Male and female immediate fear reaction to white noise in a semi‐natural environment: A detailed behavioural analysis of the role of sex and oestrogen receptors

In classical rodent anxiety models, females usually display lower anxiety than males, whereas anxiety disorders are more prevalent in women. Perhaps this contradiction is caused by the use of behavioural models with low external validity. Therefore, we analysed immediate reactions to a sudden 90‐dB white noise in a semi‐natural environment. We observed mixed‐sex groups of rats for the 60 seconds preceding noise onset and the first 60 seconds of exposure. White noise elicited fear‐specific behaviours hiding alone and huddling. It also increased exploratory and ambulatory behaviours, although only in the burrow zone farthest from the open area. Thus, in a semi‐natural environment, white noise enhanced motor activity as a product of fear‐induced general arousal. Then, we compared male and female sexual, social, exploratory and anxiety‐related behaviour, and found little sex difference. This absence of behavioural effect, also observed in other studies, might be a result of our study design, a familiar environment with an ecologically relevant social context. Fear and anxiety responses are modulated by oestrogens through the activation of oestrogen receptors α and β. Thus, in a third part of out study, we analysed how treatment with either oil, oestradiol benzoate (EB), an agonist to the oestrogen receptor α (propylpyrazoletriol [PPT]) or β (diarylpropionitrile [DPN]) influenced female behaviour. The effect of treatment was limited, both EB and PPT stimulated motor activity in the open area before white noise, probably because of sexual activity. PPT increased the probability of fleeing from the noise, and decreased the latency to do so, which is consistent with a pattern of anxiogenic properties found in previous studies. Contrary to reports in classical procedures, we failed to detect any effect of DPN on immediate fear reactions in a semi‐natural environment.


| INTRODUC TI ON
Differences in male and female behaviour have been debated for decades, and resulted in the most exotic theories based on dubious evolutionary principles. 1 Biological factors (eg, gonadal hormones or sex chromosome genes) can partly explain at least some of the gender differences, and sex is a significant risk factor for neurodevelopmental and neurodegenerative disorders. 2 Indeed, a number of psychiatric troubles are distributed along a biased sex ratio, with these including anxiety and depression, which are more prevalent in women than in men. 3 In this context, it is highly necessary to understand how sex influences our health to improve patient care and treatment. The study of male subjects has prevailed so far, even though considerable efforts have been made to include females in the last decade. In particular, the USA National Institutes of Health recently requested to consider sex as a relevant biological variable in National Institute of Health applications. 4 However, much work remains to be done, notably to adapt the statistical methods to the investigation of sex differences and sex interaction with treatment, and to report results by sex, which is rarely achieved. 5 In the past, inadequate experimental designs, either ignoring female behaviour or focusing on passive rather than active reactions, or biased data collection have also hindered discoveries in female health. 6 Fear and anxiety-related behaviours are usually assessed through a battery of classical tests in rodent models. The most commonly used are the open-field test, the Vogel test, the light-dark compartment test and the elevated plus-maze test. 7 These tests present a single experimental subject with a new, anxiogenic situation, allowing for quantification of behaviours, supposedly comparable across the tests. In these tests, females often show patterns of decreased fear compared to males. For example, in an elevated plus-maze, females have been reported to show more entries into the open arms, more distance travelled and less fear-related behaviours, such as freezing or defecation. [8][9][10][11][12] In the open field test, some data show that females cover more distance and display more rearing postures. 8,13 The proposal that females may show reduced fear in these procedures, whereas women are more at risk of developing mood disorder, is rather contradictory. However, most of the behavioural patterns collected in classical tests rely on motor activity or locomotor exploration. This ignores the fact that females sometimes display higher locomotor activity than males, regardless of the environmental context. 12,14,15 Motor activity is a potentially important confounding factor in measures of emotional status. 16 In addition, the above-mentioned classical tests suppress the social component of behaviour, despite its determining nature for highly social animals such as the rat. Indeed, social interaction has a rewarding value for rats and can induce conditioned place preference. 17 A recent review of anxiety studies in rodent models highlighted the challenge of anxiety measurements, and emphasised the need for clearer definitions of the measured variables and conditions used, to achieve greater transferability. 7 This is especially relevant to the contradiction between results obtained in female rodents and the prevalence of anxiety in women. Brunswik et al. 18,19 defined procedures from which the results are generalisable to other contexts as procedures with an external validity. In sex difference research, anxiety studies would benefit from naturalistic conditions and complex social environments.
Indeed, it has been suggested that an ethological approach could increase the translational value of animal models, particularly by incorporating group-housed animals. 20 Semi-natural environments are particularly suitable for this purpose and have already been used to study fear reactions, [21][22][23] as well as sexual behaviour in both sexes. 24,25 Previous studies conducted in our laboratory have looked into the expression of fear in females rats hosted in a semi-natural environment 26 and more specifically into the differential role of oestrogen receptors in emotional responses in this environment. 27,28 Indeed, the variability in male and female fear and anxiety-related behaviours is considered to rely, at least partly, on the main female hormone oestradiol. 29 This steroid modulates behaviour differently depending on the environmental context. 30 In particular, oestrogen receptor (ER)α and ERβ, present in both male and female mammals, 31,32 have different effects on fear reactions. ERα has shown anxiogenic properties in several anxiety models. A selective ERα agonist increased defecation and time spent grooming in an elevated plus-maze. 33 By comparison, reducing the expression of this receptor in the medial preoptic area alleviated indicators of fear and anxiety in the open field and the light/dark box, 34 suggesting that the activation of this receptor is anxiogenic. In parallel, when tested in a semi-natural environment, females with a reduced number of ERα in the ventromedial nucleus of the hypothalamus showed almost no huddling behaviour when exposed to aversive white noise, and they recovered fast from white noise exposure. 28 By contrast, activation of ERβ has consistently led to anxiolytic effects in the elevated plus-maze 33,[35][36][37] and in the open field. 33 In a semi-natural environment, females treated with an ERβ agonist showed a distinct profile in response to aversive situations, 27 whereas females with a reduced number of ERβ in the central amygdala showed a pattern of increased anxiety, including increased risk assessment and decreased food consumption. 28 Thus, oestradiol plays an important role in the modulation of fear and anxiety reactions in females, through the differential activation of ERα and ERβ, which could partly explain the sex difference in anxiety prevalence. Even though there are data available showing that ER agonists modulate anxiety responses in males, 33,35 we limited the present study to an evaluation of their role in females.
Most classical anxiety tests present the experimental subject with an anxiogenic situation but not with a discrete external, fearful stimulus. To this effect, we decided to use white noise, a widely used stressor in anxiety studies. Even though it is not a standard part of the rat natural habitat, loud noise is part of the anthropogenic disturbances that can be faced by urban animals, such as the rat. Experiments previously conducted in our laboratory showed that white noise was highly aversive to the rats, efficiently eliciting classical fear-and anxiety-related behaviours. [26][27][28] These and other anxiety studies analysed behaviour expressed over the entire duration of the test (ie, sustained anxiety). Immediate fear and anxiety reactions (ie, phasic anxiety) might be more informative and are worthy of special attention. There is evidence showing that phasic and sustained anxiety responses depend on different neural systems 38,39 and that they are differently modulated by drugs. 40 Because our earlier studies of the role of oestrogen receptors in fear and anxiety responses were limited to sustained anxiety, we aimed to analyse their importance in phasic anxiety. Furthermore, the fear responses of males were ignored in the earlier studies. Here, we also report data from males. Based on video recordings from a previous experiment, 27 we made a detailed ethological analysis of immediate behavioural reactions of multi-male, multi-female groups of rats housed in a semi-natural environment. Detailed analyses of the spatial distribution of behavioural activity were also made. In typical anxiety tests such as the elevated plus-maze, the dark/light choice procedure or the open field, the differential use of space is used as an indicator of fear or anxiety. 41 Therefore, we also determined the localisation of each behavioural activity in the semi-natural environment.
We focused on the 60 seconds preceding the onset of a 90-dB white noise, as well as the first 60 seconds of exposure to it.
Ovariectomised females were administered oestradiol or a selective ERα or ERβ agonist. Because white noise can be expected to induce fear, and since the ERα has been reported to be anxiogenic in such situations, we predicted that an ERα agonist would enhance fear reactions. Because the ERβ is generally believed to be anxiolytic, we predicted that an ERβ agonist would reduce fear responses. The effects of oestradiol, acting on both receptors, were difficult to predict. The male subjects were left intact. Indeed, there is evidence showing that conditioned fear responses are not altered by castration. 42 It may be assumed that this also is the case for unconditioned fear.
In the present study, we carefully examine noise-, sex-and treatment-effects on behaviour. The results will provide a better understanding of sex differences and the relative contribution of ERs in phasic anxiety responses, in a procedure with external validity.

| Subjects
Forty female and 30 male Wistar rats (mean ± SEM weight was 278.7 ± 2.7 g and 352.9 ± 4.8 g at the beginning of the experiment, respectively) were obtained from Charles River WIGA (Sulzfeld, Germany). Females were ovariectomised under isofluorane anaesthesia within 15 days after arrival, and 14 days prior to the beginning of the experiment, in accordance with the established surgical procedure. 43 Rats were housed in same-sex pairs in standard Makrolon® IV cages (Tecniplast, Buguggiate, Italy) from their arrival to their introduction into the semi-natural environment (ie, for approximately 30 days). During this period, water and food (RM1; Special Diets Services, Witham, Essex, UK) were available ad lib. The temperature was maintained at 21 ± 1°C and the relative humidity at 55 ± 10%. Lights were set on a reversed 12:12 hour light/dark photocycle (lights on 11.00 pm). The ventilation system in the animal facility produced an ambient noise of approximately 40 dB. All experimental procedures employed in the present experiment were approved by the Norwegian Food Safety Authority and were in agreement with the European Union council directive 2010/63/EU.

| Apparatus
The semi-natural environment used in this study has been described in detail earlier. 24,44,45 Rats typically live in burrow systems surrounded by a large area described as the home range. 46,47 To approximate the natural conditions, we provided the rats with a complex Rats aggregate in multi-male, multi-female colonies, with a smaller proportion of male members than female ones. 46,48 Thus, groups of four females and three males were hosted in the semi-natural environment, allowing for the expression of a large range of social behaviours.

| Treatment and hormones
To evaluate the potential role of oestrogens in female fear reactions, we employed four groups of ovariectomised females: one treated with oil only (ie, no stimulation of oestrogen receptors). Another group was given EB in a dose sufficient for inducing full behavioural oestrus, simulating the oestrous phase in intact females. A third group was given an agonist selective for the ERα and a fourth group was given an agonist selective for the ERβ. In this way, we could compare females in a state similar to diestrus (oil treated group) with females in a state similar to oestrus (EB treated group), In addition, we could determine the possible role of each of the ERs. It should be noted that all groups received progesterone, which is an important part of the endocrine environment in natural oestrus. Progesterone by itself may have actions on general activity, fear responses and other behaviours. 29,49 By treating all groups with progesterone, we eliminated, or reduced, the confound between oestradiol and progesterone effects that otherwise would have occurred.
Oestradiol benzoate (EB) and progesterone (P) (both from Sigma-Aldrich, St Louis, MO, USA) were administered s.c. at a dose of 18 μg kg -1 and 1 mg per rat, respectively. The hormones were dissolved in peanut oil (Den norske Eterfabrikk, Oslo, Norway), with an injection volume of 1 mL kg -1 for EB and 0.2 mL per rat for P.
The oestrogen receptor agonists propylpyrazoletriol (PPT; ERα) and diarylpropionitrile (DPN; ERβ) were obtained from Tocris Bioscience (St Louis, MO, USA). PPT is selective to ERα, with a 410-fold preference compared to ERβ, and with a relative binding affinity of 50% compared to oestradiol. 50 DPN is selective to the ERβ, with a 72-fold preference compared to ERα, and with a relative binding affinity of 18%. 51 PPT and DPN reach their maximum serum concentration approximately 30 minutes after s.c. injection and have a half-life of 6.0 ± 0.03 hours and 8.2 ± 1.7 hours, respectively. 52 Both PPT and DPN were dissolved in undiluted dimethyl sulphoxide (DMSO; Sigma-Aldrich) right before s.c. injection, and were administered at a dose of 10 mg kg -1 body weight in a volume of 1 mL kg -1 , on two consecutive days. The rationale for using the agonists at these doses is elsewhere. 27 The injection did not cause any significant necrosis at the injection site. The acute toxicity of DMSO has been reported to be low, 53,54 and adverse effects are found only at doses far superior to the amount administered here. Undiluted DMSO was also used in an earlier study on the effects of ER agonists on sexual behaviours, and no difference between DMSO and sesame oil vehicle was reported. 55 Therefore, we did not consider it justified to add an additional vehicle group to control for unlikely effects of DMSO.

| Procedure
The floor of the semi-natural environment was disinfected and cov-

| Design
Ten groups of seven rats (three males and four females) unknown to each other before the experiment were run in the semi-natural

Non-social and maintenance behaviors
Resting alone; f,d Rat rests, laying at a distance longer than one body length to another rat Eating food; f.d Self-explanatory Self-grooming and scratching; f,d Self explanatory

Fear-and anxiety-related behaviors
Hide alone; f,d, a Rat lays still with head down and legs under its body in a corner or a nest box, at a distance longer than one body length to another rat Rat lays still with head down and legs under its body in a corner or a nest box, at a distance shorter than one body length to another hiding rat. Several rats can hide together in a stack Note: This behavior is also described as 'risk assessment' in our previous studies.
Abbreviations: f, frequency; d, duration; l, latency; o, occurrence. a Behavior appears only after the onset of white noise. Behaviors in italics were rarely observed, thus not included in the statistical analysis.
environment. The video recording started when the animals were introduced at 1.00 pm on day 0. Recording was then continuous for a period of 8 days, when the experiment was terminated. The rats were left undisturbed for 5 days. On day 5, females were injected with either EB, PPT, DPN or peanut oil at 9.00 am. On day 6, the treatment was repeated at the same time, with the exception of the females having received EB the previous day who got administered oil. On day 7, all females received P at 9.00 am ( Table 1). The males remained untreated. The noise started on day 7 at 4.55 pm and stopped 15 minutes later. The behaviours analysed here were recorded during the minute preceding white noise onset and the first minute following it.

| Behavioural observations
We used the observer xt, version 12.5 (Noldus, Wageningen, The Netherlands) for behavioural scoring by an observer blind to treatment. We used a refined ethogram based on that used in a previous study, improved with detailed exploratory and fear-related behaviours ( Table 2). The frequency and duration of each behaviour pattern was recorded. This made it possible to calculate the mean duration of each behavioural episode. For each behaviour, we specified the individual initiating it, the individual to whom the behaviour was directed when relevant, and the zone of the semi-natural environment in which the behaviour was performed.

| Data preparation and statistical analysis
For the evaluation of the effects of white noise and for the sex comparison, the four experimental female groups were collapsed into one female group, which we compared with the male group. When data satisfied criteria for parametric analysis according to Shapiro-Wilk's test, we used two-way ANOVA for repeated measures on one factor. The between factor was sex (male or female) and the within factor was noise exposure (before or during). Post-hoc tests were not necessary because both factors had only two modalities.
When data did not satisfy criteria for parametric analysis, noise effect was analysed with Wilcoxon tests, with both sexes collapsed.
When a noise effect was detected, we proceed to analyse the effect of noise for each sex separately with Wilcoxon tests. The ob- Probability to flee from the noise at its onset was analysed with binomial tests.
In addition, we analysed how sex and noise affected the localisation of behavioural activity in the semi-natural environment. First, we grouped the observed behaviours in six categories according to our ethogram ( Table 2): anxiety-related, exploratory, non-social, pro-social, anti-social and sexual behaviours. The sum of the duration of the behaviours included in each category was determined for each of the seven zones in the semi-natural environment ( Figure 1B). The six categories were then used as dependent variables in one-factor non-parametric multivariate analyses of variance (nparMANOVA). 56 The factor was sex. Each zone was analysed separately, before and during noise. In all these tests, an F approximation was used to de- , where x is the number of multiple comparisons accounted for and P u is the uncorrected P value.

| Co-occurrence analysis
Chronological scoring of behavioural activity allowed for the visualisation of clusters of temporally associated behaviours, and therefore how experimental manipulations might have altered the structure of behaviour. This was achieved via an analysis of co-occurrence. This method has been described earlier. 26,28 We used a moving window of four behaviour patterns and determined how often one behaviour pattern occurred together with another in the same window. This is defined as a co-occurrence.
The window moved, by steps of one behaviour pattern, over the entire individual record. Treatment or sex and noise condition (before or during) were also included in the matrix. Descending hierarchical classification was used to identify clusters of related behaviour. 57,58 The descending hierarchical classification is based on the probability for an item to be proportionally more present in a cluster than it is in the entire data set, as evaluated by chi-squared analysis. Each item is permutated from one cluster to the other to test the robustness of the classification, until statistically independent profiles of items appear. 59 Clusters can therefore be interpreted as groups of individuals and behaviours significantly more co-occurring together than with items of another cluster, as visualised using the Fruchterman-Reingold algorithm. 60 The criterion for including elements in their respective cluster is a higher frequency of co-occurrence compared to the average oc-

| White noise immediate effect on male and female behaviour
Here, we only present statistical data concerning the effects of noise. Comparison between sexes and interactions between sex and noise are reported subsequently.

| Pro-and anti-social behaviours
Resting with other rats only occurred before white noise onset (data not shown). Exposure to white noise had no effect on the frequency of any rat sniffing male conspecifics (Z = 0.783, P = .434).
Looking at the effect of noise in each sex, exposure to white noise increased the frequency of female sniffing male conspecific (Z = 2.546, P = .022, Bonferroni correction 2 , P u = .011), although noise had no effect on males sniffing other males (Z = 1.485, P = .276, Bonferroni correction 2 , P u = .138) (Figure 2A). Exposure to white noise increased the frequency of any rat sniffing a female conspecific (Z = 2.139, P = .032). This did not appear when looking at each sex separately: exposure to white noise had no effect

| Fear-related and non-social behaviours
The frequency of alertness posture was higher during white noise exposure (F 1,68 = 43.614, P < .001) ( Figure 4A). The behaviours hiding alone and huddling only appeared during white noise. By contrast, resting alone only occurred prior to white noise ( Figure 4B). Finally, rats showed more self-grooming before white noise onset than after (Z = 2.496, P = .013) ( Figure 4C).

| Prosocial behaviours
There was no sex difference in the frequency of resting with another rat before white noise onset (t 68

| Antisocial behaviours
There was no sex difference with regard to the frequency of noseoff directed to females, neither before (U = 582, P = 1, Bonferroni

| Exploratory behaviours, locomotion and activity spatial distribution
There was no sex effect on the frequency of sniffing the floor Sex did not influence the frequency of walking (F 1,68 = 0.007, P = .931) and we found no interaction between sex and noise exposure for this behaviour (F 1,68 = 2.208, P = .142) ( Figure 2D). Similarly, we found no effect of sex on running, neither before (U = 515, P = .222, Bonferroni correction 2 , P u = .111), nor during white noise (U = 514, P = .562, Bonferroni correction 2 , P u = .281) ( Figure 2E).

| Fear-related behaviours
Male and female probability to flee from the noise at its onset did not differ (Binomial test, P = .468). The latency to flee from the noise was no different between males and females (Z = 0.688, P = .491) (data not shown). There was no main effect of sex on the frequency of alertness posture (F 1,68 = 0.784, P = .379), and no interaction between sex and noise (F 1,68 = 0.241, P = .625) ( Figure 4A). Hiding alone was displayed by males more often than by females (Z = 2.064, P = .039) ( Figure 4B). Sex had no effect on huddling frequency (t 68 = 0, P = 1) ( Figure 4B).

| Localisation of male and female behavioural activity
Before exposure to white noise, we found an effect of sex on the time spent displaying the six different behavioural categories in the lower burrows (F 3.37, 224.16 = 2.291, P = .019). Univariate analyses revealed that males displayed more prosocial behaviours than females in this zone (U = 810.5, P = .019, Bonferroni correction 6 , P u = .003).
This was associated with a 67.54% probability for this behavioural category to be displayed by a male in this area compared to randomly selected behavioural categories by either sex, according to the relative effects reported by the nparMANOVA. No sex difference appeared in the other behavioural categories (all P = 1, Bonferroni correction 6 , all P u > .225). The nparMANOVA comparing sex differences among behavioural categories was non-significant in all other zones before exposure to white noise (all P > .271) ( Figure 5A).
Finally, during exposure to white noise, multivariate analysis reported a sex effect in the nest boxes (F 1.81, 120.17 = 4.715, P = .013) but not in the other zones (all P > .080). However, univariate analyses failed to detect any significant sex effect on the behavioural categories displayed in the nest boxes (all P > .136, Bonferroni correction 6 , P u > .023), even though relative effects reported that exploratory behaviours occurring in the nest boxes had a 62.08% probability of being displayed by a female compared to randomly selected behavioural categories by either sex (Figure 5B).

| Co-occurrence analysis
Male and female rats appeared in two different clusters before white noise onset. Males were associated with most exploratory and ambulatory behaviours, with prosocial behaviours as well as with self-grooming. Female rats were associated with all anti-social behaviours, resting behaviours, and with the alertness posture ( Figure 6A).
During exposure to white noise, male rats appeared in a distinct cluster only including sniffing female conspecifics and nose-off to other males. The cluster of behaviours associated with female rats showed a more extensive behavioural repertoire, including other pro-and anti-social behaviours, and all exploratory, ambulatory and fear-related behaviours ( Figure 6B).

| Pro-and antisocial behaviours
During exposure to white noise, the frequency of sniffing a male conspecific differed between the treatments (χ 2 = 8.101, df = 3, P = .044). Females treated with PPT sniffed males more frequently than those treated with oil (P = .016) and EB (P = .011) ( Figure 7A).
We did not observe any other difference between treatment groups in social behaviours before or during white noise (all P > .392).
Additionally, female treatment did not affect the frequency of being sniffed by other rats (all P > .505), nor that of receiving nose-off (all P > .533) (data not shown).

| Exploratory behaviours, locomotion and spatial distribution of activity
There was no difference between the treatments in the total number of transitions between zones of the semi-natural environment No other difference between treatment groups was found in exploratory behaviours and activity spatial distribution (all P > .076).

| Fear-related behaviours
Only PPT-treated females showed a high probability to flee from the noise at its onset (Binomial test, P = .019); other treatment groups did not differ from the mean flight probability (all P > .227). In addition, the latency to flee from the noise was different between the groups (Χ 2 = 9.064, df = 3, P = .028). PPT-treated females had a shorter latency to flee from the noise than Oil-(P = .004), EB-(P = .021) and DPN-treated females (P = .023). Other groups did not differ from each other (all P > .490) ( Figure 7D). We found no other treatment effect on fear-related behaviours (all P > .392).

| Other behaviours
There was no difference between the treatment groups for antisocial behaviours (all P > 0.076) and non-social behaviours (all P > .426) (data not shown).

| Localisation of female behavioural activity
The nparMANOVA used to compare treatment effects within each of the seven zones of the semi-natural environment before exposure to white noise reported significant differences between treatments in the lower burrows (F 10.89, 130.65 = 2.291, P = .014). However, univariate tests did not show any significant effect of treatment within any behavioural category (all P > .141, Bonferroni correction 6 , P u > .024). The nparMANOVA was not significant for any other zone (all P > .058) ( Figure 8A). During exposure to white noise, multivariate analysis did not find any significant treatment effect in any of the zones of the semi-natural environment (all P > .266) ( Figure 8B).

| Co-occurrence analysis
Before white noise onset, females treated with Oil, PPT and DPN appeared in the same cluster associated with most pro-and anti-social behaviours, as well as with resting alone. Females treated with EB formed a separate cluster including all exploratory and fear-related behaviours, nose-off to males and self-grooming ( Figure 9A). During exposure to white noise, females treated with EB remained in a separate cluster including the exploratory behaviour 'sniffing the floor', and nose-off to males. The PPT group also formed a distinct cluster, associated with fear-related and exploratory behaviours. Finally, the Oil and DPN groups belonged to the same cluster with the fear-related behaviour 'hiding alone' and prosocial behaviours directed to males and females ( Figure 9B).

| D ISCUSS I ON
The effects of white noise and the sex comparisons are summarised in Table 3.  that there is no habituation to the aversive stimulus during an exposure of that length. It may also be noted that some behaviours, such as huddling, returned to pre-noise levels within 1 minute after noise offset.

| Immediate reaction to white noise
Visits to the open area recovered more slowly, with recovery needing more than 5 minutes. 28 The latter observations indicate that the white noise did not cause a lasting fear or anxiety reaction. procedures. It is also noteworthy that white noise did not produce any freezing response in our procedure. Indeed, freezing was so unusual that it could not even be analysed. In traditional procedures, freezing is a prominent response to white noise. 68 Similarly, fox odour causes freezing in several tests 69,70 but not in the semi-natural environment. 26 We have suggested that phenomena such as social buffering 71 or a sense of controllability, 72,73 present in the semi-natural environment but absent in traditional tests, can explain the low incidence of freezing in the former. Not finding an expected response may be as informative as finding an unexpected response. In this particular case, it shows that a fear or stress response depends on the social or physical context. Because of these and similar observations, we propose that data from the semi-natural environment have larger generalisability to natural contexts than data from other procedures.

| Sex difference in immediate fear reactions
Both before and during the noise, there were few sex differences ( Table 3). One of the few differences was that the males sniffed females more than the females did before the noise. Curiously, they also sniffed the other males more than the females did. It appears that the males were more sociable than the females. This coincides with earlier data from the social interaction test. 74  before noise exposure is similar to that observed in completely different procedures. During the noise, males preferred sniffing females, whereas females showed no preference for sniffing a particular sex. Similarly, the females hid less alone than the males did. It thus appears that, with these exceptions, males and females behave in a similar way when exposed to white noise in the seminatural environment.

The similarity between the sexes in behavioural responses
to fear coincides with the similarity in the endocrine response.
Although no data are available from the semi-natural environment, corticosterone and adrenocorticotrophic hormone are released after white noise of approximately the same intensity as used in the present study in both sexes. [77][78][79] Unfortunately, males and females were evaluated in different studies, making direct sex comparisons impossible, although it is evident that both sexes show a robust, endocrine stress response when exposed to white noise. In this context, it is important to note that there are sex differences with regard to the regulation of the corticotrophin-releasing factor response to stress, which may lead to increased stress sensitivity in females.80 However, even though such differences are likely, they do not appear to alter the immediate response to white noise. Indeed, the present data, together with earlier studies of sex differences, suggest that such differences are much more prominent with regard to sustained stress than to phasic stress.  Note: Behavioural expression was measured in frequency ( F ) or in duration in seconds ( S ). When possible, the effects of sex and noise were analysed by a two-ways ANOVA for repeated measures on one factor. Otherwise, the effect of sex was analysed by Mann-Whitney tests, and that of noise by Wilcoxon tests. In the case of a significant sex effect, the effect of noise was analysed in each sex separately; otherwise, both sexes were collapsed in the analysis. Sex differences are indicated in bold. Noise effect is indicated by up and down arrows representing increased (up) and decreased (down) display of the examined behaviour, compared with the period preceding noise. Noise effect on female (⇓ ⇑) and male (↑ ↓) behavioural expression. ND = behaviour pattern not displayed. Any behaviours that are not here did not show any noise or sex effect.
test procedures and novel statistical analyses are required before any conclusion can be reached. 81 Another review also found many inconsistent observations. In some tests, females appeared more reactive than males to anxiety-provoking situations; in others, they were less reactive than males. 82 The lack of consensus concerning possible sex differences appears to persist. One recent study in male and female rats illustrates this. 12 The results showed that females no sex difference with regard to approach to a conspecific confined in a cage. 12 This differs from the reliable sex difference found when direct physical interaction is possible (see above). It appears that sex differences often are limited to specific tests. One reason for the persistent confusion may be the use of tests lacking external validity.
Only the future will tell us whether the semi-natural environment can offer more consistent data.

| Effect of ERs on female immediate fear reactions
Several of the neural responses to stress are modulated by oestro- Oestrogens also affect serotonergic functions 84  The limited effects of the ER agonists (and of EB itself) on the responses to white noise may appear to contradict the many earlier studies reporting their anxiolytic or anxiogenic effects. However, the effects of ERs on anxiety are complex, with the ERα considered to be anxiogenic in certain contexts, whereas the ERβ is always anxiolytic. 86 Because endogenous oestrogens are acting at both receptors simultaneously, it is extremely difficult to predict the net effect of oestrogen actions. It is possible that the many effects of the administration of selective ER agonists are purely pharmacological. This notion is supported by recent data. A carefully conducted study in male and female rats failed to detect any effect of the oestrus cycle on behaviour in the elevated plus-maze, in the open field test, or in the social interaction test. 12 These data would certainly speak against any functionally significant role of oestrogens for the behaviour displayed in these tests. This conclusion is reinforced by data from a study performed in male and female mice lacking either the ERα or the ERβ. Neither of these mice were different from wild-type in the open field, light/dark choice test or in the elevated plus-maze. 87 The studies outlined above strengthen the notion that it is difficult to formulate founded hypotheses concerning anxiolytic and anxiogenic effect of oestrogens. However, in our previous study 27 of sustained fear or anxiety during noise exposure, we found that PPT enhanced the probability for escape from the noise and reduced the latency to escape. Furthermore, in the co-occurrence analysis, PPT formed a separate cluster associated with fear-related behaviours. This was interpreted as an anxiogenic effect. 27 Also, in the present study, we found that PPT formed a cluster separate from oil and DPN during but not before noise exposure, and that the behaviours in the PPT cluster were mostly related to a fear reaction. It appears that the ERα agonist heightened fear responses already during the first min of noise (present study) and that these responses persisted during the entire exposure (ie, during sustained fear or anxiety). 27 The entirely negative results obtained in previous studies 12,87 are difficult to explain. They do not coincide with the results of either the present study or those of our previous studies in which we also found an anxiogenic effect of the ERα in fear-inducing contexts. 27,28,88 We propose that the semi-natural environment is more appropriate for detecting subtle effects than the classical anxiety tests.

| CON CLUS IONS
One of the essential elements in the present study is that we evaluated behaviour displayed in response to a sudden, aversive stimulus in a familiar environment, in rats living in a mixed sex group. In the most commonly used procedures for studying anxiety, the experimental subject is introduced into a novel, often aversive, situation.
Thus, reactions to novelty are superimposed on possible reactions of fear. Furthermore, the experimental subject is tested alone, whereas it is known that rats are gregarious, and that group-living is an integral part of their natural habitat.
The importance of these essential elements is not known, although their presence should assure external validity in the brunswikian sense, whereas their absence should reduce that validity, making generalisations between experimental procedures risky and translational relevance limited.
The immediate responses to an aversive stimulus, or phasic anxiety, are similar in male and female rats. Moreover, these immediate responses are similar to those recorded during a long period of noise exposure. Thus, there is no habituation to the aversive stimulus. It has been speculated that phasic and sustained anxiety have different neurobiological bases. Phasic anxiety should be mediated by the central nucleus of the amygdala, whereas sustained anxiety is assumed to be mediated by the bed nucleus of the stria terminalis. 38 This may well be the case for anxiety produced in other procedures, particularly conditioned anxiety or fear responses, 39 although it does not appear to apply with respect to the response to a strongly aversive stimulus in a familiar environment. It is quite unlikely that different neural systems should provoke highly similar behavioural responses. Indeed, the different functions of the central nucleus of the amygdala and the bed nucleus of the stria terminalis in fear and anxiety reactions has been questioned. 89 The present data do not contradict this proposal.
The anxiogenic action of the ERα was confirmed, whereas the purported anxiolytic action of the ERβ failed to appear. Indeed,

CO N FLI C T O F I NTE R E S T S
The authors declare that they have no conflicts of interest.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/jne.12902.

DATA AVA I L A B I L I T Y
The data that support the findings of this study are available from the corresponding author upon reasonable request.