ABSTRACT

This paper presents a synthesis of communal roosting behavior from several insect orders: Lepidoptera, Hymenoptera, Odonata, Coleoptera, and Diptera. Insect communal roosting can be typified as 1) facultative, 2) diapause, and/or 3) circadian. Site selection typically is influenced by species-specific habitat preferences: substrate, microclimate needs, proximity to water and food resources, and minimal human disturbance. Some species have shown fidelity to particular roosting sites on a daily, nightly, or yearly basis. Flights to the roosts tend to be associated with overcast conditions, dropping temperatures, and precipitation. Nocturnal roost arrival occurs asynchronously an hour or two before sunset, with some species engaging in pre-roosting assemblies near the roost sites. In contrast, morning departures tend to occur synchronously and under lower light levels than existed during roost arrival; however, this may depend on species. Behaviors within the roost usually consist of relocating and/or jostling for positions. Roost demographics (population, sex, age) can vary widely on a daily and seasonal basis. Both mono-specific and multi-specific roosts occur. Functional and adaptive significance of communal roosts varies according to species and roost type. Five hypotheses are reviewed: protection from predation, population regulation, social function, food information exchange, and thermoregulation.
 
 

 INTRODUCTION

Communal roosting occurs when multiple insects of one or more species assemble in close proximity to one another to rest for at least several hours. Roosts may be diurnal/nocturnal, seasonal/permanent, and mono-/multi-specific (Waller and Gilbert 1982). Roosts are typically located away from feeding and oviposition sites (Moore 1953, Larsen 1973, Miller 1989, but see Chemsak and Thorp1962). Roost populations vary from a few individuals to many hundreds of thousands. Insects exhibit two basic resting patterns within the roost: 1) loose aggregations with space between the roosting individuals (e.g., Joseph 1982, Greig and DeVries 1986, DeVries et al. 1987); and 2) dense aggregations with individuals in physical contact with one another (e.g., Evans and Linsley 1960, Thoenes 1994, Alcock 1998).

Communal roosting has been documented in butterflies, moths, dragonflies, bees, and wasps, for over a century. Despite the prevalence of the phenomenon, no detailed synthesis exists in the literature. An obstacle to such a synthesis is the wide variety of terms used for the phenomenon (e.g., nocturnal habits, Penn 1950; dragonfly dormitories, Gambles 1971; gregarious roosting, DeVries et al. 1987; and sleeping aggregations, Alcock 1998). Another obstacle is researchers' tendency to cite examples from only the orders they are studying. Synthesizing roosting information across insect orders could provide useful generalities. This paper will summarize six important aspects of insect communal roosting: types of roosting, site selection, site fidelity, roost arrival, roost departure, roost behaviors, and roost demographics. In addition, five hypotheses of the functional and adaptive significance of communal roosting will be reviewed.

 Types of Roosting

DeVries et al. (1987) proposed and defined three types of communal roosts for Lepidopterans: circadian, diapause, and facultative. These three roost types are applicable to other orders, as evinced by the examples provided below. Broadening the use of these categories provides a useful organizational tool for reviewing the literature and for doing future research. Circadian roosts are aggregations daily formed at the same site for purposes of sleep. This roost type typically involves sedentary and long-lived species with a limited learned home range. Heliconius butterflies are the most well known example of circadian roosting. This type of roosting has also been well documented in dragonflies (Parr and Parr 1974, Miller 1989, Rehfeldt 1993), wasps (Joseph 1982), beetles (Pearson and Anderson 1985), and moths (Greig and DeVries 1986).

Diapause roosts are aggregations formed in response to seasonal rhythms; individuals are in quiescence for at least several weeks and tend to be active for only short periods during the day. Break-up of the roost is influenced by indices of seasonal change: temperature, day-length, and/or humidity. These roosts are typically used year after year. Several Lepidopterans exhibit this type of roosting (e.g., Danaus plexippus and Smyrna karwinskii). Joseph 1982 suggests that Chalybion bengalense also form diapause roosts.

Facultative roosts are aggregations formed during migration; individuals and sites change in a short period of time, usually every 24 hours. Many Lepidopterans exhibit this type of roosting (e.g., Danaus plexippus and Libytheana carineta). Corbet (1984) documented migrating Anisopterans roosting among several trees during a one-night stopover in Uganda. Site Selection

Species vary in their preferred roost substrate: caves (Celaenorrhinus fritzgaertneri), herbaceous vegetation (Anax junius and Coenagrionids), dead shrubs (Idiomelissodes duplocincta and Chalybion tibiale), fine twigs of dead vines (Heliconius spp.), horizontal leaves (Odontochila spp. and Calopteryx haemorrhoidalis), guava trees (Bactrocera dorsalis), or trees (Danaus plexippus and various Anisopterans). However, there is some plasticity to these preferences. For example, Muyshondt and Muyshondt (1974) report diurnal roost sites on lava walls as well as tree trunks for Smryrna karwinskii.

In addition to substrate, several other factors may influence site selection: (1) suitable micro-climate (Linsley 1962, Mallet 1986, Leong 1990, Leong et al. 1991, Frey et al. 1992, Neubauer and Rehfeldt 1995, Stark 1995); (2) proximity to a water source (Parr and Parr 1974, Leong et al. 1991, Neubauer and Rehfeldt 1995); (3) proximity to a food source (Waller and Gilbert 1982, Neubauer and Rehfeldt 1995); (4) safety from predators (Pearson and Anderson 1985, Dennis 1986); and (5) relatively little human disturbance (Frey et al. 1992, Greig 1986, Kellicott 1890, Mallet 1986). While freedom from human disturbance is important for several species, some insect species will roost in areas of human activity. Young and Carolan (1976) documented use of a Heliconius roost by a roadside. In addition, human-made structures have provided useful roost substrate. Dennis (1986) commented that ranch fencing provided protected roost sites beneath its rails for the butterfly Lasiommata megera. Fraser (1944) reported that up to 16 dragonflies (Bradinopyga germinata) hung in rows nightly from the cords of a verandah blind. Jayakar and Mangipudi (1964) reported the frequent use of a chain from a lavatory cistern as a roost site of the wasp Chalybion bengalense. Schoenly and Calabrese (1983) noted a roost of 50-100 Chalybion californicum in a bell. Once species establish a site (circadian roost), they must find their way back to it daily. Three researchers (listed below), using similar types of manipulation, found that different species use different cues.

Alcock (1998) found that male bees regularly clustered together on a few stems, though many other stems were available. When roost stems were experimentally cut and moved 25-50 cm into other shrubs and replaced by stems of similar characteristics, the bees rejected the replacement stems in nine out of twelve trials and actively sought out the original stems. Alcock speculated that an odor cue (applied by sleeping bees) attracted the bees to specific roost sites. Miller (1989) manipulated a popular dragonfly roosting branch in a tree by cutting it and taping it at the same height 1 meter away. He replaced the original branch with another of similar size. Despite the fact that this roost site had been regularly used by up to 70 dragonflies, none roosted on either the original or the replacement branch. Dragonflies only returned to this site after the original branch had been returned. Miller speculated that dragonflies locate roost sites by visual cues.

Jones (1930) removed roost twigs from a Heliconius roost. These twigs were moved 10 feet away and tied to foliage at the same height. Upon return to the roost, several butterflies paused upon the original twigs in their new location. However, they roosted that night in the original roost location (in twigs adjacent to the experimentally removed twigs). In another experiment, Jones introduced foreign twigs to a roost site; the butterflies readily accepted them. Based on these two manipulations, Jones speculated that butterflies return to their roosts by "place memory" and not by scent.

 Site Fidelity

 Many insect species exhibit circadian roost fidelity by daily returning to the same roost site. Alcock (1998) experimentally demonstrated fidelity to a particular shrub by relocating 30 marked bees 130 meters from their roost site; 53% found their way back to their roost site by the following evening. Linsley (1962) reported that parasitic anthophorids and megachilids tended to return to the same stem night after night. Benson and Emmel (1973) studied two roosts of the Nymphaline butterfly Marpesia berania and also found strong evidence of roost fidelity when marked butterflies regularly returned to the same location over a five-month period. Similar findings are also reported by Miller (1989) in a population of libellulids (Potamarcha congener) which regularly used nine roosting sites for up to 70 days; individually marked dragonflies returned to the same site for at least 23 successive nights. Heliconius butterflies also tend to use the same roost site every night (Turner 1971, Benson 1973). Poulton (1931) reported that marked individuals of Heliconius charitonius used the same twig for several nights. However, there is some variability to circadian roost fidelity. Poulton (1931) wrote of "constant shift and interchange" of Heliconius butterflies among three roosts that were within a hundred yards of each other. A particular marked butterfly made eight changes between two of the roosts during 26 days of observation. These observations are consistent with Young and Thomason's (1975) findings that some individuals were faithful to roosts while others were less so. Parr and Parr (1974) noted that Nesciothemis nigeriensis nightly varied roost sites in herbaceous vegetation. In addition to daily fidelity within a season, several species will use the same sites year after year. The most notable example is of monarchs yearly using the same branches for roosting (Frey et al. 1992); they will use the same sites for at least ten years (Leong 1991). The bee Idiomelissodes duplocincta has been reported to use several of the same roost shrubs in successive years (Alcock 1998). Likewise, Joseph and Lahiri (1989) documented the use of the same group of trees for over 5 years by communally roosting dragonflies (Potamarcha congener). The same pine branch was used for at least five years by wasps of the genus Steniolia (Evans and Gillaspy 1964). Greig and DeVries (1986) suggest that roost sites are important, based on the observation of diurnal roost fidelity in consecutive years by the moth Ocalaria spp. DeVries et al. (1987) reported diurnal roost fidelity to a cave for the skipper Celaenorrhinus fritzgaertneri in consecutive years.

 Some research indicates that roost fidelity can be altered. Young and Thomason (1975) documented accidental damaging of perch sites at a Heliconius roost, which resulted in total abandonment of the roost site. They also documented that the presence of several birds at the roost site one evening caused the butterflies to roost elsewhere that evening. Neubauer and Rehfeldt (1995) provided possible evidence that roost use by damselflies was altered when female orb web spiders (Argiope brunichii and Larinioides folium) were experimentally introduced into the roost. Young and Carolan (1976) found that Heliconius charitonius was "quite fluid with regard to the fidelity" at a roost site adjacent to a road; they speculated that fidelity may be influenced by the permanency of the roost site itself. Young (1978) also speculated that roost fidelity may be influenced by phenotypic differences in mobility, age, location of adult food sources and oviposition sites, and weather.

 Roost Arrival

 Most research on roost arrivals has occurred in the vicinity of the roost itself; therefore, little information is available on the routes/stages taken by insects from feeding and oviposition sites to roost sites. There is, however, some evidence of staging (e.g., pre-roosting assemblies in the vicinity of the roost), arrival behaviors, and the influence of weather on arrival times.

Miller (1989) observed the dragonfly Potamarcha congener gathering and perching 10-20 meters from the roost site about 60 minutes before sunset. From here, the dragonflies would enter the roost site as light faded. The most detailed account of roost arrival from feeding and oviposition sites comes from Parr and Parr (1974). They studied the dragonfly Nesciothemis nigeriensis, characterizing its flight to the roost site as fast, direct, and often high. This libellulid species is unusual in that it leaves for its roost early in the day (1500). Around 1800, they leave their pre-roosting assembly and arrive at the roost. Rau (1945) described what might be considered a pre-roosting assembly in the dragonfly Anax junius. About 50 of these dragonflies gathered and flew in a small area above the roost site for about 15 minutes before singly dropping out into the grass to roost in a loose aggregation. Jones (1930), commenting on a similar pre-roosting flight, wrote, "[A]t sunset, the air is filled with butterflies in swirling flight," which slowly dropped, one by one, to the roost.

Typically, bees and wasps begin to gather an hour or two before sunset (Evans and Linsley 1962, Joseph 1982). Alcock (1998) found that male bees (Idiomelissodes duplocincta) typically begin arriving at roost sites around 90 minutes before sunset and "swirl back and forth through the shrub before settling on a perch." This flight around the roost site upon arrival is also typical of Heliconius charitonius, which, upon arrival, slowly explores the vicinity of the roost before settling on a perch (Jones 1930). Latecomers of both these species go directly to the roost and perch. Freeman and Johnston (1978) observed a distinctive roost arrival behavior in the Sphecid wasp (Sceliphron assimile). First, males began to decrease their flight speed and increase their frequency of settling, forming small groups low in the foliage around 1700 EST. Then pre-reproductive females entered the area and settled in small groups on the tops of vegetation. On some nights, males would join these females. At about 1730, older females appeared, making rapid "zigzag flights" followed by a slower flight until they settled.  While arrival at the roost site typically occurs one to two hours before sunset, this pattern can be altered by weather conditions (e.g., cloud cover, rain, temperature, and wind). Aculeate hymenopterans (Linsley 1962), sphecid wasps (Jayakar and Mangipudi 1964), odonates (Parr and Parr 1974, Hansan 1976, Miller 1989), nymphaline butterflies (Benson and Emmel 1973), Satyrids (Dennis 1986), and Heliconiids (Young and Carolan 1976, Mallet 1986) have all been documented as roosting earlier on cloudy days than on clear days.

 Roost Departure

 Although information on roost departure is slight, there is evidence that roost type influences whether departures differ from or are similar to arrivals. For nocturnal circadian roosts, insects tend to depart from roosts synchronously, whereas they tend to arrive asynchronously. Departure from a Heliconius roost usually occurs over a short time period, with individuals leaving singly or in small groups (Jones 1930, Benson and Emmel 1973, Young and Thomason 1975, Young 1978); their arrivals tend to be asynchronous. On numerous occasions, Freeman and Johnston (1978) observed the sudden departure of the sphecid wasp Sceliphron assimile from its communal roost.

At diurnal circadian roosts, both arrivals and departures tend to be synchronous. DeVries et al. (1987) report that departure and arrival times were "highly synchronized" for skipper butterflies. The period of time between the first and last individual to enter the cave averaged 2-3 minutes. Greig and DeVries (1986) reported similar synchrony in a diurnal roost of Ocalaria spp. There is evidence that insects at facultative roosts also exhibit synchrony in both arrivals and departures. Corbet (1984) reported a synchronous departure and arrival of migrating Anisopterans (Hemianax ephippiger, Philonomon luminans, Tramea basilaris).

 At circadian roosts, there is mixed evidence of light intensity influencing departure times. Heliconiids are reported as leaving the roost around sunrise (Jones 1930, Young and Carolan 1976). Miller (1989) observed Anisopterans wing-whirring (to prepare for departure) 40-45 minutes after sunrise. Overcast mornings (Young and Thomason 1975) and rainfall (Young and Carolan 1976) can delay roost departures for hours. Evans and Linsley (1962) observed that solitary bees and wasps begin to stir as soon as the sun hits the roost, though they do not depart until one to two hours later. There is also evidence that bees do not leave roosts until well after sunrise (Joseph 1982, Alcock 1998).  As with arrivals, observations of departures have been primarily within the vicinity of the roosts; hence, few data are available on characteristics of the departing flight. Jones (1930) observed direct flight of Heliconiids away from the roost. Short flights into adjacent areas to bask in the sun have also been observed (e.g., butterflies, Mallet 1986; Anisopterans, Miller 1989).

 Behaviors in the Roost

It is generally believed that once settled in the roost the insects remain motionless during the night/day (Rau and Rau 1916). The time prior to settling, however, is either characterized by no interactions (Young and Borkin 1985, Greig and DeVries 1986), slight repositioning (Muyshondt and Muyshondt 1974), or more typically, as a time of jostling for positions and relocating within the roost.

Alcock (1998) observed much "pushing and kicking" as perched and newly arriving bees jostled for positions on the stems. This activity ceased after sunset as the bees grabbed the substrate tightly with their mandibles. Thoenes (1994) also observed this shifting of positions along branches prior to sunset in megachilid bees. Corbet (1984) reported considerable rustling and jostling among migrating odonates at a facultative communal roost before they finally rested at twilight.

Joseph (1982) observed wasps (Chalybion bengalense) relocating within the roost as individuals either exchanged places or sought new locations by flying. Gambles (1971) observed that the damselfly Lestes virgatus frequently changed perches; relocations were so rapid that it became impossible to keep track of individuals.

Hansan (1976) reported that odonates (Palpopleura lucia lucia and Acisoma panorpoides inflatum) usually shifted from perch to perch, made occasional feeding flights, and sometimes groomed their heads with their legs and shook their bodies at roost sites. Roosting individuals responded to intruding latecomers by elevating all four wings. In 247 of these encounters, roosting lucia were displaced in only 5% of the encounters. Roosting inflatum were displaced in 3% of 174 encounters. Additional roost behaviors have been described for Heliconius butterflies. Mallet (1986) and Young and Carolan (1976) have observed a behavior termed "fanning" in Heliconius erato, H. melpomene, and H. charitonia. Fanning consists of brief hovering approaches of roosting conspecifics. "Clutching" occurs when a butterfly grasps a perch or the wings of a perched butterfly. Recipients of these behaviors often respond by clinging to their perch while flapping their wings vigorously to fend off the intruder. In 597 instances of fanning and clutching, Mallet observed the fanner usurping the recipient's twig in only 10 instances. In 6 of these 10 instances, the usurper did not remain on its "won" twig but flew off to fan other roostmates. Mallet speculated that fanning is required for species recognition on the roost sites. Miller (1989) observed similar behaviors in dragonflies around sunset. Before settling, these individuals would briefly hover above conspecific roostmates in a behavior suggestive of "fanning." The roosting dragonflies responded by fluttering their wings weakly (at < 10 Hz), producing an audible sound. A behavior similar to "clutching" was also observed. As numbers of roosting dragonflies increased, there was much jostling for position and relocating within the roost, during which time some dragonflies perched on the abdomen of a perched dragonfly. The recipient responded by flexing its abdomen until the intruder left. Miller speculated that these behaviors induced settling.

Sexual interaction among sexually mature individuals at a roost site is uncommon (Benson and Emmel 1973, Young and Thomason 1975), but there are exceptions. Mass mating in overwintering monarchs is reported around the time of the roost break-up, when individuals are no longer in a state of reproductive diapause (Tuskes and Brower 1978). In early morning, males of Sympetrum depressiusculum seek out roosting females in order to form tandems (Miller et al. 1984, Rehfeldt 1993). Parr and Parr (1974) recorded an unusual instance of mating in Nesciothemis nigeriensis at a roosting site.

 Roost Demographics

Facultative and diapause roosts typically contain thousands of insects, but for any given roost these numbers can fluctuate widely during a season (e.g., Common 1954). Circadian roosts are generally more modest in size (e.g., 60-70 individuals, DeVries et al. 1987; 30 individuals, Greig 1986), but they also will fluctuate in size during a season. At a circadian roost, three to nine tiger beetles gathered nightly at a communal roost (Pearson and Anderson 1985); similar numbers have been found for damselflies (Gambles 1971). At a circadian roost of the male bee Idiomelissodes duplocincta, numbers fluctuated between a few to hundreds (Alcock 1998).  Some roosts may be limited to mostly one age class (e.g., immature dragonflies, Miller 1989), a structure which can change during a season. For example, Waller and Gilbert (1982) documented a Heliconius butterfly roost consisting of worn and intermediate butterflies early in the season; later in the season, it consisted of multi-age classes (also see Mallet 1986). Neubauer and Rehfeldt (1995) observed three damselfly roosts, two dominated by adults and the other by immatures. They speculated that age-class separation was due to the proximity of the roost to oviposition sites and territories used by adults. Hence, immatures roosted communally away from oviposition sites.

  There is much evidence of males and females roosting in a 1:1 sex ratio (sphecid wasps, Jayakar and Magipudi 1965, Freeman et al. 1978, Jones 1982; aculeate hymenoptera, Linsley 1962; damselflies, Gambles 1971; nymphaline butterflies, Benson and Emmel 1973; monarchs, Brower et al. 1977; acraeine butterflies, Carpenter 1931; hairstreak butterflies, Courtis 1988; and skipper butterflies, DeVries et al. 1987). A few reports exist of one sex predominating at a roost. For example, males were predominant at several Heliconius roosts (Young and Carolan 1976, Mallet 1986) while females were predominant at several nymphalline roosts (Muyshondt and Muyshondt 1974). Miller (1989) observed nine dragonfly roosts, some of which showed a biased sex ratio for males and others for females. Some roosts will contain only one sex class (Callan 1985, Thoenes 1994, Alcock 1998).

 Roosts may be composed of a single species (mono-specific) (see Benson and Emmel 1973, Callan 1985, Thoenes 1994, and Alcock 1998) or multiple species (multi-specific). In many instances of multi-specific roosting, species are in the same order. For example, Evans and Linsley (1960) and Linsley (1962) documented 21 species of wasps and 15 species of bees at a roost in Arizona; both authors observed stratification among the species within the roosts. Young and Borkin (1985) documented three species of Hamadryas butterflies roosting together. Similarly, Carpenter (1931) observed four species of acraeine butterflies roosting together in east Africa. Coutsis (1988) noted two species of hairstreak butterflies at a roost in northern Greece. Larsen (1973) counted 14 species of butterflies roosting communally along a small rivulet over a distance of 30 meters. Though Heliconius butterflies typically form mono-specific roosts, when in multi-specific roosts the other species are three times as likely to be co-mimics (Mallet and Gilbert 1995). Additionally, Gambles (1971) observed two species of damselflies using the same roost. Occasional multi-specific roosts were documented among tiger beetles (Odontochila spp.) (Pearson and Anderson 1985). Multi-specific roosts have also been observed in migrating dragonflies (Hemianax ephippiger, Philonomon luminans, Tramea basilaris) (Corbet 1984).

 Functional and Adaptive Significance

 The functional and adaptive significance of insect communal roosting is not fully understood; however, several hypotheses have been proposed. Currently, no single hypothesis is able to explain communal roosting for all insect species. More than one advantage may be responsible for the evolution of this behavior.

 Protection from Predators  This hypothesis states that communal roosting provides protection from predators. There are four mechanisms: enhanced aposematic advertisement, enhanced vigilance, dilution effect, and/or cryptic advantages.

 Communal roosting has been reported for unpalatable groups of butterflies such as heliconiines, acraeines, and danaines (Benson and Emmel 1973). This massive group warning display can facilitate rapid learning in naive birds whose feeding areas may include roosting sites (Poulton 1931, Tuskes and Brower 1978). Another type of chemical defense may be enhanced by roosting arrangement; roosting tiger beetles position themselves on leaves so their posteriors face outward, which may enhance the effect of their chemical defense (benzaldehyde and benzoyl cyanide) against nocturnal predators (Pearson and Anderson 1985). Other insect species avoid predators by a dilution effect: predation risk is lowered for those individuals in a group. During two summers, Alcock (1998) documented 50 attacks on the male bee Idiomelissodes duplocincta by assassin bugs (Apiomerus flaviventres). Bees were not vigilant with this predator, and in many instances the adjacent struggles of a victim had no effect on the nearby roosting males. Alcock suggested that communally roosting bees may obtain anti-predator benefits through the dilution effect. Monarch butterflies are thought to benefit from the dilution effect as well. Monarchs that overwinter in large roosts are better protected against bird predators than those roosting in smaller colonies (Calvert et al. 1979), because concentration of bird predation at the periphery of the roost decreases relative exposure.

Communal roosting in dragonflies may reduce nocturnal predation via increased vigilance. Miller (1989) noted that dragonflies present at a roost were wary; if one flew off, others also dispersed. The presence of other dragonflies might then act as an early warning system. Joseph and Lahiri (1989) observed early morning bird predation at a dragonfly roost; when a dragonfly was captured, its roostmates produced a "synchronous audible wing whirring sound" as they flew off. The authors speculated that this behavior served an anti-predator function. Freeman and Johnston (1978) grasped a roosting wasp (Sceliphion assimile) with forceps; its struggles aroused its roostmates which then flew off. Roosting monarchs, when disturbed, fall from their roost perches in near synchrony ("cascading"). This group behavior may be an adaptation to "confuse and disorient" predators, thereby increasing the possibility of escape (Calvert 1994). Tuskes and Brower (1978) observed several chickadees pecking at monarchs within a roost which caused dozens of butterflies to cascade; in all instances, the chickadee fled as the monarchs cascaded from their perches.

 For a few species, communal roosting may enhance the ability to go undetected. In El Salvador, communal roosts of the butterfly Smyrna karwinskii give the appearance of dried moss or lichen clusters (Muyshondt and Muyshondt 1974). Clench (1974) found that Colias and Phoebis formed communal roosts near yellowing leaves and/or other yellow butterflies. One "communal roost" located by Clench turned out to be mere yellow leaves. Roosting bees typically occupy the ends of dead branches, giving the plant the appearance of having a flower or seed head. Linsley (1962) commented on this "strange sight," and Alcock (1998; Fig. 1) included a photograph in which roosting bees give the plant the appearance of having seedheads. Freeman and Johnston (1978) speculated that communal roosts of this type may not "satisfy the search image of a flying predator."

Population Regulation  Jayakar and Mangipudi (1965) and Joseph (1982) believed communal roosting would make individuals more vulnerable to detection by predators. They instead favored the explanation that roosts serve the epideictic purpose of allowing roostmates to assess population density and to disperse accordingly (see Wynne-Edwards 1962). This explanation would not apply to multi-specific roosts and has largely fallen out of favor among biologists (Freeman and Johnston 1978, Allen and Young 1982). Social Function  The social function hypothesis proposes that communal roosting serves a social function by synchronizing reproduction and dispersal. This is especially evident in diapause roosting. For example, a social stimulation function is suggested for monarchs due to the synchronous mating response in mid-February; after this event, the colony soon disperses (Brower et al. 1977). DeVries et al. (1987) suggested that the diurnal communal roosting of the skipper butterfly Celaenorrhinus fritzgaertneri allows "synchronization and concentration" of reproduction and dispersal when environmental conditions are favorable for larval development. Similarly, Muyshondt and Muyshondt (1974) speculated that communal roosting during aestivation is beneficial for Smyrna karwinskii because the sexes remain together, optimizing the chance of early copulations when weather conditions are favorable for larval development. This function is also suggested for dragonflies (Miller 1989).  Communal roosting of the dragonfly Sympetrum depressiusculum appears to have the function of keeping males and females in close proximity for early morning tandem formation (Miller et al. 1984, Rehfeldt 1993). While this type of social function appears to be uncommon among roosting odonates, there is evidence of it in other orders. For example, North American fly-catcher wasps (Steniolia obliqua) gather in dense communal roosts, in which females form the center of the roosting ball and males cling to the outside. Mating is frequently observed in and about the roost, suggesting a mating function (Andrewes, in Joseph 1982). Likewise, Evans and Gillapsy (1964) observed mating in digger wasps (Steniolia spp.) and speculated that the main function of roosts was social. Similarly, Stark (1995) observed large numbers of female oriental fruit flies (Bactrocera dorsalis) roosting on the undersides of leaves in the upper canopy. These roosting females were later joined by males; mating ensued before nightfall. Chemsak and Thorp (1962) observed males of the solitary bee Melissodes robustior roosting in cosmos flowers. This roost site was preferred during the time when females began gathering pollen from the cosmos. The authors speculated that this change in roosting behavior may be due to the flower being used as a mating site.  Heliconius butterflies exhibit another type of social function, recruiting newly emerged conspecifics to the roost (Gilbert and Waller 1982). The recruitment may enhance breeding success of individual butterflies (Young 1978).

 Food Information Center  The food information center hypothesis (Ward and Zahavi 1973), developed to explain communal roosting in birds, states that information pertaining to the locality and abundance of food resources is communicated at communal roosts. Successful foragers may be detected by differences in behavior or appearance (e.g., pollen load variation was noticeable on roosting individuals; Young 1978) and then followed the next morning to resources. Waller and Glibert (1982) found that roost membership of Heliconius charitonia was closely tied to resource use. They speculated that information was shared at roosts to help individuals locate the rather inconspicuous nectar sources. Following behavior, however, was only observed in the evening and not in the morning.

 Mallet (1986) made three arguments against this hypothesis. First, Heliconius melpomene, which feeds on a very inconspicuous food source (Psiguria plant), rarely communally roosts. Second, for communally roosting species, similar home ranges could be due to the limited number of resources within flying distance to the roost, rather than due to information shared at the roost. Third, no evidence of morning following behavior has been observed in Heliconiids. It is important to consider that Mallet's third argument is based on a prediction appropriate for birds and may not be the mechanism by which insects share food information in the roost.

 Thermoregulation  This hypothesis proposes that communal roosting enhances thermoregulatory benefits via the physical structure of the site and/or the presence of other insects. Corbet (1963) suggested that "local pockets of high temperatures" are responsible for the communal roosting of some species. Some bees which communally roost in protected areas (e.g., those that can provide insulation such as crevices) may obtain an enhanced thermoregulatory benefit by raising the ambient temperature of the group (Rayment in Freeman and Johnston 1978). This hypothesis is rarely cited in the literature.

 Conclusions

By synthesizing most of the available literature on insect communal roosting, several of the following generalities have emerged. Insect communal roosting can be typified as 1) facultative, 2) diapause, and/or 3) circadian. Site selection typically is influenced by species-specific habitat preferences: substrate, microclimate needs, proximity to water and food resources, and minimal human disturbance. Some species have shown fidelity to particular roosting sites on a daily, nightly, or yearly basis. Flights to the roosts tend to be associated with overcast conditions, dropping temperatures, and precipitation. Typically, nocturnal roost arrival occurs asynchronously an hour or two before sunset, with some species engaging in pre-roosting assemblies near the roost sites. In contrast, morning departures tend to occur synchronously and under lower light levels than existed during roost arrivals; however, this may depend on species. Behaviors within the roost usually consist of relocating and/or jostling for positions. Roost demographics (population, sex, age) can vary widely on a daily and seasonal basis. Both mono-specific and multi-specific roosts occur. The functional and adaptive significance of insect communal roosting varies according to species and roost type and is not fully understood; however, a number of hypotheses have been proposed: (1) protection from predators, (2) population regulation, (3) social function, (4) food information exchange, and (5) thermoregulation. Currently, no single hypothesis is able to explain communal roosting for all insect species, and it is likely that more than one advantage may be responsible for the evolution of this behavior.

LITERATURE CITED

Alcock, J. 1998. Sleeping aggregations of the bee Idiomelissodes duplocincta (Cockerell) (Hymenoptera: Anthophorini) and their possible function. Jour. Kansas Ent. Soc. 71: 74-84.

Askew, R. R. 1982. Roosting and resting site selection by coenagrionid damselflies.Adv. Odonatol. 1: 1-8.

Benson, W. W., and T. C. Emmel. 1973. Demography of gregariously roosting populations of the nymphaline butterfly Marpesia berania in Costa Rica. Ecology 54: 326-335.

Brower, L. P., W. H. Calvert, L. E. Hedrick, and J. Christian. 1997. Biological observations on an overwintering colony of monarch butterflies (Danaus plexippus,Danaidae) in Mexico. Jour. Lepid. Soc. 31: 233-242.

Callan, E. McC. 1985. A sleeping aggregation of Chalybion tibiale (Hym., Sphecidae)in South Africa. Ent. Mon. Mag. 121: 25-26.

Calvert, W. H., L. E. Hedrick, and L. P. Brower. 1979. Mortality of the monarch butterfly (Danaus plexippus) at five overwintering sites in Mexico. Science 204:847-851.

Calvert, W. H. 1994. Behavioral response of monarch butterflies (Nympalidae) to disturbances in their habitat-a group startle response? Jour. Lepid. Soc. 48:157-165.

Carpenter, G. D. H. 1931. Acraeine butterflies congregating in a small area for the night's rest. Proc. Roy. Entomol. Soc. 6:71.

Chemsak, J. A., and R. W. Thorp. 1962. Note on the sleeping habits of males of Melissodes robustior Cockerell. Pan-Pac. Ent. 38: 53-55.

Clench, H. K. 1970. Communal roosting in Colias and Phoebis (Pieridae). Jour. Lepid. Soc. 24: 117-120.

Common, I. F. B. 1954. A study of the ecology of the adult bogong moth, Agrotis infusa (Boisd.) (Lepidoptera: Noctuidae), with special reference to its behaviour during migration and aestivation. Aust. J. Zool. 2: 223-262.

Corbet, P. S. 1984. Orientation and reproductive condition of migrating dragonflies(Anisoptera). Odonatologica. 13: 81-88.

Corbet, P. S. 1962. Biology of dragonflies. Witherby, London.

Coutsis, J. G. 1988. Hairstreak roosting habits. Entomol. Rec. Jour. Var. 100: 54.

Dennis, R. L. H. 1986. Selection of roost sites by Lasiommata megera (L.) (Lepidoptera: Satyridae) on fencing at Brereton Heath Country Park, Cheshire, U.K. Nota Lepid. 9: 39-46.

DeVries, P. J., J. Schull, and N. Greig. 1987. Synchronous nocturnal activity and gregarious roosting in the neotropical skipper butterfly Celaenorrhinus fritzgaertneri(Lepidoptera: Hespeeriidae). Jour. Linn. Soc. 89: 89-103.

Evans, H. E., and E. G. Linsley, Jr. 1960. Notes on a sleeping aggregation of solitary bees and wasps. Bull. Southern Cal. Acad. Sci. 59: 30-37.

Evans, H. E., and J. E. Gillaspy. 1964. Observations on the ethology of digger wasps of the genus Steniolia (Hymenoptera: Sphecidae: Bembicini). Amer. Midl. Nat. 72:257-280. Fraser, F. C. 1944. Diurnal and nocturnal resting habits of Bradinopyga geminata (Rambur) (Odonata: Libellulidae). Ent. Mon. Mag. 80: 76-77.

Freeman, B. E., and B. Johnston. 1978. Gregarious roosting in the sphecid wasp Sceliphron assimile. Ann. Entomol. Soc. Am. 71:435-441.

Frey, D., K. L. H. Leong, D. Fredericks, and S. Raskowitz. 1992. Clustering patterns of monarch butterflies (Lepidoptera: Danaidae) at two California central coast overwintering sites. Ann. Entomol. Soc. Am. 85: 148-153.

Gambles, R. M. 1971. Dragonfly dormitories. Niger. Fld. 36: 166-170.

Greig, N., and P. J. DeVries. 1986. Observations on the Diurnal gregarious roosting of Ocalaria Sp. (Noctuidae) in Costa Rica. Jour. Lepid. Soc. 40: 124-126.

Hassan, A. T. 1976. Studies on the roosting behaviour of Palpopleura lucia lucia (Drury) and Acisoma panorpoides inflatum Selys (Anispotera: Libellulidae). Odonatologica. 5: 323-329.

Jayakar, S. D., and R. S. Mangipudi. 1965. Dormitories of Chalybion bengalense Dahlb. (Hymenoptera: Sphecidae) J. Bombay Nat. Hist. Soc. 61: 708-711. Jones, F. M. 1930. The sleeping heliconias of Florida. Nat. Hist. 30: 635- 644.

Joseph, K. J. 1982. The roosting behaviour of Chalybion bengalense (Dahlbom) (Sphecidae: Hymenoptera). Entomon. India. 7: 403-405.

Joseph, K. J., and A.R. Lahiri. 1989. The diel patterns of communal roosting behaviour in Potamarcha congener (Rambur) (Anisoptera: Libellulidae). Adv. Odonatol. 4: 45-52.

Larsen, T. B. 1973. Communal roosting among butterflies in the Lebanon. Ent. Scand. 4: 299-301.

Leong, K. L. H. 1990. Microenvironmental factors associated with the winter habitat of the monarch butterfly (Lepidoptera: Danaidae) in central California. Ann. Entomol. Soc. Am. 83: 906-910.

Leong, K .L. H., D. Frey, G. Brenner, S. Baker, and D. Fox. 1991. Use of multivariate analyses to characterize the monarch butterfly (Lepidoptera: Danaidae) winter habitat. Ann. Entomol. Soc. Am. 84: 236-267.

Linsley, E. G. 1962. Sleeping aggregations of aculeate Hymenoptera-II. Ann. Entomol. Soc. Am. 55: 148-164.

Mallet, J. 1986. Gregarious roosting and home range in Heliconius butterflies. Nat. Geo. Res. 2: 198-215.

Mallet, J., and L. E. Gilbert. 1995. Why are there so many mimicry rings? Correlations between habitat, behaviour and mimicry in Heliconius butterflies. Biol. Journ. of the Linnean Soc. 55: 159-180.

Miller, A. K., P. L. Miller, and M. T. Siva-Jothy. 1984. Pre-copulatory guarding and other aspects of reproductive behaviour in Sympetrum depressiusculum Selys at rice fields in southern France (Anisoptera: Libellulidae). Odonatologica 13: 407-414.

Miller, P. L. 1989. Communal roosting in Potamarcha congener (Rambur) and its possible functions (Anisoptera: Libellulidae). Odonatologica 18: 179-194.

Moore, N. W. 1953. Population density in adult dragonflies (Odonata: Anisoptera). Journ. Anim. Ecol. 22: 344-359.

Moss, A. M. 1933. The gregarious sleeping habits of certain ithomiine and heliconine butterflies in Brazil. Proc. Roy. Ent. Soc. 7: 66-67.

Muyshondt, A., and A. Muyshondt, Jr. 1974. Gregarious seasonal roosting of Smyrna karwinskii adults in El Salvador (Nymphalidae). Jour. Lepid. Soc. 28: 224-229.

Neubauer, K., and G. E. Rehfeldt. 1995. Roosting site selection in the damselfly species Calopteryx haemorrhoidalis (Odonata: Calopterygidae). Entomol. Gener. 19: 291-302.

Parr, M. J., and M. Parr. 1974. Studies on the behaviour and ecology of Nesciothemis nigeriensis Gambles (Anisoptera: Libellulidae). Odonatologica 3: 21-47.

Pearson D. L., and J. J. Anderson. 1985. Perching heights and nocturnal communal roosts of some tiger beetles (Coleoptera: Cicindelidae) in Southeastern Peru. Biotropica 17:126-129.

Poulton, E. B. 1931. The gregarious sleeping habits of charitonia L. Proc. Roy. Ent. Soc. 6: 4-10.

Rau, P., and N. Rau. 1916. The sleep of insects; an ecological study. Ann. Entomol. Soc. Am. 9: 227-274.

Rau, P. 1945. The night habits of the dragonfly, Anax junius Dru. Jour. Comp. Psychol. 38: 285-286.

Rehfeldt, G. E. 1993. Heterospecific tandem formation in Sympetrum depressiusculum (Selys) (Anisoptera: Libellulidae). Odonatologica. 22: 77- 82.

Schoenly, K., and D. M. Calabrese. 1983. An aggregation of Chalybion californicum(Hymenoptera: Sphecidae) in a bell. Ent. News. 94: 145-146.

Stark, J. D. 1995. Nocturnal behavior of oriental fruit flies and melon flies (Diptera: Tephritidae) and associated parasitoids in a commercial papaya growing region on Kaua'i, Hawai'i. Proc. Hawaii. Entomol. Soc. 32: 149-151.

Thoenes, S. C. 1994. Description of a sleeping aggregation of male Chalicodoma chilopsis (Cockerell) (Hymenoptera: Maegachilidae). Pan-Pac. Ent. 70: 328- 330.

Turner, J. R. G. 1975. Communal roosting in relation to warning colour in two heliconiine butterflies (Nymphalidae). Jour. Lepid. Soc. 29: 221-226.

Tuskes, P. M., and L. P. Brower. 1978. Overwintering ecology of the monarch butterfly,Danaus plexippus L., in California. Ecol. Entomol. 3: 141-153.

Waller, D. A., and L. E. Gilbert. 1982. Roost recruitment and resource utilization:observations on a Heliconius charitonia L. roost in Mexico (Nymphalidae). Jour. Lepid. Soc. 36: 178-184.

Wynne-Edwards, V. C. 1962. Animal dispersion in relation to social behavior. Edinburgh and London: Oliver and Boyd.

Young, A. M., and J. H. Thomason. 1975. Notes on communal roosting of Heliconius charitonius (Nymphalidae) in Costa Rica. Jour. Lepid. Soc. 29: 243-255.

Young, A. M., and M. E. Carolan. 1976. Daily instability of communal roosting in the Neotropical butterfly Heliconius charitonius (Lepidoptera: Nymphalidae: Heliconiinae). Jour. Kansas Ent. Soc. 49: 346-359.

Young, A. M. 1978. A communal roost of the butterfly Heliconius charitonius L. in Costa Rican premontane tropical wet forest (Lepidoptera: Nymphalidae). Ent. News 89: 235-243.

Young, A. M., and S. S. Borkin. 1985. Natural history notes for some Hamadryas butterflies (Nymphalidae: Nymphalinae; Aegeronini) in northwestern Costa Rica during the tropical dry season. Jour. Lepid. Soc. 39: 229-235.