The Condor 92~405-412 0 The CooperOrnithological Society 1990

COMPARISON OF ROOST USE BY THREE SPECIES OF COMMUNAL ROOSTMATES

DOUGLAS W. MORRISON Department of BiologicalSciences, Rutgers University,Newark, NJ 07102

DONALD F. CACCAMISE Department of Entomology, Cook College,Rutgers University,New Brunswick,NJ 08903

Abstract. Our previous studies of communally roosting European Starlings (Sturnus vulgaris)revealed that each fed daily for months on its own “diurnal activity center” (DAC) and commuted to a variety of nearby and distant roosts. Since these observations contrast sharply with the predictions of most foragingexplanations for communal roosting, we wanted to determine if this DAC-centered roostingpattern occurredin other communally roosting species.In this study we used radiotelemetry to monitor feeding and roosting sites used by starlings,Common Grackles (Quiscuh quiscula),and American Robins (Turdus migratorious),three speciesthat share communal roosts in central New Jersey. Our goals were to determine (a) whether avian speciesthat roost together use the roosts in similar ways, and (b) when and why individuals change roosts. patterns were similar in all three species;individuals fed daily on their DACs for many weeks. Roosting patterns were similar for gracklesand starlings;individuals switchedamong nearby and distant roosts. In contrast, robins always roosted near their DACs, changingboth roosts and DACs at the end of the breeding season. rates at roostswere extremely low and did not explain the use patterns of large and small roosts. We argue that (a) DACs and DAC-centered roostingare probably widespreadamong communally roosting species,and (b) DAC-based individuals select roosts primarily on the basis of their proximity to good sourcesof food. Key words: Communal roosting:foraging; diurnal activity center;predation; European Starling; Stumus vulgaris; Common Grackle; Quiscalusquiscula; American Robin; Turdus migratorious.

INTRODUCTION individuals actually used a variety of roosts but Several investigators have hypothesized that returned to feed on their own “diurnal activity communal roosting by evolved to facilitate center” (DAC, an area of l-2 km?) day after day foraging in various ways. For example, roosting for months. Contrary to the assumptions of ear- in a central location can reduce average com- lier models, foraging and roosting were centered muting costs when food patches are short-lived around a stable feeding area (DAC) rather than (Horn 1968). Communal roostsmight also serve around a roost. as “information centers” at which individuals Foraging and roosting behaviors of starlings can learn about new or better food sourcesby change in midsummer. During most of the year, following successfulroostmates (Ward and Za- starlingsfeed exclusively on their DACs and join havi 1973). These and other “roost-centered” small (50-500 bird) communal roostsnearby (< 2 models (e.g., Weatherhead 1983) are based on km away). During late summer and fall, starlings the assumption that the roost is a relatively stable begin to commute to rooststhat are much larger base of operation from which individuals exploit (5,000-50,000 birds) and more distant (4-12 km). less stable food sources. These larger roosts do not form because of a A very different picture of communal roosting shortageof acceptable roosting sites (Lyon and emerged when roostmates were individually Caccamise 198 l), and they are not premigratory marked (Morrison and Caccamise 1985). Radio- aggregations,because they form months before tracking studies of communally roosting Euro- migration (Caccamise et al. 1983). pean Starlings (Sturnus vufguris) revealed that When using distant roosts, starlings continue to feed primarily on their DACs, but also stop briefly at food sourcesnear the distant roosts.By 1Received 26 July 1989. Final acceptance 7 Feb- roosting overnight near food patches far from ruary 1990. their DACs, starlingscan reducecommuting costs

t4051 406 DOUGLAS W. MORRISON ANDDONALD F. CACCAMISE

to supplemental food patchesby as much as 50% from cotton shoelace(Morrison and Caccamise (Caccamise and Morrison 1986). Recent obser- 1985). After an initial period of adjustment (6- vations of individual starlings en route to and 48 hr), the bird’s behavior was not noticeably from roosts support this “patch sitting” aspect affectedby the transmitter; the radio-taggedbirds of communal roosting (Caccamiseand Morrison flew and foragednormally for months. The range 1988). Is the DAC-centered roosting pattern of of the transmitters was greater than 1.0 km for starlings unique or do other communally roost- birds roosting in trees and less than 0.5 km for ing specieshave similar patterns? birds foraging on the ground. Transmitter life To investigate the relationship of communal averaged 90 (SD = 15) days. roosting to DACs, we undertook a comparative To determine whether the birds had DACs study of starlings, Common Grackles (Quisculus (i.e., spatially clustereddiurnal sightings),we used quiscula),and American Robins (Turdus migra- an automobile with a rooftop antenna to locate torious),three speciesthat sharecommunal roosts each bird during the day, 6 days a week, at ran- in central New Jersey (Caccamise and Fischl domly predetermined times between 06:OO and 1985). This is the first study to monitor both 18:O0. Attempts to locate the birds were suc- roosting and foraging behavior of heterospecific cessfulover 85% of the time for all three species. roostmates. By radio-tagging individuals cap- This successrate was somewhat lower than in tured from the same feeding area, we controlled previous studies(Morrison and Caccamise 1985) for the effectsof heterogeneity. We found becauseof a lack of experienced field assistants that DAC-based foraging behavior is not unique in 1987. to European Starlings and may be widespread For interspecific comparisons, spatial cluster- among communally roosting species.We argue ing was quantified as the mean distance from that DAC-based birds select among suitable each diurnal sighting to the use-weighted center roosting sitesprimarily on the basis of their prox- (i.e., mean x and y coordinates) of all diurnal imity to good feeding areas. sightingsmade on that bird. SeparateDACs were recognizedwhen all sightingsin one cluster were METHODS >2 km from all sightingsin the other cluster. In We radio-tagged nine European Starlings, nine practice, the few cases of multiple DACs in- Common Grackles, and nine American Robins, volved separations of several kilometers. all adults, in central New Jerseyduring the local We located the roost usedby each radio-tagged roosting season (June-November) of 1987. All bird six nights a week. Attempts to locate roost- starlingsand grackleswere postbreedingindivid- ing sites were successfulover 93% of the time. uals. Five of the robins were captured toward the The birds could almost always be found by cen- end of their breeding seasonand may have been susingafter dark the many traditionally usedroost reproductively active for the first 1O-20 days of sites in the study area. Distances from DACs to the 46-102 days (X = 77 days) that they were communal roosts were measured from the use- tracked (Bovitz 1990). weighted center of the bird’s DAC. All birds were captured during the day in the To describe long-term patterns of roost use, horticultural display garden of Rutgers Univer- we restrictedour analysisto the 17 birds for which sity, New Brunswick, New Jersey. Robins were we had greater than 40 days of observations: six mist-netted; starlings and grackleswere captured starlings (four males, two females), four grackles using decoy traps and baited walk-in traps. Birds (three males, one female), seven robins (four selectedfor radio-taggingappeared in goodhealth, males, three females). Radio contact with the with well-maintained plumage and body weights other 10 birds (X = 20 days) was much shorter. above the mean for that species and sex. We Four had transmitters with defective batteries; taggedand releasedall birds at the site of capture after only l-2 weeks in the field, their pulse rates 53 hr after they were trapped. The birds sub- increased almost two-fold, symptomatic of a sequentlyfrequented an urban-suburban area that voltage-drop in lithium batteries. The other six was primarily residential and commercial (8 1%) transmitters lost their antennas, reducing their but contained a few agricultural fields (4%) and effective range to < 100 m. woodlots (2%) (Fischl and Caccamise 1985). Estimates of predation rates were based on the The transmitter package (< 5 g) was attached fates of 63 adult starlings tracked during four to the back of the starling (7 l-85 g), grackle (95- summer roosting seasons(1983-l 986). Each bird 131 g) or robin (73-84 g) with a “vest” made was assignedto one of five categories:eaten, dead AVIAN COMMUNAL ROOSTING 401 but not eaten, battery expired, premature trans- munal roosts at a total of 15 different sites. The mitter failure, and unknown. The eaten category three specieswere frequently found roosting to- may have overestimated predation rate because gether. The six starlings(n = 487 nocturnal sight- it may have included birds eaten after dying of ings) and the four grackles (n = 279 nocturnal other causes.The battery expired category was sightings)joined roosts known to include both assignedonly when the signal was lost (a) after starlingsand gracklesover 97% of the time. Rob- the expiration date calculated on the basis of ins (n = 462 nocturnal sightings) roosted with battery capacity and transmitter current drain, starlings and grackles at least 57% of the time. and/or (b) after a significant (1.5 to 2.0-fold) Of the 15 roosts, 12 were minor roosts of increase in transmitter pulse rate, indicating a ~2,000 birds. One site (R-88) was occupied by voltage drop in lithium batteries. Premature an intermediate-sized flock that peaked at 3,000- transmitter failure was assigned only when the 5,000 in late August. Two sites-(R-42 and R-97) bird had been visually sightedwith a nonworking were occupiedby major roostingflocks of 1O,OOO- transmitter. The unknown categoryincludes un- 35,000 birds. Major roosts have formed at the confirmed transmitter failures and birds that latter two sites every roosting season for many moved out of the study area. years (Caccamiseet al. 1983). In past years, R-42 Interspecificcomparisons were made using one- has peaked as high as 65,000-125,000. way ANOVAs (GLM procedure of PC-SAS). The two major roosts (R-42 and R-97) were Duncan’s multiple range test (P < 0.05) was used the only distant rooststo which our radio-tagged to compare species’ means. birds commuted; i.e., the only roosts used that were > 4 km from their DACs. Therefore, in this RESULTS study major roost and distant roost are synon- FIDELITY TO DIURNAL ACTIVITY ymous. CENTERS Although all individuals used more than one Foraging grackles and robins showed starling- roost, most birds (11 of 17) used only nearby like fidelity to DACs. The diurnal sightingsmade roosts (e.g., Fig. 2E, F). All seven robins, one of on individuals of all three specieswere spatially four grackles and two of six starlings roosted clustered.The mean distance of diurnal sightings within 3 km of their DACs 100% of the time. from the use-weightedcenter of the DAC (shown On average, robins used roosts that were signif- as circle radii in Fig. 1) were not significantly icantly closer to their DACs (1.0 km) than did different for starlings (0.87 km), grackles (0.88 starlings(3.9 km)andgrackles(3.6 km; F= 10.94, km), and robins (0.48 km; F = 1.93, P = 0.18). P = 0.001). Daily sightings suggestedthat starlings and Six individuals (three starlings and three grackles moved around more than robins, but grackles)used distant roosts (>4 km from their the mean distancemoved between sightingsmade DACs). Distant roosts were used for extended on successivedays was similar for starlings (0.54 periods (50-90% of the time) by three starlings km), grackles(0.55 km), and robins (0.36 km; F (e.g., Fig. 2A) and two grackles (e.g., Fig. 2C). = 1.16, P = 0.34). One starling (Fig 2B) used distant roosts on only Most of the birds had a single DAC (e.g., Fig. five of 78 nights: one roost (R-42) 9 km west of l A-C). All six starlings and three of the four its DAC and one roost (R-97) 8 km south of its grackles returned daily to a single DAC for the DAC. entire observation period. In contrast, four of the seven robins shifted to new DACs 7-9 km to the ROOST SWITCHING south at the end of the breeding season(e.g., Fig. In 1,228 bird nights of observation, we recorded lE, F). The only other case of DAC shifting in- 133 casesof roost switching in which the bird volved grackle female 207 (Fig. 1D); she used used a roost different from the roost it had used three distinct DACs in succession,separated by the previous night. The average distance moved distances of 6.5 and 5.5 km, respectively. from night to night was significantly greater for grackles (0.85 km) than for starlings (0.39 km) USE OF ROOSTS and robins (0.18 km; F = 7.82, P = 0.005). Although all 17 radio-taggedbirds were captured Grackles not only switched roosts frequently in the same foraging area, they commuted to a (15.1% of the time), but almost always (39 of 4 1 variety of different roosts. During the June-No- cases)switched to roosts >2 km from the pre- vember roosting season,the 17 birds joined com- vious night’s roost(X = 5.74 km). Robins switched 408 DOUGLAS W. MORRISON AND DONALD F. CACCAMISE

A. STARLING (male) - 413 8. STARLING (maim) - 111

R_4(15)q~3km& R_4(4)qnAco*5km&

C. CRACKLE (maim) - 201 D. CRACKLE (female) - 207

R-79 (13) R-14 (2)

R-99 (5)

OAC 1.3 km

DAC 1.1 km R-4 (23) +J

t t .O DAC 1.2 km N O! Q - R-97 (39)

E. ROBIN (female) - 404 F. ROBIN (male) - 408

R-93 (20) DAC 0.1 km DAC 0.1 km n-93 (24) -+=J

* R-45 (1)

R-99 (2)

DAC 0.9 km R-97 (36) t t N DAC 0.6 km N % -&I R-97 (49) 07 Or

FIGURE 1. Maps showing relationship of communal rooststo diurnal activity centers(DAC, circles) used by starlings, grackles, and robins. The number by each DAC circle is the average distance (km) of all diurnal sightingsfrom the use-weighted center of the DAC. R-numbers refer to roosts. Numbers in parenthesesand width of lines indicate how many nights the bird was found at each roost. DAC-centered roosting is evident in all three species.

almost as frequently (12.3% of the time), but only in DAC location. Most birds (three grackles,sev- rarely (four of 56 cases)did they switch to a roost en starlings, and three robins) returned daily to >2 km away (X = 1.44 km). Starlings switched a single DAC. lessfrequently (7.8%) but always (36 of 36 cases) Of the five casesof DAC shifting (four robins to roosts greater than 2 km away from the pre- and one grackle), all were accompanied by a ma- vious night’s roost (K = 5.15 km). jor switch in roost. Prior to the change, the four In all three species,individuals switchedroosts robins foraged daily in the Rutgers Display Gar- (n = 133 cases)much more frequently than they dens and roosted together in a small (200-bird), changed DACs (n = 5 cases).Accordingly, most all-robin roost < 1 km away. Over a l-week pe- roost switchescould not be attributed to a change riod in mid-July, at the end of the breeding sea- AVIAN COMMUNAL ROOSTING: 409

LO* A. STARLING (male) - 413 B. STARLING (male) - 111

8” c 0.

P 4.

8 2.

0 I 1 I 0, * I I 1 2EJUN 25JUL WUG lescp 16ul DBJM OaJUL asUL MUI; 1esEP

C. GRACKLE (male) - 201 “jD.GRACKLE (female) - 207+ 1

0 # I I iJ 04Auc a7Affi1wuL OaAlE 016Ep 2asw 18ocT

LO- E. ROBIN (female) - 404 F. ROBIN (male) - 408 8.

FIGURE 2. Seasonalpatterns in the distancebetween diurnal activity center(DAC) and communalroosts usedby individual starlings,grackles, and robins.

son, all four abandoned this roost. Two of them in l-3 km of R-97. The only other bird that (females 404 and 412: Fig. 1E and 2E) moved shifted its DAC was gracklefemale 207 (Fig. 1D). 1.5 km south to a roost (R-48) of several hundred She twice moved her DAC and roost simulta- robins and starlings, but continued to feed in the neously, first 6 km S to a DAC near R-97 and Display Gardens. After two and five nights at then 5 km NW to a DAC near R-42 (Fig. 2D). R-48, they switched to R-97, a major roost of 30,000 starlings, grackles, and robins, 8 km to PREDATION RATES the south. The other two robins (males 407 and During the roosting seasons of 1983-1986, 63 408; Fig. 1F and 2F) alsojoined R-97 after using adult starlings were radio-tracked for a total of unknown roosts and foraging areas for 5 and 6 4,665 bird days. Seven of thesebirds were found days. All four robins establishednew DACs with- eaten or partially eaten; e.g., one bird found on 410 DOUGLAS W. MORRISON AND DONALD F. CACCAMISE a tree branch was likely eaten by a bird of prey, We had speculatedthat DAC fidelity in star- and two were found with cat-like tooth marks in lings might improve the chances of securing a the transmitter coating. The eaten category may nest hole the next spring (Morrison and Cacca- have overestimated predation because some of mise 1985) because competition for nest holes thesebirds could have been eaten after dying for begins in winter (Kessel 1957), and some starling other reasons.Relevant here is that none of the DACs contained the bird’s nesting site. How- seven suspectedpredations occurred at roosts. ever, a nesting site is not required for this be- All occurred during the day. havior to manifest itself. Robins maintained DAC-centered foraging patterns even after shift- DISCUSSION ing their DACs far from their breeding areas. Four of seven robins changed the location of The overall pattern of foraging and roosting was their DACs at the end of the breeding season. similar for starlings, grackles, and robins. Indi- DAC-shifting also occurred in gracklesand star- viduals fed daily on their DACs and commuted lings, but lessfrequently. One grackle shifted her to a variety of nocturnal roosts. These similari- DAC twice. No starling shifted its DAC in this ties may reflect the similar feeding used study, but we observed DAC shifting by five of by these species. During the spring and early 63 adult starlings tracked prior to 1987. One summer, all three speciesforage primarily for soil male and two females moved their DACs just invertebrates on commercial and residential after breeding, and two females shifted in late lawns, habitats that have a relatively homoge- summer (Morrison and Caccamise 1985, un- neous distribution in our study area (Fischl and publ. observ). Caccamise 1985). During this period, the birds DAC shifting might be a responseto a decline use only small roosts near their DACs. As the in food availability on the original DAC or may summer progresses,diet composition begins to reflect the dietary change from invertebrates to change, becoming almost entirely plant material fruits and grains. Robins may also move to more by autumn in starlings (Fischl and Caccamise protective habitats (from lawns to forests) in 1987), grackles (Maccarone 1985), and robins preparation for their molt in August-September (Wheelwright 1986). These foods are found in (Bovitz 1990). habitats that are more heterogeneouslydistrib- Whatever its adaptive significance, fidelity to uted, e.g., orchards and grain fields. The change DACs may be found in many communally roost- to plant material coincides with the formation ing species.Our studiesof starlings,grackles, and of larger roosts (Fischl and Caccamise 1987). robins are the first to quantify this behavior, but Interspecific differencesin this DAC-centered similar fidelity to feeding areas is suggestedin pattern were primarily differencesof scale. Star- the descriptions of foraging by wing-tagged star- lings and grackles had similar-sized DACs and lings (Feare 1984) Red-winged Blackbirds Age- commuted to both nearby and distant roosts. luiusphoeniceus (Johnson 1979), Cattle Egrets Robins had somewhat smaller DACs and used Bubulcusibis (Siegfried 197 l), and Great Blue only nearby roosts.These differencesmay reflect Herons Ardeuherodius (Krebs 1974).DACs will more subtle, unmeasured differencesin the spa- probably be found in other speciesas more stud- tial and temporal distribution of food patches. ies are done using marked individuals. Recognition that DACs are widespread in DIURNAL ACTIVITY CENTERS communally roosting birds is an important mile- In all three species,individuals returned to feed stone. Traditionally, models of communal roost- on their DACs day after day for many weeks. ing have been simple extensions of central place Fidelity to a DAC may bestow a number of se- models intended to describe colonial nesters. lective advantages. Returning to a familiar area These models assumedthat the nocturnal roost- is known to increase foraging successin starlings ing site was stable and that food patches sur- (Tinbergen 198 1). Frequenting a familiar feeding rounding the roost were short-lived. The as- area might also decreasethe risk of diurnal pre- sumption that roostsare more stablethan feeding dation (Feare 1984, p. 216). Our observation that areas is reasonable for nesting birds tending im- all on adult starlings occurred away mobile eggs and nestlings. However, this as- from a roost suggeststhat diurnal predation is sumption is violated by many speciesthat roost an important selective force. communally in the nonbreeding season. AVIAN COMMUNAL ROOSTING 411

COMMUNAL ROOSTS Actual differences in predation rates at large and small roostsare extremely small and do not By roosting communally, individuals probably explain the roost-usepatterns documented in this gain both foraging and antipredator benefits. The study. Instead, roosts appear to be selectedpri- evidence for antipredator benefits is indirect. For marily for their proximity to good sources of example, communal rooststypically form in sub- food. strates that are relatively inaccessible to preda- tors (Lack 1968). In theory an individual’s pre- ACKNOWLEDGMENTS dation risk should be lower in larger groups We are especially grateful to Joseph Fischl for assis- (Hamilton 197 1, Pulliam 1973). This hypothesis tance with the fieldwork and data analysis. We also holds for some foraginggroups (e.g., Powell 1974, thank M. B. Decker, P. Bovitz, and K. D’Angiolillo for help in the field. This is contribution no. 55 from but see Lindstrom 1989). The hypothesis has the Department of Biological Sciences,Rutgers-New- never been tested for roosting groups. ark. This is New Jersey Agricultural Experiment Sta- The roost-use patterns that we observed can- tion publication D-081 32102-90, supported by state not be explained by predation alone. If antipre- funds and the U.S. Hatch Act. dator benefits are greater at larger roosts, then LITERATURE CITED birds should converge to form larger and larger roosts(Sibly 1983), or roost size should stabilize Bovrrz, P. 1990. Relationship of foraging substrate at a point where antipredator benefits and for- selection and roosting in American Robins and EuropeanStarlings. M.S.thesis, Rutgers Univ., New aging costsbalance (Giraldeau and Gillis 1985). Brunswick, NJ. The differences assumed to exist in predation CACCAMISE,D. F., L. A. LYON, AND J. FISCHL. 1983. rates at large and small roostsmust be extremely Seasonalpatterns in roosting associationscom- small, becauseactual predation ratesat our roosts posedof starlingsand Common Grackles.Condor 84:474-481. were very low. None of our recorded predations CACCAMISE,D. F., AND J. FISCHL. 1985. Patterns of on starlings occurred at a roost. In addition, ma- associationof secondaryspecies in roosts of Eu- jor and minor roosts were active concurrently, ropean Starlingsand Common Grackles. Wilson and individuals switchedback and forth between Bull. 97:173-182. them. Roost switching has also been observed in CACCAMISE,D. F., ANDD. W. MORRISON. 1986. Avi- an communal roosting: implications of “diurnal starlings on their overwintering grounds, but no activity centers.” Am. Nat. 128:191-198. explanation for the switching was given (Heister- CACCAMISE,D. F., ANDD. W. MORRISON.1988. Avi- berg et al. 1984). an communal roosting:a test of the “patch sitting” As DAC-based foragers,starlings, grackles, and hypothesis.Condor 90:453458. FEARE,C. J. 1984. The Starling. Oxford Univ. Press, robins do not routinely use communal roosts as Oxford. information centers (sensu Ward and Zahavi FISCHL,J., ANDD. F. CACCAMISE.1985. Influence of 1973). Nevertheless, foraging considerations re- habitat and seasonon foraging flock composition main essential for understanding how these in the EuropeanStarling (Sturnusvulgaris). Oeco- speciesselect communal roosts.On most nights, logia 67:532-539. FISCHL,J., ANDD. F. CACCAMISE.1987. Relationships all three speciesjoined roosts in protected sites of diet and roostingin the EuropeanStarling. Am. close to their DACs, probably to reduce the risk Midl. Nat. 117:395-404. of nocturnal predation while minimizing com- GIRALDEAU,L. A., AND D. GILLIS. 1985. Optimal muting costsback to the DAC. group size can be stable: a reply to Sibly. Anim. Behav. 331666-667. Distant roostswere used only by starlings and HAMILTON,W. D. 1971. Geometry for the selfish grackles,probably to reduce commuting coststo herd. J. Theor. Biol. 3 1:295-311. distant food patchesbeing used as supplemental HEISTERBERG,J. F., C. E. KNITTLE,0. E. BRAY,D. F. feeding areas.We did not follow gracklesen route Morr, AND J. F. BRESSER.1984. Movements of to distant roosts,but starlingsusing distant roosts radio-instrumentedblackbirds and Europeanstar- lings among winter roosts. J. Wildl. Manage. 48: usually stop at orchards, grain fields, or other 203-209. supplemental food sources near those roosts HORN, H. S. 1968. The adaptive significanceof co- (Caccamiseand Morrison 1988). Roosting over- lonial nestingin the Brewer’s Blackbird (Euphagus night near food patches far from the DAC re- cyanocephah). 49:682-694. JOHNSON,R. J. 1979. Foraging distribution, habitat duces commuting costs to these supplemental relationships, and bioenergetics of roosting and feeding areasby up to 50% (Caccamiseand Mor- flocking red-winged blackbirds in central New rison 1986, 1988). York. Ph.D.diss., Cornell Univ., Ithaca, NY. 412 DOUGLAS W. MORRISON AND DONALD F. CACCAMISE

KESSEL, B. 1957. A study of the breeding biology of POWELL,G.V.N. 1974. Experimental analysis of the the EuroneanStarlina (Sturnus vukaris L.) in North social value of flocking by starlings(Sturnis vul- America. Am. Midc Nat. 58:25%331. garis) in relation to predation and foraging.Anim. KREBS,J. R. 1974. Colonial nestingand socialfeeding Behav. 22:501-505. as strategiesfor exploiting food resourcesin the PULLIAM,H. R. 1973. On the advantagesof flocking. Great Blue Heron (A&a hero&as). Behaviour 5 1: J. Theor. Biol. 38:419-422. 99-134. SIBLY,R. M. 1983. Optimal group size is unstable. LACK, D. 1968. Ecological adaptations for breeding Anim. Behav. 31:947-948. in birds. Methuen, London. SIEGFRIED, W. R. 1971. Communal roosting of the LINDSTR~M,A. 1989. Finch flock size and risk of cattle egret. Trans. R. Sot. S. Afr. 39:419-443. hawk predation at a migratory stopover site. Auk TINBERGEN,J. M. 1981. Foraging- - decisions in star- 1061225-232. lings (humus vulgarisL.) Ardea 69~1-67. LYON, L. A., AND D. F. CACCAMISE.1981. Habitat WARD. P.. ANDA. ZAHAVI. 1973. The imnortance of selection by roosting blackbirds and starlings: certain assemblages of birds as “information management implications. J. Wildl. Manage. 45: centres” for food finding. Ibis 115:517-534. 435-443. WEATHERHEAD,P. J. 1983. Two principal strategies MACCARONE,A. D. 1985. Use of foraging substrates in avian communal roosts. Am. Nat. 121:237- by starlingsand gracklesin relationtofood avail- 243. abilitv. Ph.D.diss.. Rutgers Univ., Newark, NJ. WHEELWRIGHT,N. T. 1986. The diet of American MORRISON,D. W., A&D I? F. CAC~AMISE..1985. Robins: an analysisof U.S. Biological Survey rec- Ephemeral roosts and stable patches?A radiote- ords. Auk 103:710-723. lemetry study of communally roosting starlings. Auk 102:793-804.