Soaring and Non-Soaring Bats of the Family Pteropodidae (Flying Foxes, Pteropus Spp.): Wing Morphology and Flight Performance

Because thermal convective air currents are absent at night, bats cannot usually make use of thermal flight. Soaring is used by many birds to save energy when patrolling in search of food and travelling across country. In slope soaring, which is used by many petrels and larids (Pennycuick, 1960), the bird glides in a region of rising air caused by the upward deflection of wind over a slope, such as a hill, forest edge or wave, etc. In thermal soaring, the bird climbs in circles on a rising column of air (‘thermal’).

The height gained allows the bird to glide to some destination or to a different thermal to regain the height. This type of soaring is used by many large species, such as storks, eagles, cranes, vultures and buzzards, during migration and foraging. Gliding flight is the main component of soaring. It is inexpensive compared with active flight because the flight muscles produce mostly isometric force to keep the wings horizontal, opposing the aerodynamic forces and, hence, supporting body weight.

The largest pterosaurs were probably good gliders and soarers; their wings appear to have been direct natural counterparts of hang gliders with a high aspect ratio (wing span divided by mean wing chord) and cylindrical camber (Pennycuick, 1975). Although the wing skeleton of bats is very delicate, several biomechanical arrangements make their wings rigid and suitable as a gliding surface (Norberg, 1972). Some microchiropteran bats glide for very short periods during flap-gliding, but their muscle fibres are not adapted for the longer periods of isometric contraction (as in small birds) that are needed to keep the wings in an outstretched position (see Norberg, 1990). Pipistrellus pipistrellus (Vespertilionidae, 5 g) has been observed to include very short gliding phases (<1 s) during flight (Thomas et al., 1990), Lavia frons (Megadermatidae, 26 g) uses gliding for some seconds (U. M. Lindhe Norberg, personal observation) and Rousettus aegyptiacus (Pteropodidae, 140 g) has been trained to glide for up to 1 min in a wind tunnel (Pennycuick, 1971b). In contrast, some larger megachiropteran bats (Pteropus spp.) on some volcanic oceanic islands frequently use thermal or slope soaring during foraging flights to save energy. Soaring is an energy-efficient adaptation that allows diurnal flight in extremely hot climates where the bats are at risk of hyperthemia.

By being active during the day, P. samoensis in American Samoa Polynesia has doubled its potential foraging time and expanded into a niche where there is no competition with other bat species (Thomson et al., 1998). This species uses soaring during foraging flights. These behavioural adaptations may prove of significant benefit to the animals’ survival by minimizing energetic flight costs, especially in times of stress when fruit resources are scarce, such as after hurricanes or during droughts.

In this investigation, we compare soaring with non-soaring Pteropus species on American Samoa and the Comoro Islands to determine whether there are any differences in gross wing morphology that can be directly related to adaptations for soaring and gliding, and thus for the minimization of the energy costs of flight. We used morphological data for theoretical calculations of the gliding and soaring performances in both the soaring and non-soaring species. We also compared the circling performance of the species, which is important not only for soaring in thermals but also for manoeuvrability.

Using aerodynamic theory for animals, which all have moveable wings, always introduces errors. Wind-tunnel and respiration measurements also involve various errors, as do calculations based on field observations. Aerodynamic theory has been developed over several decades to fit animal flight, and results from wind-tunnel and field studies on flying birds, bats and insects have been used to improve the equations for theoretical calculations. As long as the same theory is used for all species in an investigation, the results are very valuable for a comparison, although the real values in the flying bats may be somewhat different from those obtained theoretically.

We compared two species pairs from two different islands, of which one species in a pair uses soaring flight and the other does not. In American Samoa, Pteropus s. samoensis often slope-soars, while P. t. tonganus uses mostly continuous flapping flight. A non-soaring subspecies of P. samoensis on Fiji, P. s. nawaiensis, is included in this investigation for comparison. If particular adaptations for soaring behaviour were important, one would expect that the two different subspecies, where one uses soaring and the other does not, should have different flight morphology, even though they are closely related.

The second species pair investigated roosts on the Comoro Islands and includes the soaring P. livingstonii and the non-soaring P. seychellensis comorensis. We also included a single species from Madagascar, P. rufus, which uses soaring flight (E. Long, personal communication), and one from Palau, P. mariannus pelewensis, which is reported to use gliding or flap-gliding, but not usually soaring flight (G. Wiles, personal communication). The silhouettes of the species are shown in Fig. 1.

Since the bats in this investigation are endangered and some are difficult to catch, we were only able to measure the wing span and wing area of a few specimens. Body masses (M) were obtained from these species and from published values (e.g. Brooke, 1997). Wing dimensions and body mass are given in Tables 1–3. The mass of adult males was used to standardise comparisons among species. When the masses of the specimens from which wing data were taken (in some cases from subadults, e.g. P. samoensis and P. tonganus) differed greatly from the mean values for adult male specimens, the wing measurements obtained were scaled isometrically to fit the masses used. Although wing span and wing area in megachiropteran bats both increase slightly faster with body mass (slopes 0.35 and 0.72, respectively; Norberg and Rayner, 1987) than predicted from isometry (0.33 and 0.67, respectively), the intraspecific scaling is usually close to isometry (U. M. Lindhe Norberg, personal observation). Body mass and wing measurements for P. s. nawaiensis refer to a single adult female, since data on adult male mass for this species are lacking.

Wing area (S, the area of the two wings, the tail membrane and the body between the wings) and wing span (b) were measured from photographs of living specimens with outstretched wings. Great care was taken to keep the leading edges of the arm wings and parts of the hand wings along a straight line. The following variables were calculated (defined in Norberg and Rayner, 1987; Norberg, 1990): aspect ratio (AR=b2/S), wing loading (Mg/S, where M is body mass, g is the acceleration due to gravity, and Mg is therefore body weight), wingtip length ratio (T_l, length of hand wing/length of arm wing), wingtip area ratio (T_s, handwing area/armwing area) and wingtip shape index [I=T_s/(T_l−T_s)] (Tables 1, 2). Wing measurements independent of body mass were also calculated. Regression equations for megachiropteran bats (Norberg and Rayner, 1987) were used to calculate these mass-independent indices for wing span (b/M^0.35), wing loading (RWL; (Mg/S)/gM^0.33) and aspect ratio (AR/M^0.11).

Wing measurements were also scaled up or down to fit a body mass of 0.5 kg for all species (the mean for all species in this investigation) to determine how soaring performance is related to wing form irrespective of the body size of the bats. In this way, we can trace wing adaptations to flight behaviour. Differences in wing measurements were tested using Student’s t-test for small samples. The morphological data were used to calculate theoretical glide polars and circling envelopes for the soaring and non-soaring species.

The curve relating the minimum sinking speed V_s,min to the radius of turn r is the circling envelope. The minimum sinking speed and the corresponding forward speed V are obtained from the glide polar for the particular species. Successive values of the bank angle ϕ are then selected, and the corresponding sinking speed V_st and radius of turn r are calculated from equations 18 and 19.

P. samoensis roosts singly or in pairs and forages primarily in the same area in which it roosts (Cox, 1983; Banack, 1996). It is active both during the day and at night and regularly utilizes soaring in thermals during the day to move over long distances searching for food (Cox, 1983; Banack, 1996; Pierson et al., 1992a, 1996). It is routinely seen soaring with outstretched wings that are held rigidly or with little movement for minutes at a time and has been documented in continuous soaring flight for in excess of 3 min, during which time the animals travelled several kilometres and out of sight (A. P. Brooke, unpublished data).

The Fijian sub-species P. s. nawaiensis is rare, hunted by indigenous Fijians and extremely reclusive. There is no good documentation of diurnal flight, but this may be a reflection of the rarity of the bats in Fiji, rather than behavioural differences. We placed this species in the non-soaring category, but we do not pay much attention to it in the comparisons.

P. tonganus roosts in large colonies and is typically active during the night. It commutes long distances between roosting sites and foraging areas using flapping flight. It has never been observed in soaring flight in many hundreds of hours of field work by numerous observers (Pierson et al., 1992b; Morrell and Craig, 1995). However, Ruth Urtzurrum (personal observation) has recently seen P. tonganus soaring very occasionally. We suspect that, as the population increases, more P. tonganus will appear during the day.

When the flapping flights of P. tonganus and P. samoensis are compared, P. tonganus appears to fly faster and usually flies in a straight line. Both have a wingbeat frequency of approximately 3.0 Hz. P. samoensis appears to be more manoeuvrable and to move in and out of the vegetation, but this may be because it can be observed in the daytime, whereas this behaviour of P. tonganus is more difficult to observe at night.

P. livingstonii tends to roost on steep-sided valleys at altitudes between 450 and 1000 m on the Comoro Islands and is dependent on the natural forest found at higher altitudes in these islands, whereas P. seychellensis generally roosts at lower altitudes and is less dependent on this habitat (Action Comores, 1992; Trewhella et al., 1998). Observations at roosts show that P. livingstonii is more diurnal in its activity than P. seychellensis, with bats leaving the roosts in the early afternoon, several hours before P. seychellensis tend to leave their roosts. Individual P. livingstonii were seen taking off from a roost situated approximately 50 m below the ridge of a valley, to gain height by soaring in circles in the valley, before gliding off to foraging sites. Similar soaring behaviour was seen at several roost sites. Observations of the two bat species in flight show that P. livingstonii flies with a significantly slower wingbeat frequency in normal flapping flight than P. seychellensis (2.2 Hz compared with 3.2 Hz). Previous observations (Waters et al., 1988) recorded similar differences in the flight behaviour, noting that, although P. seychellensis was seen gliding, ‘the majority of flight time was spent flapping, with the wings appearing to move loosely, and bending considerably at the wrist joint’. In P. livingstonii, the periods spent gliding were interspersed with slow rigid flaps, with the wings bending only very slowly at the wrist on the downstroke.

P. rufus in northeastern Madagascar have been observed to glide in windy conditions and to soar in thermals when these are available (E. Long, personal communication). P. mariannus are typically seen flying in the early to mid-morning hours and again before dusk, and only rarely in the middle of the day. Much of their flying occurs close to tree-top level, so they have little opportunity to use thermals, but they have been observed to use gliding or flap-gliding (G. Wiles, personal communication).

Birds and bats with different foraging behaviour, and therefore flight behaviour, have different wing sizes and shapes (Norberg and Rayner, 1987). Flying animals with long, high-aspect-ratio wings and hence low wing loadings, such as emballonurids, hipposiderids and noctilionids among bats and swifts, swallows and many seabirds among birds, have rather slow and inexpensive flight and mainly use continuous foraging flights in open spaces. Foraging within vegetation requires shorter wings, but a large wing area and low wing loading for slow and manoeuvrable flight. The low aspect ratio of the wings of this category of bats and birds makes their flight more expensive than in the high-aspect-ratio group, so these species are often perchers or spend much of their time on the ground or in vegetation during foraging (clinging, climbing, hanging; Norberg and Rayner, 1987; Norberg, 1994, 1995). Bats and birds with high-aspect-ratio wings and high wing loadings (such as molossid bats and ducks, mergansers and alcids among birds) have faster and still more expensive flight. Birds that soar over land, e.g. storks, eagles, cranes, vultures and buzzards, typically have short broad wings of low aspect ratio and with broad wingtips, whereas birds that soar over the sea, such as gulls and albatrosses, usually have long narrow wings with pointed tips, a high aspect ratio and a low wing loading (Pennycuick, 1971a, 1983; Norberg, 1995). Birds using cross-country soaring as their main means of locomotion during migration usually have higher wing loadings than other soaring birds; this favours fast flight. The short wing is adapted to the requirements of take-off from the ground by flapping in the absence of wind. Short wings are also more manoeuvrable, since they permit the animals to roll and yaw at higher angular velocities for a given wingtip speed (Thollesson and Norberg, 1991; Tucker, 1993), and also more practicable for flight among vegetation. Sea birds have fewer requirements for manoeuvrable flight than land birds, and the most efficient means of reducing drag for these birds is to have long wings with pointed tips.

There are therefore two wing forms adapted to soaring flight, the ‘long soaring wing’ and the ‘short soaring wing’; the long soaring wing is adapted to soaring in open habitats with little need for manoeuvrability, and the short soaring wing is adapted to soaring plus manoeuvrable flight and to take-off from the ground or from vegetation (Pennycuick, 1983).

Bats do not have the ability to split the wingtips into winglets to increase lift and reduce induced drag, as do birds with the short soaring wing. But the same drag-reducing effect can be achieved by elongation of the wings and by pointed wingtips (as in the long-winged soaring birds). Pointed wingtips also increase agility (roll acceleration and hence manoevrability) at fast speeds, whereas broad wingtips increase agility at slow speeds (Thollesson and Norberg, 1991). Broad wingtips also result in a larger wing area and hence reduce wing loading.

Pteropodid bats fly part of the time close to vegetation in the roosting and foraging areas, and fly in open spaces primarily during travel between these areas. Therefore, these bats cannot have wings that are too long, and no bats have wings as long as those we find in the long-winged soaring birds. Instead, their wings are broader and their wing loadings lower to permit slow and manoeuvrable flight. But low wing loadings also are obtained by maintaining a low body mass.

Gliding is the main component in soaring flight, so similar adaptations are required for soaring as for gliding flight. We would predict that the gliding/soaring species have a lower sinking speed during gliding, a higher glide ratio and a smaller radius of turn with low sinking speed than their non-soaring coexisting species, all being favoured by a low wing loading. In this way, they can exploit thermals even when these are weak. Therefore, we also predict that the soarers should have a lower body mass, longer wings and broader wingtips, and thereby larger wing areas, lower wing loadings, lower wingtip area and length ratios and a larger wingtip shape index than the non-soaring species. However, the ability to maintain isometric tension in the flight muscles needed to keep the wings outstretched during gliding (and soaring) is dependent on muscle size and hence on body size, so gliding and soaring bats cannot be too small (the size limit in birds is approximately 200 g). Furthermore, because rounded wingtips also increase agility (higher roll acceleration) in slow-flying bats, this is not necessarily an adaptation for gliding/soaring flight. In summary, a low wing loading is the main adaptation to increase soaring performance.

Body mass and wing data for the species from American Samoa are given in Table 1. The soaring species Pteropus samoensis is approximately 22 % smaller (lower body mass) than the non-soaring P. tonganus and has a broader wing and a larger wing area in relation to body size (Table 1). The size-scaled wing spans are almost the same in the two species. P. samoensis has an approximately 8 % lower aspect ratio, approximately 23 % lower wing loading and approximately 16 % lower relative wing loading than P. tonganus (Table 1). Although P. tonganus has a narrower wing than P. samoensis, its wingtip shape index is approximately 14 % larger than in P. samoensis. The wingtip area index is the same in the two species, but P. samoensis has an approximately 8 % larger wingtip length ratio than P. tonganus.

If P. tonganus is scaled down to have similar body mass to P. samoensis, it turns out to have approximately 19 % higher wing loading than P. samoensis. Aspect ratio and relative wing loading are unaffected, since these measurements are independent of body mass. In summary, the soaring species is smaller and has lower a wing loading and lower relative wing loading than the non-soarer, as predicted for gliding/soaring behaviour.

The single female of the non-soaring subspecies P. s. nawaiensis on Fiji is approximately 17 % smaller than females of P. samoensis (mean body mass 0.42 kg), has a slightly shorter wing span and smaller wing area in relation to its size, and hence a slightly higher wing loading (compare the data for the P. samoensis-sized P. s. nawaiensis). Although this female is smaller than adult males of P. samoensis, it has a similar wing loading and aspect ratio; however, the relative wing loading is slightly higher and its wing tip is more rounded than that of P. samoensis (Table 1).

On the Comoro Islands, we find the reverse with respect to body size of the soaring and non-soaring species (Table 2): the soaring P. livingstonii is approximately 27 % larger than the non-soaring P. seychellensis. But, in spite of its larger size, it has a slightly lower wing loading (2 %) and 10 % lower relative wing loading (P<0.1) than P. seychellensis. Its wing span is slightly longer in relation to body size (P<0.1) and its aspect ratio is 20 % higher (P<0.05) than those of P. seychellensis (Table 2). Both wingtip length ratio and wingtip area ratio are lower in P. livingstonii than in P. seychellensis. The wingtip length and area ratios are smaller in P. livingstonii (P<0.001), but the wingtip shape index is similar to that of P. seychellensis, indicating slightly more rounded tips in the former species. This means that the handwing makes up a smaller part of the total wing area in the former species, thereby minimizing the area reduction. In summary, the longer wing span and larger wing area compensate for the greater body mass in the soaring species, thereby decreasing its wing loading and relative wing loading.

Table 3 shows measurements for P. mariannus and P. rufus. The gliding/flap-gliding P. mariannus is the smallest of all the species investigated. Its wing span index is the second longest of the species (only P. livingstonii has a larger span index) and its relative wing loading is low, as in the soaring P. samoensis (Table 4). Aspect ratio is high, as in the non-soaring P. tonganus, and the wingtip is relatively pointed (high wingtip area and length ratios and low wingtip shape index). The soaring bat P. rufus is the largest of the species in this study.

Table 4 summarizes wing-shape dimensions normalized to body size (relative measurements) with the bats grouped both according to their flight behaviour and as a species-pair comparison for American Samoa and the Comoro Islands. Note that P. mariannus is placed as a flap-glider/soarer and P. s. nawaiensis as a non-soarer. Comparing soaring bats with non-soaring bats (upper part of Table 4), there are too few data to show any significant differences. Only the wing-span indices do not overlap; the soarers (including P. mariannus) have a slightly larger wing-span index (1.38) than the non-soarers (1.31; P<0.1). The non-soaring P. tonganus and P. s. nawaiensis both have larger relative wing loadings, aspect ratio index and wingtip shape index than any soaring species, but the non-soaring P. seychellensis has values more similar to those for the soarers. Wingtip area ratio is slightly smaller in the soarers (0.58) than in the non-soarers (0.66), but at very low significance level (P<0.1). The wingtip shape index shows lower value for the glider/soarers (0.81) than for the non-soarers (1.12), but the difference is not significant (Tables 1–4).

The bottom part of Table 4 gives a pairwise comparison of the soarer versus non-soarer on American Samoa and the Comoro Islands. The soaring species on each island has longer or slightly longer wings and lower absolute and relative values of wing loading despite the opposite differences in body mass between the members of each pair compared with the non-soaring partner. More species pairs with similar trends would be required to obtain significant differences. There are no particular trends with regard to wingtip length and area ratios or wingtip shape index.

In a species-pair comparison of soaring and non-soaring bats on American Samoa and the Comoro Islands, the soarers have lower wing loadings and lower relative wing loadings than the non-soarers, despite opposite size differences among the pairs. Wing span is similar or slightly longer in the soarers. When comparing the soaring bats from different islands as a group with the non-soaring species, there are tendencies towards longer wings and lower wing loadings in relation to body mass in the soarers than in the non-soarers. This agrees with our hypotheses on morphological adaptations for improved soaring ability and hence minimization of flight energy costs. However, there are few significant differences, which may be partly due to the small sample size.

A theoretical model was used to determine the effects of body size and wing shape on the soaring performance in the various species. In accordance with differences in wing length and wing loading in the soaring versus non-soaring species, the gliding and circling performances are slightly better in the soaring than in the non-soaring species.

Fig. 4 shows the glide polars and Fig. 5 the circling envelopes for P. samoensis and P. tonganus in American Samoa. Theoretically, the soaring P. samoensis has a slightly lower minimum sinking speed V_s,min (0.64 m s⁻¹ at a gliding speed V of 8.5 m s⁻¹) than has the non-soaring P. tonganus (0.66 m s⁻¹ at V=9.7 m s⁻¹) and a 23 % smaller turning radius (12.8 m versus 16.7 m at a 35 ° bank angle) (Table 1). Furthermore, the ‘best glide speed’ V_bg is 18 % lower in P. samoensis than in P. tonganus (approximately 9.8 and 11m s⁻¹, respectively). The glide ratio at V_s,min is larger in P. tonganus than in P. samoensis (14.7 and 13.2, respectively), contrary to the predictions. When scaled down to the size of P. samoensis (M=0.406 kg), the minimum sinking speed, best glide speed and turning radius in P. tonganus become 0.63 m s⁻¹ (at V=9.2 m s⁻¹), approximately 10.5 m s⁻¹ and 15.0 m, respectively. P. s. nawaiensis (grouped here as a non-soarer), scaled to the size of P. samoensis, has similar morphological properties and calculated soaring performance to P. samoensis (Table 1).

Fig. 6 shows the glide polars and Fig. 7 the circling envelopes for P. livingstonii and P. seychellensis on the Comoro Islands. The soaring P. livingstonii has a minimum sinking speed that is 11 % lower than in the non-soaring P. s. comorensis (0.62 and 0.70 m s⁻¹ at V_g=8.2 and 8.3 m s⁻¹, respectively) but only a slightly smaller turning radius (11.9 and 12.2 m, respectively, at a bank angle of 35 °) (Table 2; Fig. 6). But, when the latter is scaled up to the size of P. livingstonii, its sinking speed and turning radius become still larger than in P. seychellensis (18 % and 16 %, respectively). The glide ratio at V_s,min remains larger in P. livingstonii than in P. seychellensis (13.2 and 11.8, respectively), as predicted.

The partly soaring, or gliding, P. mariannus has, because of its small size, the lowest V_s,min (0.57 m s⁻¹ at V=8.0 m s⁻¹), V_bg (9.0 m s⁻¹) and turning radius r at a 35 ° bank angle (11.4 m) of all species (Table 3). Its glide ratio at V_s,min is 14.0. When scaled up to the size of the soaring P. samoensis (M=0.406 kg), the corresponding measurements are V_s,min=0.59 m s⁻² (at V=8.4 m s⁻¹), V_bg=9.5 m s⁻¹ and r=12.6 m, which all are slightly lower than the corresponding values for P. samoensis (0.64 m s⁻¹, 9.8 m s⁻¹ and 12.8 m, respectively).

The soaring P. rufus is the largest of the species (M=0.765 kg), being more than twice as heavy as P. mariannus. It has a rather low wing loading for its size (Tables 3–5), which favours the soaring performance; however, because of its large size, its minimum sinking speed (0.70 m s⁻¹ at V=9.0 m s⁻¹) and turning radius (14.4 m at 35 ° bank angle) are still high compared with those of the other soaring and non-soaring species (Table 3). The glide ratio at V_s,min is 12.8. Thus, there are no differences in the glide ratios between the gliding/soaring and non-soaring species. The mean value of the glide ratio for the gliding/soaring species is 13.3, and it is 13.5 for the non-soarers.

Since the bats in this investigation vary in body size, a comparison of their adaptation to soaring performance can be difficult to interpret, but this can be done when the wing shape measurements are scaled to fit the same body mass. Table 5 shows the wing span and wing loading of the species scaled to a body mass of 0.5 kg (the mean mass of the species). These data were used to calculate the glide polars (sinking speed against gliding speed) and circling envelopes (sinking speed against radius of turn) in Figs 8 and 9. The glide polars (Fig. 8) differ between soarers and non-soarers, but overlap. The gliding/soaring species have, however, significantly lower minimum sinking speeds (V_s,min) than the non-soarers (P<0.05). There is less overlap between the soarers and non-soarers for their circling envelopes (Fig. 9), although the turning radii at particular bank angles as well as the limiting minimum radii are quite similar for the two groups (Table 5).

Theoretically, and as a result of the lower wing loadings, the gliding/soaring species have, in relation to their body sizes, lower or slightly lower minimum sinking speeds, best glide speeds and turning radii than the non-soarers. The glide polars for the two flight categories overlap, but at low gliding speeds (airspeeds) the sinking speeds are lower for the glider/soarers than for the non-soarers, and at higher gliding speeds the sinking speeds for the gliding/soaring species become higher. These relationships hold even when P. s. nawaiensis, about whose flight behaviour we know little, is excluded. The circling envelopes show larger displacement, although the turning radii are similar at similar bank angles. The gliding/soaring species should be able to make tighter turns and at lower sinking speeds than the non-soarers, which is in accordance with our predictions. There is slight overlap only at higher bank angles, which are disadvantageous for the bats to use. These displacements are distinct in spite of the small differences in wing morphology between the gliding/soaring and non-soaring species.

We predicted that the gliding/soaring species should have a lower body mass, longer wings and lower wing loadings than the non-soaring species to increase their soaring performance and thus minimize energetic flight costs. We also predicted that the glider/soarers should have a lower sinking speed during gliding, a higher glide ratio and a smaller radius of turn with a smaller sinking speed than their non-soaring co-existing species, all being favoured by a low wing loading.

Richmond et al. (1998) made a comparative analysis of the wing morphology and flight behaviour of P. tonganus and P. samoensis, and found that these species are morphologically and ecologically similar, despite differences in flight behaviour. They calculated wing area from the two wings and added 10 % for body area (according to Norberg and Rayner, 1987) when measurements of body area were lacking. Our masses and wing measurements are, however, quite different from theirs. It is difficult to avoid errors when making wing measurements, and great care has to be taken when stretching the wings to keep them in the right position for wing-span and area calculations. Except that we used larger bats for our calculations (our mean body masses were 7 % larger for P. samoensis and 21 % larger for P. tonganus), differences in making the wing measurements may be the reason for the differences between the two studies. The wing spans in the bats in our study were 16 % longer in P. samoensis and 40 % longer in P. tonganus, and the corresponding measurements for wing area were greater by 32 % and 68 %, respectively. Both wing span and area for the bats in the analysis of Richmond et al. (1998) are too low for the body masses given, and the wing loadings apparently become too high.

We found that there were small, non-significant differences in wing form and flight performance between the gliding/soaring and non-soaring bats investigated, which may indicate adaptations to increase soaring performance. In the species-pair comparison, there are similar trends, but more pairs would be required to obtain significant differences.

In spite of the opposite relationships between body size and aspect ratio in the soaring and non-soaring species pairs on the two islands, their soaring and circling performances show similar trends for the comparison between soarers and non-soarers. The glider/soarers have significantly lower minimum sinking speeds than the non-soarers and slightly lower best glide speeds than their non-soaring partners, but the glide ratios are similar in the two groups. When comparing the soaring and non-soaring pairs on American Samoa and the Comoro Islands, the soaring species also have smaller turning radii than their non-soaring partners.

The glide polars and circling envelopes of the bats lie in similar regions to those for birds and pterosaurs (see, for example, Norberg, 1990, Figs 5.2a, 6.4c). When the bats are scaled to the same size, there are clear differences between the glide polars, and particularly in the circling envelopes, for the glider/soarers and the non-soarers. This is very interesting, especially since the morphological differences in the wing forms are relatively small. This means that even small morphological differences can cause obvious differences in flight performance.

Since most of the differences are small and at low significance levels, we must conclude that diurnality in the soarers is the crucial factor permitting soaring and is probably the ultimate factor for selection towards morphological adaptations for better soaring performance. By flying during the day as well as at night, the bats can increase their foraging time. Soaring allows diurnal flight in extremely hot climates with little energy costs of flight, and these behavioural adaptations may prove of significant benefit to the animals’ survival by minimizing energetic flight costs and thereby maximizing net energy gain. This is particularly important in times of stress when fruit resources are scarce.

This work was financed by the Swedish Natural Research Council (U.M.N.), the Federal Aid to Wildlife Restoration Act administered by the US Fish and Wildlife Service (A.P.B.) and the following sponsors: Bat Conservation International, British Airways Assisting Conservation Nature Roots Group, British Ecological Society, Fauna and Flora International, Jersey Wildlife Preservation Trust, Leverhulme Trust, The Mammal Society, The Marshall Aid Commemoration Commission, The People’s Trust for Endangered Species and The Royal Geographic Society (W.J.T.). We are indebted to Gary Wiles and Emma Long for sending us wing tracings and information about P. mariannus and P. rufus, respectively. We would also like to thank all Action Comores personnel, especially P. F. Reason and K. M. Clark, for providing the data on the Comoro Island bats, and two anonymous referees for valuable comments on the manuscript.