Interspecific Comparisons of Endothermy in Honey-Bees (APIS): Deviations from the Expected Size-Related Patterns

Several recent studies have demonstrated that the ability of endothermic insects to maintain a thoracic flight temperature at a level above the ambient temperature is strongly size-dependent (reviewed by Bartholomew, 1981). The flight temperature observed in an endothermic insect is the result of a balance between the rate of metabolic heat production from the flight muscles and the rate of heat loss. Both processes are partially influenced by body size. In general, the rate of total heat loss decreases with body size, but more slowly than does the rate of total heat production, and so the gradient in flight between thoracic (flight muscle) temperature (T_th) and ambient temperature (T_a) is usually greater for large insects than for small insects.

Previous studies of the relationship between body size and flight temperature have been carried out on related species of dragonflies (May, 1976a), bumble bees (Heinrich & Heinrich, 1983), euglossine bees (May & Casey, 1983), beetles (Bartholomew & Heinrich, 1978), moths (Bartholomew & Heinrich, 1973; Casey, 1976; Bartholomew & Casey, 1978; Casey & Joos, 1983), and several other insects (e.g. May, 1976b). Within each group and, in general, between groups, strong size effects like those just decribed are observed (with some exceptions attributable to morphological or physiological differences among taxa, and to other factors to be discussed). The Asian honey-bees (genus Apis), the workers of which range in mass from 118 mg for A. dorsata, to 44 mg for A. cerana, to 23 mg for A. florea (see Table 1), present an ideal system for examining the relationship between body mass and flight temperature. All three are abundant throughout tropical Asia, are relatively easy to study, and differ in size enough for physiological differences dependent upon size to be readily apparent. Furthermore, all three fall within a size range (body mass <150 mg) in which the strongest body size effects are observed (Bartholomew & Heinrich, 1973, 1978). Most important, a great deal is already known about the basic biology of Apis from numerous studies of the European bee, A. mellifera, including several on the energetics and thermoregulation of individual workers (Esch, 1960, 1964, 1976; Bastian & Esch, 1970; Heinrich, 1979a,b, 1980a,b). These studies provide both a body of techniques and a solid baseline for comparing and interpreting the differences exhibited by the Asian species.

We have measured flight temperature in the Asian honey-bee species over the range of T_a from about 15 to 30°C. A comparison of the data with previous measurements of A. mellifera reveals that honey-bees as a group deviate from the pattern of temperature differences that would normally be expected for insects with such size differences (see Seeley, 1985). The paper has three main emphases. First, we present the interspecific differences that we have measured in thoracic flight temperature, and compare these with data from A. mellifera. Second, we describe an analysis of the physiological mechanisms that may account for the differences, and for the departure of Apis as a group from the size-dependent relationship seen in other groups of endothermic insects. Third, we briefly discuss the ecological and evolutionary implications of the interspecific patterns.

The three Asian honey-bees differ strikingly not only in body size but also in nesting behaviour (see Table 1). Apis cerana builds nests consisting of 5–7 parallel wax combs suspended from the roof of an enclosed cavity in a tree or in rocks, much like A. mellifera. Apis florea and A. dorsata, however, build nests in the open, each one consisting of a single comb which is protected from predators and from climatic fluctuations by a thick curtain of interlinked bees. Apis florea hides its small nests (each the size of a dinner plate) in dense vegetation, while A. dorsata suspends its enormous (1–1·5 m²) sheets of comb from heavy branches, overhanging rocks, or buildings.

We worked at three locations in Thailand during stays of 10 months and 1 month, for the first and second authors, respectively. We measured the flight temperature of A. cerana and A. florea in Khao Yai National Park (Nakhon Ratchasima province) in December 1984, as bees foraged for nectar and pollen on a ground plant, Mimosa sp. In the northern city of Chiang Mai in January 1985, we measured T_th of A. cerana and A. dorsata as bees arrived at the nests. The A. cerana colony was in an apiary hive and the A. dorsata nest was suspended from a ledge overhanging a third-floor balcony of an office building. We also used this A. dorsata colony for measurements of flight speed. In June-September 1985 at Suwan Farm (National Corn and Sorghum Research Centre; Pak Chong, Nakhon Ratchasima province), we made extensive measurements of T_th with all three species as bees arrived at artificial feeding stations. Apis florea and A. cerana were trained to these stations from colonies that had been captured in nearby orchards; A. dorsata were from a colony that nested on a tall tree on the farm, and whose foragers to our good fortune discovered one of the feeding stations and came daily in large numbers for 4 months. Other studies carried out at Suwan Farm included measurements of flight speed in A. florea and A. cerana, passive cooling rates in all three species, and average worker body mass, as well as dissections of the circulatory system in each species.

The data that we present for A. mellifera came from three sources: first, previous studies of flight temperature and energetics by Heinrich (1979a,b, 1980a,b); second, measurements of flight speed and body mass by Seeley (1986) ; third, measurements of cooling rate and wing-loading that we made with bees from the same colony used by Seeley (1986).

Bees were captured for weighing as they landed at a feeding station, and so the whole-bee mass measured for each species is that of a virtually unloaded forager. This must be borne in mind when interpreting the flight temperatures of bees returning to the nest from foraging trips, since the added weight of a full honeycrop might require a greater power output, and hence greater heat production. Thoracic mass is used for many comparisons. The thoracic muscles generate virtually all of the heat produced in flight, and most convective heat loss is from the thorax (see May, 1976b). We weighed thoraces after cutting off the head, abdomen, wings and legs. In all cases the mass recorded is the average for a sample of 25–50 bees (or thoraces) that had been weighed to the nearest 0·01 g. Honey-bees within a given colony vary little in body mass.

We measured all temperatures with a digital thermocouple thermometer (Bailey Bat-12) using Type K (Copper-Constantan) thermocouple needle probes. Two probes were used: Bailey MT 29/1 and MT 29/1B. The two had the same outer diameter (0·033mm) but differed in response time (time constants: 0·025 and 0·015s, respectively); there were no observable differences in the data obtained with each probe. We checked the calibration of each probe against a mercury thermometer.

For consistency we compared T_th of bees in a similar behavioural state, i.e. arriving at the nest or at a feeding station, or foraging on Mimosa. For each bee we recorded the highest temperature (to the nearest 0·1 °C) seen after the probe was inserted into the thorax until its tip was approximately at the centre. Bees foraging on Mimosa were caught in flight with an insect net and stabbed with the probe through the net within 5 s of capture. When capturing bees at the nest we identified foragers to be caught when still several metres out on their approach. We grasped A. cerana foragers directly from the hive entrance with a gloved hand. With A. dorsata and A. florea we swept bees out of the air with the net stretched taut across the ring, then grasped the bees between a gloved thumb and forefinger; they could usually be stabbed within 5 s, and if not they were discarded without recording their temperature. Our procedure worked well with A. dorsata. However, because of the small size and rapid movements of A. cerana and A. florea, some of the data from these species may have come not from returning foragers but from bees on orientation or defecation flights near the nest. Moreover, A. florea colonies were very sensitive to the movement of the net, and returning foragers were often difficult to discriminate from attackers; hence we collected few data from A. florea at the nest.

Bees arriving at an artificial food source, which provided an anise-scented 2·25 moll⁻¹ sucrose solution, were easily caught with a pair of foam-padded, flat foreceps and stabbed within 2–3 s of landing. The feeder consisted of a small jar that was inverted on a grooved Plexiglas base from which bees sucked the syrup (see von Frisch, 1967). We found that bees landing at a highly congested feeder (<50 bees around the 16 cm circumference of the jar) sometimes had a slightly higher T_th than those at a sparsely attended feeder, so we have only compared bees that experienced the same degree of congestion. We also tried only to take temperatures when the air was still. Thus, the only forced convection experienced by the bees arose from the flight itself, which for all species covered a distance of 25–100 m. The feeders were always in the shade, and for each bee we measured T_a in the shade immediately after recording T_th. Although we had hoped to record T_th over the entire range of T_a that the bees would be likely to experience in Thailand (14–40°C), we were engaged with other projects during the hottest months (April-May), and thus regrettably can report no data for bees flying at T_a > 32°C.

Thoracic temperatures were compared using Student’s Z-test when variances did not differ, and a modified Z-test when variances were found to differ significantly (Sokal & Rohlf, 1981).

We measured the cooling rates of freshly killed (in HCN) bees with a thermocouple probe inserted into the thorax from the rear so that the tip was near the centre; the probe lead was attached to a platform, so the impaled bee was suspended motionless a few centimetres above the table top. Then we heated the thorax with a focused beam from an incandescent microscope lamp. When T_th just exceeded T_a+22·0°C we quickly removed the heat source. The thorax started to cool immediately, and as T_th reached exactly T_a+22·0°C, we began recording it at 4-s intervals while it continued to equilibrate towards T_a. One person dictated temperature readings (to the nearest 01°C) into a tape recorder as the other announced the time intervals from a digital stopwatch. To investigate the insulating qualities of the thoracic hair, we completely shaved the thoraces of some bees with a small scalpel before measuring their cooling rate. In all cases we recorded T_a just prior to mounting the bee on the probe and just after removing it. If T_a had changed by more than 0·3 °C during the measuring period we did not use the data.

We performed these experiments indoors in Thailand. We had hoped to measure the intrinsic cooling rate of each species in still air, for comparison with previous studies of A. mellifera and other insects. However, to control T_a in the laboratory we turned on a small air-conditioner, which introduced the possibility that forced convection contributed to the measured cooling rates. The measuring apparatus was as far as possible from the air-conditioner (3 m), and no air movement could be detected near the probe, but our preliminary measurements of A. mellifera in Thailand showed a slightly greater cooling rate than had previously been measured (Heinrich, 1980b), perhaps because of forced convection. Later in the US we made more measurements of A. mellifera, half in completely still air and half with a very slight but steady (speed unknown) draught blowing across the probe from a small fan in the housing of our light source. The change in the cooling rate produced by this air movement, which was certainly greater than that occurring in the laboratory in Thailand, was measurable, but not so considerable as to raise doubts about forced convection as a major confounding variable in the measurements of the cooling rates of the Thai bees.

Flight speed in insects generally increases with body size, and hence power output (Lighthill, 1978), and so may be useful for inferring the relative rates of heat production in the different-sized honey-bees. With each species we trained a small group of bees, using standard techniques (see Dyer, 1985), to fly to a feeder offering a scented 2·25 mol I⁻¹ solution of sucrose. We labelled the bees individually and then moved the feeder gradually to a measured distance from the hive (200–1000 m). We chose 1–4 of the labelled bees at a time to measure their flight speeds both outward from the nest and homeward from the food. One person was stationed at the hive and the other at the feeder; both held stopwatches. With A. florea, a visual signal from one person indicated to the other that a particular bee had just departed, and the person receiving the signal timed the interval elapsed until the bee was seen at its goal. The same bee was then timed in the same fashion on its return. When several round-trip times had been recorded for several bees, the feeder was moved to a different distance, and the series of measurements was repeated (usually with different bees, since there is often some turnover in the pool of foragers at a feeder). The times at each distance may include some time spent hovering at the feeder and at the hive, but this is factored out by calculating a regression of flight time against distance over different distances (Seeley, 1986). Because flight speed on the homeward trip often differs from that on the outward flight, we have used the speed calculated from the total time in flight over the round trip.

With A. cerana and A. dorsata we used a different timing system. The two observers started their digital stopwatches simultaneously, and then, once stationed at the food and the nest, simply listed the precise arrival and departure times of each of 3–4 bees over the course of an hour or so. When collated, these lists yielded a set of flight times that showed a remarkable consistency for different bees, and for different trips by the same individuals, overa given distance. This consistency, together with our observations of the flight behaviour of all the species, indicated that the bees flew directly along a straight path between the nest and the food. Most measurements were made in still air. Hence, our measurements of ground speed probably closely reflect the air speed of the bees. The only possible major source of error is in the measurement of some of the distances over which the A. dorsata flights were timed. Because the colony we used was on the congested outskirts of a city, the straight-line distance to the food from the nest could not be measured directly for distances over 600 m. Instead we estimated it by measuring as precisely as possible the sides of the right-angled triangle of which the flight distance was the hypotenuse. The maximum error in this estimation, given the precision of the measuring tape and the compass used to sight the bearings, was ±10m (<2%).

Another correlate of power output in flying animals is wing-loading, the force per unit area of wing surface (Lighthill, 1978). To estimate wing area we photocopied the right set of wings of 8–10 bees from each species, enlarging the images 141 %. We cut out these images and weighed them to the nearest 0–001 g. We calculated the area of one set of wings, which was then doubled to obtain the total wing area per bee, by weighing the image of 400 mm² of graph paper that had been photocopied with the same enlargement onto the same paper as the wings. Wing-loading (in N m⁻²) is computed according to the formula M_b-g/A, where M_b is body mass in kg, gris the acceleration due to gravity (9·8ms⁻²) and A is the surface area of the wings in m². Because the values used for body mass and wing area represent the averages of massed samples of known numbers of bees, and not the arithmetic means of bees measured individually, no estimate of variance is possible. However, because honeybee foragers are monomorphic in size (with negligible variance), it is fair to assume that our values apply to any individual in the population.

In A. mellifera, anatomical features of the aorta have been implicated in the restriction of heat loss from the thorax as blood warmed by the flight muscles flows posteriorly into the abdomen (Heinrich, 1979b). To examine whether the Asian species have these features, we dissected the entire aorta in each from its origin at the heart (in the dorsal sinus of the abdomen) through the petiole and into the thorax. The legs and wings were trimmed off a live bee and it was planted in beeswax so that the dorsal side was in view. 1–2 μl of a warm mixture of gelatin (10%) in India ink (Mosse, 1980) was injected by means of a microcapillary pipette into the dorsal sinus of the abdomen, which was opened between the third and fourth abdominal tergites. If the heart was still beating it rapidly took up the solution and pumped it forward into the aorta, which then became visible through the abdominal cuticle. After about 1 min we killed the bee in situ by covering it with several ml of 95 % ethanol. We cooled the bee in a freezer for about 2 min, and then removed it to perform the dissection. The congealed ink mixture usually filled the aorta completely, making its convolutions clearly visible.

Although the range in worker body size among the Asian honey-bees has long been known (see Seeley, Seeley & Akratanakul, 1982), we report the first measurements of the body masses of foragers. Our data show a five-fold range in body mass across four species measured. Table 1 shows these measurements, as well as the thoracic masses of each species and other distinguishing features. A log-log regression of thoracic mass against whole-body mass in the four species reveals a near-perfect correlation (r≈ 1·0). The scaling exponent is 1·09, indicating that all four species have pro-portionately the same flight muscle mass.

The data show that A. florea has the smallest temperature gradient at a given T_a, as one would expect of the smallest bee. However, A. dorsata consistently exhibits a slightly smaller gradient than A. cerana, despite its larger size. Fig. 1 shows plots of T_th measured over a range of T_a for each species as bees arrived either at the nest or at an uncongested feeding station. The first data we collected were those of bees arriving at the nest (Fig. 1A,C,E). These revealed the pattern clearly, but we worried that the differences in the techniques required to capture the returning foragers (A. cerana were grasped directly by a gloved hand, whereas A. dorsata were caught first with a net, a slower procedure) could have allowed A. dorsata to cool more before being measured. Because of this concern, and because of the difficulty encountered in capturing A. florea, we later made extensive measurements at the feeding stations, using identical techniques with all three species, albeit over a smaller range of T_a. The same patterns were apparent, and in fact the temperature gradients measured for A. cerana and A. dorsata were similar to those measured at the nest over the same range of T_a.

Fig. 2 is based on the same data, processed to permit a more explicit comparison. We divided the range of T_a over which T_th was recorded into a series of 1 °C intervals, separated from one another by 1°C (except for the lowest intervals for A. cerana and A. dorsata in Fig. 2A, which we chose because the data in them would not have been included by the lowest interval in the series established by the rest of the distribution). For the data in each interval we computed the mean T_th and T_a that had been recorded. Note that the data for A. florea returning to the nest were so few that we have simply plotted the T_th of individual bees along with the means of A. dorsata and A. cerana. The temperature gradients for each species were approximately 4–5°C for A. florea, 9–12°C for A. dorsata, and 11–16°C for A. cerana. All paired interspecific comparisons but one show statistically significant differences in the mean temperature gradients at a given T_a. The lone exception, the comparison between A. cerana and A. dorsata arriving at the nest at a T_a of 27–28°C, can probably be attributed to a small sample size; when compared arriving at a feeder at the same T_a, the two species differed at the 0·001 level of significance.

Thus, the data reveal a striking pattern of differences in T_th—T_a, the most surprising feature of which is that the largest species, A. dorsata, consistently exhibited a smaller T_th-T_a than A. cerana in a similar behavioural context, although it outweighs A. cerana by a factor of nearly 3. In fact, our measurements showed that A. dorsata also flies with a smaller T_th—T_a than A. mellifera, which Heinrich (1979b) observed to exhibit a gradient of about 15 °C at T_a = 20–25°C. Depending upon the race of A. mellifera being considered, A. dorsata is 30–80% larger in mass.

Observations of bees flying at T_a<19°C illustrate another aspect of this unexpected pattern of difference in T_th among the Asian honey-bees. When we arrived each day around 06.00 h to study the A. dorsata colony in Chiang Mai, T_a was frequently as low as 15 °C, and on these occasions we noticed dozens of bees resting on vegetation and on the walls of the building where the colony was nesting. We called these ‘stranded bees’ because many seemed unable to fly even when vigorously beating their wings upon being prodded by us. Some had pollen, and thus clearly had become stranded after returning from foraging flights. Moreover, even while some bees were stranded others would lift off when prodded, and at all times there was a steady traffic of foragers coming and going from the colony, all clearly able to fly.

Our evidence suggests that the stranded bees had landed because they were unable to maintain a minimum body temperature for flight. First, there were never any stranded bees at T_a< 19·0°C, and on most days none were seen at T_a< 18·0°C. Furthermore, on several occasions while catching bees at the nest, we saw some returning foragers lose altitude suddenly when T_a fell from just above to just below 18·0°C. These bees often missed the nest and landed on the balcony; when prodded into flight they could sometimes gain altitude momentarily, but then drifted slowly downwards. When T_a increased again to exceed 18·0°C this tendency of bees to lose altitude was not seen. We measured T_th of bees that missed the nest or appeared stranded. As shown in Fig. 3, those that were losing altitude or were already stranded and unable to lift off upon being prodded had T_th > 27°C, whereas bees that could lift off virtually all had T_th<27·5°C. We suggest that the bees were forced to land when their rate of heat production failed to compensate for the rate of convective heat loss; some of these bees were able to warm themselves while resting (free from the forced convection experienced in flight) and to attain a T_th27°C. Only four bees caught in flight had T_th > 27°C, further indicating that this temperature marks the threshold above which continuous flight is possible. This is virtually identical with the minimum T_th that A. mellifera requires to sustain level flight (Heinrich, 1979b).

The decrease in T_th with T_a for A. dorsata, together with the inability of some bees to maintain the minimum T_th for flight at T_a≈ 17°C, suggests that A. dorsata’s T_th is highly dependent upon T_a, and essentially unregulated at low T_a. Although the mean temperature excess recorded at this T_a was about 12°C, most of the data are concentrated at ≈T_a+10°C, the same excess measured at higher T_a. The mean is skewed upwards by a few bees with high T_th (well above 30°C). Given the uncertain histories of the bees measured -how long in flight, whether flying in sunshine or shade, or in windy or calm conditions, all factors that affect T_th -we need not conclude that these few bees were thermoregulating at higher T_th in flight. It is more likely that they had landed during their foraging trip in order to warm up (Heinrich, 1979b), just as the stranded bees seemed to have done.

Apis cerana’s T_th is clearly also dependent upon T_a, but the greater T_th—T_a of this species in flight allows it to fly at lower T_a. We observed A. cerana in flight at T_a as low as 14·5 °C, and there were never any stranded bees. The T_th–T_a of A. cerana at lower T_a exceeded 15 °C, which would allow foragers easily to maintain a flight temperature of 27°C (if that is indeed A. cerana’s minimum T_th) even at T_a>14·5°C. The climate did not permit a test of the limits of A. cerana’s endothermic performance at low T_a. This species’ T_th–T_a was smaller at high T_a than at low T_a (Figs 1, 2), which could be attributed either to enhanced T_th-T_a at low temperatures (perhaps by stopping to warm up) or to reduced T_th–T_a at higher T_a. In any event it is clear that A. cerana’s flight performance is not constrained at the ambient temperatures which affect A. dorsata. In fact, A. cerana’s T_tj, while foraging on Mimosa (essentially hovering) at T_a = 17–23 °C indicated an ability to sustain a gradient of 18–20 °C when relatively free of the effects of forced convection (data not shown). If 27°C is A. cerana’s minimum T_th for flight, then these data suggest that this species may be able to fly at T_a>10°C, perhaps by stopping intermittently to warm up.

Apis florea also shows a strong dependence of T_th upon T_a, and this species’ T_th^_T_a of 4–5°C at T_a = 25–30°C would imply that it should be even more affected than A. dorsata by low temperature. Indeed, our observations of A. florea colonies in the morning indicated that foragers do not leave the nest until T_a≈20°C (x̄ = 20·1 ± 1·5°C; N = 4 days). On the other hand, at some colonies we recorded departures of foragers at cooler T_a values, and also observed foragers in the field (on Mimosa) at T_a as low as 19°C. The T_th of these foragers was never below 27°C, and therefore often exceeded T_a by more than 5°C. Again, the lack of forced convection as these bees hovered while collecting from Mimosa could account for the greater T_th –T_a. Alternatively, the bees might land occasionally to warm up at low T_a. Unfortunately our only extensive measurements of T_th in A. florea were performed during a season when T_a never fell below 23 °C, and so we have not yet tested fully the limits of A.florea’s performance at low T_a. However, these various observations imply that the threshold for A. florea’s flight is T_a≈20°C, but that T_th–T_a might have to be elevated slightly (presumably by stopping intermittently) to permit flight at this T_a.

In the data presented so far there are two examples of intraspecific variability in endothermic performance; first, the variance in A. dorsata’s T_th at low T_a, and the variable tendency of the foragers to become stranded; second, the higher T_th–T_a recorded for A. cerana and A. florea when foraging on Mimosa than when caught at the end of a rapid flight. This variability does not obscure the interspecific patterns that we have observed, but we note it now for the sake of completeness. Another source of variability was seen at artificial feeders. When feeders were heavily congested with bees the T_th values recorded were often slightly higher than when the feeders were uncongested. The differences were significant for A. cerana, but not for A. dorsata, perhaps because the sample sizes for the latter were not large enough (there was no time to check for the same phenomenon in A. florea). Table 2 shows this pattern at narrow ranges of T_a where measurements were made on the same day, alternating between species and between congested and uncongested feeders to make the comparison explicitly. Three explanations for the pattern that come to mind are : (1) bees at a congested feeder have to hover upon their arrival at the food before they can find an opening in the crowd, and so may warm up in the absence of forced convection; (2) because we controlled congestion primarily by regulating food quality, a higher level of motivation among bees collecting the rich food in a congested feeder may result in a higher T_th ; (3) bees preparing to enter the fray at a congested feeder may elevate their T, in anticipation of the considerable effort that will be required to get to the food and avoid being pushed away by other bees. Since all three species may try to ‘rob’ from the food stores of other conspecific and heterospecific colonies (personal observations), the congested feeders may have a natural analogue. In any event, the difference in T_th observed in these two conditions will have to be considered, along with the other examples of variability, in coming to a full understanding of the patterns of endothermy in these species.

As a first step towards understanding why the normal relationship between body size and flight temperature is not observed in these species, we measured the rate at which freshly-killed bees of each species cooled passively after being heated to about 22·0°C above T_a. Fig. 4 shows a set of representative cooling curves for individual bees of the three Asian species plus A. mellifera, the latter for comparison with previous measurements (Heinrich, 1980b). The curves all show the exponential decay of T_th–T_a that is typical in such an experiment. If T_th is plotted on a logarithmic scale against time the data fall on a straight line. The slope of the regression line fitted to these data is proportional to the cooling rate (May, 1976a,b; Bartholomew, 1981), which has units of s⁻¹, and is the value λ in the function T_th⁻T_a = (T_th–T_a)e^−λt. This function describes the exponential decrease over time, t, of the temperature gradient T_th⁻T_a from a higher starting gradient T_th–T_a (Peters, 1983). The heated bees equilibrated with T_a within 3–5 min, but the regression was only calculated for the first 100s for each bee (72s for A. florea, because of its faster cooling rate), because the data deviated sharply from linearity as T_th approached T_a. In contrast to some previous studies (see Bartholomew, 1981), the curves obtained for intact (unshaved) bees did not show a transiently abrupt initial drop in T_th–T_a before declining at the steady-state cooling rate. Perhaps this is because we heated bees slightly above the intended starting gradient (22·0°C), and began measuring T_thvs time only once the bees had cooled to this point. On the other hand, we did sometimes observe a transient drop for shaved bees, but only for the first 4-s interval; in such cases we excluded this first interval in calculating the slope of the regression line. Table 3 gives the mean slope calculated from the cooling data of each species, for both intact and shaved bees.

Over a large range of body sizes, cooling rates and other energetic variables can be related to body mass by an equation of the form y = ax^b, where x is mass (in kg, to use S.I. units), and y is the dependent variable in question. The values a and b are derived empirically from the y-intercept and slope of the least-squares regression fitted to the log-transformed data. A strong linear relationship implies that most of the variation in the dependent variable can be attributed to scaling with body size, rather than to adaptive modification of the variable.

Fig. 5 shows a regression of the cooling constants (mean slopes of cooling curves) of the three Asian bees against body mass. The slope of this line is –0·467, which compares well with the value obtained for other species of tropical bees (May, 1976b). Clearly these measurements have shown qualitatively that the normal body size relationships hold for the passive cooling rates of the Asian species’, but the value of these data in describing the bees quantitatively depends in part on how much error crept into our measurements. As noted earlier, we had some concern that our determination of the cooling rates could have been biased by forced convection that arose from air movements in the room where we made the measurements. The magnitude of the effects of forced convection might not be equal for bees of all sizes, but instead might depend upon factors such as insulation and relative surface area, which could scale allometrically with body mass. Our concern was increased when preliminary measurements of A. mellifera indicated a faster cooling rate than Heinrich (1980b) had measured earlier with this species. In explicit comparisons of A. mellifera’s cooling in still air and moving air (speed unknown, but faster than the draughts to which the Asian bees were exposed, which could not be felt), we found that cooling constants of bees in a draught fitted the regression established by the other species slightly better than those calculated from bees exposed to no draught (Fig. 5). Hence, our measurements of the Asian bees might have slightly over estimated the cooling rates in still air. However, both measurements of A. mellifera fall close to the line relating the Asian species, as does the previous measurement of A. mellifera by Heinrich (1980b). Even if there is a slight systematic bias in our measurements, none of the differences is large enough to obscure the strong relationship between cooling rate and body size. There fore the unexpectedly low flight temperature of A. dorsata relative to A. cerana cannot be attributed to a distortion of this normal relationship.

Such a strong relationship in intact bees is persuasive, but the insulating body hair covering a bee would have the greatest effect on heat loss when the air movement is great (Church, 1960), as in flight. Our measurements of T_th in flight were mainly taken from rapidly flying bees, and so it is important to examine more directly the effect of insulation on cooling. We shaved the thoraces of bees and measured cooling rates in the same way as before. As shown in Fig. 6, which plots the difference between the cooling constants measured for shaved and unshaved bees as a function of body size, the body hair has a greater effect on heat loss in larger bees, at least in relatively still air. Church (1960) observed a similar pattern in bumble-bees and moths. Because the effect of insulation would only be magnified in rapidly moving air, this pattern strongly reinforces the conclusion that the convective loss of heat generated in flight is greatly diminished in A. dorsata, as compared with the smaller species, owing to its size. Another point to notice in passing is that the data for A. mellifera in Fig. 6 fall well above the regression defined by the Asian species, indicating a greater importance of insulation. This is to be expected given the temperate origin of the race of A. mellifera that we studied.

To test the possibility that an unusually high rate of power output by A. cerana and A. mellifera compensates for their faster cooling rates and results in a T_th–T_a that is higher than A. dorsata’s, we used our measurements of flight speed to provide a qualitative estimate of relative power output in flight. (We lacked the equipment to measure the metabolic rates of flying bees.) Table 4 shows the flight speeds recorded for the Asian species, as well as data from Seeley (1986) for A. mellifera. These are derived from regressions of average round-trip flight time (t) against round-trip flight distance (d) for bees visiting feeding stations at various distances from the hive.

The reciprocal of the slope is the flight speed. The regression equations for the Asian species are as follows: A. florea, t = 5·5+0·162d (r = 1·0); A. cerana, t = 7·0 + 0·137d (r = 1-0); A. dorsata, t= 18·7+0·139d (r = 0-99). The equation for A. mellifera is t= 17·8 + 0·128d (r = 0-99) (Seeley, 1986). Normally flight speed scales linearly with body mass to approximately the 0·17 power (Lighthill, 1978; Peters, 1983; Calder, 1984), but there is no such relationship in Apis. Both A. cerana and A. mellifera exceed A. dorsata’s speed in flight. Because a higher flight speed in a smaller bee implies a disproportionately high rate of power output -i.e. a greater mass-specific metabolic rate in flight -it follows that an elevated rate of heat production could account for the higher T_th of A. cerana and A. mellifera in flight.

Wing-loading is another functional correlate of power output (Casey, 1976). We calculated it for the three Asian species and for bees from the same A. mellifera colony whose flight speed had been measured (Table 4). Partly because it includes a mass term, wing-loading tends strongly to scale with mass in related animals, whether they are birds, bats or insects (Lighthill, 1978). However, this is evidently not so in Aprs, and once again A. cerana and A. mellifera surpass A. dorsata in a trait related to power output. Previous studies of flight temperature in moths, beetles and other insects have shown that wing-loading is a good predictor of flight temperature, sometimes better than mass itself (Dorsett, 1962; Bartholomew & Heinrich, 1973, 1978; Casey, 1976; Casey & Joos, 1983). The proximate reason for this is that insects identical in mass but differing in total wing area require different wing beat frequencies -and hence different rates of power output -just to stay aloft. Because the cooling rates are similar, this difference in power output results in dissimilar flight temperatures. This relationship between wing-loading and flight temperature is upheld dramatically in the honey-bees (Fig. 7A,B). Thus we have additional evidence that a disproportionately high rate of power output in A. cerana and A. mellifera in comparison with the other species can account for the interspecific pattern of flight temperatures.

In principle it is possible to compute directly the metabolic heat production (HP) of a flying insect in thermal equilibrium, using the equation HP = C(T_th–T_a), where C is the heat conductance obtained from the cooling rate. In practice, various uncertainties confound a rigourous determination of HP with such a simplified expression (Kammer, 1981; May & Casey, 1983). For example, our cooling data were collected in relatively still air, while the temperature gradients probably reflected in part the effects of forced convection during flight. Moreover, air movement may have a greater cooling effect on small insects, though this might be roughly balanced by their relatively less effective insulation (Fig. 6). Despite these and other concerns, we have calculated HP for each of the bees to compare them on a mass-specific basis. Normally, mass-specific rates of metabolism and heat production scale linearly and inversely with mass, but Table 5 shows that the two intermediate sized species, A. cerana and A. mellifera, have similar high rates of heat production, whereas the smallest and largest species, A. florea and A. dorsata, have similar low rates. The calculations for A. mellifera show clearly that we overestimated the actual metabolic rate; the average of several published values, all from direct measurements, was 484Wkg⁻¹ (Seeley, 1986). However, if we can assume that our over-estimation is approximately consistent for each species, then our calculations give at least a qualitative estimate of the relative rates of heat production. They imply not only that A. cerana and A. mellifera are similar in producing heat at a high rate in flight, but also that A. dorsata and A. florea are more similar physiologically than had been immediately apparent.

Further analysis of wing-loading also reveals this dichotomy between the two pairs of species. According to the principle of geometrical similarity, wing area should scale with mass²/³. Because wing-loading is proportional to mass divided by wing area, it should scale with mass¹/³. Except for animals specialized for hovering, this is roughly true over a large range of body sizes, and it is true in particular for bees and flies, which can be described by the same allometric expression (Greenewalt, 1975).

For animals described by the same function, the quotient of wing-loading and the cube root of mass should yield a constant, which in the case of flies and bees comes out to about 250 (Lighthill, 1978). Table 5 shows that the values for A. cerana and A. mellifera are both considerably higher than 250, while those for A. dorsata and A. florea are both nearer to 250.

A priori it seemed possible that the relatively low flight temperature of A. dorsata might be accounted for if it were relatively inefficient at physiologically retarding heat loss from the thorax. In A. mellifera very little heat passes from the thorax to the abdomen as blood warmed by the flight muscles flows posteriorly through the petiole. The most likely reason for this is that the heat from this blood is transferred in a counter-current exchange to blood being pumped anteriorly through the aorta (Heinrich, 1980b). The aorta is highly convoluted in the region that passes through the petiole, and the convolutions probably prolong the passage of cool blood flowing forward, allowing more heat to be transferred to it. This preventsA. mellifera at least from shunting excess heat to the abdomen to lower their T_th, as bumble-bees do at high T_a (Heinrich, 1976). It seemed reasonable to hypothesize that inA. dorsata the convolutions are not as developed as inA. mellifera and A. cerana, and so relatively more heat is allowed to escape from the thorax to the abdomen. However, in dissections of the Asian species we could find no clear evidence to support this hypothesis. In all three, as inA. mellifera, the aorta makes 8–10 alternating hairpin turns as it passes through the petiole. There may be a slight tendency for A. dorsata’s aortal loops to be shifted posteriorly relative to the petiole (perhaps resulting in less contact of warm blood with the convoluted region), but a difference of such a small degree would have to be subjected to a physiological analysis, given all of the other ways in which rates of heat transfer could be affected anatomically.

Moreover, we made some measurements of abdominal temperature (T_ab) as bees landed at feeding stations. Apis cerana’s T_ab was actually slightly higher than A. dorsata’s, indicating a greater heat flux from the thorax, or at least a flux that was no less in magnitude relative to the rate of heat production than/I. dorsata’s,. There may well be some contribution of anatomical differences in the circulatory systems to the differences in T_th among the species, but it seems likely that it will turn out to be small in comparison to the contribution of the differences in power output.

In previous studies of the dependence of T_th on T_a and on body size, different patterns have emerged for large and small endothermic insects; the body mass ‘threshold’ separating these broadly overlapping patterns ranges from 100 to 300 mg, depending upon the taxonomic group. For very large insects, which we have not discussed, T_th tends to be independent of T_a for individual insects of a given species (Bartholomew & Heinrich, 1973, 1978; May, 1976a). Furthermore, T_th is often independent of body size if the species is polymorphic in size (e.g. Bartholomew & Heinrich, 1978; Heinrich & Heinrich, 1983). That is, there seems to be a particular ‘set point’ for the species (or even for different species within a genus), a narrow window of T_th within which the animals regulate their body temperature (Heinrich, 1981). Usually, however, comparisons on a broader taxonomic scale show that larger species or larger families maintain higher set-points (Bartholomew & Heinrich, 1973, 1978).

In smaller insects (<100–300mg), there is usually a strong dependence of T_th on T_a for a given individual (e.g. May, 1976a), just as we have measured in the Asian honey-bees and Heinrich (1979b) measured in A. mellifera. Furthermore, among the insects in this smaller size range, larger individuals (of the same or of different species) have a higher T_th at a given T_a, because of the size-dependent thermal equilibrium of heat production and passive heat loss discussed at the outset. However, exceptions to this latter generalization have been documented, even in previous studies of honey-bees. Heinrich (19796) found two races of A. mellifera – A. m. adansonii and A. m. mellifera to have the same average T_th–T_a, even though the latter is 30% larger in mass than the former. To explain this pattern Heinrich reasoned that A. m. adansonii would have to have a greater rate of heat output sufficient to overcome the greater rate of cooling that would be expected on the basis of its smaller size. Our data show that this explanation probably applies to the interspecific pattern that we have uncovered in Apis. The cooling rate does scale linearly and inversely with body mass, as studies of other insects have shown, but only the flight temperature of the smallest species, A. florea, compares with that of the other honey-bee species as one would expect from its relative size. By contrast, the mass-specific power output of both A. cerana and A. mellifera appears to be considerably higher than one would expect from their size, resulting in a higher T_th than A. dorsata’? despite their smaller size. Thus, although flight temperature in each honey-bee species is strongly dependent upon T_a, as in other small endothermic insects, there has been some departure among the species (or between races within one species) from the usual size-dependent pattern of physiological differences that underlies differences in body temperature.

We can only make preliminary remarks regarding the extent to which actual thermoregulation plays a part in the physiology of the Asian honey-bees. Foragers of A. mellifera in continuous flight regulate T_th only at high T_a, when excess heat has to be dumped (Heinrich, 1979a, 1980a,b). Apis mellifera can fly at T_a as low as 10°C, but only by stopping intermittently to warm up (Heinrich, 19796). Our measurements suggest that the T_th of the Asian species is probably also unregulated at low T_a, and that the lower limit of T_a at which each species can sustain continuous flight depends primarily upon its characteristic T_th–T_a, and upon the minimum T_th at which its flight muscles can generate the needed lift. The elevated T_th–T_a seen in some individuals of all three species at low T_a could result from bees stopping to warm up, as is true of A. mellifera. However, the Asian honey-bees should be studied under more controlled conditions at low T_a. Even if thermoregulation turns out to play little part in determining T_th in continuous flight, there may well be other size dependent differences among the species, such as in the minimum T_th necessary for flight (May, 1976a).

All three Asian species forage at T_a as high as 40°C (personal observations). At this T_a, if the lethal body temperature is 46°C, as is generally true of insects, both A. cerana and A. dorsata, but not A. florea, would either have to stop flying to cool off, as bumble-bees do (Heinrich, 1975), or, more likely, to dump heat in flight, as A. mellifera does (Heinrich, 1979a, 1980a,b). Apis mellifera begins to depress its temperature excess when T_a > 25 °C by extruding a drop of water or nectar from its mouth and losing heat through evaporative cooling. The slight decrease in T_th–T_a seen in A. cerana and A. dorsata at T_a > 24°C as compared to lower T_a could be interpreted as evidence of this cooling process, but obviously more observations of all three species are needed at higher T_a levels and in a context where the flight itself can be monitored continuously.

Now we turn briefly to a consideration of the possible evolutionary basis for the differences in flight temperature among the four honey-bee species. Our mass specific comparisons of worker attributes (Table 5) revealed a hidden pattern: the four species seem to sort out into two pairs which differ more from each other than species within a pair differ. Elsewhere (F. C. Dyer & T. D. Seeley, in preparation) we explore in detail the possible ecological implications of this pattern, but for now we would merely like to point out that this dichotomy in forager physiology, which is yielded through two very different sorts of analysis, corresponds to the dichotomy in nesting behaviour exhibited by these four species (see Table 1). The fast-flying, high-powered bees both live in enclosed cavities, whereas the slow-flying, low-powered bees both live on exposed combs that are protected by curtains of interlinked worker bees. The protective benefits of the curtain (insulating the brood against heat loss at low T_a, shading it from direct sunlight, ventilating to dissipate excess heat, and detecting and responding to potential predators; Seeley et al. 1982; F. C. Dyer & T. D. Seeley, in preparation) probably derive more from the sheer numbers of workers than from their activity levels. Consequently, colonies of the open-nesting bees, A. florea and A. dorsata, have almost five times as many workers relative to the number of cells in the comb they protect than the cavity-nesting species have (Seeley et al. 1982). An excess of workers could be reared in a given number of cells either by shortening the development time from egg to adult or by extending adult longevity. The development time in all four species is similar -about 18-22 days (Kapil, 1959; Sandhu & Singh, 1960; Qayyum & Ahmad, 1967; Seeley, 1985). Hence, one may predict that workers of the open-nesting species live longer than those of the cavity-nesting species. Our observations of A. florea and A. cerana in Thailand (F. C. Dyer & T. D. Seeley, in preparation) support this prediction. If there is a causal link between the physiological tempo of workers and their longevity, as scaling considerations (Pearl, 1928; Calder, 1984) suggest, then one might speculate that the increased longevity required to maintain a permanent excess of workers to protect the nest has constrained A. dorsata and A. florea workers to a low-tempo existence, whereas A. cerana and A. mellifera, for whatever reason, have evolved a high-tempo existence and a high turnover of workers in the colony. Differences in worker tempo could in turn influence both the foraging strategy and reproductive strategy of colonies (Oster & Wilson, 1978). Clearly, a great deal more work is needed to define the implications of these physiological differences with regard to the colony demography and energetics of each species, and to investigate further the ecological differences among the species.

We see two general implications in these results. First, they re-emphasize the importance of taking a critical look at body size as a constraint on physiology when seeking to explain ecological and physiological differences among closely related species. There is no denying that empirical scaling relationships allow robust generalities to be made across a wide range of body sizes (Peters, 1983; Calder, 1984), and these generalities often allow specific predictions or assumptions concerning the correlates of size differences. The Asian Apis do uphold some predictions; e.g. larger species carry larger loads of food (F. C. Dyer & T. D. Seeley, in preparation). However, they also demonstrate that general scaling relationships based on body mass alone may fail to predict qualitative physiological differences even within a closely related group of species. As in some other insect groups (Bartholomew & Heinrich, 1973, 1978; Casey & Joos, 1983), wing-loading may be a more suitable predictor of flight temperature in honey-bees than body mass, but cannot itself be predicted from mass, and has no value in interpreting other massspecific energetic parameters such as heat loss. Perhaps wing-loading is one route by which mass-specific performance of bees in flight has been altered in evolution, or perhaps it has changed as a consequence of selection on other morphological or physiological parameters. In any event, as Calder (1984) has pointed out, deviations from the general scaling trends can often provide unusually clear examples of adaptive differences among species as against differences based on non-adaptive factors. As yet it is uncertain exactly what ecological factors underlie the interspecific patterns we have observed in the energetics of honey-bees, but at least we now have a more complete picture of what the interspecific differences are, based on direct empirical studies instead of on untested extrapolations from other species.

The second implication arises from the possible parallel between the divergence in nesting behaviour and the divergence in forager physiology among the four species. This pattern, if it is real, raises the question of how energetic constraints operating at the colony level might interact with energetic constraints affecting individuals. Already it is clear that worker size itself has been an extremely malleable trait in the evolution of honey-bees. There is a five-fold range in worker body mass but less than a two-fold range in the size of queens and drones from A. florea to A. dorsata (Seeley et al. 1982; unpublished data). However, physiological variables often found to scale with body size have been shaped somewhat independently of size in workers of the different honey-bee species, and we have argued that demographic constraints on the colony related to nesting behaviour may be important factors shaping these variables. In support of the possibility that body size differences do account for some of the interspecific physiological differences, we note that species that have similar nesting behaviour do exhibit the qualitative patterns usually associated with body size differences: A. dorsata flies faster and with a higher T_th than A. florea, and the same pattern holds for A. mellifera and A. cerana. Of course some of the differences between the latter two could be attributed to the climatic differences between the regions to which they are adapted, but it is also possible that colony-level and individual-level constraints acting together have produced divergent size-dependent trends. For example, interspecific differences in worker size may have evolved according to differences in defence strategies (A. flore a’s inconspicuousness vs A. dorsata’s aggressiveness and strength; Seeley et al. 1982) or in climate (temperate A. mellifera vs tropical A. cerana), while physiological variables related to tempo and longevity, and normally scaling with size, may vary only within a range set by the demographic requirements of the colony. Whether or not these are the actual factors underlying the interspecific patterns that we have uncovered, an important question now is not only why the species of Apis have evolved the worker sizes that they have, but also how the size-dependent constraints upon an individual’s interactions with the environment are functionally interlinked with the ecological challenges faced by the colony.

We are deeply indebted to Josh L. Schein for his committed and able technical assistance throughout the study and to Dr Pongthep Akratanakul for invaluable scientific and logistical help. Others who helped with various stages of the work include Ian Cooke, Chirayus Laohawanich and Somsak Junhom. We also thank Dr Porn Roong-jang, Director of Suwan Farm, for providing facilities during the project, and the Thai Royal Forest Department for permission to work in Khao Yai National Park. C. L. Craig, R. A. Levien and two reviewers made helpful suggestions on the manuscript. Supported by NSF grant BNS84-05962 and by a Seesel Anonymous Grant (to FCD).