The Auk 122{l):108-120, 2005© The American Ornithologists' Union, 2005.Printed in USA.

ALARM CALLING IN SRI LANKAN MIXED-SPECIES BIRD FLOCKSEBEN GOODALE' - AND SARATH W . KOTAGAMA-

^Graduate Program in Orgaiiismic and Evolutionary Biology, University of Massachusetts, Amherst,MassachuseHs 01003, USA; and

'Department of Zoology, Univerf^ity of Colombo, Colombo 3, Sri Lanka

ABSTRACT. —Vocal alarm calls are important to the vigilance and likely the orga-nization of mixed-species flocks, biif community-wide studies of alarm calling inflocks are lacking. We investigated which species alarm-call, and the characteris-tics of their calls, in a large flock system of a Sri Lankan rainforest. We recordednaturally elicited alarm calls during .several attacks by Accipitcr hawks and whilefollowing flocks for 10 h. We then artificially elicifed alarms by throwing a stick fothe side of the flock, in a total of 70 trials af 30 flock sites. The Orange-billed Babbler{TiirLioides rufescens) was the most frequent caller to both the artificial and naturalstimuli, followed hy the Greater Racket-tailed Drongo (Dicniriis paradiseus). Severalother species also called, and multiple species often called to the same stimulus (in23 trials, and in all of the hawk attacks). The species differed in their rapidity ofresponse and in their sensitivity to different natural stimuli. Calls of the gregarioushabbler usually provided a first, unreliable warning of an incoming threat, whereaslater calls of other species emphasizxd the seriousness of the threat. We suggest thatbirds in mixed-species tlocks may be particularly aware of aerial predators for tworeasons: (1) a "numbers effecf," whereby nongregarious species are more aware ofpredafors when surrounded by large numbers of other species; and (2) an "informa-tion effect," whereby species differ in the information available in their alarm calls,leading to an accumulation of information in a mixed-species flock. Received 2 May2003, accepted 25 Aii^^iisl 2004.

Key words: Accipitcr hawks, alarm calls, community ecology, Dicrurus pnradiseus,mixed-species flocks. Orange-billed Babbler, predator vigilance. Greater Racket-tailed Drongo, Sri Lanka, Tiirdoides rufescens.

Llamadas de Alarma en Bandadas Mixtas de Aves en Sri Lanka

RFSUMFN.—Las llamadas de alarma son importantes en la vigilancia yprobablemente en la organizacion de las bandadas mixtas, pero existen pocosestudios sobre llamadas de alarma en bandadas a nivel de comunidad. Investigamoscuales especies presentan llamadas de alarma y las caracteristicas de sus llamadasen un ainplio sistema de bandadas en una selva lluviosa de Sri Lanka. Registramosllamadas de alarma emitidas naturalmente durante varios ataques llevados a cabopor rapaces del genero Accipiter y mientras seguiamos bandadas por un periodo de10 h. Luego inducimos artificialmente llamadas de alarma arrojando una vara sobreel costado de la bandada, en un total de 70 pruebas en 30 sitios. Tiirdoides rufescensfue la especie que llamo con mas frecuencia en relacion con los estimulos artificialesy naturales, seguida por Dicrurus paradiseiis. Otras varias especies tambien llamaron,y multiples especies comiinmente reaccionaron con el mismo estimulo (en 23pruebas y todos los ataques de Aecipiter). Las especies difirieron en su velocidadde respuesta y en su sensibilidad a diferentes estimulos naturales. Las llamadas de

'li-mail: egoodaleC.'bio.umass.cdu

108

January 2005 Alarm Calling m Mixed Flocks 109

T. rufescens usualmente brindaron una primera serial de alerta no confiable sobre laaparicion de una amenaza, mientras que las llamadas postcriorcs de otras especiesenfatizaron la seriedad de la amenaza. Sugerimos que las aves en bandadas mixtaspodrian estar particularmente atentas a depredadores aereos por dos razones: (1)un "efecto del ntimero," en el cual especies no gregarias estan mas atentas a losdepredadores cuando estan acompanadas por una gran cantidad de individuos deotras especies; y (2) un "efecto de intormacion," en el cual las especies difieren enla informacion disponible en sus llamadas de alarma, generando la acumulacion deintormacion en una bandada mixta.

VOCAL ALARM CALLS are an important compo-nent of vigilance in mixed-species bird tlocks.The evidence continues to mount that mixed-species flocks are in large part an adaptation toreduce predation (Powell 1985, Thiollay 1999),and that increased vigilance is an importantcomponent of the benefit (Pulliam 1973, Elgar1989). Although birds can use the movementsof other species as alarm cues (e.g. Lima 1995),vocal alarm calls are crucial in low-light environ-ments such as rainforests (Terborgh 1990). Alarmcalls, and responses to them, have heen observedin many mixed-species flock systems (e.g. Morse1970, MacDonald and Henderson 1977, Munn1986, Ficken 1990), though only a few stud-ies have set out to document the alarm-callingbehavior of most or all species in a flock system(Gaddis 1980, Wiley 1980, Sullivan 1983).

It has also been hypothesized that alarmcalls are an important factor in the organiza-tion of flocks, with calling species surroundedby noncalling ones (Gaddis 1980, Greig-SmithI9HI). Two systems of such asymmetricalalarm-calling have been described. In the firstsystem, alarm calls are made by the most gre-garious species in the flock. Gregarious speciesmay produce vocalizations like alarm callsthat are costly to the caller but beneficial to thecaller's kin (Maynard Smith 1965, Gaston 1977,Sherman 1977) or the caller's mate in a flock(Witkin and Ficken 1979, Hog.stad 1995). Otherspecies in flocks eavesdrop on the gregariousspecies, as has been shown for the DownyWoodpecker {Picoiiics piihesccns), which lowersits vigilance when associating with flocks ledby Black-capped Chickadees (PoecHe alriccipil-lus) and Tufted Titmice (Baeolophus bicohr)(Sullivan 1984, 1985). Such eavesdropping onalarm calls may be widespread (Gaddis 1980)and may account for the fact that gregariousspecies often play an important role in flocks(Moynihan 1962, Hutto 1994). In the second

system, a sallying (fly-catching) species is theprimary alarm caller (Bell 1983, Munn 1986,Diamond 1987, Terhorgh 1990); sallying speciesare highly vigilant, because they are constantlyscanning for prey while perched (Munn 1984).In both systems, many noncalling species takeadvantage of one or two calling species. A moremutualistic system, in which several speciesalarm-call, has not previously been described.

Here, we present a study of alarm calls in alarge and complex mixed-species system of aSri Lankan rainforest. The flock system includesboth kinds of potential alarm-callers: highlygregarious species, such as the Orange-billedBabbler (hereafter "babbler"; scientific names ofspecies are given in Table 1), and sallying spe-cies, including the Greater Racket-tailed Drongo(hereafter "drongo") (Kotagama and Goodale2004). We hypothesized that other specieswould eavesdrop on babblers and not call if theinformation available in babbler calls was suf-ficient for them and if alarm-calling was costly.We observed, however, that babblers, drongos,and several other species alarm-called duringattacks by raptors. To determine the quality ofthe information available in these species' calls,we recorded natural alarm calls and artificiallyelicited alarm calls by throwing a stick to theside of the flock. We then measured how reliablethe calls were (i.e. what percentage of naturalalarms were made to raptors), and how oftenand quickly they called. We also measured theacoustic structure of the calls, and we use theconnection between the form and fimction ofpredator-related calls (Marler 1955, Klump andShalter 1984) to suggest how costly the calls wereand what benefits the birds gained by calling.

METHODS

Study site and species.—The Sinharaja WorldHeritage Reserve (6°2rN, 80°21'E) is located in

no GOODAI.F AND KoTAGAMA [Aiik, Vol. 122

flie humid southwestern lowland of Sri Lanka.Vegetation within the reserve is broadly classifiedas evergreen dipterocarp rainforest (Gunatillekeand Gunatilleke 1981). Our study area, near theSinharaja Research Center in the northwesternsector of the forest (between 450 and 600 m abovesea level), was logged in the 1970s.

Mixed-species flocks on the study area havebeen studied since 1981; they average 11 speciesand >40 individuals (Kotagama and Goodale2004). The babblers and drongos are by far themost frequent species, present in -90% of flocks.The babbler and another member of the familyTimaliidac, the Ashy-headed Laughingthrush(hereafter "laughingthrush"), are highly gre-garious, averaging 16 and 7 individuals perflock, respectively; most other species averagebetween 1 and 3 individuals per flock. Thesebird flocks also involve squirrel species, a phe-nomenon common on the Indian suhcontinent(Henry 1971, AH and Ripley 1987).

Recordings of naturnlly elicited alarm calls.—Over several years, we opportunisticallyrecorded occasions when an avian predatorflew directly into the flock. The primary avianpredators we encountered were three speciesof Accipiiter hawks: the Shikra {A. baditis), theCrested Goshawk {A. trivirgatiis), and the BesraSparrowhawk (A. virgatiis).

To systematically study alarm calls, werecorded 10 h of flock vocalizations from 12separate flocks throughout the northwest-ern sector of the reserve, in August 2001 andFebruary 2002. We noted when an alarm callwas made as a response to a stimulus (a raptoror a large or fast-moving nonraptorial bird) orwas followed by alarm behavior (a scattering ofbirds or a sudden silence). On listening to thetapes, we scored alarm calls by the more precisedefinitions given below.

Sampling design for artificiaUy elicited calls.—Between June and August 2001, we artificiallyelicited alarm calls from birds. To encounterflocks, we walked a circuit of -15 km composedof the main road to the forest as well as footpaths,completing a full circuit every two days. To avoidoverestimating sample size, we divided the cir-cuit into sites where flocks were repeatedlyencountered and used tlock sites as the unit ofreplication. Sites were defined to be >250 m fromother sites (with 150 m allowed in two excep-tional areas, where neighboring flocks were oftenseen <250 m apart). During previous work in the

reserve, we documented that flocks have stablehome ranges and that flocks seen >250 m apartare highly likely to include different birds. Thetotal number of tlocks thus defined was consis-tent with our previous estimates of flock density.For example, in our fieidwork between 1996 and1998, we regularly encountered nine flocks on a3.5-km stretch of road (Kotagama and Goodale2004), and that same stretch included 10 sites inthe present study. Overall, we performed 70 tri-als at 30 flock sites.

Experimental procedure. —Because we noticedthat any fast aerial movement elicited alarm callsfrom birds, we artificially elicited alarm callsby throwing sticks to the side of flocks. Stickswere thrown -10 m high and -10 m distant, sothat they passed -3 m from the closest bird. Weused a Marantz PMD 222 cassette recorder anda Sennheiser ME62 microphone with a Telingaparabola to record the birds' baseline activityfor 30 s before throwing the stick, and for afurther 30 s after a throw. Because a stick some-times failed to elicit a response, we threw threesticks consecutively, with 30 s to 1 min betweenthrows; in analyses, we used the response to thefirst stick that elicited an alarm call. We also tooknotes on how many individuals of each specieswere present in the 10-m^ area located betweenthe place from which the stick was thrown andwhere it hit the vegetation.

Definition of an alarm call—We defined analarm call as a change in a bird's vocalizationsin the 10 s after the natural or artificial stimulusas compared with the 30-s period before (notethat this operational definition may differ fromtraditional definitions of alarm calls based onfunction or acoustic structure, e.g. Klump andShalter 1984; see below). Some species werealways silent during the 30-s baseline period,and an alarm call was simply the beginningof a vocalization. However, several specieswere often vocal during the baseline period.For those species, we defined an alarm call as achange in note type (descriptions of note typesare included in the results).

Analysis of characteristics of alarm calls.-Recordings were analyzed using the sound-analysis program AVISOFT, version 3.9e (R.Specht, Berlin, Germany). Recordings weredigitized at 22.05 kHz, 16-bit sample rate,and sonograms were constructed with a fastFourier transform (FFT) size of 1,024 points.For both the naturally and artificially elicited

January 2005 Altiriii Ctj!liii;i in Mixcii 111

calls, we measured which species called first.For artificially elicited calls, we also measuredduration of alarm calls (including calls ofall individuals of a species) and rapidity ofresponse (measured as the time elapsed afterthe stick was thrown).

We made acoustic measurements on thefirst note of artificially elicited calKs, where anote was defined as an uninterrupted vocal-ization. We constructed a power spectrimi oithat note and measured the frequency withstrongest amplitude ("peak frequency"), andthe frequency bandwidth, defined as the dif-ference between the minimum and maximumfrequencies that had amplitudes within -15 dBof the peak frequency (Pudos 20G1). For 41 of 91calls thus analyzed, a constant band of cricketnoise between 4.5 and 6 kHz interfered with themeasurements. For those calls, we determinedthe extent of the cricket peak in the amplitudespectra before the note and subtracted that fromthe spectra of the note, using the followingrelationship (measurements in volts; R. Spechtpers. comm.): square root (signal) = square root(signal + cricket) - square root (cricket).

Statistical riud/i/s/s.—To characterize the unitsof replicatitm for artificially elicited calls, wescored the species on their responsiveness ineach trial, and determined their scores for eachtlock site. A species scored "1" if some individu-als called, and "0" if individuals were presentnear the stick and did not call, even though abird of another species called (we refer to suchbehavior as "keeping silent"). If there were noindividuals of that species near the stick, or if nobird called during a trial, the species was givenno score and not included in the mean for theflock site. Similarly, we determined note-typesused in alarms within sites; a species scored "^"if it used the note-type being counted and "0" ifit used a different note-type. These methods ofaveraging within sites resulted in fractional val-ues (e.g. Table 1, column 5). We also determinedthe characteristics of the alarm calls of a specieswithin flock sites.

We analyzed only those species that werepresent near the stick in at least five flock sites.To compare the characteristics of alarm callsamong species, we used a general linear modelin SAS, release 8.00 (SAS Institute, Cary, NorthCarolina). Variables that did not conform toassumptions of homoscedasticity (as measuredby a modified Levene test; Neter et al. 1996)

or normality (as measured visually by normalprobability plots) were transformed. We used alog transformation or, if the variable includedonly integers, a square-root transformation(Sokal and Rohlf 1993). Mtiltiple comparisonswere made using the Tukey H5D method. Weused a G-test for independence with Williams'scorrection when analyzing proportions (Sokaland Rohlf 1995).

RESULTS

NOTE-TYPE CATEGORIZATION AND ALARM-CALL

DEFINITIONS

Alarm calls were unambiguous, and all spe-cies used predominantly a distinct note-typeas an alarm, responding similarly to naturaland artificial stimuli. We describe here the notetypes we identified for each species (illustratedin Fig. 1) and document which ones were usedas alarm calls.

Ashy-hcaiicd Lau^'^liiiii^thnish. —This specieswas constantly vocal in the baseline period, pro-ducing "pi'ig" notes, which had strongest ampli-tudes ~4 kHz with a suppressed fimdamental at~2 kHz, and "iau^h" notes, which were of broadfrequency spectrum and of duration >0.20 s. Allalarm calls {n = 5 naturally elicited calls, ii = 5artificially elicited calls) consisted of a switchfrom those note-types to a series of "cak" notes,each of which had a broad frequency spectrumand duration of -0.10 s. "Cuk" notes were fol-lowed by periods of silence.

B!ack-)mpcd Monarch.—This species (hereafter"monarch") was usually silent during the base-line period. Alarm calls usually (5 of 7 nah.irallyelicited calls, 6.5 of 8 artificially elicited calls)consisted of "zee" notes, which had frequen-cies sweeping from 2 kHz up to 7 or 8 kHz.Occasionally, when the monarch was producing"zee" notes in the baseline period, it switchedafter the stimulus to a "trill" call, which consistedof a series of notes sweeping from 2 to 4 kHz.

Greater Racket-tailed Drongo. —The drongo'scalls, and the notes within them, were highlyvaried. We distinguished only between callsthat started with notes that had strong ampli-tude between 4 and 6 kHz ("high-pitchednotes") and calls that did not. Drongos wereusually silent during the baseline period andusually (16 of 16 naturally elicited calls, 13.5of 15 artificially elicited calls) started alarm

112 GOOOAI F AND KOTACAMA [Auk, Vol. 122

HJ

vt/101

^QJ

ra

i-QjX3

I -

1_raIM

raX_gt n

QJIMQJ^I/)1.-*J

XXra

o

QJucra

Xra1-1

01

ra

cie

Qj

roC

a;t n

3B

u uO C

5;U

ifi > C_ _ * - QJ

ra 2 ^

Qj ^

i^\ Q.J .pM

° 3 X

ly. ra

ra " QJ•n T i 4=

TJC

i

u T;

c —I

in ffl

O O)

be

< - ^ ii• ra p .2

c .2

:;= o C

p IT, O p pT-H fvj (— t ^ LO

in p in p

CL,

ra

C •

oZ

^ c

_QJ T3 •f P .5

n: ra

o5 : TJ- = : QJ

aj ora X(

U — C•n ra bc

-a D-, t -

O O S :S

-W >M

nC•

«

-arau

•a

~ c .E— I"

i s

5 ;- ,r

_aj

'iiiJ ^

flo

0

(1

s

uli

E

-3"m

CC

-a

"3£In

laniiarv20{15 Alarm CalUiig in Mixed 113

L-Thrush

t i

Ping L-Thrush: Laugh L-Thrush Cak

4-

M 2

o

£ 8D"

4

2

Monarch: Zee Monarch: Trill Drongo

t

f f f f A'-r

Babbler: Babble Babbler: R

H

Chatter

i i,«, \i . 1

Babbler:

; < ; * ;

S

i

Chatter

.Hi;;.

Babbler: Cuk Bulbul: Weep Bulbul: Loop

I f.. A

1 2 3 4 5Time (s)

FIG. 1. Note-types of vocal flock participants. Alarm calls were defined as a change in note-type following the stimulus, and all species predominantly used a distinct note-type as an alarm(the "cak" note of the Ashy-headed Laughingthrush, the "zee" note of the Black-naped Monarch,phrases of the Greater Racket-tailed Drongo starting with a high-pitched note, the "cuk" note of theOrange-billed Babbler, and the "weep" note of the Yellow-browed Bulbul).

114 GOOUALE AND KOTACiAMA [Auk, Vo!, 122

calls with their high-pitched notes, followedby phrases (defined as a series of notes withno interval >0.50 s) that were repeated severaltimes. We measured the end of the drongo callas the last note of a phrase or series of identicalphrases introduced by a high-pitched note (inFig. 1, the call lasts for 2.5 s).

It should be noted that drongos are capableof mimicry, and vve have found, in fieidworksubsequent to the present study, that drongosmimic the alarm calls of other species whenthey themselves appear to be alarmed. We donot believe Ibal such mimicry poses a seriousproblem for the identification of alarm calls,because it is rare in comparison with drongos'use of their own alarm notes (E. Goodale andS. W. Kotagama unpubl. data) and because ouraural scoring ot" alarm calls was consistent withour field notes on the direction from which thealarm call came and the spatial position of thebirds. Drongo mimicry is often identifiable,because mimicked notes are typically repeatedand included in sequence with regular drongonotes, and can have harmonic patterns or otherfeatures unusual for the species that was mim-icked (we documented three occasions whenthe drongo used mimicry in the present study).

Orniigc-hillcii Babbler.—This species was con-stantly vocal in the baseline period, producingmostly "babble" notes distinguished by a shortchevron-shaped beginning to the note andan ending of falling frequency. Less commoncall-types included "regular chatter," a macbine-gun-like series of evenly spaced notes witb ashort interval between notes (0.09-0.12 s) and"staccato chatter," a similarly fast-paced seriesof notes that were unevenly spaced. Alarm callsusually (28 of 31 naturally elicited calls, 21 of21 artificially elicited calls) consisted of a switchfrom tbose call types to "cuk" notes, unevenlyspaced notes with a relatively long intervalbetween notes (0.20-0.40 s). "Cuk" notes werethen followed by a period of silence.

Yellow-browed Bulbul.—Thi^ species (hereafter"bulbul") was usually silent during the baselineperiod. Alarm calls usually (five of six naturallyelicited calls, seven of nine artificially elicitedcalls) consisted of "loeep" notes, which rosein frequency and were produced in an unpre-dictable order. We distinguished "weep" notesfrom tbe "loop" call, whicb was a series of notesalways presented in a predictable order andrepeated regularly, much like a song.

NATURALLY ELKITKD ALARM CALLS

Stimuli that elicited alarnn^. — Aceipiter hawkswere involved in five of the six attacks werecorded (tbe sixth attack was by a CrestedSerpent Eagle [Spllonii^ cheela], though tbeseeagles are not known to prey on birds; Henry1971). In their attacks, the bawks flew low(5-10 m above ground) directly through theflock (see similar observations by Morse [19731).In one of fhe Accipiter attacks, however, we didnot record the first several seconds of the alarm;we do not include that attack in the analysis. Inthree of the AccipiSer attacks, the hawks wereseen making repeated flights through the tlock,several minutes apart, wbich resulted in severalbursts of alarm-calling per attack (though weanalyze only tbe first burst of an attack).

In tbe 10 h of systematic recording, therewere 36 alarm-call episodes in which at leastone species called. Of the 27 episodes in wbichwe observed the stimulus that elicited the call,21 included a large or fast-tlying nonraptorialbird, and 6 a raptor (though no direct attackswere observed during that period).

ComiiiO}Uic::^fl of alarm calls.—The species dif-fered considerably in their responsiveness tothe natural stimuli. Babblers called the most,drongos were next, and birds of the otherspecies mentioned above called less commonly(Table 1, column 1). Birds of six additionalspecies called just one time eacb and are notanalyzed. Commonness of calling may havebeen a function of the number of individuals ofeach species (see Table 1; for artificially elicitedalarms, tbe number of individuals near the stickwas known).

Order of alarm calls. —Botb babblers and dron-gos tended to call first. Babblers called first in 21of the 31 times they called, and drongos calledfirsf in 12 of the 17 times they called. Other spe-cies called later: monarcbs were never first inthe seven times they called, and bulbuls werefirst in only two of tbe six times tbey called.Tbose last two species called first significantlyless often than did babblers and drongos (C.,, =11.94, two-tailed P< 0.002).

Comparisoti of calls elicited hy raptors and bynoiiraptorial f'/>rfs.—Babblers made many morealarms to birds that were not raptors than didthe otber species (Table 1, column 3). Babblerswere thus less reliable than birds of theother species considered together (G . = 7.73,

Alarm Cullin;^ in Mixed Flocks 115

two-ttiilcd P < 0.012) or drongos in particular(G . = 5.16, two-tailed P < 0.044).

Significantly more species responded it the.stimulus was a raptor (average = 2.4 species call-ing; ;i = 11) than it the stimulus was a non-raptdrialbird (average = 1,5; n = 21; two-sample ^test formeans, t = 2,71, df = 30, two-tailed P < 0.013). Themost species responded in the four attacks hyAccipiters (average = 3.0 species calling).

Duration of alarm calls was also much longerwhen the stimulus was a raptor than when thestimulus was a nonraptorial hird. Alarms lastedan average of 48,0 s in response to raptors {n =10), and an average of 5.3 s in response to otherbirds {il = 16; two-sample (-test for means, t =5,23, df = 24, two-tailed P < 0.0001). Responseduring the four Aceipiter attacks was very long(average = 88.7 s). Much of the duration ofalarms in response to raptors was attributableto drongos, which were the last birds to stopcalling in 7 of 10 responses to raptors (including3 ot" the 4 responses to Aecipiter attacks).

A R T I I ICIALLY ELRllEn CALLS

Connnoiuiess of alarm calls.—Tht^ species dif-fered in the number of alarm calls they made to

artificial stimuli, and the pattern was similar tothat seen in response to natural stimuli. Babblerscalled the most frequently; drongos were next;and bulbuls, monarchs, and iaughingthrushescalled at a lower rate (Table 1, column 5).SqLiirrels (including two Fiinanihuliis spp. thatwe did not distinguish in our observations) alsocalled to the artificial stimulus, switching fromsilence to a high-pitched series of notes in all{u = 7) of their calls. As in the natural observa-tions, birds of several species sometimes (23 of70 trials) called to the same presentation of thepredator model (see Fig. 2 for an example). Birdsof two species, the Red-faced Malkoha and theMalabar Trogon, never called (Table 1, column 6),despite their frequent presence in the tlocks.

The number of calls a species produced was inpart a consequence of the abundance of the spe-cies within flocks. To produce a measure of a spe-cies' propensity to call, we divided the numberof times a species called by the number of timesit kept silent. By that measure, the babbler had alower propensity to call than several other spe-cies (Table 1, column 7; more exact measures ofthe number of alarm calls per individual cannotbe made, because we were unable to determinehow many individuals of a species called).

>.oc3

cr

£ 2

LL

Babbler; Cuk

Time (kHz)

FIG, 2, An example of a multispecies alarm in response to a stick throw. Calls of four species areevident in the sequence; the stick was thrown at 0 s. Elements of the original sonogram are tracedmanually for emphasis.

116 GOODALE AND KOTAGAMA [Auk, Vol. 122

Order and rapidity of alarm (."fl//s. —Babblersand drongos tended to call first, as in the natu-ral observations, though babblers were moreoften (18.4 of 21 calls) and drongos less often(8.3 of 15 calls) first. Monarchs called first inonly 2,6 of the 8 times they called, and bulbulsin 2.5 of the 9 times they called, significantlyless often than babblers and drongos (G , =9.11, two-tailed P< 0.006).

The order of the calls was a consequence ofthe rapidity of response, which differed signifi-cantly among the species (f = 6.51, df = 5 and 59,P < 0.0001). Specifically, bahblers responded sig-nificantly more rapidly than monarchs (TukeyHSD multiple comparisons, P < 0,01), drongos{P < 0.025), and bulbuls (P < 0.05). Rapidity ofresponse was correlated with the numbers ofindividuals of each species that were near thestick: the more individuals nearby, the quickerthe response (Fig. 3A). The same trend can beseen within babblers, the species with the larg-est sample size: the more babblers present, theshorter the response time (Fig. 3B).

Another factor that may have influenced therapidity of response was the noise created bythe stick when it hit the vegetation. Babblerstended to alarm-call before the noise (16.0 of 21calls), but the other species mostly called after it(32.8 of 44 calls for the species combined; com-parison with babblers, G , = 14.93, two-tailedP < 0.0003). Indeed, both squirrels (0 of 7 calls)and monarchs {0 of 8 calls) never called beforethe stick noise. That audio component made theartificial disturbance quite different from that ofany actual threat, and we discuss its significancebelow.

Acoustic eharaeteristics of alarm calls.—Duration, peak frequency, and bandwidth ofcalls varied significantly among species. Ofthose variables, duration varied least, thoughsignificantly (F = 2.45, df = 5 and 59, P < 0.044),with a tendency for alarm calls of drongos tobe longer than those of babblers or laughing-thrushes (Fig. 4A).

The species differed more strongly in theacoustic characteristics of their calls. Peakfrequency was significantly different amongspecies (F = 15.31, df = 5 and 58, P < 0.0001),and several of the multiple comparisons weresignificant, particularly those involving thehigh-pitched calls of monarchs or the low-pitched calls of bulbuls (Fig. 4B). Frequencybandwidth also differed significantly among

0 4 8 12 16

# Conspecifics near predator model

0 4 8 12 ie

# Conspecifics near predator model

FIG. 3. Rapidity of response was related to thenumber of individuals present near the artificialstimulus. (A) Species with many individuals nearwhere the stick fell responded more rapidly (f =24.32, df = 1 and 4, P <0.008, (-=0.86). (B) A'simi-lar trend is shown by the most numerous spe-cies, the Orange-billed Babbler, including thedata from all sites in which the babbler called(f = 13.15, df - 1 and 19, P < 0.002, r - 0.41),

species (F = 11.83, df = 5 and 50, P < 0.0001), andseveral comparisons were significant, particu-larly between the richly harmonic calls of bab-blers and Iaughingthrushes and the more tonalcalls of bulbuls (Fig, 4C).

DISCUSSION

More species make alarm calls in these SriLankan tlocks than in any other reported flocksystem. We hypothesized that the gregariousbabblers would be the primary alarm-callersin the system, because of their numerical domi-nance and the high probability that many of thebabblers in a flock are closely related, whichmakes them more likely to produce costly sig-nals that benefit their kin (Maynard Smith 1965;see Gaston |1977] and Zahavi |1990] for studiesof congeneric species). Although that speciesdid call most frequently and rapidly, other spe-cies also called, and often several species calledat once. To understand why the other speciescalled, we first look at the cost of alarm-callingby examining the acoustical structure of the

January 2005 Alarin Calling in Mixed Flocks 117

10

FIG, 4. Alarm-call characteristics differedamong species. Species with the same letterwere not significantly different, as determinedby Tukey HSD multiple comparisons. Numberof flock sites at which calls were recorded isindicated in parentheses. Calls varied amongspecies in (A) duration (comparisons significantat a = 0.10), (B) peak frequency (comparisonssignificant at a = 0.01, except that betweenMonarch and L-Thrush, significant at a = 0,05,and that between L-Thrush and Bulbul, signifi-cant at a = 0.10), and (C) bandwidth (compari-sons significant at a =0.01, except those between

calls. We then discuss the differences amongspecies in their alarm-calling behavior, focusingon the unreliability of the babbler as one reasonwhy other species needed to call.

lt should first be noted that our definition ofan alarm call, as simply a change in vocali ^ationpattern following a stimulus, is operational; thequestion arises whether the changes are specificto the predator context. We believe that ourdefinition is appropriate for an initial surveyof birds in mixed-species flocks. Because birdsresponded with the same changes in vocaliza-tion repeatedly in many trials, those changesappear to be context-specific.

The calls recorded in the present study are,however, not typical of alarm calls made to aer-ial predators. Alarm calls to aerial predators areusually of high pitch and narrow bandwidth,which makes it difficult for avian predators todetect or localize them (Marler 1955, Klumpand Shalter 1984, Klump et al. 1986, Jurisevicand Sanderson 1998). Calls like those recordedhere (no species had peak frequencies >6 kHz,and several species had frequency bandwidths>3 kHz) are more often associated with callsmade to terrestrial predators or during distressor mobbing situations, when signals advertisethe location of the predator to recruit otherbirds to mob or attack it (Marler 1955, Curio1978, Jurisevic and Sanderson 1994). Yet thecalls in our observations were all made to mov-ing aerial stimuli, and the call notes evoked bythe artificial disturbance were also elicited byreal Aecipiter attacks.

One explanation of the acoustical propertiesof alarm calls in this system is that they aremade during a stage in a predator attack whenthe cost of alarm-calling is low. We noticedthat Aecipiter hawks sometimes flew straightthrough tlocks several times. Thus, a bird thatobserved a hawk moving out of the flock afterthe initial attack may have had a low cost of call-ing but a high benefit in attracting the attentionof conspecifics to the danger in the area. Such achange in costs is consistent with a distinctionbetween the very first calls during an attack—which were closer to theoretical expectations for

L-Thrush and Drongo, and Squirrel and Bulbul,which were significant at a = 0,05).

118 GOODAI.FL AND KOTAGAMA [Auk, Vol, 122

alarm calls—and the later, more conspicuouscalls that followed. The first calls includedthe short calls of the babbler species, followedby silence, and fhe high-pitched beginning ofdrongo calls; whereas the later calls consistedof longer and repeated phrases by drongos andbirds of several other species (see Fig. 2). Twoother sfudies have also noted that hawk atfackscan be followed by prolonged volleys of calling(Morse 1970, Jurisevic and Sanderson 1998),which could cither confuse the hawk or servenotice that if has been seen (Caro 1995). Giventhat calls made in such sifuations seem inferme-diafe in funcfion befween alarm calls and moh-bing calls, if is nof surprising fhaf fhe acousticalstructure may be intermediate as well,

Species differed significantly in fheir rapid-ify of response fo the artificial disturbance andtheir sensitivity to different natural stimuli. Asexpected, the babbler was the most frequent andrapid alarm-caller in the system. Thaf is largelyan effect of ifs abundance in flocks: a fhreat wasdefected more quickly when there were morebabblers present (see Fig. 3). The observationthat awareness rises with fhe number of indi-viduals in a group is consistent with manystudies on group vigilance (reviewed by Elgar1989), Babblers were nof reliable, however,because they made twice as many calls fo non-predators as they did fo predafors in our naturalobservations (see Haftorn 2000 for a descriptionof similar unreliable calling by fhe Willow Tif[Parus montanus]). Indeed, babblers appear fobe simple "motion defectors," sfimulated byany fast movement—a conclusion supported byfhe artificially elicited alarms when the babblersusually called before fhe sfick made confacfwifh the vegetation. On an individual level,babblers also are nof highly sensitive to aerialthreafs, given fhaf fhey kepf silent in responseto sticks more than any other species. Overall,then, fhe informafion available in bahhier callsis far from perfect.

Monarchs and hulbuls were usually precededin calling by babblers buf continued calling,probably to emphasize a serious threat. The lafe-ness of fhose species' calls may be a direct con-sequence of fheir small numbers of individualsin flocks; hofh species called firsf at leasf once foeifher nafural or arfificial sfimuli, so if is unlikelythey were waiting for ofher species fo call.Rather, they may have been responding to somestimulus other than the initial movement. In the

artificially elicited trials, the sound of fhe sfickhitting the vegetation may have been unusualenough to stimulate those species into calling(thaf may explain the behavior of fhe squirrelsalso). Similar movemenf hy nonpredaceousbirds would not have fhaf audio componentand were mostly disregarded by the monarchsand bulbuls. In the naturally elicited alarms,those species' contribution resulted in the highernumber and length of the calls made to realpredators, fhus reinforcing fhe inifial informa-fion given by fhe babblers about fhe threat.

It also appears fhaf, beyond merely amplify-ing bahhier calls, drongos may be more sensifivefo initially perceiving true threats. Like the callsof monarchs and bulbuls, drongos' long, drawn-out phrases after the initial alarm emphasizedfhe encounfers with raptors. In contrast to thosespecies, however, drongos called more to realpredator stimuli than did babblers, despite hav-ing far fewer individuals per flock, and dron-gos tended to call as rapidly as babblers fo fhenatural sfimuli. The drongo's sensitivity maybe aftribufable to its need fo scan before sal-lying (e.g. Munn 1984), fhough fhe connectionbetween sallying and alarm-calling is nof caus-afive; Malabar Trogons mostly sally for preyand do nof call at all. The sensitivity of drongosfo real fhreats, and the conspicuousness of fheircalls (qualifatively, drongos are louder than theofher species), make drongo vocalizafions agood source of informafion about aerial threatsfor ofher species. Thaf drongos also occasion-ally mimic the alarm calls of ofher birds (seedescripfion of their vocalizafions above) mayfurther extend the influence of fheir calls onofher species.

The differenf informafion available in fhedifferenf species' calls is one reason why birdsin mixed-species flocks may be more aware ofpredators than birds in single-species groups.Morse (1977) firsf hypothesized fhat predatorvigilance might be a benefit of mixed-speciesassociafions, arguing thaf species used differ-enf parts of the vegetation and thus were awareof threats coming from different locations.Two ofher reasons why birds in mixed-spe-cies flocks may be more aware of predaforsare suggested Here. First, there is a "numberseffecf": predators are detected more rapidlywhen there are more individuals, and somenongregarious species will be more awareof predators when associating with large

2005 AInrii! in Mixed Flocks 119

numbers of heferospecifics fhan they could bein fheir small, single-species groups. Scci>nd,there is an "informafion effecf": species differin fhe characterisfics of fheir calls and fhus inthe information provided in them; therefore,associating with more species may increase abird's total source of information. Both of thesehypotheses assume that species are respond-ing to heterospecific calls, an assumption fhatmust be tested with playback experimenfs (e.g,Munn 1986).

ACKNOWLEDGMENTS

We thank B, E. Byers and D. E. Kroodsmafor their assistance in all phases of this study.E,G, thanks P, S. Ashton, C V. S. and I. A. U. N.Gunatilleke, N. E. Pierce, and the late D. R, Griffinfor their help in getfing him fo Sri Lanka andliving in fhe counfry. The Sri Lanka ForesfDepartment provided permission and facili-ties at the Sinharaja Research Station. K. T.Jayarathna and P. A. Jayarathna were excellentfit-Id assistants, and U. M. Goodale and B, W. M.Wijesinghe provided invaluable help in the field.We fhank ]. Podos and R. Specht for advice onmeasuring the frequencies of calls. Earlier draffsof this manuscript were impri>ved by fhe sugges-tions of B. E. Byers, U, M. Goodale, R, L. Hutfo, S.Jolinson, D. I. King, P B, Mclntyre, S. M. Smith,K. A. Sullivan, and two anonymous reviewers.The study was made possible by financial assis-fance to E.G. from a National Science Foundationpredocforal fellowship, a Sigma-Xi grant-in-aid,and a Woods Hole scholarship for field research.

i-: CITED

A[.i, S,, AND S. D. RiPLEY. 1987. CompacfHandbook of the Birds of India and Pakistan,Together wifh Those of Bangladesh, Nepal,Bhutan and Sri Lanka, 2nd ed. OxfordUniversify Press, New Delhi.

BELL, H. L, 1983. A bird communify of lowlandrainforest in New Guinea. 5. Mixed-speciesfeeding flocks. Emu 82:256-275.

CARO, T. M, 1995, Pursuit-deterrence revis-ifed. Trends in Ecology and Evolufion 10:500-503.

CuRro, E. 1978. The adaptive significance of avianmobbing. I. Teleonomic hypotheses andpredictions. Zeitschrift fiir Tierpsychologie48:175-183.

DIAMOND, J. 1987. Flocks of brown and blackNew Guinea birds: A bicoloured mixed-spe-cies foraging association. Emu 87:2(M-211.

EiGAR, M. A. 1989. Predator vigilance andgroup size in mammals and birds: A criticalreview of the empirical evidence. BiologicalReviews of tbe Cambridge PhilosophicalSociety 64:13-33.

FiCKEN, M. S, 1990. Acoustic characteristics ofalarm calls associated with predation risk inchickadees. Animal Behaviour 39:400-401.

GADDIS, P. 1980, Mixed flocks, accipiters, andanfipredafor behavior. Condor 82:348-349.

GASTON, A. J, 1977. Social behaviour wifhingroups of Jungle Babblers {Tunioitics stria-tus). Animal Behaviour 25:828-848.

GREIC-SMIIEI, P. W. 1981. The role of alarmresponses in the formation of mixed-spe-cies flocks of heafhiand birds. BehavioralEcology and Sociobiology 8:7-10.

GRiMMtrr, R., C. 1NSKU>P, AND T. INSKIPP. 1999.A Guide to fhe Birds of India, Pakisfan,Bangladesh, Bhutan, Sri Lanka and theMaldives. Princeton University Press,Princefon, New Jersey,

GUNATFLLEKE, C. V. S., AND I. A. U. N.GuNATiLLEKE. 1981. The florisfic composi-tion of Sinharaja —A rain foresf in Sri Lankawith special reference to endemics and dip-terocarps. Malayasian Forester 44:386-396.

HAFTORN, S, 2000, Contexts and possible func-tions of alarm calling in fhe Willow Tif,Parus niontaniis: The principle of 'betfer safefhan sorry'. Behaviour 137:437-449.

HKNKV, G, M. 1971. A Guide to the Birds ofCeylon, 2nd ed. Oxford University Press,Oxford,

HOGSTAD, O, 1995. Alarm calling by Willow Tits,Parus moutauus, as mafe invesfmenf. AnimalBehaviour 49:221-225.

HUTTO, R. L. 1994. The composition and socialorganization of mixed-species flocks in atropical deciduous forest in western Mexico.Condor 96:105-118

JURISEVIC, M. A., AND K. J. SANDI-RSON, 1994.Alarm vocalizafions in Ausfralian birds-Convergent characteristics and phyloge-nefic differences. Emu 94:67-77.

JuRisi'Vic, M. A., AND K. J. SANDERSON. 1998.Acoustic discrimination of passerine anti-predafor signals by Ausfralian raptors.Australian Journal of Zoology 46:369-379.

KLUMP, G, M., E. KREIZSCHMAR, AND E. CURIO.

120 GOODALE AND KOTAGAMA [Auk, Vol. 122

1986. The hearing of an avian predatorand its avian prey. Behavioral Ecology iindSociobiology lS:317-323.

KLUMP, G. M., AND M. D. SHALTER. 1984. Acoustic

behaviour of birds and mammals in tbepredator context. I. Factors affecting tbestructure of alarm calls. II. The functionalsignificance and evolution of alarm calls.Zeitscbrift fiir Tierpsychologie 66:189-226.

KOTAGAMA, S. W., AND E. GOODALE. 2004. The

composition and spatial organization ofmixed-species flocks in a Sri Lankan rainfor-est. Forktail 20:63-70.

LTMA, S. L. 1995. Collective detection of preda-tory attack by social foragers: Fraught withambiguity? Animal Behaviour 50:1097-1108.

MACDONALD, D. W., AND D. G. HENDERSON.

1977. Aspects of the behaviour and ecologyof mixed-species bird flocks in Kashmir. Ibis119:481-491.

MARLER, P. 1955. Characteristics of some animalcalls. Nature 176:6-8.

MAYNARD SMITH, J. 1965. Tbe evolution of alarmcalls. American Naturalist 99:59-63.

MORSE, D. H. 1970. Ecological aspects of somemixed-species foraging flocks of birds.Ecological Monograpbs 40:119-168.

MORSE, D. H . 1973. Interactions between titflocks and sparrowhawks Accipiter nisiis.Ibis 115:591-593.

MORSE, D. H. 1977. Feeding behavior and preda-tor avoidance in heterospecific groups.BioScience 27:332-339.

MovNiHAN, M. 1962. The organization andprobable evolution of some mixed-speciesflocks of Neotropical birds. SmithsonianMiscellaneous Collections vol. 143, no. 7.

MUNN, C. A. 1984. The bebavioral ecology ofmixed-species bird flocks in Amazonian Peru.Pb.D. dissertation, Princeton University,Princeton, New Jersey.

MUNN, C. A. 1986. Birds that 'cry wolf. Nature391:143-145.

NETER, J., M. H. KUTNER, C. J. NACHTSCMEIM,

AND VV. WASSERMAN. 1996. Applied LinearStatistical Models. WCB/McGraw-Hill,Boston.

PODOS, J. 2001. Correlated evolution ofmorphology and vocal signal structure inDarwin's finches. Nature 409:185-188.

POWELL, G. V. N. 1985. Sociobiology and adap-tive significance of interspecific foragingflocks in the Neotropics. Pages 713-732 iuNeotropical Ornithology (P. A. Buckley, M.S. Foster, E. S. Morton, and R. S. Ridgely,Eds.) Ornithological Monographs, no. 36.

PULLIAM, H. R. 1973. On the advantages offlocking. Journal of Theoretical Biology 38:419-422.

SHERMAN, P. W. 1977. Nepotism and the evolu-tion of alarm calls. Science 197:1246-1253.

SOKAL, R. R., AND F. J. ROHLF. 1995. Biometry, 3rd

ed. W. H. Freeman, New York.SULLIVAN, K. A. 1985. Selective alarm calling

by Downy Woodpeckers in mixed-speciesflocks. Auk 102:184-187.

SULLIVAN, K. A. 1984. The advantages of socialforaging in Downy Woodpeckers. AnimalBehaviour 32:16-22.

TERBORGH, J. 1990. Mixed flocks and polyspecificassociations: Costs and benefits of mixedgroups to birds and monkeys. AmericanJournal of Primatology 21:87-100.

THIOLLAY, J.-M. 1999. Eret]uency of mixed-spe-cies flocking in tropical forest birds andcorrelates of predation risk: An intertropica!comparison. Journal of Avian Biology 30:282-294.

WILEY, R. H. 1980. Multispecies antbird societiesin lowland forests of Surinam and Ecuador:Stable membership and foraging differences.Journal of Zoology (London) 191:127-145.

WiiKiN, S. R., AND M. S. FICKEN. 1979. Chickadeealarm calls: Does mate investment pay divi-dends? Animal Behaviour 27:1275-1276.

ZAHAVI, A. 1990. Arabian Babblers: The questfor social status in a cooperative breeder.Pages 105-130 in Cooperative Breedingin Birds: Long-term Studies of Ecologyand Behaviour (P. B. Stacey and W. D.Koenig, Eds.). Cambridge University Press,Cambridge, United Kingdom.

Associate Editor: R. L. }-lolbcrton