Stress-induced hyperthermia (SIH) is characterised by a rise in body temperature in response to a stressor. In endotherms, SIH is mediated by the autonomic nervous system, whereas ectotherms must raise their body temperature via behavioural means by moving to warmer areas within their environment (behavioural thermoregulation). A recent study suggested that zebrafish (Danio rerio), an important model species, may move to warmer water in response to handling and confinement and thus exhibit SIH, which, if accepted, may have important practical and welfare implications. However, an alternative hypothesis proposed that the observed movements may be produced by avoidance behaviour rather than behavioural thermoregulation. Investigating the claims for SIH in zebrafish further, we conducted two experiments that extend the earlier study. The first experiment incorporated new conditions that considered fish behaviour in the absence of thermal variation, i.e. their null distribution, an important condition that was not performed in the original study. The second was a refined version of the experiment to reduce the numbers of fish and aid movement between areas for the fish. In contrast to the previous study, we saw no effect of handling or confinement on preference for warmer areas, and no evidence for SIH in either experiment. Instead, we observed a short-lived reduction in preference for warmer areas immediately post-stress. Our work suggests that zebrafish may not experience SIH, and claims regarding fish consciousness based on SIH may need to be revised.
An important component of animal research explores how animals react to, or cope, with stress. Although stress is often not always easily defined (Koolhaas et al., 2011; Levine, 1985), in a broad sense stress can be considered to be a response of the body to a noxious stimulus (Koolhaas et al., 2011). Understanding the causes of and responses to stress at behavioural and physiological levels remains a clear focus for experimental work across animal species (McEwen, 2007; Schulte, 2014; Koolhaas et al., 2011, 1999). Understanding responses to stress is also fundamental from a welfare perspective, not least because, ethically, we are bound to maintain animals in the healthiest environment possible, but also because adequate conditions and mitigation of stress in captivity is essential for obtaining reliable and biologically relevant data (Newberry, 1995). Laboratory conditions and handling procedures are well known to cause stress (Balcombe et al., 2004), and understanding the sources of and response to stress of animals maintained in captivity has been and will be an important component of research (Moberg, 1985).
Stress-induced hyperthermia (SIH), also referred to as behavioural or emotional fever (Boltaña et al., 2013; Briese and Cabanac, 1991; Rey et al., 2015) is one such reaction to stress. SIH is characterised by a rise in core body temperature, mediated – in endotherms – by the autonomic nervous system (Olivier et al., 2005). In species that exhibit SIH, any stressor will result in hyperthermia (Olivier et al., 2005), for example, in animals with hierarchical social contexts, individuals may exhibit SIH in response to being defeated (Cunningham et al., 2017). While there is still debate as to whether stress-induced increases in temperature should be classified as a fever or hypothermia (Mohammed et al., 2014), because of the consistency of the response and relative ease of measurement, SIH is used as a sign of anxiety in laboratory studies of many endothermic animals (Bouwknecht et al., 2007). However, among ectotherms, including most fish species, SIH requires behavioural thermoregulation, where an individual must to move to a warmer environment to achieve a rise in body temperature. If ectotherms – including the widely used zebrafish (Danio rerio) (Graham et al., 2018; Lawrence, 2007; Spence et al., 2008; White et al., 2017) – were found to exhibit SIH, this would have implications for many experimental studies of behaviour and physiology on these species, especially studies of stress where zebrafish are a common model species (Egan et al., 2009; Marcon et al., 2018; Vindas et al., 2017; White et al., 2017).
While fish have long been reported to show behavioural thermoregulation in response to being infected with pathogens (Reynolds et al., 1976), until recently, no evidence for SIH has been found (Cabanac and Laberge, 1998). Several studies that explicitly test for SIH in ectotherms focus on the suggestion that SIH is considered to provide behavioural evidence for consciousness in non-human animals (Briese and Cabanac, 1991; Cabanac and Laberge, 1998; Rey et al., 2015). Indeed, the lack of evidence for SIH in fish and amphibians has been used to suggest that they lack the capacity for consciousness that other vertebrates are capable of (Cabanac et al., 2009). Other researchers argue that SIH, or lack of it, is at best a contentious diagnostic for consciousness for any species (Allen, 2013; Droege and Braithwaite, 2015). Reservation in using SIH as a test for consciousness is especially important given that there is still uncertainty regarding the underlying physiological and neurological mechanisms that govern the relationship between temperature change within the body/brain and behaviour. A rise in body/brain temperature can be observed in response to situations that are not necessarily stressful, for example in sexual interactions in rats (Mitchum and Kiyatkin, 2004). Additionally, there is still debate as to whether SIH corresponds to a physiological fever or hyperthermia, even in well studied endotherm species (Mohammed et al., 2014).
The subject of SIH in ectotherms and a suggestive link to consciousness in fish was raised in a recent study suggesting that zebrafish exhibit a preference for warmer water following a mild stressor (Rey et al., 2015). The authors provided some evidence that in a thermal gradient zebrafish exhibited a tendency to move to warmer waters after handling and temporary confinement. However, other researchers have been critical of the authors’ analyses and interpretation, and suggested that avoidance of and movement away from confinement area, rather than a preference for a warmer area, would have produced similar movement patterns and cannot be ruled out as an alternative hypothesis for the observed behaviour (Key et al., 2017). Related to this, distinguishing between active thermoregulation and a change of thermal preference within such a gradient can be especially difficult in ectotherms. Evidence of active thermoregulation in ectotherms requires a comparison between the observed distribution of individuals in a thermal gradient and the distribution of individuals in the absence of a thermal gradient, and/or a comparison with non-regulating individuals, as per the null model initially stressed by Heath (1964) and more recently highlighted (Anderson et al., 2007; Hertz et al., 1993).
Thus, the question of whether zebrafish exhibit SIH is still unresolved and to further address this, we conducted two experiments. In our first experiment, we extended the experimental procedure used by Rey et al. (2015) in an attempt to replicate that study, but included a no-gradient condition designed to examine whether movement away from the area where the confinement took place was affected by presence of a thermal gradient or not. In our second experiment, we used an alternative approach to explore the same question (do stressed zebrafish prefer warmer areas?). Our new setup allowed us reduce some of the complications introduced by the design of the first experiment by allowing freer movement between areas, using fewer individuals in a group and with more discrete differences in temperatures between areas.
MATERIALS AND METHODS
Subjects and husbandry
Adult AB strain wild-type short fin zebrafish (Danio rerio), from the St Andrews zebrafish population were used for this study. Fish were housed in mixed sex groups of 36 in 54 l tanks (60×30×30 cm) in the fish laboratory at the University of St Andrews, Scotland. The fish were kept at 26.5±0.4°C on a 12 h light:12 h dark photoperiod cycle and fed twice a day on a mixed commercial flake food (Tetramin tropical flake, Tetra) and freeze-dried bloodworms. Water quality indicators (dissolved oxygen, ammonia, nitrite, pH and temperature) were monitored weekly, and water changes were used to maintain pH, ammonia and nitrite levels within or below recommended levels as per Lawrence (2007) (pH within range 7–8.5; ammonia 0 mg l−1, nitrites 0 mg l−1). Individual fish were used once only and returned to stock tanks afterwards for use in future experiments. Fish were fed in their housing tanks, with a final feed delivered 2 h before moving them into the experimental tank. Fish were never fed in the experimental tank.
All procedures performed were in accordance with the ethical standards of the University of St Andrews and methods used were approved by the University of St Andrews Animal Welfare and Ethics Committee (AWEC). No procedures required UK Home Office licensing. No fish died or suffered apparent ill health during this study, and all individuals were retained in the laboratory for use in future experiments.
Experiment 1: Thermal gradient (avoidance or SIH?)
A major focus of this experiment was an attempt to reproduce the results reported by Rey et al. (2015) – which was itself based on Boltaña et al. (2013). The tank setup and procedures we used therefore followed those detailed in those studies. The chief difference was that we stressed all groups of fish and added a control treatment where fish were stressed and tested in the absence of a temperature gradient to examine whether movement away from a chamber might be due to avoidance of the confinement chamber rather than directed movement towards warmer areas to raise core body temperature. As hypothesised by Key et al. (2017), avoidance behaviour may account for or confound behaviour in a thermal gradient. If fish move away from the confinement chamber in a similar manner to those fish in a condition without a thermal gradient, this would suggest avoidance behaviour motivates zebrafish and would require additional controls to reveal whether fish do use behavioural thermoregulation as a response to stress.
The tank was positioned on the bottom level of a purpose-built fish rack in the laboratory and measured 180×43×35 cm. Five transparent Plexiglas partitions were used to create six equal chambers (30×43×35 cm). Each Plexiglas partition had a hole, 3 cm in diameter, 10 cm from the bottom, allowing for movement of fish between chambers. Chambers were set up to have minimal complexity and to be as identical as possible to reduce the chance that fish moved between chambers or showed preference for specific chambers based on the physical structure of any chamber. Each chamber had a layer of gravel (Betta ‘light’ gravel) 1 cm thick, an airstone (all with equal air pressure) in the rear left corner to enable water mixing and a transparent Plexiglas 4-mm-thick lid. Each chamber had a maximum volume of 36 l; to ensure water volume in each chamber was similar to that in the Rey et al. (2015) design, the water level was lowered to 20.5–22 cm above the gravel such that each chamber had approximately 31 l of water. A thermostat-controlled heater (200 W Tetra) was added to each chamber to aid in the maintenance of a stable temperature, with the heater set to the average temperature of the chamber. Each thermometer was held in place with standard suction cups and positioned such that the red ‘on’ light was not visible. The two end chambers differed from the others in that one had an extra internal thermometer and the other had the in- and out-flow pipes of the chiller. See Fig. S1 for a schematic diagram of the setup used.
Lighting consisted of a continuous strip of standard white aquarium LEDs that were positioned 35 cm above the tank running lengthways across the entire tank. Opaque black plastic sheeting was used to block one side of the tank, behind which the observers monitored the fish and used a laptop to control the camera and record videos without disturbing the fish. An ELP webcam (2 megapixel USB) was used to record the videos and positioned 55 cm away from the tank such that the entire tank was visible.
The temperature gradient was established and maintained by heating the chamber at one end and cooling the chamber at the other end of the tank. Heating was achieved with a 350 W thermostat-controlled heater set to 35°C. Cooling was achieved with a pumped chiller (HAILEA HC-150A) circulating and chilling water externally and cycling it back into the chamber. We note that as in the original study, the term ‘thermal gradient’ may not be exactly appropriate, as water movement between chambers may have been restricted enough such that the chambers may be better described as discrete areas of temperature differences with limited exchange between chambers.
This study had two different treatments: (1) the ‘gradient’ treatment had a thermal gradient with a ∼15°C difference between the first and sixth chamber with temperatures per chamber ranging from ∼35°C to ∼20°C; (2) the ‘control’ treatment had no thermal gradient, each chamber had temperature of ∼28°C but differed slightly between groups and over time (Table 1). In this treatment, the large heater and the chiller were turned off.
Temperature probes (Aquarium, digital thermometer) were placed in each chamber and used to record temperatures for each chamber for the duration of each trial, and a handheld thermometer (ATP, MultiThermo) was used to regularly check the temperature at different depths within each chamber during tank setup and before and after each group was tested.
The experimental procedure followed that used by Rey et al. (2015). The tank was drained and refilled with aged water and left to reach the stable temperatures in each chamber as per the pre-randomly assigned experimental treatment – either with a gradient, or no gradient. A group of 12 fish from the same housing aquaria were netted and introduced to the experimental tank between 16:30 h and 17:45 h on the evening prior to the test and left to acclimatise for 15–16 h. In each case, the fish were introduced into the same chamber (chamber 3) of the experimental tank. The following day between 10:00 h and 11:00 h, a pre-test video of 1 min duration was recorded, to record a snapshot of the distribution of fish prior to any disturbance, which is assumed to be an estimate of chamber occupancy and preferred temperature of unstressed fish. All the fish in each group were then caught and placed in a white net (12×12×10 cm, hereafter referred to as the ‘confinement net’) placed in chamber 3. Fish were kept in the confinement net for 15 min. After this confinement period, the test phase was initiated by re-releasing all the fish into chamber 3. The test period lasted 2 h for each group starting from release with continuous video recording. We tested 8 groups of 12 fish each per treatment and used individual fish only once, these sample sizes were based on the original study performed by Rey et al. (2015) and the criticism by Key et al. (2017).
Video playback was used to record the distribution of fish within the experimental tank, with a count of number of fish in each chamber every 15 min starting immediately after the fish had been released from the confinement net, time 0, for 2 h. As per Rey et al. (2015), these samples were scan samples, but the video was paused and rewound as often as needed to get accurate numbers of fish in each chamber at each specific 15 min interval. Videos were scored by an assistant who had no knowledge of the hypothesis or treatment.
All analyses were performed using R version 3.4.2 (https://www.r-project.org/). We used two approaches to analyse the effect of treatment (temperature gradient or no temperature gradient) on the distribution of fish across chambers over time.
In the first approach, we fitted a model to incorporate the variation in distribution of fish over time using the mgcv package in R (https://cran.r-project.org/web/packages/mgcv/index.html). The response variable was a proportion (number of fish in each chamber, up to 12 fish), therefore a binomial generalised additive mixed model was used to examine variation in the distribution of fish. Treatment, time and chamber number were the fixed effects in the model, and a random term incorporating the repeated measurements within a group of fish was added. To accommodate for spatial autocorrelation between contiguous chambers, a spherical function was added with a gradient specified by the chamber's number. The variables ‘Time’ and ‘Chamber number’ were standardised to have a mean of 0 and s.d. of 1. Models were fitted by restricted estimation maximum likelihood, and diagnostic plots were examined for homogeneity of variance and residual normality. A likelihood ratio test indicated that the random effect improved model fit (P<0.001). The best model was selected based on Akaike information criterion (see Script with R code).
Given the potential temporal and spatial autocorrelation inherent to the experimental design proposed by Rey et al. (2015), a simpler approach was also considered where we used a chi-square test of independence to assess whether the proportion of fish across chambers varied between treatments. In this approach, we ignored the potential temporal and spatial autocorrelation inherent to the experimental design and tested at discrete times: 15, 30, 60 and 120 min after exposure to stress.
Experiment 2: Alternative approach with discrete thermal areas
Here, we used a simpler design to allow for freer movement of fish between the tank compartments, and reduced the number of fish used in each group in an attempt mitigate potential issues related to social attraction of larger numbers of fish (Cooper et al., 2018; Faustino et al., 2017; Graham et al., 2018; Schroeder et al., 2014; White et al., 2017). We did not test fish individually, but used three fish per group for ethical considerations as zebrafish are gregarious in nature and have a strong preference for conspecifics (Al-Imari and Gerlai, 2008; Miller and Gerlai, 2011; Schroeder et al., 2014). Different chambers were set up to have significantly different thermal properties and we examined the preference for a warmer area. We used an approach similar to that of several studies of preference, specifically other studies of zebrafish preference of between environmental conditions in captivity (Kistler et al., 2011; Schroeder et al., 2014).
The tank (100×25×25 cm) was positioned in the observatory side-room of the fish lab with the same environmental condition as described for Experiment 1 above. Gravel 0.5 cm deep covered the bottom of the tank but there was otherwise no physical structure in the tank apart from two immersion heaters and the two tank divisions.
The tank was partially divided into three areas, left, right and middle compartment; using two transparent Plexiglas partitions (4 mm thick), each partition was raised 4.5 cm above the gravel such that there was no barrier to movement between areas at that depth. The two side compartments were the areas of focal interest and were kept as identical as possible. Each side compartment was 25×20×20 cm and had a single 50 W immersion heater attached to the glass 3 cm above the gravel. See Fig. S2 for schematic diagram of the setup used.
Prior to each trial, we randomly selected one side as ‘warm’ (29°C) and the other ‘cool’ (24.5°C), each side was used as the warm area an equal number of times. The middle compartment was larger than the other two areas and left bare, with no heater, as it was intended to act as a ‘standard’ option area with a temperature of 26.5°C, close to the temperature in which the zebrafish were normally housed (26°C). Temperatures for each treatment are given in Table 2. An ELP webcam (2 megapixel USB) was situated above the tank such that the top down view of the tank and the fish could be recorded.
Prior to testing each group, the experimental tank was drained and refilled with water from a holding tank where it had been aged for at least 48 h. The immersion heaters in each side were set either on or off according to which side was designated as warm and indicator lights for each thermometer were masked. After the tank had achieved a stable thermal condition (as per temperatures shown in Table 2) three fish from the same housing tank were transferred into the middle compartment of the tank.
Once in the tank, the fish were left to acclimatise to the tank for a minimum of 15 h, after which the fish were exposed to one of two experimental treatments: (1) control – fish were left undisturbed; (2) treatment – all fish in the group were captured and confined in a 15×10 cm net within the middle compartment of the tank for 15 min before being released back into the middle compartment. Videos recorded fish movements for the following 2.5 h in both treatments. For this experiment, we tested 10 groups of 3 fish each per treatment and used individual fish only once, our similar sample sizes were based on previous studies of preference for specific areas within aquaria in small groups of zebrafish (as per Schroeder et al., 2014).
Video playback was used to record the distribution of fish within the experimental tank, with a count of number of fish in each chamber every 5 min starting at 1 min after the start of the trial for 2.5 h (a total of 31 counts per group). Counts were scan counts of fish per compartment, and the video was paused and rewound as often as needed to get accurate numbers of fish in each chamber at each specific 5 min interval. Trials were scored and fish in each compartment at each interval were counted by a hypothesis-naïve assistant.
Descriptive data analysis suggested that mean preferences for warmer water were lower in groups that were netted (treatment) compared to those groups that were not netted (see figure below and Fig. S3). A break-point analysis was used to identify distinct time periods where a shift in the preference index scores occurred. We calculated the breakpoints by fitting a loess smoothing function to the mean preference index time series for the net treatment and control group respectively. The breakpoints were calculated as the inflection points where the first-differenced smoothed preference index changed sign (Tomal and Ciborowski, 2017 preprint). This analysis was also conducted for the mean preference index plus and minus one standard deviation in order to create upper and lower bounds for time that indicates a shift in preference index. A distinct inflection point (mean, range) was found for the treatment (fish subjected to netting and confinement) at 46 min (range: 32–48 min).
We tested for differences in means in the overall time period and also in the period before and after the inflection point, set at 45 min, as the closest count interval to the mean inflection point at 46 min. As the data were not normal (Shapiro–Wilk test W=0.77514, P-value<0.001), we used the non-parametric Wilcoxon rank sum test to compare mean preferences for the warmer compartment across the two treatments.
Raw data from both experiments are presented in Table S1.
Experiment 1: Thermal gradient (avoidance or SIH?)
This experiment compared groups of fish exposed to either a thermal gradient or no gradient. The best model showed no effect of treatment on the proportion of fish across chambers (Table 3). While the model we retained contained an interaction between chamber and time (Table 4), fish stayed predominantly in the chamber where they were first released regardless of the treatment, and only over time started to move to other chambers. Our alternative analysis also showed no effect of treatment on fish distribution across time: (chi-square test of independence at 15 min (=0.66, P=0.71), 30 min (=0, P=1), 60 min (=1.08, P=0.58) and 120 min (=4.8408, P=0.09) post-stress (Fig. 1).
Experiment 2: Alternative approach with discrete thermal areas
This experiment compared groups of fish exposed to either a handling stress (net and confinement) or left undisturbed. Overall there was a minor but significant difference in preference, with netted fish spending less time in the warmer area than control fish (Wilcoxon rank sum test, W=53,800; P=0.006; Fig. 2), see also Fig. S3. This statistical difference can be attributed to behaviour during the initial 45 min of the trials, as estimated from breakpoint analysis, during this period the groups exposed to the treatment (netted and confined) showed a significantly lower preference for the warm area than control groups (Wilcoxon rank sum test, W=6519.5, P<0.001). After the first 45 min there was no difference in preference for the warmer area between treatment and control groups (Wilcoxon rank sum test, W=22,802; P=0.529).
Fish employ behavioural thermoregulation in multiple contexts including physiological and immune responses (Boltaña et al., 2013; Reynolds et al., 1976; Ward et al., 2010). In this study, we asked whether zebrafish exhibited behavioural thermoregulation as part of a SIH response, as previously reported by Rey et al. (2015). We observed no evidence for SIH; across two experiments, the stressed fish showed no preference for warmer areas.
In the first experiment, we aimed to reproduce the findings of Rey et al. (2015) and then account for potential confounding effects of chamber avoidance. Chamber avoidance may occur if the fish attempt to move away from the area in which they experienced stress, and any apparent preference for warmer water might therefore represent greater movement by stressed fish (interacting with an avoidance of colder water) rather than a specific preference for warmer water (Key et al., 2017). The key difference in our approach compared with that of Rey et al. (2015) was that all groups of fish were exposed to stress and that we performed both a thermal gradient and a no gradient treatment. Our results suggested that fish movement between chambers after stress was not affected by a thermal gradient. Additionally, there was no evidence of avoidance of the confinement chamber as predicted by Key et al. (2017) as movement away from the confinement area (chamber 3) was slow and occurred in low numbers. Regardless of treatment, occupancy remained higher in chamber 3 than any other chamber when compared with pre-stress conditions. Had we observed greater movements away from the confinement chamber, as we expected given Rey et al.’s results, we intended to run further controls with different chambers within the tank being used as confinement areas. Specifically, in this or future designs, we recommend that fish are stressed and released in each of the different chambers to ensure any change in distribution across chambers is a result of thermal preferences changing and not due to differences in the chamber or temperature effects on behaviour.
Given the results from the first experiment, we postulated that the effects of the stressor on fish movement may have been due to restricted movement between chambers (as the connecting holes were quite small) and/or the effects of social attraction between fish. Social buffering of stress response and tendency to remain with the group is especially likely given that zebrafish are highly social and been have shown to shoal in response to certain stressors (Al-Imari and Gerlai, 2008; Faustino et al., 2017; White et al., 2017). These issues may have overridden any effects of aversion to the chamber and or individual preferences for warmer water and led to us the second experiment.
In Experiment 2, despite having much smaller groups and less restricted movement, we again found no evidence for an increased preference for a warmer area in stressed fish. Indeed, our results offer evidence to the contrary: for an initial period post-stress, zebrafish show a reduced preference for warmer areas. Specifically, in the ∼45 min post-stress, stressed fish spent less time in the warmer compartment than unstressed fish did. This time period corresponds with physiological responses to stress in zebrafish, where cortisol levels peak within 10 min of stress, and remain high for about 40 min post-stress (Ramsay et al., 2009).
Behavioural assays of thermal preferences in fish can be difficult, with nuanced effects on thermoregulatory behaviour by inter-individual differences further complicated by the nature of highly social species (Al-Imari and Gerlai, 2008; Cooper et al., 2018; Faustino et al., 2017; Miller and Gerlai, 2011). Any additional studies exploring SIH may benefit from adapting methods such as those used in other studies (Cooper et al., 2018; Killen, 2014; Petersen and Steffensen, 2003) and focus on individual fish (as also suggested by Rey et al., 2015). Our results, however, are clear in that zebrafish did not show a generalised SIH response to netting and confinement; fish in our study showed no preference for warmer areas after this stress in either experiment. Our finding is consistent with earlier studies that explored the area of behavioural fever in fishes and showed no preference for warmer areas as a response to stress (Cabanac and Laberge, 1998; Reynolds et al., 1976).
Beyond the lack of evidence for SIH, our results also suggest that zebrafish may actually show a reduced preference for warmer areas after stress. While this was unexpected, given the findings of Rey et al. (2015), we believe it is in line with the many studies that focus on the stress response in zebrafish. These studies have shown that zebrafish exhibit clearly defined responses to stress with well catalogued identifiable behaviour (Kalueff et al., 2013), such as obvious changes in locomotor behaviour – erratic swimming and/or freezing – in response to stress (Egan et al., 2009) and impaired learning abilities (Gaikwad et al., 2011; Kim et al., 2009). Indeed, these responses are so well described that they are used as assays of anxiety (Cachat et al., 2010). More pertinently, stress has been shown to disrupt spatial navigation in zebrafish (Gaikwad et al., 2011) and although it is supposition at this stage, we suggest that this is the likely cause of our results. Specifically, recently stressed zebrafish may be less able to navigate to preferred thermal conditions because of disruption of normal swimming behaviour and/or some level of impairment in their memory or ability to respond to thermal cues. Exploring the mechanism(s) behind this may be a question worthy of follow-up research, but may require alternative systems for testing more precise changes in thermal preferences. Research on other ectotherms, notably frogs (Tattersall and Boutilier, 1999), however, has shown that one likely mechanism may be anapyrexia. Both endotherms and ectotherms show anapyrexia, where there is reduced body temperature after exhausting exercise and/or hypoxic conditions (Seebacher and Franklin, 2005) and in ectotherms, this results in behavioural hypothermia (Tattersall and Boutilier, 1999; Wagner et al., 1999). Zebrafish in the ‘stressed’ conditions in our experiments may have been exhibiting anapyrexia as they recovered from the capture and confinement, which may have exhausted them.
While our study is limited to laboratory studies, from an ecological perspective, evidence of SIH in fish may raise questions. Many studies have shown that a common behavioural response to the presence of predators in fish is reduced swimming activity and/or lowered metabolism (Kopack et al., 2015; Tang et al., 2017). Moving to warmer areas in response to a stressor, as is required to exhibit SIH, would be at odds with this anti-predator response in many fish.
Our study cannot support the contention that zebrafish exhibit SIH. Moreover, zebrafish may actually show an initially reduced preference for warmer areas, spending less time in warmer areas compared with undisturbed fish in periods immediately post-stress. While we are strong proponents for ensuring and improving the welfare of fish held in captivity, we argue that any claims for SIH in zebrafish, along with further claims for fish consciousness based on the current evidence for SIH (Rey et al., 2015) may need to be re-evaluated.
We thank Jack Rayner and Will Schneider for acting as hypothesis-naïve video scorers. Thanks to the St Andrews fish journal club, Anne Magurran, and Chris Todd for useful discussion. Thank you also to Luke Rendell for providing encouragement and freedom to complete the study. We are grateful to two reviewers who provided helpful comments that greatly improved the manuscript.
Conceptualization: N.A.R.J., T.M., F.B., M.M.W.; Methodology: N.A.R.J., T.M., F.B., M.M.W.; Software: N.A.R.J.; Validation: T.M., F.B.; Formal analysis: N.A.R.J., T.M., F.B.; Investigation: N.A.R.J., T.M., M.M.W.; Data curation: N.A.R.J.; Writing - original draft: N.A.R.J.; Writing - review & editing: N.A.R.J., T.M., F.B., M.M.W.; Visualization: N.A.R.J., T.M., F.B.; Supervision: M.M.W.; Project administration: M.M.W.
N.A.R.J. was supported by a studentship from The Fisheries Society of The British Isles.
The authors declare no competing or financial interests.