Introduction

Body temperature (T b) is a key variable that is actively regulated in endotherms despite being influenced by multiple environmental factors that affect heat production and dissipation (Schmidt-Nielsen 1990). If animals generate heat at a greater rate than it can be dissipated, the unavoidable result is hyperthermia (abnormally high T b), that can be immediately fatal or more frequently impacts future survival and reproduction (Speakman and Król 2010). Therefore, physiological or behavioural strategies that elicit hyperthermia are intriguing phenomena. For example, fever induced by bacterial and viral infections is argued to be an adaptive elevation of the set-point at which T b is regulated, that enhances immunocompetence at the expense of potential acute and long-term costs of hyperthermia (Kluger et al. 1996).

Endotherms routinely express a rise in T b in response to handling stress. In fact, a standard procedure for recording T b, handling and insertion of a rectal probe, leads to a rise in T b within minutes (Borsini et al. 1989; Briese and De Quijada 1970). Such a rise in T b in response to handling and other stressors is variously referred to as emotional hyperthermia (Briese 1995), emotional fever (Cabanac and Briese 1992), or stress-induced hyperthermia (Van der Heyden et al. 1997). This phenomenon has been recorded in humans (Briese 1995) and various other mammal species (Clark et al. 2003; Lee et al. 2000; Meyer et al. 2008; Oppermann and Bakken 1997; Parr and Hopkins 2000; Pedernera-Romano et al. 2010; Yokoi 1966). The stress-induced rise in T b can be provoked by various factors such as handling, open-field tests, cage changes, social defeat (Hayashida et al. 2010), and even in anticipation of psychological stressors (Kluger et al. 1987). A standard laboratory procedure has been established to evaluate the stress-induced rise in T b in singly housed laboratory mice that involves sequential recording of two rectal T b spaced 5–10 min apart and calculating the difference between these (∆T b) (Van der Heyden et al. 1997). Given its simplicity, low-cost, wide application on model organisms (Bouwknecht et al. 2007), and its use as a “read-out” parameter of stress (Vinkers et al. 2010), it is surprising that the stress-induced rise in T b has only been studied once among free-ranging animals in the field (Meyer et al. 2008).

There has been considerable debate whether the stress-induced rise in T b is a hyperthermia or a fever (Oka et al. 2001; Vinkers et al. 2008). According to the Commission for thermal physiology of the international union of physiological sciences (2001), hyperthermia describes a T b above the range specified for the normal active state of the species and can either be forced (implies a failure of thermoregulation, as total heat production exceeds the capacity for heat loss) or regulated (implies a fever, a rise in the set-point at which T b is regulated accompanied by supportive changes in thermoeffector activities). Evidence interpreted as consistent with the stress-induced rise in T b being a form of fever includes shivering and skin vasoconstriction coinciding with a rise in T b (Briese and Cabanac 1991) and the lack of a relationship between the stress-induced rise in T b and ambient temperature (T a) (Briese and Cabanac 1991; Long et al. 1990). Another set of evidence in support of the stress-induced rise in T b being a form of fever are studies showing that it is inhibited by antipyretic drugs (Kluger et al. 1987). However, Vinkers et al. (2009) recently reported that the stress-induced rise in T b and infection-induced fever are two distinct processes mediated largely by different neurobiological mechanisms. In the absence of empirical evidence in favour or against one of these scenarios, this phenomenon should be neutrally described as a stress-induced rise in T b.

If the stress-induced rise in T b is an adaptive mechanism involved during the “fight or flight” response, as previously suggested as an argument that it is a fever (Cabanac 2006; Oka et al. 2001), then individual differences in the stress-induced rise in T b represent a biologically relevant level of variation (Wilson 1998). However, to our knowledge, the extent of consistent individual differences in (i.e. repeatability of) the stress-induced rise in T b has never been explicitly studied, which is surprising given the longstanding interest of physiologists in individual variation (Berteaux et al. 1996; Chappell et al. 1996; Garland and Bennett 1990; Hayes et al. 1998; Labocha et al. 2004; Speakman et al. 2004; Szafranska et al. 2007). Repeatability is a key aspect of any trait that we assume to be a product (or target) of selection, as it places an upper limit on the possible heritability (Falconer and Mackay 1996). Individual variation can be seen as the raw material upon which natural selection acts, but also as the result of natural selection itself (Wilson 1998). In order to understand the evolutionary origins of the stress-induced rise in T b, a first step is therefore to evaluate the degree of repeatability (Lessells and Boag 1987).

In this study, we investigated individual differences in the stress-induced rise in T b, as well as individual and environmental correlates of the stress-induced rise in T b, in a wild population of individually marked eastern chipmunks (Tamias striatus). We quantified the stress-induced rise in T b as the ∆T b recorded by a rectal T b probe inserted in the first minute (T b1) and after 5 min (T b5) of a usual field handling procedure. We predicted that the stress-induced rise in T b would be repeatable across multiple captures on the same individual, indicative of consistent individual differences in the stress-induced rise in T b. We also evaluated potential correlates of the stress-induced rise in T b, including season, time of day, age, sex, and reproductive status, as well as T a and M b reflecting heat dissipation potential and field metabolic rate (FMR) reflecting heat production.

Materials and methods

Study population and animals

In 2008 and 2009, we monitored an individually marked population of free-ranging eastern chipmunks in the Ruiter Valley Land Trust (Sutton Mountains, Québec, 45°05′N, 72°26′W). The study site contained 228 Longworth traps distributed in a grid pattern over 25 ha of mature American beech (Fagus grandifolia) forest (Munro et al. 2008). We systematically trapped the entire grid every week from early-May through early-October. Traps were baited with peanut butter around 0800 hours and visited at intervals of 2 h until sunset. At first capture, individuals were permanently marked with numbered ear tags (National Band and Tag Company 1005-1) and a Trovan® PIT tag inserted in the inter-scapular region. For all captures, we noted trap location, M b (measured with a 300 g Pesola scale, ±1 g) sex, and reproductive status (males scrotal or abdominal, females lactating or non-lactating). We assigned minimum known age to individuals according to whether they were first caught as a juvenile or as an adult (+1), as in Careau et al. (2010). We differentiated juvenile chipmunks from adults based either on an initial capture within a month following emergence when M b was <80 g or, for individuals >80 g when first captured, on the absence of a darkened scrotum or developed mammae (Bennett 1972). We obtained measurements on juveniles as well as adults aged up to 4-year-old (since the monitoring of the population started in 2004). We divided the year into three seasons: spring (May and June), summer (July and August), and autumn (September and October). Animals could not be captured during winter hibernation. Hourly T a data were obtained from two meteorological stations located at 32 and 37 km from our site (North of the site: Lac Memphremagog 45°16′00″N, 72°10′00″W; West of the site: Frelighsburg 45°03′01″N, 72°51′41″W; Environment Canada; http://climate.weatheroffice.gc.ca). Both stations gave T a estimates that were highly correlated with each other and to a weather station located on the middle of our grid that was operational only in 2006 (Lac Memphremagog vs. Frelighsburg stations: r = 0.96, n = 3719; onsite station vs. Lac Memphremagog: r = 0.97, n = 3719; onsite station vs. Frelighsburg: r = 0.98, n = 4366; correlations estimated for each hourly estimate available for each station from May 1 to October 1 2006). We averaged T a recordings at both stations and associated each T b recording to the closest corresponding hourly T a recording. Although our T a variable ignores factors like solar radiation and wind speed, whose impacts are likely important on small mammals, the canopy of the dense forest at our site might have reduced the relative impact of these factor. The “noise” included by these factors makes our test of the effect of T a on the stress-induced rise in T b conservative.

Body temperature (T b) and stress-induced rise in T b

We recorded rectal T b over two summers (2008 and 2009) directly in the field, following a slightly modified version of the standard procedure for evaluating the stress-induced rise in T b in singly housed mice (Van der Heyden et al. 1997). All T b were taken by one person (VC) to minimize variation. Upon arriving at a trap with a closed door, a timer was started, the trap was lifted, and the chipmunk was transferred into a handling bag and taken by hand. It was possible to rapidly expose the anus of the chipmunk by pinching firmly the fur in the nape of the neck and holding one hind leg. A thermocouple probe was inserted 10 mm deep into the rectum (Omega engineering, model HH203A) and maintained until the display reached a stable value (~20 s). All initial T b records (T b1) used in the analysis were taken within 60 s of lifting the trap, otherwise the record was excluded. After the initial T b was taken, the animal was put back in the handling bag for pit tag and ear tag identification, measurement of M b, and assessment of reproductive status (see above). A final reading of temperature (T b5) was taken exactly 5 min after the trap had been first lifted. After each test, the probe was disinfected with 95% ethanol. We calculated the elevation in T b induced by handling as T b5 – T b1 and hereafter refer to this variable as ∆T b.

Ninety percent of T b recordings occurred between 1000 and 1800 hours, a period during which the diurnal pattern of T b is relatively stable at 39°C in captive chipmunks (Levesque and Tattersall 2010). We recognise that the best protocol would have included repeated T b recordings at 10-min interval, as in singly housed laboratory mice (Van der Heyden et al. 1997). However, handling can be highly stressful for wild animals and some chipmunks become lethargic (or faint death) when handled over prolonged period (VC, pers. obs.). It has been shown in singly housed laboratory mice that ~80% of the rise in T b is reached within 5 min of handling (Van der Heyden et al. 1997). Therefore, our protocol represents a necessary compromise between the length of capture under field conditions that required quick release of the animals and the full expression of the stress-induced rise in T b. We currently do not have the data to suggest that an interval of 5 min is appropriate for assessing the magnitude of the stress-induced rise in T b in chipmunks. If the peak value of the stress-induced rise in T b is obtained later than 5 min after the onset of handling, then the significance of our ∆T b measure may not be its magnitude but instead the speed at which the response occurs during stress, which itself may also be relevant from an evolutionary point of view. Our results must therefore be tentatively interpreted until a comprehensive study is performed using temperature-sensitive transmitters or data loggers implanted into the abdominal cavity of animals to obtain a complete and continuous tracing of T b before, during, and after capture.

Field metabolic rate (FMR)

We measured FMR using the doubly labelled water (DLW) technique (Butler et al. 2004; Lifson et al. 1975; Speakman 1997). This technique estimates the carbon dioxide (CO2) produced by a free-ranging animal based on the differential washout of injected hydrogen (2H) and oxygen (18O) isotopes. It has been previously validated by comparison with direct calorimetry in a range of small mammals and provides an accurate measure of FMR over periods of several days (Speakman et al. 1994). The DLW technique has also been previously used successfully on eastern chipmunks (Humphries et al. 2002).

The method we used implied two captures. On the initial capture, chipmunks were injected intra-peritoneally with 240 μl of DLW (37.78 and 4.57% enriched 18O and 2H). Following injection, subjects were held in the trap for a 1-h equilibration period (Speakman and Król 2005). Then, a first blood sample was collected via a clipped toenail for initial isotope analysis. Chipmunks were then released at the site of capture and recaptured, weighed and bled 1 to 3 days later, at as close as feasible to periods of 24-h interval, and a final blood sample was taken to estimate isotope elimination rates. Taking samples over multiples of 24-h period minimise the substantial day-to-day variability in FMR (Berteaux and Thomas 1999; Speakman et al. 1994). The range of absolute deviation from 24 h was 5–180 min (25th percentile = 24 min; median = 40 min, 75th percentile = 62 min). In 2008 and 2009, a total of 5 and 3 animals were blood sampled without prior injection to estimate background isotope enrichments of 2H and 18O (method C in Speakman and Racey 1987). Capillaries were flame sealed immediately after sample collection. Capillaries that contained the blood samples were vacuum distilled and water from the resulting distillate was used to produce CO2 (methods in Speakman et al. 1990) and H2 (methods in Speakman and Król 2010). The isotope ratios 18O:16O and 2H:1H were analysed using gas source isotope ratio mass spectrometry (Optima, Micromass IRMS and Isochrom μG, Manchester, UK). Samples were run alongside three lab standards for each isotope (calibrated to International standards) to correct delta values to ppm. Isotope enrichments were converted to values of FMR using a single pool model as recommended for this size of animal (Speakman 1993). We assumed evaporation of 25% of the water flux (equation 7.17 in Speakman 1997) which minimizes error in a range of conditions (Visser and Schekkerman 1999).

The FMR data analysed here consist in a subset of 21 individuals (18 lactating females, 2 non-reproductive females, and 1 reproductive male) for which we applied the ∆T b protocol on the final captures, just before the blood sampling started. The mean ∆T b in this subsample was not significantly different from the complete sample (mean ± SD = 0.32 ± 0.45°C; F = 0.09; P = 0.76).

Data analysis

We tested the effects of year (2008 or 2009), season (spring, summer, and fall), time of day, time of day2, T a, age, sex, reproductive status, and M b on T b1, T b5, and ∆T b using multiple regression linear mixed models in R 2.7.1 (http://www.r-project.org) with the package ‘nlme’. We included individual identity (ID) as a random effect to estimate the significance of fixed effects while accounting for pseudoreplication resulting from repeated measurements. Model selection was performed using backward procedures, sequentially removing the least significant term from the model based on its P value until all remaining terms were significant at α = 0.05. Although the mean T a differed among seasons (mean ± SD, spring = 19.6 ± 4.8°C, range: 11.0–27.8°C; summer = 21.3 ± 3.5°C, range: 14.8–27.5°C, fall = 18.3 ± 4.9°C, range 5.5–27.3°C), the high variance present within seasons (attributed to day-to-day and hour-to-hour variations) allowed us to test the effects of T a and season simultaneously without colinearity problems.

Stress-related variables are routinely affected by acclimation (habituation effects), and the stress-induced rise in T b is no exception (Cabanac and Briese 1992; Thompson et al. 2003). To test whether chipmunks acclimated to the trapping and T b recording procedures, we included the order of tests as an additional fixed effect. We also included the number of time each chipmunk was captured before each T b recording (mean ± SD = 36 ± 35 captures, range: 0–140). Finally, we calculated the time elapsed between any two T b recordings on an individual. The first T b record for each individual was given an arbitrary value of 740, which corresponds to twice the longest period elapsed between two recordings in our dataset. We never applied the temperature-recording procedure on any chipmunks prior to this study.

Once the fixed effects structure was determined, we calculated individual repeatability by calculating the proportion of variance attributed to ID as a random effect. We assessed the significance of random effects using log-likelihood ratio tests (LRT). A LRT test compares the log-likelihoods of a “full” model that includes ID as a random component versus a “reduced” model without ID (Pinheiro and Bates 2000). The test statistic is equal to twice the difference in log-likelihoods between the two nested models and assumes that it follows a χ2-distribution with one degree of freedom (df). We also estimated repeatability of ∆T b using intraclass correlation coefficient (τ) calculated on individuals with repeated measures only (Lessells and Boag 1987). In addition, we tested whether the rise in T b was significant using a paired t test with the number of degrees of freedom equal to number of individuals measured.

Because we could not control for the time spent in the trap before we handled chipmunks, we tested whether the variability in ∆T b was primarily due to variations in T b1 or mostly due to variation in T b5. We expected T b1 or T b5 to be more strongly correlated to ∆T b if the stress of trapping or handling is more important. Statistical analysis of the effect of FMR on ∆T b was assessed by including julian day and M b as covariates in a multiple regression model.

Results

We recorded T b1 and T b5 294 times on 140 individuals over two complete summers (Fig. 1). T b1 was normally distributed and ranged from 37.7 to 41.1°C (mean ± SD = 39.3 ± 0.6, Fig. 2). T b1 was not influenced by season, time of day, time of day2, age, order of test, and time since last capture (Table 1), but varied annually and was lower in males than in female and in reproductive than non-reproductive animals (Table 1). T b1 was negatively related to the number of time the individual was captured before T b recording (Table 1). T b1 was also positively related to T a (Fig. 3a) and M b (Fig. 3b). After accounting for these significant sources of variation, T b1 was significantly repeatable with 31% of its variance attributed to individual differences (Table 1).

Fig. 1
figure 1

Frequency distribution of the number of records per individual included in this study on free-ranging eastern chipmunks

Full size image
Fig. 2
figure 2

Final body temperature (T b5; after 5 min of handling) as function of initial body temperature (T b1; within 1 min of handling) in free-ranging eastern chipmunks. The line represents the 1:1 ratio (T b5 = T b1)

Full size image
Table 1 Linear mixed effects models for initial body temperature (T b1; within 1 min of handling), final body temperature (T b5; after 5 min of handling), and stress-induced rise in T b (∆T b = T b1 − T b5) in relation to environmental and individual variables in free-ranging eastern chipmunks
Full size table
Fig. 3
figure 3

Influence of ambient temperature (T a; left panels) and body mass (right panels) on (a and b) initial (T b1), (c and d) final (T b5) body temperature, and (e and f) stress-induced rise in Tb (∆T b = T b5 − T b1) in free-ranging eastern chipmunks. See Table 1 for statistical significance. A slight jitter was introduced in all panels to show overlapping dots

Full size image

T b5 was normally distributed and ranged from 37.6 to 41.3°C (mean ± SD = 39.6 ± 0.7, Fig. 2). Similarly to T b1, T b5 was not correlated to time of day, time of day2, age, and time since last capture (Table 1). In contrast to T b1, T b5 did not vary annually and but differed among seasons (lower in mid-summer than spring and fall), and was not affected by the number of previous capture but decreased with order of test (Table 1). Also in contrast to T b1, the effects of sex and reproductive status were marginally non-significant (Table 1). T b5 was positively related to T a (Fig. 3c) and M b (Fig. 3d). After accounting for these significant sources of variation, T b5 was significantly repeatable, with 25% of its variance attributed to individual differences (Table 1).

Rectal temperature rose between T b1 and T b5 in 74.5% of the captures (N = 219 out of 294; data-points located above the line in Fig. 1). ∆T b was negligible (i.e. less than our precision, 0.1°C) in 8.8% of the captures (N = 26) and negative in 16.7% of the captures (N = 49; data-points located below the line in Fig. 2). Overall, ∆T b was normally distributed and ranged from −0.9 to 1.8°C (mean ± SD = 0.3 ± 0.4°C). According to a paired t test, the rise in T b was highly significant (t = 11.09, df = 140, P < 0.001). ∆T b was more strongly correlated to T b5 (r = 0.43, P < 0.0001) than T b1 (r = −0.16, P = 0.007).

∆T b was lower in 2008 than in 2009 and varied seasonally (lower in summer relative to spring and fall; Table 1). The effects of the variables “order of test” and “number of captures prior to T b recording” were non-significant on ∆T b, whereas the variable “time since last T b recording” was marginally non-significant (Table 1). Because of the stronger effects of T a and M b on T b5 than T b1, ∆T b was positively and significantly correlated with T a (Fig. 3e) and M b (Fig. 3f). After accounting for these significant sources of variation, ∆T b was significantly repeatable, with 40% of its variance attributed to individual differences (Table 1). Figure 4a shows the extent of individual variation in ∆T b for 50 chipmunks with 2 and 3 repeated records. Individuals with 4 (n = 10) and 5 or more (n = 10) repeated measures are shown in Fig. 4b, c, respectively.

Fig. 4
figure 4

Individual variation in the stress-induced rise in body temperature (∆T b, final minus initial body temperature, adjusted for year, season, ambient temperature, and body mass) in free-ranging eastern chipmunks with (a) two or three, (b) four, and (c) five to 11 repeated measurements [total n = 224 on 70 individuals τ = 0.28 (F 69,154 = 2.24, P < 0.001)]. A slight jitter was introduced in all panels to show overlapping dots

Full size image

The mean FMR (±SD) was 216.2 ± 64.6 kJ/day (range = 124.5–398.3 kJ/day) among the subsample of 21 individuals for which we quantified ∆T b on the final capture. FMR was not a significant predictor of T b1 (estimate ± SE = 0.53 ± 1.51, t = 0.35, P = 0.73) but tended to be positively correlated with Tb5 with julian day and M b included in the model as a covariates (estimate ± SE = 2.83 ± 1.45, t = 1.95, P = 0.07). FMR was a significant positive predictor of ∆T b (estimate ± SE = 2.30 ± 0.74, t = 3.11, P = 0.006) with julian day and M b included in the model as covariates (Fig. 5).

Fig. 5
figure 5

Stress-induced rise in body temperature (∆T b) against field metabolic rate (FMR) in 21 free-ranging eastern chipmunks. Statistical analysis of the effect of FMR on ∆T b was assessed by including julian day and body mass (M b ) as a covariates in the model, but for illustrative purposes, here FMR values are plotted as residuals from the significant relationship between with M b [r 2 = 0.24, P = 0.022, Log10 (FMR, in kJ per day) = −0.93 + 1.65 Log10(M b)]

Full size image

Discussion

This study, representing the first exploration of individual variation in the stress-induced rise in T b in a wild population, confirms that individuals consistently differ in the magnitude of their stress-induced rise in T b (i.e. ∆T b is repeatable) and that the stress-induced rise in T b is affected by correlates of heat production and dissipation such as M b, T a, and energy expenditure.

Studying the stress-induced rise in T b in the field rather than the laboratory introduces opportunities as well as limitations. Studying an abundant, marked field population facilitated a relatively large sample size (Fig. 1; 294 ∆T b records on 140 individuals), which enabled the detection of consistent individual differences in the stress-induced rise in T b. These individual differences have not been a focus of most prior research on this topic, but represent a promising avenue to explore the evolutionary causes and consequences of the stress-induced rise in T b (Berteaux et al. 1996; Chappell et al. 1996; Garland and Bennett 1990; Hayes et al. 1998; Labocha et al. 2004; Speakman et al. 2004; Szafranska et al. 2007). However, studying the stress-induced rise in T b in the field introduces additional sources of variation that are more easily avoided in captivity. For example, beyond the captures included in the current analyses chipmunks have been trapped and handled multiple times for multiple purposes such as initial marking, pit tag insertion, open-field test (Montiglio et al. 2010), doubly labelled water technique (this study, Bergeron et al. 2011), and respirometry (Careau et al. 2010). Interestingly, the more an individual has been captured the lower its T b1 was, suggesting that chipmunks may habituate to being trapped (and thus had lower T b1). In contrast, T b5 was not related to number of captures prior to T b recording but instead decreased with the number of time their T b was recorded. Although the relationship was marginally non-significant, ∆T b was positively related to time elapsed since last T b recording, suggesting that frequent, regular captures lead to habituation and attenuated the stress-induced rise in T b, while infrequent captures do not attenuate the stress-induced rise in T b. These results attest for the complexity of habituation trends to previous and concurrent handling events involving a mixture of overlapping and unique procedures realised over different temporal frames. Nevertheless, this potential contributor to unexplained variation in our results did not preclude detection of significant individual differences in the stress-induced rise in T b and correlations with T a, M b, and FMR.

An additional source of variation is that we could not control the amount of time (up to 2 h) the individuals spent in the traps before handling. Therefore, our results should be interpreted with care as long as the time course of trap-stress-induced rise in T b remains unknown. One potential explanation for why some values of ∆T b were negative could be the differential effects of the stress of being in a trap on T b1 versus the stress of being handled on T b5 (see above). However, it is notable that negative ∆T b was not preferentially associated with an elevated initial T b1 (see Fig. 2). In fact, the range of T b1 (37.7–41.1°C) is only slightly higher than the range of diurnal T b recorded in captive chipmunks at the same latitude (36.3–40.9°C; Danielle Levesque, personal communication). If time spent in the trap would be the explaining factor for negative values, then the T b1 of these observations would be higher than normal (already stressed in the trap), which seems not to be the case in our study. Interestingly, T b1 is less affected by T a than T b5, suggesting that handling causes the chipmunks to regulate T b less tightly. However, even after controlling for T a and stress-related variables, ∆T b was repeatable, which indicates that some individuals are more likely to express negative ∆T b than others (see Fig. 2). It may be possible, then, that negative values are included in the natural range of ∆T b observed in response to the temperature-recording procedure. In some rat strains, negative ∆T b can be apparently induced by restraint depending on time of day and the type of restraint (mesh vs. plexiglas restrainers) and is thought to be a physiological response to severe overheating microenvironment (McGivern et al. 2009).

The most parsimonious explanation to our findings that the stress-induced rise in T b increases as a function of T a, M b, and FMR is that it represents a forced hyperthermia (i.e. an incidental consequence of heat production that exceeds heat dissipation) rather than an active hyperthermia (i.e. fever, an increase in the set-point at which T b is regulated). Under this interpretation, the occurrence of repeatable individual variation in the stress-induced rise in T b would arise from differential propensity to over-heat in stressful circumstances associated with consistent individual differences in physiology, morphology (beyond M b, which was accounted for as a covariate) or behaviour affecting heat production and dissipation during handling. Our findings that the stress-induced rise in T b differed between years and seasons could be related to physiological and behavioural variations occurring at the population level. For example, eastern chipmunks are known to become much less active in summer, a phenomenon referred to as the summer “lull” (Elliot 1978; Yahner 1978). In our population, resting metabolic rate increases in fall compared to spring and summer (Careau, unpublished data). If these seasonal changes at the population level are somehow linked to propensity to overheat during captures, they could explain the annual and seasonal differences we observed in ∆T b.

Given the debate that occurred as to whether the stress-induced rise in T b represents an uncontrolled rise in T b versus an increase in the set-point at which T b is regulated, we also tentatively interpret our findings under the fever scenario. If the stress-induced rise in T b represents an increase in the regulated set-point (Cabanac 2006), why, then, ∆T b is positively correlated with T a and M b? We suggest the possibility that the value of the set-point varies according to heat dissipation needs/capacity (Speakman and Król 2010). Although, most obviously, hyperthermia can be an outcome of insufficient heat dissipation, counter-intuitively, an up-regulated T b can also be a means to facilitate heat loss. By increasing the thermal gradient between T b and T a, an up-regulated set-point increases heat flux. If the capacity to dissipate heat is an important limit to animal performance during brief period of intense activity (Speakman and Król 2010), then individuals with higher heat load should regulate T b at a higher set-point than individuals with lower heat load. This should be the case for animals exposed to higher T a and/or have higher FMR compared to those exposed to lower T a and/or have lower FMR. Animals with lower surface-to-volume ratio (e.g. higher M b) should also have higher heat load than those with higher surface-to-volume ratio (e.g. lower M b), as long as T a is lower than T b, which probably is the case for most endotherms in most parts of the world during most months of the year (Humphries and Careau 2011). Interestingly, it was recently found that infection-induced fever was modest or even absent in mice at low T a in response to the same stimuli that induce fever at thermoneutral conditions (Rudaya et al. 2005). Two previous studies (Briese and Cabanac 1991; Long et al. 1990) reporting that SIH was independent of T a interpreted their result as in favour of the stress-induced rise in T b being a rise in the set-point (Oka et al. 2001). Briese (1992) later reported that the stress-induced rise in T b was higher in a cold (8°C) than a warm (31°C) environment, whereas we obtained the opposite result (∆T b greater in warm T a). Although there is no clear empirical consensus emerging on the effect of T a on the stress-induced rise in T b, the logic we proposed above implies that the independency of this phenomenon versus T a does not constitute per se an argument in support of the view that it is an adaptive rise in set-point.

The study of individual differences in the stress-induced rise in T b could serve as a bridge between this laboratory paradigm and other sub-disciplines where inter-individual variation is already considered as an important ecological and evolutionary characteristic of wild populations [population biology (Bolnick et al. 2003), epidemiology (Lloyd-Smith et al. 2005), endocrinology (Williams 2008), behavioural ecology (Réale et al. 2007), and energy metabolism (Careau et al. 2008)]. Although identifying the mechanisms involved in the stress-induced rise in T b are likely best explored in captivity, the study of this phenomenon into a natural context will yield novel insights about environmental effects and fitness consequences that affect the expression and evolution of the stress-induced rise in T b in wild populations (Costa and Sinervo 2004). Our demonstration of consistent individual differences in the stress-induced rise in T b is an important first step in defining the upper limit of its potential heritability (Falconer and Mackay 1996). Given the stress-induced rise in T b is speculated to be an adaptive mechanism involved in the “fight or flight” response (Cabanac 2006; Oka et al. 2001), then the obvious next step is to evaluate selection on it and its heritability in the wild. Additional promising avenues for future research, in captivity or the field, include relating the stress-induced rise in T b to personality traits (e.g. boldness, aggressiveness, and exploration) or coping styles (Agren et al. 2009; Carere and van Oers 2004), as well as conducting a more thorough evaluation of energetic correlates than was accomplished here.