Women experience significant fluctuations of ovarian hormones over the menstrual cycle. Most notably, 17-β estradiol and progesterone levels change periodically (Becker et al., 2005). During the follicular phase at the beginning of the menstrual cycle estradiol and progesterone levels are low. Estradiol is rising until reaching its peak right before ovulation and abruptly decreasing with ovulation. During the luteal phase, progesterone is rising coinciding with a second yet smaller increase of estradiol, with both hormones declining during the late luteal phase reaching the initial low levels during menstruation. To prevent pregnancy and facilitate safe family planning, millions of women rely on hormonal contraceptives such as oral contraceptives (OCs) during their reproductive years (United Nations [UN], 2020). OCs typically contain ethinyl estradiol (synthetic estrogen) and progestin (synthetic progesterone) that effectively suppress endogenous estradiol and progesterone levels and thus ultimately prevent ovulation (Petitti, 2003). Evidence is accumulating that endogenous as well as synthetic ovarian hormones impact women’s socio-affective processing, including facial emotion recognition (Derntl et al., 2008a; Hamstra et al., 2014, 2015, 2017; for reviews see: Montoya and Bos, 2017; Lewis et al., 2019; Pahnke et al., 2019; Gamsakhurdashvili et al., 2021a).

For human communication, the perception and correct interpretation of facial expressions plays a major role. In a fast and direct way, facial expressions project the emotional state of a person to be perceived during social interactions (Horstmann, 2003). Among other functions, facial expressions of emotions can serve as eminent approach- or avoidance signals (Marsh et al., 2005). In naturally cycling (NC) women, several studies revealed superior facial emotion recognition in follicular compared to luteal NC women (Derntl et al., 2008a,b, 2013; Guapo et al., 2009; Rubin et al., 2011; for reviews see: Osório et al., 2018; Gamsakhurdashvili et al., 2021a). However, there are also some studies not finding any menstrual cycle effects on female’s emotion recognition (Rubinow et al., 2007; Zhang et al., 2013; Kamboj et al., 2015). These inconsistencies could potentially be explained by different levels of progesterone in the luteal NC women, as all studies that did not report a menstrual cycle effect measured women either during the early or late luteal phase in which progesterone levels are relatively lower than in the mid-luteal phase (Gamsakhurdashvili et al., 2021a). Within the follicular phase, first studies have not found a difference in emotion recognition skills between early follicular and late follicular (i.e., peri-ovulatory) NC women (Guapo et al., 2009; Zhang et al., 2013) except for fear recognition, for which women during the peri-ovulatory phase showed a better performance (Pearson and Lewis, 2005).

In some of these studies endogenous estradiol and progesterone levels were related to emotion recognition performance. Across cycle phases, estradiol was positively associated with facial recognition accuracy of fear (Pearson and Lewis, 2005) and sadness (Hamstra et al., 2017), whereas it was linked with lower performance in anger (Guapo et al., 2009) and disgust recognition (Kamboj et al., 2015). Contradictory findings were reported with respect to neutral face recognition, as it was positively linked to estradiol in one study (Hamstra et al., 2017), but negatively in another study (Shirazi et al., 2020). This incongruency could possibly be due to the inclusion of women in different cycle phases marked by different degrees of estradiol fluctuations as well as levels. In the early follicular phase, estradiol is comparatively low and stable, whereas in the peri-ovulatory phase levels are higher and rapidly fluctuating day by day. For progesterone, lower levels were linked to higher rates of misclassifying emotional faces as neutral (Derntl et al., 2008a; Kamboj et al., 2015). When including multiple cycle phases, progesterone was associated with an increased bias for negative emotions shown by higher recognition rates (Maner and Miller, 2014) as well as longer response times (Kamboj et al., 2015). However, progesterone levels have also been negatively linked with emotion recognition performance across cycle phases and specifically when only considering the luteal phase (Derntl et al., 2008a,2013). Therefore, the measurement timepoint in the luteal phase may indeed determine whether a higher sensitivity for negative emotions or a general lower face recognition rate can be detected compared to other cycle phases.

Like the midluteal phase, the hormonal milieu in OC-users is marked by a progestogen dominance as high doses of progestogens are needed to inhibit ovulation (Lovett et al., 2017). Therefore, it is not surprising that basic as well as complex facial emotion recognition was repeatedly found to be impaired in OC-users compared to NC women (Hamstra et al., 2014, 2015, 2017; Pahnke et al., 2019). These findings hold especially for negative emotions including anger, sadness, disgust (Hamstra et al., 2014, 2015, 2017). However, there are studies not reporting differences in emotion recognition performance between OC-users and NC women (Radke and Derntl, 2016; Gamsakhurdashvili et al., 2021b), including a large-scale study (n = 395; Shirazi et al., 2020). Interestingly, androgenicity of pill type seems to play no role in the impaired emotion recognition of OC-users (Pahnke et al., 2019). Regarding the modulatory role of endogenous and synthetic ovarian hormone levels not much is known, as previous studies have only measured endogenous but not exogenous ovarian hormone levels in blood or saliva samples. Since exogenous hormones pass the blood brain barrier and bind to hormone-receptors in brain regions involved with socio-emotional processing (Toffoletto et al., 2014; Barth et al., 2015; Louw-du Toit et al., 2017; Rehbein et al., 2021), including them in analyses could aid in shedding light on underlying mechanisms of facial emotion recognition during OC-use.

The aim of this study is to elucidate hormone-based differences in emotion recognition more closely by the incorporation of exogenous in addition to endogenous ovarian hormones. To assess the roles of estrogens and progestogens on facial emotion recognition largely independently, we included three groups of women with different hormonal states: (1) NC-women during the early follicular phase with low concentrations of estradiol and progesterone, (2) NC-women during the peri-ovulatory phase with high estradiol and low progesterone concentration, and (3) women actively taking combined OC-pills, with medium estrogen and high progestogen concentration. Based on previous literature on OC-and menstrual cycle-related differences, we hypothesize that: (1) OC-users show impaired emotion recognition relative to NC women (see for reviews: Osório et al., 2018; Gamsakhurdashvili et al., 2021a), and for NC women, we hypothesized that: (2) Women in the peri-ovulatory phase show enhanced fear recognition compared to early follicular NC women (Pearson and Lewis, 2005), whilst there is no evidence for an altered fear recognition in OC compared to NC women.

We aim for a systematic investigation of hormone-related effects on female’s facial recognition performance. Therefore, we ran explorative analyses with regards to – especially synthetic – ovarian hormones. In addition, affective state supposedly impacts the recognition of valence-congruent emotions but impairs performance for valence-incongruent facial expressions (Schmid and Schmid Mast, 2010). Moreover, current affective state has been associated with fluctuations of ovarian hormones (Reed et al., 2008; Ocampo Rebollar et al., 2017). To account for a possible interplay of affective state and ovarian hormones on emotion recognition performance, we not only exploratively checked for relations of affective state with emotion recognition performance in different hormonal states, but also to ovarian hormone levels.

To investigate hormone-related differences in facial emotion recognition, we used a quasi-experimental, cross-sectional study design.

Emotion recognition and other socio-emotional processes have been repeatedly suggested to be associated with fluctuations of endogenous sex hormones as well as with the intake of synthetic ovarian hormones (see reviews: Montoya and Bos, 2017; Osório et al., 2018; Lewis et al., 2019; Gamsakhurdashvili et al., 2021a). However, studies are not entirely conclusive, and the underlying mechanisms remain largely unclear. Therefore, our aim was to systematically investigate the role of hormonal status in facial emotion recognition by linking performance not only to endogenous hormones in NC women, but for the first time also to the more dominant exogenous hormone levels in OC-users. Here the use of the highly recommended LC-MS method for hormone determination is a major strength of this study. Furthermore, we investigated associations to other emotional processes which could impact emotion recognition such as negative and positive affective state.

Overall, women during the early follicular phase, independently of negative affective state differences among groups, showed specific deficits in recognizing neutral faces by misjudging neutral faces as sadness or anger. Furthermore, in a direct comparison peri-ovulatory women, as expected, recognized fearful faces significantly better than early follicular women. There were no significant group-related differences in emotion recognition response times. Endogenous and exogenous sex hormones were, not linked to overall or neutral recognition performance. During the early follicular phase low progesterone levels were linked to higher negative affective state. Notably, during the peri-ovulatory phase progesterone levels were negatively associated with positive affective state.

Contrary to our expectation and previous literature (Hamstra et al., 2014, 2015, 2017; Pahnke et al., 2019), we were not able to replicate inferior emotional face recognition performance in OC-users compared to NC women. Even though studies use the same tasks for basic (i.e., VERT as in the present study) or complex (i.e., Reading the mind in the eye task) emotion recognition, they reveal mixed results. In line with our findings, Radke and Derntl (2016) found no OC-related impairment in basic emotion recognition. Furthermore, the up-to-now largest study on hormonal contraceptives and complex emotion recognition also failed to find any significant differences (Shirazi et al., 2020). This incongruency in findings could be due to the interplay of OC-use with other potential modulatory factors. For instance, Hamstra et al. (2016) found only a significant impairment in emotion recognition relative to NC women in OC-users with a certain genotype of mineralocorticoid receptor (MR-haplotype 1/3). In the present study, we found negative affective state to play a role in neutral face recognition of OC-users. The worse their affective state was, the more likely women misclassified a neutral expression as sadness or anger (i.e., increased negativity bias). The lack of finding any significant associations between endogenous and exogenous hormone levels with emotion recognition performance in OC-users further corroborates the view that OC-related effects may be complex and mediated rather than straightforward.

Regarding menstrual cycle phases, in a direct comparison, we replicated previous findings indicating superior fear recognition in the peri-ovulatory phase compared to the early follicular phase with significantly lower endoE2 levels (Pearson and Lewis, 2005). Interestingly, this superior fear recognition does not translate into increased fear processing in peri-ovulatory women. In fact, high levels of endoE2 have been linked to enhanced fear extinction, whereas low levels (i.e., in the early follicular phase) were associated with enhanced fear conditioning (Montoya and Bos, 2017). Moreover, high levels of estradiol were previously associated with lower disgust (Kamboj et al., 2015) and anger recognition (Guapo et al., 2009), accordingly peri-ovulatory women had lower accuracy in recognizing these facial expressions than early follicular women, however, these differences were only descriptive and did not reach significance. In the early follicular group, disgust recognition accuracy was positively associated with negative affective state. This is congruent with the notion that affective state may enhance emotion recognition of valence-congruent emotions (Schmid and Schmid Mast, 2010). In OC-users and peri-ovulatory women the negative affective state may have not been pronounced and variable enough to reveal such associations.

Independent of negative affective state, early follicular women were significantly worse than OC-users and peri-ovulatory NC women in recognizing neutral faces. The neutral faces were mostly misclassified as being sad or angry instead. However, there were no significant differences in response times. Therefore, suggesting that fNC women were not aware of their difficulty in recognizing these faces correctly, as if they were uncertain, response times should be longer. In previous studies, low endoP levels were associated with a higher number of stimuli falsely classified as neutral (Derntl et al., 2008a), faster response times in correctly identifying neutral faces (Kamboj et al., 2015), and higher amygdala activation during neutral face processing (Derntl et al., 2008b). Therefore, from these studies we could have expected enhanced neutral face processing of early follicular women, as here endogenous progesterone is low. However, instead we found a greater negativity bias (i.e., misjudging neutral as negative expressions) in this group, which was however not related to ovarian hormone concentrations. This incongruency could be explained by the different menstrual cycle phases included in the studies. The previous studies (Derntl et al., 2008a,b; Kamboj et al., 2015) pooled follicular and luteal women to generate hormone correlations. Therefore, these findings could be rather driven by the inclusion of luteal women with high progesterone levels. Furthermore, the early follicular phase was largely underrepresented in the follicular groups of the previous samples, making a comparison of the previous studies with the present study difficult. Negativity biases in neutral or ambiguous face recognition have been repeatedly implicated in individuals with affective disorders, (social) anxiety and other mental problems (Richards et al., 2002; Leppänen et al., 2004; Yoon and Zinbarg, 2008; Mier et al., 2014; Münkler et al., 2015; Gutiérrez-García and Calvo, 2017; Peschard and Philippot, 2017). These biases or overinterpretations could contribute to the difficulties in social interactions and relations in these individuals. Our analyses, however, revealed no link between affective state and (social) anxiety measures with the negativity bias in the early follicular group. Considering, that the fNC women had no elevated or clinically relevant levels on these scales, these null finding may not be surprising. Nevertheless, fNC women could have felt more menstrual discomfort and pain, which is not evaluated by the PANAS or the STAI, causing a greater precaution in processing of neutral facial expression to account for their increased vulnerability. Indeed, there is evidence of negative interpretation biases associated with pain (Khatibi et al., 2015; Heathcote et al., 2016). Therefore, future studies are needed to examine more closely the possible link between menstrual pain/discomfort and negative interpretation biases.

Independent of emotion recognition, our findings support previous literature reporting a link between menstrual cycle and affective state (Reed et al., 2008; Ocampo Rebollar et al., 2017). As similarly shown by Ocampo Rebollar et al. (2017), the negative link between progesterone and positive affective state in the peri-ovulatory phase implies that pre-ovulatory women have more positive affective state which decreases as ovulation comes closer and progesterone levels start rising. In the early follicular phase, however, lower levels of progesterone were linked with worse affective state. Since progesterone levels in this phase are already low, even lower concentrations could lead to an interruption of the mood stabilizing effects of its metabolite allopregnanolone, by reducing its effect (through lower concentrations) on the GABAeric system (Chen et al., 2021). However, these findings ought to be interpreted with caution, given the generally low levels of progesterone in the follicular phase.

In this study we have only included women using OCs but excluded women using other (hormonal) contraception methods such as intrauterine devices or vaginal ring. To fully capture the impact of (hormonal) contraception on emotion recognition and more general socio-emotional abilities, future studies should systematically investigate their effect. Furthermore, our study investigated women’s emotion recognition in a cross-sectional design, comparing different women with different hormonal status once. However, a longitudinal design enabling a within-subject comparison would be beneficial to better characterize the impact of endogenous and exogenous hormones on behavioral outcomes. Additionally, statistical power could be improved this way, without really having to increase the sample size (Gonzales and Ferrer, 2016). Another downside of the sample size per hormonal status group in addition to non-linearity issues of the data was that no mediation analyses using affective state could be carried out to investigate the interplay of sex hormones and affective state on emotion recognition (minimum size per group n = 72 for medium effects; Fritz and Mackinnon, 2007). Finally, in this study we only measured emotion recognition of basic emotions. Since in real-life, emotion recognition of complex next to basic emotions plays a major role, the inclusion of complex emotions in the study design could have raised ecological validity.

With the current study we shed some light on the role of different hormonal conditions (i.e., OC-use, early follicular and peri-ovulatory phase) in emotion recognition abilities of women. Our results suggest that women in their early follicular phase show both, elevated negative affective state as well as a negativity bias in perceiving neutral faces (i.e., neutral misjudged as sadness or anger), which may impair their success in social interactions. Furthermore, in a direct comparison peri-ovulatory women showed better fear recognition accuracy. Generally, we were not able to replicate OC-related impairments in emotion recognition performance. More importantly, we also found no significant links between endogenous and exogenous hormone levels with emotion recognition, suggesting a more complex mechanism by which emotion recognition is possibly influenced by hormonal contraception. Thus, the study motivates more research to better understand how different hormonal conditions do impact women’s social life, and ultimately their mental health. A better understanding of these processes is necessary to provide gynecologists and potential users with details on potential consequences of hormonal contraceptives on female social cognition.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Ethics Committee of the Medical Faculty of the University Tübingen. The patients/participants provided their written informed consent to participate in this study.

A-CSK and BiD designed the study and supervised data collection. BiD and IS-P helped with the methodological setup. A-CSK and ADB collected data. BiD, IS-P, and JAB were involved in the planning of data analysis and interpretation of data. BeD carried out all hormone detection analyses under the supervision of ML. A-CSK performed data analyses and wrote the manuscript. All authors contributed to the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.