Seasonal Patterns in Roosting Flocks of Starlings and Common Grackles

Condor 85:474&31 G The Cooper Ornithological Society 1983

SEASONAL PATTERNS IN ROOSTING FLOCKS OF STARLINGS AND COMMON GRACKLES

DONALD F. CACCAMISE LINDA A. LYON

AND JOSEPH FISCHL

ABSTRACT. -Communal roosting is often a regional phenomenon that involves wide-ranging and long-lasting relationships among associations.We examined roosting behavior on a scale sufficiently large to detect regional and seasonal patterns. For five roosting seasons(June-November), we studied the population dynamics of all roosting flocks of European Starlings(Sturnus vulgaris)and Com- mon Grackles (Quiscalus quiscula) located within a l,OOO-km2census area in central New Jersey. Roosts were active from 3-20 weeks and ranged in size from 2,000 to over 100,000 individuals. The total roostingpopulation (TRP) in “major” (>2,000 birds) flocks increased through early summer, generally achieving maximum size in mid-August when the largest number of roosts was active. When TRP was largest, size of major roosts varied greatly (range 2,000-100,000 individuals). Through late summer and early fall, size and number of major roosts and TRP declined. By late fall few major roosts were active, but those remaining were large (> 30,000). Movements of individual birds (radio-tagged) suggested that changesin size of TRP resulted largely from exchangeof the local population between small, “minor” roosts (largely undetected and not included in roost censuses)and major flocks. Current hypotheses concerning the functional basis of communal roosting do not adequately explain patterns of roosting behavior that we observed. -

Communal roosting behavior occurs in birds Starlings (Sturnus vulgaris) and Common of diverse taxa and habitats. Important func- Grackles (Quiscalus quisculu)within a l,OOO- tional role(s) are suggestedby the often large km2 censusarea. We investigated patterns of flock sizes and long roosting seasons.Several activity at individual roosts and population explanationsfor communal roostinghave been dynamics on a regional basis over five con- offered (Lack 1968, Siegfried 197 1, Tast and secutive seasons.Our objectives were to quan- Rossi 1973, Ward and Zahavi 1973, Weath- tify the seasonalpatterns in the size of indi- erhead 198 3) and some tests have been per- vidual roosting flocks, to determine the formed (DeGroot 1980, Loman and Tamm interrelationships among these flocks, and to 1980, Fleming 198 l), but for most speciesthe analyze the spatial and temporal distribution functional role of this behavior remains poorly of the roosting population on a regional basis. understood. Our goal was to provide an adequate descrip- Few quantitative data are available for ex- tive basis for testing current hypotheses ex- amining this behavior on a scale(temporal and plaining communal roosting behavior. spatial) sufficiently large to detect regional and seasonalpatterns. Previous studieshave either STUDY AREA AND METHODS examined activity at individual roosting as- Field work was conducted during the local sociations (e.g., Hamilton and Gilbert 1969, roosting seasons(June-November) of 1977- Swingland 1976) determined the distribution 198 1. A region within the Piedmont and inner- of roosts for a given speciesduring only part coastal plain provinces (Robichaud and Buell of the roosting season (e.g., Marples 1934, 1973) of central New Jersey (Fig. 1) was rig- Eastwood et al. 1962, Hein and Haugen 1966), orously censused(one to three times per week) or provided few data (e.g., Brown 1946, Gadgil for all major blackbird roosts: i.e., those con- and Ali 1976). These investigations have gen- taining at least 2,000 birds. The minimum size erally neglected regional relationships among of “major roosts” (>2,000) represented the roosts or seasonal variations in roosting be- smallest aggregationregularly detected in our havior. surveys. We counted some smaller roosts, but We examined the post-breeding population only at sites where they had formerly num- dynamics of all roosting flocks of European bered over 2,000 birds. Roost sites were usu-

[4741 PATTERNS IN ROOSTING FLOCKS 415

KEY

0 ROOST SITE IDENTIFICATION NUMBER

URBAN AREAS !P

0 369 km

FIGURE 1. Map of the census area indicating location of the roost sites. Numbers inside circles are roost site identification numbers.

ally located during our surveys by following ulation estimates,as roosting activity was then flightlines of birds returning to roosts in the erratic. We developed our field methods in evening. Some “minor roosts” (~2,000 birds) 1977 (Lyon 1979). were found when radio-tagged birds went to Eighteen starlings and eight grackles were them, but the small size and inconspicuous- equipped with radio transmitters (AVM In- ness of minor roosts made them difficult to strument Company) in 1980-l 98 1. Birds were detect and we made no effort to find them. We captured at roost sites using either mist-nets did not include minor roosts in size estimates atop 15-m poles or decoy traps in foraging of the roosting population. areas. Five starlings and three grackles were The size of the censusarea (approx. 1,000 lost soon after release and two starlings and km*) was defined initially by the limits of our one grackle were found dead (one predatory ability to conduct thorough censuses.After es- loss, one poisoned, one accidental). The 14 tablishing the distribution of roosting flocks remaining birds were monitored three to five (1977), we confined our searchesto the area times weekly on foraging areas during the day bounded by lines connecting the seven out- and at roosts at night for periods ranging from ermost roost sites. 13-139 days (mean 82 days). Active roosts were visited every 7-14 days. In order to estimate roost size, one to four RESULTS observerswere stationed along flight lines near roosts. Birds were counted by speciesas they PATTERNS OF ROOSTING BEHAVIOR arrived in the evening or departed in the mom- Roost sites first became active in early June ing. The total roosting population (TRP) rep- (Fig. 2). The number of active sites increased resentsthe sum of all birds in the censusarea through mid-August, then declined until the using major roosts (>2,000 birds). TRP was end of the local roosting season(early Novem- calculated at five-day intervals by summing ber). Individual roosts were active for 3-20 sizes of individual flocks that were active at weeks (Fig. 2). Generally the same sites were the beginning of each period. Counts during used in successiveyears and their heaviest use inclement weather were not included in pop- occurred over similar dates. Some sites were 476 D. F. CACCAMISE, L. A. LYON AND J. FISCHI

abandoned (sites 4, 12, 14, 22, 25) after their vegetation was cleared, but others (sites 2, 6, 18, 20, 33, 35, 36) were abandoned without apparent physical changesin the sites or their surroundings.The size of individual flocks was usually similar among years, but several of them grew over the years from intermediate to very large (sites 21, 23, 42). We witnessed the opposite trend as well (sites 4, 6, 21). All roosts were composed primarily of starlings and grackles(Fig. 3). Starlingswere generally more abundant, although either species could represent more than 80% of an individual flock during the early growth or late decline phasesof the activity cycle. Only once did we encounter a flock in which one speciescom- posed 99% of the population: at the beginning of the 1980 roosting season,site 36 was composed mainly of juvenile starlings. We also counted 11 secondary species roosting com- munally at these sites, but their roles were small, either becauseof their low numbers or their relatively brief stay (Caccamiseand Fischl, unpubl. data). The ratio of starlingsto grackles(Fig. 3) varied seasonally and independently of TRP for all roosts in the censusarea. Values were high early in the season because starlings began roosting first, but grackles joined the roosts while the TRP was still relatively small. Star- lingsgenerally exceededgrackles through most of each season,although sometimes(especially in late fall) grackles became more abundant. The proportion of starlings was much higher in 1978 than in any other year. We were not able to determine whether the grackle population was actually depressedin 1978 or the censusarea did not include the major grackle roosts. The dynamics of total roost population were similar each year of the study with respect to both maximum size (2 1O,OOO-270,000 birds) and seasonalpattern of growth and decline (Fig. 3). In each year (including 1977, Lyon 1979) except 1979, the TRP was largest in mid-Au- gust. In 1979, TRP grew more slowly and the maximum size was reached much later (3 Oc- tober). Peaking so soon before migration, the population also declined more rapidly than in other years. The number of active roosts in the census area increasedfrom June through mid-August (Fig. 2). The increase in TRP through this pe-

10 20 30 10 20 30 9 19 29 9 19 29 9 19 29 7

JUNE JULY AUG. SEPT. OCT. NOV. tification number (seeFig. 1) is locatedat the date of largest flock size. Shapeof border around roost site identification FIGURE 2. Activity periods (solid lines) of individual number indicates roost size class: circle 12,000; square roost sites for each year of the study. Each roost site iden- 2,000-10,000; triangle lO,OOO-30,000;hexagon >30,000. PATTERNS IN ROOSTING FLOCKS 477

250 10.0

200 2.0

150 1.0

z 100 0.5 8 r 50 0.1 z X Y i 0

p 250 10.0 G

5 200 2.0 2 p ’ z B 150 1.0 5

$ 100 8 0.5 F 2 g 50 0.1 8 i? : < 250 10.0 g & + 200 2.0 B b 150 1.0 ci Ii F zi 100 0.5 ii

50 0.1

10.0 200 2.0 150

1.0 100

50 0.5

0.1

lb 20 30 10 20 30 9 li 2.9 a 10 28 0 13 28 7 JUNE JULY AUG SEPT OCT NOV FIGURE 3. Total number of starlingsand grackles(solid circles)and relative speciescomposition (open squares)of all roosting flocks (ratio of starlings/grackles)within the censusarea. Shaded areas highlight ratios close to equality.

riod resulted from not only the establishment of roosting birds was in large roosts. Roosts of of new sites but also the enlargement of indi- the smaller size categorieswere always active vidual roosts once they were active. Never- through the early season,and at times collec- theless,the peak in number of sites generally tively accounted for more of the total roosting coincided with the peak in TRP. population than the largest size category (Fig. The distribution of birds among roosts of 4). different sizeschanged seasonally (Fig. 4). Ear- Following the peak in TRP, the size of in- ly in the season,large roosts (> 30,000 birds) dividual roosts began to decline, eventually increasedmore rapidly than small ones. Thus, leading to the abandonment of many sites(Fig. when TRP was greatest,the largestproportion 2). Few new roostsbecame active beyond mid- 418 D. F. CACCAMISE, L. A. LYON AND J. FISCHL

lf’ 1’ 1’ I’ I.I’ I~I’ I’ ”‘I~lI’ I’ ’ maximum TRP, when roosts of all sizeswere 1981 150 represented. In each year except 198 1, the increasein the l - >30,000 0 - 10,000-30,000 largest category during October was nearly ~-2000-10.000 /ll ( Ii equal to the decrease in the other categories 100 (Fig. 4). In 198 1, the largest category attained a maximum size that was about 60% greater than the number of individuals in the smaller 50 categories.

MOVEMENT OF RADIO-TAGGED BIRDS Of 14 birds that were radio-tagged successfully (10 starlings,4 grackles),five remained in the 150 sameroost through the observation period. We verified roost site changesfor seven starlings and two grackles.Three of the seven starlings 100 8 usedthe same sitesthrough the end of the local 8 roostingseason. After the main population mi- ;; grated,these individuals moved to minor roosts g 50 within the censusarea. The remaining six in- ii dividuals changedroost sites from one to five ts times during the roosting season. B Starlingsoften used minor and major roosts 9 1979 alternately. Movement from one communal 1 150 roosting flock to another was often inter- z! spersedwith one to several nights’ stopover at b a minor roost (Fig. 5, starling). Seven of the 100 radio-tagged starlings, and one grackle used minor roosts at some time during the season. Before the peak in total roost population, in-

50 dividuals generally remained in a single roosting flock. After the peak in TRP, they switched roostsmore often as the number of active sites was declining. The radio-tagged birds abandoned the declining sites,moving either to de- 100 1978 veloping major roosts or to minor roosts (e.g., Fig. 5). In 198 1 one radio-tagged starling moved 50 from a declining site (site 39) that had achieved a maximum size of 15,000 birds to a site (site 2 1) that never became larger than 8,500 birds (Fig. 5). Movement from one declining site to 10 20 30 10 20 30 9 19 29 8 18 28 8 18 28 7 another was unusual, but can be explained by JUNE JULY AUG SEPT OCT NOV consideringthe history of site 2 1 over the pre- FIGURE 4. Total number of starlingsand grackleswith- vious five years. From 1977, when we first in the censusarea according to their-distribution among encountered this site, until 198 1, site 2 1 pro- roost size classes(open circles O-2,000; triangle 2,000- gressed yearly from a moderately-sized fall 10,000; square lO,OOO-30,000;solid circle >30,000). roost to the major pre-migratory concentra- tion in the central part of the censusarea. By 198 1, it returned to relative insignificance August, but of those that did, most came to (maximum roost sizes: 1977-38,000; 1978- hold large fall flocks (e.g., sites 42, 21, 33). By 60,000; 1979-64,000; 1980-24,000; 1981- early October, few roosts were still active. Of 8,500). The radio-taggedstarling moved to this these, one to three became large pre-migratory site in early October, about when the large fall flocks, while the others dwindled until they flock had developed in previous years. The were abandoned.With the roosting population large associationdid not develop in 198 1 and (individuals occupying roosts > 2,000 birds) the starling returned to its minor roost (Fig. 5, concentrated into a few large associations,the starling) after only two nights. Soon thereafter smaller roostswere essentiallyabsent from the it appeared in the major fall roost for 198 1 censusarea. This contrastswith the period near (site 42). This site had held small flocks in PATTERNS IN ROOSTING FLOCKS 479

I I I I I I I ! I I I I I STARLING *L-A FO->L.*+L L C.GRACKLE o+~ +- 90 . t 60 @a t ./ I

10 20 30 10 20 30 IO 20 30 IO 20 30 10 JULY AUG. SEPT. OCT. NOV.

FIGURE 5. Activity cycle (flock size) of roosts used by a radio-taggedstarling and a grackle. Horizontal lines at top indicate when roosts were used by each radio-taggedbird (L indicates a minor roost, arrows indicate when the birds changedroosts). former years, but 198 1 was the first year that from immigration, from either migratory a major roost developed there. When site 42 movements or more local movements of birds was abandoned for the season,the starling re- residingoutside the censusarea. The total roost turned to its minor roost. population grew too early in the seasonto be Grackles migrated in early November. Their affected significantlyby migration. In addition, decline at site 42 (198 1) coincided with the the censusarea is large and encompasseshab- disappearanceof the radio-tagged grackle us- itats representative of central New Jersey, so ing this site (Fig. 5). A large proportion of the there is no reason to believe it is more or less starling population was resident over the win- attractive to vagrants than any other area. For ter. Five of the radio-tagged starlingsreturned these reasons,we believe that the early-season to minor roosts near the end of the season increase in TRP probably reflects changesin where they remained at least until their trans- the proportion of the local population asso- mitter batteries were exhausted. ciated with major roosts (>2,000 birds). Observations of radio-tagged birds suggest DISCUSSION that the early seasonincrease in TRP resulted SEASONAL FLUCTUATION IN TOTAL primarily from the movement of resident birds ROOST POPULATION to major roosting associations.For example, The total roost population typically grew to a an adult male grackle that was captured in its mid-summer peak and declined through late foragingarea and radio-taggedon 17 June 198 1 summer and fall. While 1979 was somewhat remained here for 17 days usinga nearby roost anomalous in that the peak occurred later in of fewer than 25 individuals. When this minor the season,the overall pattern remained sim- roost dispersed,we were unable to relocate the ilar. Changesin the size of TRP could reflect radio-tagged bird until seven days later, when changesin the size of either the resident roost- we found it at major roost 23. At this time, ing population or the population within the site 23 was approaching its maximum size for censusarea. Since the roosting seasonfollows the year (>50,000 birds). the nesting season,we consider the young of Total roost population usually began to de- the year to be members of the resident pop- cline in August, long before the major migra- ulation. Therefore, changesin the size of the tory exodus. Movements of the radio-tagged resident population would have to result from birds also suggesta largely resident adult pop- immigration or emigration. The early season ulation through this period. Of 12 radio-tagged increase of TRP is unlikely to have resulted birds under observation (5 grackles, 7 star- 480 D. F. CACCAMISE, L. A. LYON AND J. FISCHL lings) from mid-August to mid-October, only proceedingto active roosts.Often thesegroups 5 were lost possibly due to migration. How- only circled overhead before departing. Based ever, during this period many roostswere being on their departure bearing and our knowledge abandoned (Fig. 2) and therefore considerable of active roosts in the area, we usually could movement among sites normally occurred. tell where they were going. In some caseswe Some or all of the lost birds may have actually were able to follow the flightlines as the flock moved to sites undetected or beyond the cen- grew and to verify its destination. We hypoth- sus area and did not really migrate. Conser- esize that these birds had formerly used the vatively then, only a fraction of the five lost other sites and had recently abandoned them birds may have migrated. Taken together, the in favor of expanding alternative roosts. 13 radio-taggedbirds suggestthat only a small The smaller categoriesof major roosts were fraction of the total population could have mi- generally absent in the fall. In autumn, radio- grated. We conclude, therefore, that a migra- tagged starlings used both minor roosts and tory exoduswas probably not the primary cause major roosting flocks. Although our sample of for the mid-summer decline in TRP. radio-taggedgrackles is small, we have no evi- The large roosts(> 30,000 birds) that formed dence that they used minor roosts in the fall. near the end of the seasonare likely to have Thus in the fall, the local starling population resulted from the coalescence of local birds was distributed between minor roostsand large rather than a migratory influx. Some migrants major associations, while the grackles ap- may have been present in fall roosts,but based peared to be concentrated in the large major on the general decline in TRP through the fall, flocks. For starlings, association with a large their contribution was probably minor. Only roost did not necessarily lead to migration, in 198 1 did TRP increase substantially near because several of the radio-tagged birds re- the end of the season.However, with so few turned to minor roosts after the large flocks roosts active in the fall, our estimates of TRP declined. Decline of the fall flocks, however, were more susceptibleto sampling error. For coincided with migration of the grackles.The example, if a roost had formed near the edge decline proceeded rapidly for essentially all of of the censusarea, it would have been likely the grackle population left the censusarea over to attract many birds from outside the census a four- to six-day period. area. This would have inflated TRP, making the increase appear as a migratory influx. In THEORETICAL INTERPRETATIONS OF 198 1 the increase in the size of the TRP in ROOSTING October largely resulted from the growth of Formation of large roosting assemblageshas site 42, which was the major fall roost in the been attributed, at least in part, to the need for northern part of the censusarea. Birds from premigratory assembly (Michael and Chao inside the censusarea (Fig. 5), and likely from 1973) and to migratory influx (Davis 1970). outside as well, joined this flock. In our study, roosts of the largestsize category Roosts in the largest size category (Fig. 4) developed in mid-summer, long before migra- formed twice during the season:in Augustwhen tion, and these often declined in size or were TRP was largest, and in late fall just before abandoned before the large fall roosts devel- migration, when TRP was declining. While the oped. Furthermore, several radio-tagged star- large roosts (>30,000 birds) were outwardly lings joined large fall roosting flocks but re- similar during both periods, the patterns of mained in the censusarea after the migration development and the distribution of individ- period. Three starlingsremained through early ualsamong roost size classes(Fig. 4) were quite to mid-December, returning to minor roosts different. Prior to the peak in TRP, birds moved after the large roosts were abandoned in early into developing roosting flocks mostly from November. The peak in TRP occurred an- scatteredlocations, which were likely near their nually, well before the harsh winter weather. breeding sites. During this time, the number This also arguesagainst explanations basedon and size of active sites were increasing and the need to select favorable microclimates roosts representingall size categorieswere ac- (Brenner 1965, Francis 1976, Gyllinet al. 1977, tive. Kelty and Lustick 1977, Walsberg and King In the fall, large roosts(> 30,000 birds) arose 1980). from the coalescenceof individuals from both Under certain circumstances,a shortage of the decliningmajor roostsand individuals from available sites could promote communal minor roosts (five radio-tagged birds, Fig. 5). roosting. However, in our study area suitable In addition, we often saw large groups of birds roost siteswere not limiting, as all of the known moving from declining to developing sites. sites were never used simultaneously (Lyon While making counts at declining or recently and Caccamise 198 1). Also, active sites often abandoned sites, we commonly observed were near inactive sites (e.g., sites 3-7, 6-35, groups of birds briefly stopping there before 32-39). PATTERNS IN ROOSTING FLOCKS 481

Theoretical considerationsargue againstex- DAVIS, G. J. 1970. Seasonalchanges in flockingbehavior planations based on the hypothesis of protec- of starlingsas correlated with gonadal development. tion against predators (Lack 1968, Gadgil Wilson Bull. 82:391-399. DEGROOT,P. 1980. Information transfer in a socially 1972). Flocking should be beneficial at group roosting weaverbird (Quelea quelea; Ploceinae): an sizes far below even the relatively small sizes experimental study. Anim. Behav. 28:1249-1252. (1,000 to several thousand) of roosting asso- EASTWOOD,E., G. A. ISTED,AND G. C. RIDER. 1962. ciations common in spring and summer (Pul- Radar ring anglesand the roosting behavior of starlings. Proc. R. Sot. (Lond.) B 156:242-267. liam 1973). While protection is probably great- FLEMING,T. H. 1981. Winter roosting and feeding be- est in relatively large groups, it is likely that havior of Pied Wagtails Motacilla alba near Oxford, by joining such flocks individuals incur other England. Ibis 123:463-476. costs(e.g., longer commuting distances,more FRAN&, W. J. 1976. Micrometeorology of a blackbird competition). Furthermore, the anti-predation roost. J. Wildl. Manage. 40: 132-136. GADGIL, M. 1972. The function of communal roosts: hypothesis offers no apparent explanation for relevance of mixed roosts. Ibis 114:53l-533. the observed variation in size of active roost- GADGIL,M., AND S. ALI. 1976. Communal roostinghab- ing flocks either seasonallyor at a single time its of Indian birds. J. Bombay Nat. Hist. Sot. 72:7 16- in the season.If protection against predators 727. GYLLIN.R.. H. KALLANDER,AND M. SYLVEN. 1977. The were the most important factor, roosts should mi&oclimate explanation of town centre roosts of be much smaller and more consistent in size. Jackdaws.Corvus monedula. Ibis 119:358-36 1. Our results are also inconsistent with the HAMILTON,WI J., III, AND W. M. GILBERT. 1969. Starling information center hypothesis (Ward and Za- dispersalfrom a winter roost. Ecology 50:886-898. havi 1973). A basic assumption of this theory HEIN, D., AND A. 0. HAUGEN. 1966. Illumination and Wood Duck roosting flights. Wilson Bull. 78:301- is that the size of roosting associationsis di- 308. rectly related to the need for information con- KELTY, M. P., AND S. I. LUSTICK. 1977. Energeticsof the cerning location of foraging areas. We have starling (Sturnus vulgaris)in a pine woods. Ecology shownthat roostsof vastly different sizesoften 58:1181-1185. LACK, D. 1968. Ecological adaptations for breeding in occur simultaneously (Fig. 4) and in close birds. Methuen, London. proximity (Figs. 1,2). To interpret theseresults LOMAN,J., AND S. TAMM. 1980. Do roosts serve as “in- according to the information center hypothe- formation centers” for crows and ravens?Am. Nat. sis, one would have to assume that birds in 115:285-289. different-sized roosts (simultaneously active LYON, L. A. 1979. Communal roostingofblackbirds and starlings:I. Habitat selection.II. Population dynam- and nearby) perceive current food distribu- ics. M.Sc. thesis, RutgersUniv., New Brunswick,NJ. tions in very different ways. We know of no LYON, L. A., AND D. F. CACCAMISE.198 1. Habitat se- empirical evidence for this assumption. lection by roostingblackbirds and starlings:manage- Current hypotheses are inadequate to ex- ment implications. J. Wildl. Manage. 45:435-443. MARPLES,B. j. 1934. The winter star&g roostsof Great plain the patterns of roosting behavior we ob- Britain. 1923-1933. J. Anim. Ecol. 3: 187-203. served. Future studiesshould examine behav- MICHAEL,i. D., AND W. T. H. CHAO. 1973. Migration ior on a scalesufficiently large to showtemporal and roosting of Chimney Swifts in East Texas. Auk and spatial patterns on a regional and seasonal 90:100-105. basis. Such studies would provide a founda- PULLIAM,H. R. 1973. On the advantagesof flocking. J. Theor. Biol. 38:4 19-422. tion for the development of new and more ROBICHAUD,B., AND M. F. BUELL. 1973. Vegetation of satisfactoryhypotheses, thereby furthering our New Jersey.Rutgers Univ. Press,New Brunswick,NJ. understanding of communal roosting behav- SIEGFRIED,W. R. 1971. Communal roostingof the Cattle ior. Egret. Trans. R. Sot. S. Afr. 39:419-443. SWINGLAND,I. R. 1976. The influence of light intensity ACKNOWLEDGMENTS on the roosting time of the Rook. Anim. Behav. 24: 154-158. For their devoted aid with the field work we thank N. TAST,J., AND P. Rossr. 1973. Roost and roosting flights Bent, D. Brook, A. Conroy, C. Cove, C. Kopec-Korsmo, of wintering Jackdaws, Corvus rnonedula,ai Tam- P. McIsaac, R. Unnasch. We are grateful to D. Morrison Dare. Finland. Omis Fenn. 50:29-45. for sharing his expertise on radio telemetry and for pro- WAL~BE~G,G. E., AND J. R. KING. 1980. The thermo- viding valuablesuggestions on an earlier draft of the manu- regulatory significance of the winter roost-sites se- script. We also thank E. Waltz for providing comments lected by robins in easternWashington. Wilson Bull. on our manuscript. This is New Jersey Agricultural Ex- 92:33-39. periment Station Publication No. D-08230-38-82, sup- WARD, P., AND A. ZAHAVI. 1973. The importance of ported by State Funds and U.S.D.A. Regional Research certain assemblagesof birds as ‘information centres’ Funds. for food finding. Ibis 115:517-534. WEATHERHEAD,P. J. 1983. Two principle strategiesin LITERATURE CITED avian communal roosts. Am. Nat. 121:237-243.

BRENNER,F. J. 1965. Metabolism and survival time of Departmentof Entomologyand EconomicZoology, Cook groupedstarlings at varioustemperatures. Wilson Bull. College, New Jersey Agricultural Experiment Station, 771388-395. Rutgers University,New Brunswick, New Jersey 08903. BROWN,F. J. 1946. A Cheshire starling roost. J. Anim. Received 18 October 1982. Final acceptance9 May 1983. Ecol. 15:75-81.