch- ing, to day 28 post-hatching,during which chicks hit each other with their feet in a brief flurry have downy plumage), and Juvenile (day 29 to dis- of activity. persal or renesting, during which offspring have ju- An attack comprised a variable number of venal plumage). The transition date between the these attack behaviors. To represent this vari- downy and juvenile stagesis somewhatarbitrary be- ability, an intensity scorefor individual attacks 510 MARTINL. STEPHENS [Auk, Vol. 101

TABLE1. Percentagesof attacksof variousintensities as a function of attacktype (data from 1980). Typeof Intensityi Mean attack 1 2 3 4 5 6 7 > 7 n intensity Solo 11 58 9 17 1 3 1 0 103 2.52 Joint -- 8 0 30 30 11 5 16b 37 5.84 Either 8 45 6 21 9 5 2 4 140 3.40

See text for definition. Includes 3 attacks with intensities of 8, and 3 attacks with intensities of 14, 17, and 18.

was computed. Threats were assigned a score mate during ongoing attacks. The communi- of 1 and swoopsa scoreof 2. Swoopswere con- cative function is exploited primarily by males sideredmore intensethan threats,because they (Mace 1981). Males initiated most joint attacks often involved striking the victim, whereas (83% during 1980) and were thus in situations threats did not. The rare attack behaviors were in which they could recruit their mate's aid. assigneda scoreof 2. Although individuals oc- Females were often far from attacks initiated casionally performed more than one threat by males but respondedto their mate's scream during an attack, multiple threats by a single by flying to him and joining the attack. individual were difficult to distinguish from a There was no difference between males and single threat interrupted by short pauses.In- females in their tendency to perform single dividuals were therefore assigneda maximum versusmultiple swoopsduring attacks;this was of one threat per attack. the caseregardless of whether solo attacks,joint Nearly 90% of attackshad intensitiesof less attacks, or all attacks were examined (Chi- than 6 (Table 1), indicating that most attacks squaredtests, P > 0.05). Little can be said con- consistedof only a few attack behaviors. Not cerning sexual differences in uncommon attack surprisingly,solo attacksby a male or female behaviorsbecause of their rarity,but fightswere were roughly half as intenseas joint attacksby restrictedto interactionsbetween female jacan- a male and female. as and Purple Gallinules. Certain nonaggressivebehaviors performed Speciesattacked.--Jacanas attacked 15 species during attacksare noteworthy: (1) jacanasnear- of birds (excluding conspecifics) during ly alwaysinterposed themselves between their AggressionSampling (Table 2). Purple Galli- offspring and intruders; (2) during joint at- nules alone accounted for 79% of all attacks. tacks,males often led their offspring from in- This speciesopportunistically preys on jacana truders while females continued to confront the eggsand chicks(Jenni and Betts 1978;see also intruders; (3) during or just before attacks,ja- Bailey 1927,McIlhenny 1936,Beadel 1946). An canas occasionally gave alarm calls or distrac- additionaleight specieswere recordedon focal tion displays in responseto intruders; and (4) territories at least once but were never attacked jacanachicks hid themselvesin the water dur- during sampling (Table 2). Not all of these ing a few attacks.These observationsreinforce speciesare "tolerated" (Waiters 1979) by jacan- direct evidence (see below) that jacana off- as, however. Two species[Bare-throated Tiger- spring are subjectto predation by some species Heron (Tigrasomamexicanum) and Limpkin (Ar- of territorial intruders and that interspecificat- amusguarauna)] were attackedby jacanasdur- tacks are a means of reducing such predation. ing casualobservations. Also, four specieswere One of the most striking sexual differences rare (Table 2) and, therefore, provided few op- in interspecificaggressive behavior was the dif- portunities for attacks. ferential use of threats.During 1980,males used The relationship between the attack rate threats in 49% (n = 82) and females in only 12% against each species and that species' abun- (n = 101) of their attacks.Threats appear to have dance on jacana territories was examined by two functions (Jenni et al. 1975, Mace 1981): (1) letting "responsiveness"(Waiters 1980) equal aggressiondirected against intruders and (2) attack rate (here calculated as attacksper 10 h) communication directed toward mates, i.e. the divided by the mean number of individuals of scream uttered during threats summons the that speciesrecorded in SpeciesScan Samples. July1984] JacanaInterspecific Aggression 511

TABLE2. Number of jacanaattacks on each speciesof territorial intruder, abundanceof each specieson jacanaterritories, and responsivenessof jacanasto eachspecies.

Number of Number of Responsive- Species attacks(%)a intrudersb nessc Purple Gallinule (Porphyrulamartinica) 396 (79.0) 460 5.04 Pied-billed Grebe (Podilymbuspodiceps) 141 LeastGrebe ( Tachybaptusdominicus) 26(5.2) a 60 0.76 Common Moorhen (Gallinulachloropus) 23 {4.6) 161 0.84 Green-backed Heron (Butoridesstriatus) 15 (3.0) 69 1.27 (Porzanacarolina) 12 (2.4) 2 36.33 Great Egret (Casmerodiusalbus) 9 (1.8) 25 2.10 Red-wingedBlackbird (Agelaius phoeniceus) 8 (1.6) 4 12.17 TropicalKingbird (Tyrannusmelancholicus) 3 (0.6) 82 0.21 Great Kiskadee(Pitangus sulphuratus) 2 (0.4) 12 0.95 Snail Kite (Rostrhamussociabilis) 2 (0.4) 9 1.29 Least Bittern ( Ixobrychusexilis) 2 (0.4) 3 3.06 Groove-billedAni ( Crotophagasulcirostris) 1 (0.2) 24 0.25 Black-belliedWhistling-Duck (Dendrocygnaautumnalis) 1 {0.2) 85 0.07 Little Blue Heron (Egrettacaerulea) 1 {0.2) 0 5.80- Bare-throatedTiger-Heron ( Tigrisomamexicanurn) 0 9 0.00 Anhinga (Anhingaanhinga) 0 19 0.00 Limpkin (Aramusguarauna) 0 3 0.00 Black-crowned Night-Heron (Nycticoraxnycticorax) 0 1 0.00 Cattle Egret (Bubulcusibis) 0 1 0.00 Fulvous Whistling-Duck (Dendrocygnabicolor) 0 3 0.00 Mangrove Swallow (Tachycinetaalbilinea) 0 1 0.00 White-collaredSeedeater (Sporophila torqueola) 0 16 0.00 a Basedon 1,100 h of Aggression Sampling. bNumber of individuals recordedon jacanaterritories during 644 SpeciesScan Samples; each suchsample was taken midway into eachobservation period. cResponsiveness = (number of attacksper 10 h)/(mean number of individuals recorded during Species Scan Samples). a These speciesare combined, becauseI sometimescould not discern which of the two was attacked. eTo avoid division by zero in calculatingthis, I set the number of intrudersequal to one.

Before considering these scores,one should TABLE3. Percentagesof attacksinitiated by males note that they are inflated for any inconspic- and femalesagainst different species. uousspecies that was underrecordedin Species Initiator • ScanSamples. Although the magnitude of this effect is unknown, I suspectthat it is small. Speciesattacked Females Males Responsivenessvaried considerablyamong Purple Gallinule 81.3 78.2 species(Table 2). Soras (Porzanacarolina) elic- Red-winged Blackbird 3.6 0.9 ited the highest responsivenessby far, and Leastand Pied-billed grebes 0.0 7.2 Green-backed Heron 2.9 2.1 Purple Gallinules,the mostfrequently attacked Common Moorhen 4.3 5.4 species,elicited the fourth highest. Great Egret 3.6 1.2 Males and females differed in the extent to Sora 1.4 3.0 which they initiated attacks against particular Others b 2.9 2.1 Number of attacks 139 335 species(Table 3). The main differencewas the males' larger proportion of initiations against a Attack initiations among speciesare not indepen- grebes,which femalesdid not attackat all. Least dent of initiator (X2 = 19.0, df = 7, P < 0.01, frequen- ciesanalyzed). Grebes (Tachybaptusdominicus) and Pied-billed b These include Least Bittern, Little Blue Heron, Grebes(Podilymbus podiceps) were attackedonly Snail Kite, Groove-billedAni, TropicalKingbird, and when they surfaced near jacana broods. Be- Great Kiskadee. 512 MARTINL. STEPHENS [Auk, Vol. 101

32. A Purple Gallinule -6O

28 ß iI ;/ ,

_20 E

/ ',, .0

Q. Other Species 60 u• !/ E 1- 40 •

20 •. Distance Between Intruder & Jacana Nest (m) 2

Fig. 1. Frequency distribution of attacksduring incubation, as a function of distanceto jacananests, B o •' o comparedwith the percentagesof time that intruders O-10 11-20 21-30 > 30 spent at those distances(Kolmogorov-Smirnov test, P < 0.01, raw frequenciesof intruder numbersana- Distance Between Intruder & Jacana Brood (m) lyzed). Purple Gallinules accountedfor 90% of these Fig. 2. Frequencydistribution of attacksagainst attacks;therefore, the curve of time spent at different (A) Purple Gallinules and (B) other species,as a func- distanceswas restricted to Purple Gallinules. These tion of distanceto jacanabrood, comparedwith the percentageswere derived from 32 SpeciesScan Sam- percentagesof time that these speciesspent at those pies. Data are from five territories sampled during distances(In A and B, Kolmogorov-Smirnovtest, P < 1980. 0.01, raw frequenciesof intruder numbersanalyzed). "Other species" include Pied-billed Grebes, Green- backed Herons, Great Egrets, and Common Moor- hens. The percentagesof time were derived from 32 SpeciesScan Samplesin both A and B. Data are from cause females rarely escort broods (Jenni and one male jacana'sterritory. Betts 1978), femaleswere rarely in a better po- sition to respond to grebes near broods than were males. Spatial distribution of attacks.--Jacanas at- Attacks upon Sofas were the only exception tackedheterospecifics only when the latter were to this pattern of spatial specificity.This species closeto the jacanas'offspring. This spatial spec- arrived at the study site in small numbers to- ificity of attacksoccurred whether jacanashad ward the end of the study. Apparently, jacanas eggsor mobile offspringand was not a byprod- attacked them wherever they were encoun- uct of similar microhabitatpreferences of in- tered on jacana territories, regardless of their truders and jacana offspring (Fig. 1 and 2). At- proximity to jacana offspring. This pattern was tacks during pre-incubation clustered about the responsiblefor the very high responsiveness jacanas'empty nest. Thus, the Northern Jacan- to Sofas (Table 2). as that I observed defended a small area around One factor that influenced whether the male their nest or offspring against potential pred- or the female initiated an attack was their rel- ators. Those observed by Jenni and Collier ative proximity to the intruder. Data from Be- (1972) at times defended their entire territory havior and Position Sampling were used to es- against such predators. timate the distances of one focal male and July1984] JacanaInterspecific Aggression 513

TABLE4. Frequencydistribution of the distancesthat TABLE5. Frequencydistribution (expressedas per- differentspecies moved when attackedby jacanas. centages)of distancesmoved by Purple Gallinules during attacks by lone males, lone females, and Distance moved male/femalepairs.

Speciesattacked a <6 m 6-10 m >10 m Distance moved

Purple Gallinule 356 41 48 Attacker • <6 rn 6-10 m >10 m n Common Moorhen 17 3 3 Sora 1 1 6 Males 88 6 6 156 Green-backed Heron 14 2 7 Females 79 10 10 154 Great Egret 10 0 0 Pairs 72 12 16 135 Grebespp. 24b 0 0 Red-wingedBlackbird 3 2 3 a Distanceswere not independent of attacker(X 2 = 12.18,df = 4, P < 0.05, frequenciesanalyzed). aOnly speciesattacked at leastfour times are con- sidered. b Pied-billed and Least grebesinvariably sub- merged when attacked. which intruders were attacked during one breeding stage but not in the preceding or fol- lowing stages.This occurred on a focal terri- female to intruders immediately before attacks. tory where the breeding pair defended a nest The male tended to initiate attacksonly when that was eventually lost. Following nest loss, he was closer to the intruder than the female the male deserted his territory, and the area was. The female initiated attacksregardless of was soon incorporatedinto the territory of the her mate's proximity to the intruder, however female's other mate; the male and female did (X2 = 6.89, df = 1, P < 0.01). not defend the incorporatedarea from hete- Response to attack.--Intruders generally rospecifics,however. I further restricted the ducked to avoid blows from aerial attacks, and analysisto attacksupon Purple Gallinules, be- certain species, especially the larger ones, causethey accountedfor nearly all of the in- counterattacked,as noted by Jenni and Collier teractions. (1972). Purple Gallinules and Common Moor- I tallied the number of Purple Gallinulesre- hens (Gallinula chloropus)counterattacked by corded in a 20-m x 20-m plot, centered at the charging their attackers, Great Egrets (Casmer- nest, during the incubation and post-egg-loss odiusalbus) and Green-backed Herons ( Butorides periods.The recordsspanned a 30-day interval striatus)by thrusting their bills toward their in both periods. To control for a seasonal in- attackers. creasein Purple Gallinulespresent at the study Most intruders moved short distances when site, ! compared these counts to counts within attacked (Table 4). The exceptions are Red- the desertingmale's entire territory. The pro- winged Blackbirds(Agelaius phoeniceus) and, es- portion of Purple Gallinulespresent within the pecially, Soras,the two smallestspecies listed. plot was 31% lower during incubation than Purple Gallinules moved farthest when at- during the following period (X2 = 4.09, df = 1, tacked by pairs, less far when attacked by fe- n = 261, P < 0.05). This suggeststhat jacana at- males, and least far when attacked by males tacksreduce the density of intruders near their (Table 5). Femalesappeared to be more formi- nests. A similar d•nsity reduction may occur dable attackersthan males; their strikes during during the posthatchingstages but would be swoopswere sometimesstrong enough to dis- more difficult to detect,because attacks during place Purple Gallinules a full meter. Note, these stagesare not restricted to the nest area; however, that the commonestresponse of Pur- they occur wherever the nidifugous offspring ple Gallinules was to move lessthan 6 m, even happen to be at the time. when attacked by pairs. Sample sizes from Timingof aggression.--Themedian attack rates species other than Purple Gallinules are too of male/female pairswere fairly low, the high- small for separateanalyses and their distance- est rate being 4.6 attacksper 10 h (Fig. 3, "total" moved distributions are too heterogeneousfor curve). The true rates are undoubtedly higher, pooling (Table 4). as my arrival sometimescaused intruders to flee Intruders apparently avoided areas where and remain off the focal territory for varying they were being consistentlyattacked. To in- lengths of time (see Methods). The pairs' total vestigatethis, I analyzed the only situation in attackrates were highest during the downy and 514 MARTINL. STEPHENS [Auk, Vol. 101

5= TABLE6. Percentagecontribution of solo male at- tacks(M), solo female attacks(F), and joint attacks total (MF) to total attack rates during each stage of breeding (data from Fig. 3).

Fe- Males' males' total total Attacker (M + (F + M F MF MF) • MF) Pre-incubation 47 0 53 100 53 •o Incubation 36 37 27 63 64 • Downy 49 27 24 73 51 • Juvenile 62 18 20 82 * 38 ' Significant differencesbetween male and female ot totals (Wilcoxon Signed Rankstest, P < 0.05) are in- dicatedby an asterisk(raw attackrates analyzed). See _E10/// byfemaleslegend of Fig. 3 for samplesizes. pre- incubation downy juvenile the juvenilestage, although more data from pre- incubation incubation are needed. Breeding Stage Raw attack rates do not reveal the true effect of the breedingstage upon aggressive behavior, Fig. 3. Median ratesof solomale attacks,solo fe- because intruder abundances were not inde- male attacks,and joint attacksduring each stageof breeding; "total" indicatesthe sum of the median pendent of stage (Stephensunpubl. data). I rates of all three attack types. Data are from 5 pairs therefore calculatedresponsiveness scores as a during pre-incubation,10 during incubation, 11 dur- function of stage.Many pairs were sampledin ing the downy stage,and 8 during the juvenile stage. more than one stage,causing samples in differ- Data from polyandrousfemales include only those ent stagesto be statisticallydependent. A sub- attacks that occurred on the focal male's territory. set of the data from the incubation,downy, and Sample sizes,in terms of hours of sampling,range juvenile stageswas usedto determinewhether from 8.8 to 88.0, with the interquartile range being or not there was an effect of subjects(pairs) on 20.0-42.0. responsiveness.No effecton responsivenessto Purple Gallinules or to all other speciescom- bined was found (Friedman 2-way ANOVA by juvenile stages,and the samewas true for solo ranks, P > 0.05). I therefore considered data male attacks. The median rate of solo female from the samepair sampledin different stages attackswas highest during the incubation and to be independentand combinedthese data downy stages.Note that the ratesof solo male with data from pairs sampledin only one stage attacks exceeded those of females during all in order to test for a stageeffect. stagesexcept incubation. Finally, the rate of Responsivenessof males,females, and male/ joint attackswas uniform acrossstages. female pairs to Purple Gallinules was highest Thesedata are recastas percentages in Table during the downy stage(Table 7). The peakfor 6 in order to identify the contribution of solo females is especiallypronounced. The males' male attacks,solo female attacks,and joint at- fairly high responsivenessduring pre-incuba- tacks to the total attack rate during each stage tion, when no eggsor offspring were present of breeding.These percentages approximate the to defend, is noteworthy. percentagesof all attacksthat consistedof each Males and females had roughly similar re- type. Males and femaleshad equal solocontri- sponsivenessto Purple Gallinules during the butions during incubation, but, as the nesting incubation and downy stages,whereas males cycle progressed,the males' contributionsbe- were more responsivethan femalesduring the cameincreasingly greater than the females'.The juvenile stage(Table 7). The availabledata from total contributions (solo plus joint attacks)of the pre-incubation stage suggeststhat males males and females were similar during all but were more responsive than females. Across July1984] JacanaInterspecific Aggression 515

TABLE7. Responsivenessa of males, females, and TABLE8. Responsivenessof males, females, and male/femalepairs of jacanasto Purple Gallinules male/female pairs of jacanasto speciesother than during each stage of breeding. Purple Gallinules.a

Median Fe- Median Fe- responsiveness Malesb males Pairs nc responsiveness Malesb males Pairs nc Stagea Stagea Pre-incubation 5.56 1.79 6.33 5 Pre-incubation 0.00 0.00 0.00 5 Incubation 2.08 2.11 3.75 9 Incubation 0.31 0.27 0.53 9 Downy 6.49 8.65 11.57 10 Downy 0.84 0.35 1.12 11 Juvenile 4.44 * 1.25 4.93 7 Juvenile 0.68 * 0.00 0.97 8 Overall 4.69 2.68 6.33 31 Overall 0.75 * 0.15 0.88 33 a See text for definition. ßThese include all attackedspecies listed in Table bSignificant differencesbetween males and fe- 2 exceptPurple GallinuIes. males (Wilcoxon Signed Rank test, P < 0.05) are in- bSignificant differences between males and fe- dicated by asterisks. males (Wilcoxon Signed Ranks test, P < 0.05, except cn = number of males,females, or pairsduring each for juvenile stage,where P = 0.062) are indicated by stage,and the sum of thesefor overall. asterisks. a Responsivenessscores of males,females, and pairs c As in Table 7. are not independent of stage (Kruskal-Wallis tests, a Responsivenessscores of males,females, and pairs P < 0.05). do not differ significantly among stages(Kruskal- Wallis tests,P > 0.05). The apparent differencein re- sponsivenessof malesand pairs during pre-incuba- tion and other stages combined is not significant stages,the sexual difference in responsiveness [Dunn's test (Dunn 1964), P > 0.05 in both cases]. to Purple Gallinules is not statisticallysignifi- cant (Table 7). Too few data were available from other speciesto consider responsivenessto them in- identity of territorial intruders, (2) spatialprox- dividually; I therefore combineddata from this imity of intruders to jacananests or offspring, heterogenousgroup. This failed to reveal any and (3) stage of breeding. significantdifferences in the timing of respon- Speciesattacked.--Of the speciesthat were not sivenessof males, females, or pairs (Table 8). rare at the study site, Purple Gallinules elicited However, males, females, and pairs were most the highest responsivenessand accountedfor responsiveto other speciesduring the downy most attacks (Table 2). Purple Gallinules are stage,as they were to Purple Gallinules. If data known predators of jacana eggs and chicks. from other speciesare reliable, then the pattern Some of the other attacked species,or at least of responsivenessto thesespecies differs in two their closerelatives, are also predatorsof avian important ways from responsivenessto Purple eggs and chicks at times (Hunter and Morris Gallinules: (1) relatively low responsiveness 1976;Skutch 1976:444; Repenning 1977;Terres during pre-incubation and (2) similar respon- 1980:498; Welty 1982: 110). sivenessof pairs during the downy and juve- Speciesidentity was lessimportant when in- nile stages.Males and females differed signifi- truders moved rapidly toward jacanaoffspring cantly in responsivenessduring the juvenile or suddenly appeared near the offspring. In stageand overall. these contexts, jacanasseemed to attack vir- tually any species.These contextsaccount for DISCUSSION many of the attacks on speciesthat probably pose little danger to jacana offspring [Least The discussion focuses on the proximate Grebe, Pied-billed Grebe, Red-winged Black- causes of interspecific attacks, the conse- , Groove-billedAni (Crotophagasulcirostris), quences of these attacks, and the role of fe- Tropical Kingbird (Tyrannusmelancholicus) and males in offspring defense. Jenni and Betts Great Kiskadee(Pitangus sulphuratus)]. (1978) discussrelated topics. Spatialcontext of attacks.--Jacanasattack het- erospecificsonly when the latter are near ja- PROXIMATE CAUSES OF ATTACKS cana nests or offspring (Fig. 1 and 2). Because of this spatial specificity, no birds of other At least three factors influenced the proba- specieswere fully excluded from jacana terri- bility of attacksby Northern Jacanas:(1) species tories by aggression. This spatial specificity 516 MARTINL. STEPHENS [Auk, Vol. 101 probably introduced some bias into respon- vincing evidenceof the antipredatorfunction sivenessscores. For example, speciesthat pref- of attacksagainst most species.Attacks against erentially inhabited the same vegetation zones Sofaswere exceptional,however, in location that harbored jacanaoffspring probably would and timing. Jacanasapparently attempted to have inflated responsivenessscores. exclude this speciesfrom their territories and Stageof breeding.--Responsivenessto birds of were somewhat successfulat this (Table 4). The other specieswas highest during the downy habitat use and diet of Sofas are similar to those stage(Tables 7 and 8). This is when the jacanas of Northern Jacanas(Terres 1980). I therefore studied by Jenni and Betts (1978: 215) spent suspect that the function of attacks on this most time in "overt anti-predator behaviour." speciesis to defend resources,not offspring. The vulnerability of offspringto predatorssuch as Purple Gallinules undoubtedly decreasesas CONSEQUENCESOF ATTACKS the offspring mature, and this probably ac- counts for the decline in responsivenessfrom Jacanaattacks appeared to have both imme- the downy to the juvenile stages.It is lessclear diate and long-term consequencesfor offspring why responsivenessto Purple Gallinules is protection. The immediate consequence was lower during incubation than during the that intruders were prevented from getting downy stage,but this could be due to (1) the nearer to jacana offspring. In fact, the only greater conspicuousnessof the mobile chicks known casesof predation by Purple Gallinules comparedto eggs,and/or (2) the fact that eggs on jacana eggs and chicks occurred while the are incubated for a large portion of the day latter were temporarily unattended by their (Jenni and Betts1978) and are presumablysafe parents (Jenni pers. comm.). The long-term during this time, whereas chicks are brooded consequenceof attackswas a reduction in the for a smaller portion of the day, especiallyafter density of Purple Gallinules (and perhaps oth- the first few days of chick life (Jenni and Betts er intruders) near jacananests (and perhapsnear 1978, Stephens unpubl. data). the mobile offspring as well). This conclusion The high responsivenessof malesto Purple needs confirmation, however, as the support- Gallinules during pre-incubation,when no eggs ing data come from only one nest. Confirma- or offspring were present to defend, suggests tion is needed for an additional reason:the pe- that males were trying to discouragePurple riod of higher density included a higher Gallinules from inhabiting the area near the proportion of immature Purple Gallinules than males' future nest. did the period of lower density and it is thus Much theoretical and empirical work on the possiblethat the change in density was a by- temporalpattern of offspringdefense concerns product of differences in microhabitat use be- the severityof attacks(e.g. Andersson et al. 1980 tween adult and immature Purple Gallinules and references therein, Bierman and Robertson and had nothing to do with jacanaaggression. 1981) and not responsiveness to intruders. Someevidence suggests that aggressionby oth- Temporal changesin offspringdefense proba- er avian speciesreduces the density of poten- bly are bestindicated by the severity of attacks, tial predators near nests or offspring (Gorrans- rather than responsivenessto intruders, when son et al. 1975, Walters 1980). aggressionis directed against moderately to highly dangerouspredators of eggsand young, THE ROLE OF FEMALES IN OFFSPRING DEFENSE such as gulls and foxes, which tend to elicit attackswhenever they are near nests(e.g. Kruuk Female Northern Jacanasare highly unusual 1964).The presentstudy primarily concernsless among female birds in being under intense se- dangerousspecies that commonly searchfor a lection to acquire multiple mates (Jenni and variety of foods within jacana territories but Betts 1978, Stephens 1982) and to retain them are also potential predators of jacana eggs or over a long breeding season.This requires that offspring. The severity of attacksagainst these female jacanasbe free of the major time con- speciesdid not vary much from attackto attack straints involved in incubating, brooding, and (see above), at least according to my measure escortingthe young (Jenni and Betts 1978). On of severity. the other hand, participation in offspring de- The proximate causesof attacks and direct fense is compatible with simultaneous poly- evidenceof predationon jacanaeggs and chicks andry. Jacanasspend little time attacking off- by Purple Gallinules (see above) provide con- spring predators (Jenni and Betts 1978). July1984] JacanaInterspecific Aggression 517

Moreover, becausefemales are much larger than assistance;the Organization for Tropical Studies(RIAS males, they probably are more effective than Fund), the American Museum of Natural History males in preventing predation against off- (ChapmanFund), the American Ornithologists'Union spring. There is some support for this sugges- (Van Tyne Fund), NIMH (BehavioralBiology Train- ing Grant MH-15181 to S. Altrnann), the National tion. Female jacanasare known to attack con- Academyof Sciences(Bache Fund), the Universityof specific offspring despite vigorous attacks by Chicago (Hinds Fund), and NSF (Grant DEB-8113258) males (Stephens 1982, 1984); yet they are easily for financial support; N. Zuniga for assistancewith repelled by other females (Jenni and Collier fieldwork during 1981;P. Dring, D. Jenni, and F. G. 1972, Stephens 1984). Stiles for logistical advice; and D. Moskovits, M. Per- I suggestthat females,because of their larger eira, T. Price, J. Altrnann, S. Arnold, J. Fitzpatrick, M. size, are ultimately responsiblefor the fact that Howe, C. Packer,J. Walters,and anonymousreview- opportunistic predators such as Purple Galli- ers for commentson previous drafts of the manu- nules do not attack jacanaeggs or offspring in scriptand/or other aspectsof this study. I am espe- the presence of the parents. I therefore also cially grateful to Stuart Altrnann for advice throughoutthis studyand commentson severaldrafts suggest that, while it is often concluded that of this manuscript. female Northern Jacanasprovide little or no parental care (e.g. Wittenberger 1981, Halliday LITERATURE CITED 1982, Knowlton 1982, Erckmann 1983), the fe- males'role in offspringdefense is important in ANDERSSON, M., C. G. WIKLUND, & g. RUNDGREN. the life history of jacanas,even though this ac- 1980. Parentaldefense of offspring:a model and tivity is not time consuming.Although data are an example. Anirn. Behav. 28: 536-542. limited, females in other speciesof jacanasap- ALTMANN,J. 1974. Observationalstudy of behav- pear to have a similar role in offspring defense iour: sampling methods. Behaviour 49: 227-267. (Hoffmann 1949, Cunningham-van Someran BAILEY,A.M. 1927. Habits of the Purple Gallinule (Ionornismartinicus). Auk 44: 560. and Robinson 1962, Mathew 1964). BEADEL,H. L. 1946. Purple Gallinule robs nest of Offspring defense through direct attack is Green Herons. Auk 63: 87-88. common among birds (Skutch 1976). This study BIERMAN,G. C., & R. J. ROBERTSON.1981. An in- suggeststhat offspringdefense by polyandrous creasein parental investment during the breed- femalesor polygynousmales is compatiblewith ing season.Anirn. Behav.29: 487-489. polygamous breeding, when birds can become CUNNINGHAM-VAN SOMEREN,G. R., & C. ROBINSON. polygamous without deserting mates. In some 1962. Notes on the African lily-trotter, Actophi- nidifugous species,territories break down after lornisafricanus (Grnelin). Bull. Brit. Ornithol. Club 82: 67-72. broods hatch, and the families of a polygynous DUNN,O.J. 1964. Multiple comparisonsusing rank male or polyandrous female may then move sums. Technornetrics 6: 241-252. apart in searchof food. In such species,polyg- EMLEN,S. T., & L. W. ORING. 1977. Ecology,sexual amous individuals may opt to follow and de- selection,and the evolutionof mating systems. fend only one family (Hannon 1984). Even in Science 197: 215-233. these species,however, if mates remain within ERCKMANN,W.J. 1983. The evolutionof polyandry hearing range, asdo Northern Jacanas,individ- in shorebirds:an evaluationof hypotheses.Pp. uals can recruit distant mates to ongoing at- 113-168 in Social behavior of female vertebrates tacks. Few data are available on the role that (S. K. Wasset, Ed.). New York, Academic Press. polyandrous females or polygynous males play GORRANSSON,G., J. KARLSSON,S. G. NILSSON, & S. ULFSTRAND. 1975. Predation on birds' nests in in offspring defense(Jenni and Betts 1978, Pic- relation to antipredator aggressionand nest den- man 1983). A likely contrast to emerge from sity: an experimental study. Oikos 26:117-120. suchstudies is that polyandrousfemales invest GRAUL,W. D., S. R. DERRICKSON,& D. W. MOCK. 1977. less than polygynous males in offspring de- The evolutionof avianpolyandry. Amer. Natur. fense during the pre-incubation period, when 111: 812-816. femalesare forming eggs. HANNON,S. J. 1984. Factorslimiting polygyny in the Willow Ptarmigan. Anirn. Behav. 32: 153- ACKNOWLEDGMENTS 161. HALLIDAY,T. 1982. Sexual strategy.Chicago, Univ. ! thank the Costa Rican Ministry of Agriculture Chicago Press. and Ranchingfor permissionto conductthis study; HOFFMANN,A. 1949. Ober die Brutpflege des poly- the personnel of the Rodriguez Refuge and the Or- andreschenWasserfans, Hydrophasianus chirurgus ganization for Tropical Studies for hospitality and (Scop.).Zool. Jahrb. 78: 368-403. 518 MARTINL. STEVHEN$ [Auk, Vol. 101

HUNTERß R. A., & R. D. MORRISß 1976. Nocturnal REPENNING,R. 1977. Great Egret preys on sandpi- predation by a Black-crownedNight Heron at a per. Auk 94: 171. Common Tern colony. Auk 93: 629-633. SIEGEL,S. 1956. Nonparametric statisticsfor the be- JENNI,D.A. 1974. Evolution of polyandry in birds. havioral sciences. New York, McGraw-Hill. Amer. Zool. 14: 129-144. SKUTCH,A. F. 1976. Parent birds and their young. ß & 13.J. ]3ETTS. 1978. Sex differences in nest Austin, Texasß Univ. Texas. construction, incubationßand parental behav- SLATERßP. J. ]3. 1974. Boutsand gaps in the behav- iour in the polyandrousAmerican Jacana(Jacana iour of zebra finches,with specialreference to spinosa).Anim. Behav. 26: 207-218. preening. Rev. Comp. Anim. 8: 47-61. ß& G. COLLIER.1972. Polyandry in the Amer- SLUD, P. 1980. The birds of Hacienda Palo Verdeß ican Jacana,Jacana spinosa. Auk 89: 743-765. Guanacaste, Costa Rica. Smithsonian Contr. Zool. ß R. D. GAMBS,& B. J. BETTS. 1975. Acoustic 292: 1-92. behavior of the Northern Jacana.Living Bird STEPHENSßM.L. 1982. Mate takeover and possible 13: 193-210. infancticide by a female Northern Jacana(Jacana KNOWLTON, N. 1982. Parental care and sex role re- spinosa).Anim. Behav. 30: 1253-1254. versal. Pp. 203-222 in Current problems in so- ß 1984. Intraspecificdistraction displays of the ciobiology(King's College SociobiologyGroupß polyandrous Northern JacanaJacana spinosa. Ibis Eds.). New Yorkß Cambridge Univ. Press. 126: 70-72. KRUUK,H. 1964. Predatorsand antipredator behav- TERRES,J. K. 1980. The Audubon Societyencyclo- iour of the black-headedgull Larusribidundus. Be- pediaof North Americanbirds. New YorkßAlfred haviour Suppl. 11: 1-129. Knopf. MACEßT.R. 1981. Causation, function, and variation WALTERS,J.R. 1979. Interspecificaggressive behav- of the vocalizations of the Northern Jacana,Ja- iour by Long-toed Lapwings (Vanelluscrassiros- cana spinosa.Unpublished Ph.D. dissertation. tris). Anim. Behav. 27: 969-981. Missoula, Montana, Univ. Montana. 1980. The evolutionof parentalbehavior in MATHEW,D.N. 1964. Observationson the breeding lapwings. Unpublished Ph.D. dissertation.Chi- habits of the Bronze-winged Jacana(Metopidius cagoßUniv. Chicago. indicus(Latham)). J. BombayNat. Hist. Soc. 61: WELT¾,J. C. 1982. The life of birdsß third Ed. Chi- 295 -301. cago,Saunders College Publ. MCILHENNY,E.A. 1936. Purple Gallinules (Ionornis WITTENBERGER,J. F. 1981. social behavior. martinica)are predatory. Auk 53: 327-328. Boston,Duxbury Press. PICMAN,J. 1983. Aggressionby red-winged black- birds towards marsh wrens. Can. J. Zool. 61: 1896- 1899.