Introduction

Autonomic temperature regulation, defined as the involuntary thermoeffector response to heat and cold including sweating, tachypnea, shivering and non-shivering thermogenesis, and adjustments in circulatory convection to the body surface (Mercer 2001), incurs costs to a homeotherm in terms of energy for enhanced heat production or water for evaporative cooling. Complementary to autonomic thermoregulation is behavioural thermoregulation, generally considered the “cheap option”, which involves either avoidance of ambient extremes (microclimate utilisation) or postural adjustment altering energy exchange with the environment. Inhibiting behavioural responses to a thermal challenge increases the energetic cost of homeothermy (McEwen and Heath 1973) and may lead to altered body temperature patterns (Fuller et al. 1999). Optimisation of the response to thermal stress should implement the least-costly behavioural responses before autonomic shivering or sweating/panting (Satinoff 1996). While there has been considerable enquiry into operant responses to thermal stimuli in the laboratory (see Cabanac 1996; Satinoff 1996), our knowledge of the extent to which behaviour is used as a thermoregulatory effector in free-ranging homeotherms, and especially large mammals, is scant. In the opinion of some, behavioural thermoregulation, beyond shade seeking and bathing, does not play a role in large animals (Davenport 1985).

Laboratory-based operant conditioning studies have revealed that the control of behavioural responses to thermal stimuli is similar to that of autonomic responses. There is a threshold body (hypothalamic or skin) temperature for initiation of a conditioned response and an increase in intensity as body temperature deviates further. There is also evidence that these operant responses can be modified if conflicting demands are made on a behavioural common path, such as when food or water are withheld from an animal placed in the heat, and the animal is given the choice to obtain food or cool air (Rautenberg et al. 1980; Torres-Contreras and Bozinovic 1997). Free-ranging animals are faced with many simultaneous environmental stimuli, including ambient temperature, wind speed, solar radiation, long wave radiation, and ambient water vapour pressure. There is some evidence that free-ranging mammals behaviourally adjust aspects of their heat exchange with the environment to reduce autonomic thermoregulatory costs (Hainsworth 1995; Stelzner and Hausfater 1986), often avoiding thermal extremes by selecting favorable microclimates. Avoidance of climatic extremes is not an option for the black wildebeest (Connochaetes gnou), a large (130 kg) mammal that is endemic to treeless plains in the temperate grasslands (highveld) and arid shrublands (Karoo) of South Africa (Skinner and Smithers 1990). These habitats occur at moderate altitude (1,000–2,000 m above sea level) and have a climate characterised by predominantly cloudless days and nights. There is no respite from radiant heat gain or loss, or wind. The body temperature of black wildebeest shows a nychthemeral variation of less than 1°C, despite ambient globe temperature fluctuating by up to 40°C in 24 h, and direct solar radiation reaching 1,000 W m−2 (Jessen et al. 1994). Any behavioural adaptations that reduce the cost of thermoregulation should be an advantage to the black wildebeest, and contribute to their homeothermic abilities.

We have investigated behavioural thermoregulatory strategies in black wildebeest by relating changes in activity patterns (Altmann 1974) to coincident climatic variables, on five occasions over the course of one year. We address several specific questions. Firstly, are gross activity patterns affected by physical factors associated with changes in season? Secondly, are active (thermogenic) behaviours (feeding and running) less common on warmer days or, alternatively, are these activities shifted to cooler nocturnal hours on warmer days? Thirdly, is the proportion of the inactive period spent lying and standing correlated with ambient thermal conditions, since posture will affect heat transfer?

Materials and methods

Data collection

Behavioural observations were made on black wildebeest at Benfontein game farm (111 km2) near Kimberley, South Africa (28°50′S, 24°50′E), approximately 1,200 m above sea level. The climate is semi-arid with dry winters (June to August) and wet summers (December to February). Mean annual rainfall at Kimberley between 1876 and 1994 was 422±132 mm (SD), most of which occurred between October and March. The vegetation of the area is described as Kalahari thornveld invaded by Karoo (Acocks 1975), and the game farm itself consists of open thorn tree savannah surrounding a large treeless pan or calcareous tufa (approximately 6 km2). Herds of black wildebeest tended to concentrate on and around the treeless pan, which was covered in shrubs and short perennial grasses.

A herd was observed from a stationary vehicle parked between 50 and 700 m (typically about 200 m) away, using 10×50 binoculars (Nikon). Data were collected by the scan sample method (Altmann 1974) which involved recording the behaviour of each animal in a herd every 15 min. During data collection, behaviour was classified into seven categories: lying, standing, feeding, walking, running, grooming/social interaction, and drinking. No wildebeest were observed feeding while lying and so these categories were all mutually exclusive. The latter four categories accounted for only a small proportion of the behaviour in each survey, and so in most analyses they were combined into an “other” category. Because we used the instantaneous scan method, we could not distinguish the “search” component of foraging time and thus we have measured feeding/cropping time as opposed to total foraging time (sensu Owen-Smith 1992). No attempt was made to identify when animals were ruminating. Observations were recorded onto micro-cassette tape and the data transcribed later.

We conducted surveys over three consecutive 24-h periods around the time of full moon in May, August, and November 1994, and January 1995. Overcast conditions on the first 3 days of observations in November precluded night observations, so three further days of data were collected after the weather had cleared. In November and January data collection was restricted to two of the three nights because of overcast conditions. Three days of diurnal observations were also made in December 1994.

The herd under observation usually consisted of a territorial bull with adult females and a few calves (Skinner and Smithers 1990). Because only a single bull and few calves were present during observations, we only analysed the behaviour of the females. Weather data were collected near to where observations were made. Sensors measuring dry bulb temperature, relative humidity, solar radiation, wind speed, and black globe temperature were connected to a data logger (MC Systems, South Africa) which sampled every minute and recorded 15-min averages of each variable to coincide with behavioural observations. Weather data for one day in August were lost for technical reasons. Rainfall data were obtained from the Kimberley airport weather station, approximately 5 km west of Benfontein.

Statistical analyses of seasonal changes

For analysis of nychthemeral changes in behavioural activity, the day was divided into morning (observations from 0600 to 0845 hours; all times are local time, GMT + 2 h), midday (0900–1545 hours), afternoon (1600–1845 hours), and night (1900–0545 hours). This pooling facilitated interpretation of the data, and indicated the changes in behaviour across the 24-h day that we considered relevant.

In May and August sunrise was around 0700 hours and sunset at 1800 hours. In November, December, and January sunrise was at 0600 hours and sunset at 1830 and 1900 hours, respectively. Morning and Afternoon included most of the major post-dawn and post-dusk feeding bouts in all surveys. In preliminary statistical analyses, we pooled surveys on the basis of similarity of their climatic conditions and behavioural activity homogeneity into “seasons”. Namely, data for May and August surveys were called the “Cool” season and the November and December surveys were pooled into the “Warm” season. In January the feeding behaviour proved to be quite different from the “Warm” season and so that period, though climatic conditions were similar to the Warm season, was analysed separately. Data from the overcast days in November, when no night observations were possible, and for December, when no night observations were made, were included only in diurnal analyses.

Data were analysed using a hierarchical-loglinear model for multidimensional contingency tables (v. 5.01, SPSS Inc., IL, USA). Just as ANOVA can be thought of as multifactorial t-tests, hi-log linear analysis is a multifactorial χ2-squared test for categorical data. The initial process involved fitting a saturated model (including main effects and interactions between Time Period, Season, and Behaviour) which predicted perfectly the observed values for each cell. Removal of individual interactions and/or main effects reduces the goodness of fit and can be used to reveal whether each main effect or interaction term was significant. In our model, all main effects and interactions were significant since removing any of the interactions resulted in the observed values being significantly different from the expected values based on the reduced model.

In the second stage significant main effects and interaction terms (determined in the first stage) were fitted into a Logit model, with behaviour used as the dependent variable. The most easily interpreted model proved, again, to be a saturated Logit model (Likelihood ratio and Pearson’s chi-square was <0.0001, DF=0, P=1.0). Main effects or interactions were then explored by removing them from the model and obtaining expected values without the main effect or interaction of interest. With terms removed, expected values no longer matched the observed values perfectly and departures from expected were analysed using residuals. Standardised residuals may be negative or positive depending on the magnitude of the observed frequency relative to the expected frequency (Norusis 1992). A SR greater than 3.27 for a behaviour category corresponded to a probability of less than 0.001 that the observed frequency was the same as the expected frequency, within the main effect or interaction. Similarly 2.56<SR<3.27 corresponded to a P<0.01, 1.96<SR<2.56 to P<0.05, and 1.64<SR<1.96 corresponded to a P<0.10.

For example, in analysing the effect of season (shown in Fig. 2), interactions involving season were removed, and time periods were collapsed so that the expected value for each behaviour depended only on the overall proportion of each behaviour across all seasons. As such, the expected value for each behaviour was the same in each season. Residuals were then calculated and assigned a significance. The full interaction between season and time period (Fig. 4) was explored by obtaining expected values from a model with the season × time period interaction removed. As such, the expected values for each cell then depended only on the overall proportion of each behaviour in each time period and in each season. Residuals then revealed where significant departures from expected occurred.

Thermal environment and activity

The proportion of the diel or diurnal period spent engaged in each of the four behavioural categories was arcsine transformed and regressed on black globe temperature as an index of heat load. The same was done for the proportion of the day spent running. We also tested the hypothesis that the choice of lying or standing during inactive periods would be influenced by ambient thermal conditions, by regressing the proportion of the inactive time spent lying (proportion of time lying/proportion of time lying + proportion of time standing) on average daily black globe temperature.

Analysis of herd synchrony

Observing the wildebeest gave the impression that most of the herd were engaged in the same behaviour at any one scan, and that when behaviour changed, most of the herd displayed a synchronous change in behaviour. To verify this impression statistically, we normalised each scan where more than 5 animals were observed to the median herd size (13) and compiled a frequency table representing the proportion of the normalised herd (0/13, 1/13, ..., 13/13) engaged in each behaviour. This frequency distribution was compared to a binomial distribution (representing a random distribution of behaviour) constructed using the probability, from all scans, of an individual in the herd being engaged in each behaviour at each scan.

Results

Rainfall and herd size

The year preceding the study was wetter (yearly rainfall=504 mm), while the year of the study was drier (280 mm) than average. From May to October there was 4.9 mm of rain. There were small falls in November (20 mm) and December (15 mm). In January, 75 mm was recorded during the week preceding the observations. Table 1 shows the average weather for days and nights when observations were made. May and August had cool days and cold nights, while November, December, and January had warm nights and hot days.

Table 1 Summary of weather data for observation days and nights
Full size table

The size of the herd under observation varied from 2 to 60 individuals with a median herd size of 13. The smaller samples occurred mainly at night. The 25th and 75th percentiles of herd size were 3 and 20, respectively. There was no significant difference in herd size between surveys (ANOVA, F 5,12=0.822; P=0.56). There were 919 scans in total, constituting 12,770 individual recordings of behaviour.

Seasonal and nychthemeral effects

On average wildebeest spent 43% of the total time feeding, with 52% of the night and 36% of the day spent feeding (Fig. 1). The next largest behaviour category was lying, which accounted for 35% of the time, again being concentrated during the nocturnal hours. Standing occurred mainly in the diurnal hours and accounted for 15% of the total time. These three behaviour categories accounted for 93% of the total observations.

Fig. 1
figure 1

Average proportion of the diurnal (open bars), nocturnal (black bars), and diel (grey bars) periods spent engaged in each of four behaviour categories for all surveys. Statistical analysis of the proportions is incorporated into Figs. 2, 3 and 4. Bars show mean±SEM for 10 days

Full size image

Activity patterns were significantly different between seasons (Fig. 2). In the Summer post-rain season, the wildebeest lay less and fed more than expected from yearly averages. During the Cool season the wildebeest lay more, stood less, and showed the other behaviours less, while in the Warm season they lay and fed less, stood more, and showed the other behaviours more than expected. Activity was also significantly different between time periods (Fig. 3). A significantly greater proportion of the night than of other time periods was spent feeding, with most of the rest of the night spent lying. Similarly the afternoon was mostly spent either lying or feeding. During the morning and midday periods the wildebeest stood when not feeding.

Fig. 2
figure 2

The proportion of the 24-h day spent engaged in each behaviour during each season. Probabilities are for comparisons of seasons for each behaviour. A significantly greater (+) or smaller (−) proportion of time than expected according to the saturated logit model is indicated by superscripts on each mean. Number of superscript symbols indicates probability level; +/- P<0.05, ++/− − P<0.01, +++/− − − P<0.001. For example +++ on “Summer post-rain” means the herd spent a significantly greater proportion [P<0.001] of the 24-h day feeding in the Summer post-rain season than expected. Bars show mean ± SEM for the average of 6 days (Cool), or 2 days (Warm and Summer—post-rain)

Full size image
Fig. 3
figure 3

The proportion of each time period spent engaged in each behaviour. Probabilities are shown for comparisons of time periods across behaviours. Statistical results are as described in the legend of Fig. 2. For example +++ on “Midday-Stand” means the herd spent a significantly [P<0.001] greater proportion of time standing during the midday time period than expected. Bars show mean ± SEM for the average of 10 days

Full size image

The interaction between season and time period was significant (Fig. 4). During the Cool season the night was divided between feeding and lying. The morning period also was spent mostly feeding. There was less standing overall during the Cool season (Fig. 2) and such standing as the animals did employ occurred mostly in the midday (Fig. 4a). In the Warm season the animals did not feed as much as during the other seasons (Fig. 2) and the feeding that was done was concentrated to the night (Fig. 4b). When not feeding in the Warm season, the wildebeest mostly stood during the morning and midday periods, and lay more during the afternoon and night. In the Summer post-rain season the wildebeest fed for much of the day, with the largest proportion of time devoted to feeding in the night and morning (Fig. 4c). When not feeding, the animals mostly stood in the morning and midday, and spent the afternoon and night lying.

Fig. 4
figure 4

The proportion of each time period spent engaged in each behaviour in the Cool (a), Warm (b), and Summer post-rain (c) seasons. Probabilities are shown for comparisons of time periods across each behaviour for each season separately. Statistical results are denoted as described in the legend of Fig. 2. For example ++ on “Warm-Night-Feed” means the herd spent a significantly greater proportion [P<0.01] of the night period feeding in the Warm season than expected. Bars show mean ± SEM for the average of 6 days (Cool), or 2 days (Warm and Summer—post-rain)

Full size image

Thermal environment and activity

The proportion of the diel period spent lying decreased as the mean diel globe temperature increased (F 1,7=24.6, P=0.002, r 2=0.78; Fig. 5a). There was no significant change in the overall proportions of the other behaviour categories with temperature over the diel period. In concert with the reduction in lying overall, the proportion of the diel inactive period spent lying tended to decrease as the mean diel globe temperature increased (F 1,7=5.1, P=0.058, r 2=0.42). The same relationship was significant for the diurnal period, in that the proportion of the diurnal inactive period spent lying decreased as the mean diurnal globe temperature increased (F 1,15=5.4, P=0.03. r 2=0.26; Fig. 5a). In analysing the proportion of the diurnal, rather than the diel, period spent feeding, the Summer post-rain season appeared to present a special case in that much more feeding was observed than during the other seasons. With the Summer post-rain season removed from the regression analysis (see justification in Discussion) there was a significant reduction in diurnal feeding with increased black globe temperature (F 1,12 = 31.8, P<10−4, r 2=0.73). The points for the Summer post-rain days lie well outside the 95% prediction limits for the data relating feeding to globe temperature for the other seasons (Fig. 5b). The proportion of the diurnal period spent running also decreased as average black globe temperature increased (F 1,15=5.6, P=0.03; data not shown).

Fig. 5
figure 5

Plots, against average diel or diurnal (as appropriate) black globe temperature, of arcsine-transformed proportions (expressed in radians) of a diel period spent lying (open circles and unbroken regression line) and diurnal inactive period spent lying (closed squares and broken regression line), and b diurnal period spent feeding, with regression line and 95% prediction limits. Summer post-rain data (closed symbols in b) were not included in the regression analysis. Each point represents the mean for 1 day

Full size image

Social effects: behavioural synchrony

The frequency distribution of all four behaviours differed significantly from the expected binomial distribution (lying χ2 8=18,328, standing χ2 8=18,355, feeding χ2 10=28,318, other χ2 4=970; P<10−6 in all cases). By comparing the variance of the observed distribution to the variance of the expected (binomial) distribution, the type of deviation from randomness for such distributions can be assessed. A variance greater than expected indicates clumping of the observed data, which, in our case, meant an unexpected number of observations where no animals were engaged in a given behaviour, or many animals were engaged in the same behaviour (Zar 1996). The expected and observed variances, respectively, for each behaviour in these analyses were: lying 3.3, 32.4; standing 3.0, 20.9; feeding 3.8, 33.3; other 1.3, 9.7. These results support the hypothesis that the herd were in synchrony for each behavioural activity.

Discussion

A major impetus leading us to investigate behavioural thermoregulatory strategies in the black wildebeest was the observation that they are excellent homeotherms despite the absence of opportunity for microclimate selection in their habitat (Jessen et al. 1994), and especially the absence of shade-seeking, the most common behaviour of large animals in response to heat load (Ben Shahar and Fairall 1987; Blackshaw and Blackshaw 1994; Jarman and Jarman 1973; Mitchell 1977). Our results show that wildebeest behaviour did change with ambient thermal conditions in a manner implying that the animals adopt strategies to behaviourally adjust energy exchange with the environment. Season and time of day had significant effects on behaviour, notably that total feeding time was reduced in warmer weather and that most of the feeding when conditions were warmer was shifted to the cooler nocturnal hours (Fig. 4). Combining that observation of a seasonal change with the observation that diurnal feeding was reduced by increased heat load we answer question one (raised in Introduction) affirmatively, gross activity patterns were affected by climatic factors associated with change of season. The active behaviours of feeding and running were recorded less when days were warmer, and feeding was shifted to the nocturnal hours in the warmer seasons, so question two is also answered in the affirmative. That wildebeest spent the majority of their inactive time lying in cooler conditions and standing in warmer conditions, and that there was a significant negative relationship between the proportion of the inactive time spent lying and heat load, indicates that wildebeest did alter posture, when inactive, in a way associated with ambient thermal conditions, so question three is also answered in the affirmative.

Active behaviours

No-one has analysed the heat production of different behaviours in free-ranging antelope, but Malechek and Smith (1976) have for cattle. Grazing and travelling increase heat production by fivefold and threefold, respectively, above the heat production of standing idle. Since endogenous heat forms a large part of the heat load on mammals, one strategy for reducing heat stress behaviourally would be to decrease behaviours that increase heat production. We observed that the wildebeest ran less on warmer days. In December, the hottest survey, the entire herd never ran. The normal reason for free-ranging animals to run would be to escape predators. Though there was predation of smaller antelope in the habitat (Mitchell et al. 1997), we did not observe predators harassing the wildebeest at any time.

The active behaviour that occupied the majority of the wildebeest’s time was feeding. Many laboratory (Hamilton 1971; Johnson and Strack 1989; Torres-Contreras and Bozinovic 1997) and field (Belovsky and Slade 1986; Klein and Fairall 1986; Leuthold and Leuthold 1978; Leuthold 1977; Lewis 1977; Owen-Smith 1998) studies have reported an inhibitory effect of high ambient temperature on feeding. The adverse effects of heat stress on agricultural animal growth and production are well known (Cabanac 1996; Conrad 1985). Since shade-seeking generally precludes grazing, part of the decrease in feeding with increased heat load for grazing species will be a consequence of shade-seeking (Lewis 1978). However, here we have shown that the time spent feeding was reduced in warmer conditions in a mammal that is not able to seek shade. How the heat load transduced physiologically into reduced feeding is not known. Andersson (1963) suppressed feed intake by goats by heating their hypothalamus, but later Spector et al. (1968) and Hamilton and Ciaccia (1971) showed that brain heating increased feed intake in rats. The thermal signal modulating feeding need not be core temperature, since skin temperature appears to play a large role in thermoregulatory behaviour (Frank et al. 1999).

Over the year of the study, the average time that our wildebeest spent feeding was 43% of each 24-h period. That figure compares favourably to the 41% predicted allometrically for a ruminant of wildebeest size (Owen-Smith 1992), and 40% of the diurnal period reported previously for black wildebeest (Vrahimis and Kok 1993). Within that average there is concealed seasonal variation not identified in earlier reports. During the Cool season the wildebeest fed for the predicted proportion of time (43%), but they spent less time feeding during the Warm season (35%), and a greater proportion of time than predicted feeding in the Summer post-rain season (55%). Factors other than the thermal variables we measured also probably influenced feeding time. For example, forage quantity and quality decreases as the dry season progresses (Esler and Rundel 1999) and so our warmer surveys would have been associated with the lowest forage availability and quality. However, one would expect feeding time to increase when forage is poorer, so forage quality decline cannot account for the pattern we observed.

The Summer post-rain data present us with an obvious anomaly related to feeding, in that the active behaviour of feeding was increased compared to the other seasons. There are several possible explanations. The wildebeest calved in November/December and during January five calves were present and suckling in the herd observed, so presumably five of the females were lactating. Lactation is energetically the most expensive period in a mammal’s life history (Poppitt et al. 1994) and would have necessitated an increase in forage intake regardless of the ambient conditions. But since median herd size was 13, not all of the animals were lactating. It is possible that lactating females acted as focal animals directing herd behaviour (see below) but we hypothesize that the main effect was the availability of water after the first heavy rains of the wet season. At the study site, only 40 mm of rain was recorded in the 8 months before January, but 75 mm fell in the week preceding the January survey. There are several ways that this may have influenced wildebeest activity. The greater availability of water, both in ephemeral pools and in the higher water-content of forage may have allowed the wildebeest to rely more on evaporative heat loss than behavioural avoidance of heat gain to maintain homeothermy. Thus the thermally undesirable, but otherwise essential, behaviour of feeding could be assigned higher priority. We did not collect quantitative data on panting, the main autonomic heat loss mechanism dependent on water supply, but noticed it frequently during the January survey and not at all in November or December. Interestingly, Owen-Smith (1998) found that the effect of high ambient temperatures on kudu feeding differed with season, and attributed this to changes in pelage characteristics. During the dry season, feeding time by kudu was reduced on days when ambient temperature exceeded 30°C, but during the wet season feeding time was reduced only when ambient temperature exceeded 36°C. It may be that the availability of water for autonomic thermoregulation during the wet season, rather than any change in pelage, allowed the kudu to forage at higher ambient temperatures. Several other studies have noted an increase in foraging time after rain (Duncan 1975; Owen-Smith 1973). Increased preformed water is not the only consequence of rain that might affect thermoregulatory strategies since the albedo of wet ground is less than that of dry ground, with the difference being largest for bare ground (Allen et al. 1994). Wet ground reflects less direct and diffuse radiation and would reduce the radiant heat load on the wildebeest after rain.

Whether it had rained recently or not, the proportion of the night spent feeding was greater than the proportion of the day spent feeding (Fig. 4), especially in the Warm season, when over 50% of the night was spent feeding but only 20% of the day. These data support the hypothesis that active behaviours were concentrated in the cooler hours, and more so as conditions became hotter, so that it was indeed heat load, and not some other feature of the night, that was the driving force. A potential source of bias in our study was that we measured activity only during periods when the nights were moonlit, and activity on moonlit nights may be increased relative to moonless nights (Leuthold 1977). Even if moonlight did provide a bias in this manner, however, the bias would have been equal in all surveys. Thus we suggest that the strategy adopted by the wildebeest of shifting their feeding to the cooler nocturnal hours in warmer weather was not because of moonlight. Nagy and Knight (1994) have suggested that springbok (Antidorcas marsupialis) may feed preferentially at night during dry conditions because plant moisture content is highest at night, a possibility also for wildebeest.

Posture

A less obvious influence of thermal factors on the activity pattern of black wildebeest was the division of the inactive time between lying and standing. Laboratory studies have shown that some mammals adopt postures that reduce surface area in the cold and increase it in the heat (Bustamante et al. 2002). Our wildebeest lay more at night when the weather was cooler, thus reducing exposed surface area therefore heat loss. At night and during the diurnal periods of the Cool season, almost the entire inactive time was spent lying. Several studies have noted similar behaviour in other species (Berry et al. 1982; Jarman and Jarman 1973; Norton 1981).

The corollary is that, in hotter conditions, the animals would increase exposed surface area by standing (Hayasaka and Yamagishi 1990). On hotter days the wildebeest did spend a larger proportion of the inactive period standing (Fig. 5a). However, although the wildebeest stood for most of the morning and midday periods during the Warm season, they preferred to lie during the afternoons when ambient temperatures were highest (Fig. 4). That apparent contradiction is resolved by the observations of Finch (1972), who showed that in the latter part of the day the largest single component of the heat load on an antelope was thermal radiation from the ground, not from the sun directly. A lying posture in the afternoon would reduce the interception of thermal long wave and reflected solar short and long wave radiation from the ground.

Social environment

A major factor influencing patterns of wildebeest activity was the tendency for most of a herd to be engaged in the same behaviour at any one time. Such “activity synchrony” is commonly observed in social species and probably facilitates maintenance of the spatial coherence of the herd (Conradt and Roper 2003). Whether this homogeneity resulted from entrainment of activity based on a focal animal, or from similar environmental stimuli independently evoking similar behaviours in all individuals, is unknown. Synchrony of activity was obvious especially in the active behaviours of walking and running. If one individual in the herd began to run, the entire herd soon would follow. Even for inactive behaviours, though, our analysis of herd synchrony showed that the entire herd tended to alter behaviour at the same times. Since the thermal consequences of changing behaviour will depend on anatomical factors (such as body mass) and physiological factors (such as lactation), which were not the same throughout the herd, we doubt if the synchronous behaviour was driven solely by thermal factors.

Conclusion

We have shown that the behaviour patterns of a black wildebeest population varied in a manner suggesting that the behaviours were subject to a thermoregulatory drive. Increases in exogenous heat load resulted generally in a decrease in active behaviours. In cooler weather and at night in all seasons, the wildebeest mostly lay down when not feeding, thereby reducing the exposed surface area. High ambient temperatures reduced diurnal feeding activity, for which the animals compensated by feeding mainly at night during warm weather. During warmer weather the animals stood during the mornings, but lay down in the afternoons. Thermal factors were not the only determinant of behaviour, though. Feeding in our animals conformed to the “typical” ungulate feeding pattern of peaks in the morning, at midday, and in the evening, irrespective of the weather. A major difference between black wildebeest and other ungulates was that sundown initiated a feeding bout, while in other species sundown signals the end of the afternoon feeding bout. Also, most of the herd engaged in the same behaviour at any time, irrespective of the thermal consequences for individual animals.

If there were no cost to the animals in using behavioural strategies for thermoregulatory purposes, then presumably they would be implemented whenever needed. Our Summer post-rain data imply that there are costs involved, and that the advantages of thermoregulatory behaviour must be traded off against the advantages of other competing behaviours. Possible factors affecting trade-offs are the fact that standing is more energetically expensive than lying (Malechek and Smith 1976), that reduced feeding activity reduces energy intake, especially when forage quality is reduced toward the end of the dry season, and that concentration of feeding nocturnally may affect rumination time and reduce digestibility. Behavioural strategies may be favoured when the costs of autonomic thermoregulation are deemed greater than the cost of implementing changes in behaviour. When water was freely available for evaporative cooling, the disadvantage of autonomic thermoregulation was attenuated and the animals appear to have reduced reliance on behavioural strategies. Orientation behaviour may have been influenced by trade offs in a similar fashion (Maloney et al. 2005).

Despite being exposed to large nychthemeral swings in their thermal environment, and having no access to thermal refuges in the day or at night, black wildebeest show little nychthemeral variation in body temperatures (Jessen et al. 1994). The behavioural strategies we outline would contribute to this competency, and at a cost less than that of autonomic strategies. Details of their autonomic thermoregulation are discussed by Jessen et al. (1994). Supplementing the behavioural responses to environmental thermal conditions reported here, black wildebeest also use body orientation to modify solar heat loads (Maloney et al. 2005), which also contributes to their remarkable thermoregulatory abilities.