Hormonal and behavioral responses to an infant simulator in women with and without children

Abstract We investigated the impact of maternal status on hormonal reactivity and behavioral responses to an infant simulator in 117 women (54 primiparous, 63 nulliparous). The amount of affectionate touch and motherese were analyzed as behavioral measures of caregiving. Saliva was collected before and 10 min after interaction with the infant simulator to analyze oxytocin, testosterone, cortisol, and estradiol levels. Nulliparous women also provided information about their fertility motivation. Linear mixed models indicated that greater use of affectionate touch was associated with lower overall testosterone levels. Cortisol decreased in response to the interaction in both groups. In the primiparous group, the amount of affectionate touch associated inversely with cortisol levels, whereas in the nulliparous group such association was not found. Oxytocin or estradiol reactivity to the simulator did not differ between the groups, nor were these hormones associated with behavior. Higher fertility motivation in nulliparous women was related to more motherese, and lower testosterone levels. Our results indicate that the simulator elicits hormonal reactivity both in mothers and nonmothers, but the patterns of associations between caregiving behavior and hormonal levels may be partially different. These results encourage using the infant simulator to explore hormonal processes related to the transition to parenthood.

parental sensitivity Glynn et al., 2016). Currently, there are no studies comparing caretaking behavior between new parents and nonparents or examining whether caretaking behavior is similarly associated with hormone levels in these two groups.
Comparing parents and nonparents directly is important for understanding whether hormonal influences on caretaking behavior (or vice versa) might change across the transition to parenthood. In this study, we examine the behavioral correlates of oxytocin, testosterone, cortisol, and estradiol in primiparous and nulliparous women while they are in interaction with an infant simulator.

Oxytocin
The neuropeptide oxytocin is produced and segregated in the hypothalamus (Brunton & Russell, 2008) and it is essential for contractions during labor (Blanks & Thornton, 2003) and lactation (Augustine et al., 2018). Animal studies indicate that oxytocin is critical for the onset of maternal behavior (Rilling & Young, 2014). For example, in rat dams, oxytocin augments approach motivation toward pups through the dopamine network, strengthening maternal behavior (Rilling, 2013).
In humans, plasma oxytocin levels may increase already when women are in a relationship . Levine et al. (2007) observed that during pregnancy, some mothers displayed an increase in their plasma oxytocin levels, whereas others showed decreasing or constant levels throughout pregnancy. Increasing oxytocin levels during pregnancy also correlated with self-reported prenatal bonding with the fetus . After pregnancy, both maternal and paternal oxytocin levels increased similarly during the first six postnatal months . Higher oxytocin levels in maternal saliva have been associated with more positive behavior toward their 4-to 6-month-old infants (e.g., gaze toward infant, positive affect, and infant-directed speech) and higher motherinfant interaction synchrony (Feldman et al., 2011). Higher maternal plasma oxytocin levels have also been associated with more affectionate touch and eye contact when mothers engage in face-to-face interaction with their infants . Similarly, Gordon et al. (2010) noticed that higher maternal plasma oxytocin levels were associated with affectionate parenting behaviors ("motherese" vocalizations, expression of positive affect, and affectionate touch) when assessed during the first six postpartum months. No studies, however, have directly compared oxytocin levels in parents and nonparents or investigated whether peripheral oxytocin levels are associated with behavior toward infants also in nonparents. Investigating associations between oxytocin and maternal behavior in both mothers and nonmothers is important for a better understanding of whether oxytocin may be associated with caretaking behavior already before the transition to parenthood.

Estradiol
Together with progesterone, the ovarian steroid estradiol prepares the uterus for pregnancy and estradiol levels increase throughout pregnancy, peaking just before birth (Edelstein et al., 2015;Fleming, Ruble, et al., 1997;Glynn et al., 2016). Estradiol levels decrease rapidly in the postnatal period in women (Fleming, Ruble, et al., 1997). In addition to reproductive functions, estradiol is associated with individual differences in responses to emotional intimacy (Edelstein et al., 2010), but the effects of estradiol on parent-infant interaction or maternal behavior have been very little studied. One study observed that mothers with lower estradiol increase during pregnancy showed more sensitive parenting behavior 1-year postpartum (Glynn et al., 2016).
Relatedly, women with a lower prenatal estradiol increase were rated as providing more spousal support by their partner (Edelstein et al., 2017). The mechanisms underlying the somewhat paradoxical negative relations between prenatal estradiol levels and postnatal parenting outcomes are not clear, but it is possible that they are related to interactions between estradiol and other hormones such as testosterone . However, the preliminary findings are partially in line with studies on nonhuman primates, which have associated higher estradiol levels during pregnancy with less optimal maternal behavior toward offspring (Fite & French, 2000;French et al., 2004).

Cortisol
Having an important role in fetal maturation, maternal cortisol levels increase during pregnancy, remain elevated for the first postnatal weeks, and then decline to their original levels both in plasma (Fleming, Steiner, et al., 1997) and urine samples (Conde & Figueiredo, 2014).
Furthermore, cortisol levels are higher in primiparous than multiparous women (Bleker et al., 2017;Conde & Figueiredo, 2014) and this difference is partly mediated by pregnancy-specific distress in primiparas who tend to experience more stress symptoms through pregnancy than multiparas (Gillespie et al., 2018).
Higher salivary cortisol levels have been linked to higher sensibility to the own infant's body odor (Fleming, Steiner, et al., 1997) and higher self-reported sympathy to crying infant stimuli in postpartum women (Stallings et al., 2001). On the other hand, higher salivary cortisol levels, measured both cross-sectionally and longitudinally, have been associated with lower maternal sensitivity (Finegood et al., 2016;Gonzalez et al., 2012). In addition, lower salivary cortisol levels have been associated with less intrusive parenting behavior in mothers of 6-month-old infants (Mills-Koonce et al., 2009

Testosterone
Testosterone research has mainly focused on men, with results showing decreasing testosterone levels during the transition to fatherhood.
However, testosterone has an important role in female parenting as well, although the associations between testosterone and parenting might be different in women. According to the Challenge hypothesis, decreased testosterone levels facilitate investment in family life and sensitive parenting behaviors in men (Archer, 2006;Gettler, et al., 2011;Meijer et al., 2019). In line with these findings, higher testosterone levels have been associated with lower reproductive ambition in young women (Deady et al., 2006). Testosterone levels change during the prenatal period, but in a sexually dimorphic manner: testosterone levels increase during pregnancy in mothers and decrease in fathers (Edelstein et al., 2015). This suggests that results from male samples cannot be generalized to women.
Postnatally, both fathers and mothers have lower testosterone levels in general than nonparents (Barrett et al., 2013;Fleming et al., 2002;Grebe et al., 2019;Meijer et al., 2019) and lower testosterone levels are associated with better relationship quality in both men and women (Edelstein et al., 2017). In line with the Challenge hypothesis, Weisman et al. (2014) found that higher baseline testosterone is associated with lower paternal sensitivity (gaze, touch, infant-directed speech). Similarly, Fleming et al. (2002) observed that both fathers and nonfathers with lower testosterone levels expressed a higher need to respond to infant cries than men with higher testosterone levels. In women, higher salivary testosterone levels have been associated with their motivation to view cute infant faces (Hahn et al., 2015). Furthermore, the dual-hormone hypothesis suggests that the association of testosterone with behavior depends on cortisol levels: higher cortisol levels may inhibit the effects of testosterone on aggression or dominance (Mehta & Josephs, 2010). The same effect can be found in men's self-reported empathy: when basal cortisol is low, high testosterone levels predict lower empathy (Zilioli et al., 2015). However, the moderating effects of cortisol on testosterone have so far been studied mainly in men (but see Voorthuis et al., 2019).

Hormonal reactivity
Research on hormonal associations with parenting has largely focused on hormonal baseline levels. Nonetheless, hormonal levels also show short-term reactivity and the associations of such reactivity with parenting behavior are an important research target. Trajectories of hormonal baseline levels through pregnancy are well established but whether short-term hormonal reactivity to infant stimuli changes across the transition to parenthood is not known. Hormonal reactivity (measured from saliva or plasma) has been observed in response to various triggers, such as exercise (Copeland et al., 2002;de Jong et al., 2015), stressful situations (Cox et al., 2015;de Jong et al., 2015), massage (Carter et al., 2007;Riem et al., 2017), and breastfeeding (Grewen et al., 2010). Importantly, hormonal reactivity also occurs in response to infant stimuli. For example, fathers' salivary cortisol levels decreased after interacting with their children (Gettler et al., 2011) and listening to infant cry sounds has been associated with increases in oxytocin and cortisol levels in mothers (Swain et al., 2011). In addition to responses to infant crying, parent-infant touch is a significant aspect of parent-infant interaction. Skin-toskin contact has been found to increase oxytocin levels and decrease cortisol levels both in infants and parents (Cong et al., 2015;Vittner et al., 2018).
Hormonal reactivity has also been found to relate to parental behavior. Kohlhoff et al. (2017) observed that in some mothers oxytocin levels increased while in others oxytocin levels decreased or remained constant when mothers and their 3-to 4-month-old infants were presented with the still-face paradigm. Increasing oxytocin levels were associated with higher observed maternal sensitivity. Increased maternal oxytocin levels after mother-infant interaction have also been associated with more affectionate touch ) and longer duration of mother-to-infant gaze . In addition, oxytocin responses to interaction with the infant have been found to be larger in mothers with secure attachment representations as measured with the Adult Attachment Interview (Strathearn et al., 2009). In single women, attenuated estradiol reactivity to an emotionally intimate parent-infant video was linked to avoidant attachment style (Edelstein et al., 2012). In fathers, testosterone levels decreased during the Strange Situation Procedure and the amount of decrease predicted their sensitive parenting behavior when in interaction with their own 12-month-old child (Kuo et al., 2016). Lotz et al. (2022) found associations at trend level between fathers' sensitivity and both oxytocin reactivity and testosterone reactivity, with more sensitive parenting behavior during a 10-min interaction with their own 2-month-old infant related to stronger increases in fathers' oxytocin and testosterone levels.
The impact of parental status on hormonal reactivity to infant stimuli has not yet been studied. A few studies have used an infant simulator, which provides a validated and ethical method for studying both parents and nonparents in a realistic caretaking situation (Bakermans-Kranenburg et al., 2015;Voorthuis et al., 2013). The infant simulator is a doll resembling a real infant in terms of weight, appearance, and affective vocalizations. Testosterone levels in nulliparous women were found to decrease after taking care of a crying infant simulator for 30 min (Voorthuis et al., 2019). A similar effect was observed in men when they interacted with a crying but soothable infant simulator: salivary testosterone levels decreased (van Anders et al., 2012). However, this finding did not replicate in an independent sample: instead of a decrease, testosterone levels were found to stay constant when men interacted with the crying simulator (van Anders et al., 2014).
Another study, on the other hand, observed increased testosterone and cortisol levels in pregnant women following interaction with an unsoothable infant simulator (Bos et al., 2018). Postnatally, when the same participants interacted with their own infant, cortisol levels were found to decrease. Thus, cortisol reactivity may be different before and after birth or different to the mother's own infant as compared to an infant simulator. However, neither testosterone nor cortisol was related to observed parental sensitivity in mothers, whereas in fathers decreases both in testosterone and cortisol in response to interaction were related to more sensitive parental behavior (Bos et al., 2018).
The associations between hormones (and their reactivity) and caretaking behavior seem to be somewhat different in mothers and fathers, and also in parents and nonparents, which makes it important to study these groups separately.

The present study
In this study, we investigated the impact of maternal status on salivary hormonal reactivity and behavioral responses to an infant simulator.
Mothers and nonmothers have not been compared on these outcomes and, therefore, the current study will provide important data for understanding whether the associations between hormones and behavior are independent of parental status. Primiparous and nulliparous women took part in a realistic caretaking situation with an infant simulator that made both positive and negative sounds. Maternal behavior was operationalized as the amount of affectionate touch and motherese vocalizations (i.e., infant-directed speech) during the caretaking situation. First, we investigated whether the two groups differed in their hormonal reactivity toward the infant simulator. Due to the lack of earlier studies comparing hormonal and behavioral responses toward infants in parents and nonparents, we were cautious in making strong directional hypotheses. Oxytocin reactivity has not been investigated with an infant simulator, but based on research in mothers with their own infants, oxytocin levels were expected to rise at least in the primiparous group (Cong et al., 2015;Swain et al., 2011;Vittner et al., 2018). Considering estradiol, we had a more exploratory approach and did not make directional hypotheses. Based on the available evidence, we predicted that in nulliparous women testosterone levels would decrease during the interaction (Voorthuis et al., 2019), whereas in primiparous women testosterone levels were not expected to show any reactivity, in line with the findings of Bos et al. (2018). Also, according to Bos et al. (2018), one might expect differences in cortisol responses in the two groups, with a decrease in primiparous women and an increase in nulliparous women. However, as that study differed in terms of study population and simulator paradigm from the current study (i.e., pregnant women interacting with a persistently crying simulator), we did not make strong directional hypotheses regarding cortisol. Second, we investigated whether hormonal reactivity is associated with behavior toward the infant simulator. We hypothesized that positive oxytocin reactivity would be associated with greater amount of affective touch and motherese vocalizations Kim et al., 2014;Kohlhoff et al., 2017). Regarding estradiol, testosterone, and cortisol, we did not have clear a priori hypotheses due to insufficient earlier research. Based on earlier research on the dual-hormone hypothesis (Mehta & Josephs, 2010), we also analyzed whether the association between testosterone and maternal behavior was dependent on baseline cortisol levels. Third, we investigated whether the possible associations between hormones and behavior were different in the two groups. Again, we adopted an exploratory approach due to the nonexistent previous research comparing parents and nonparents during interaction with an infant simulator. Finally, in nulliparous women their attitudes and desires toward reproduction (i.e., fertility motivation; Brase & Brase, 2012;Deady et al., 2006) might influence their interaction with the simulator or their hormonal reactivity. Thus, we investigated whether in nulliparous women self-reported fertility motivation was associated with behavior and hormonal levels as well as hormonal reactivity in response to the infant simulator.

Participants
The participants were part of the TransParent project, which investigates changes in processing infant cues during the transition to parenthood. The study protocol was reviewed by the Ethics Committee of the Tampere  interactions with an effect size of 2 p = .064 and larger. As a reward for participation, the participants received one movie ticket and course credit if necessary. Descriptive statistics are presented in Table 1.

Procedure
The participants were called before the laboratory visit. During the call, menstrual cycle and hormonal contraceptive use were screened.  At the beginning of the laboratory visit, participants received information about the study, signed an informed consent, and completed a short questionnaire, which was followed by the first saliva sample give the impression of a successful caregiving experience.
After the interaction with the simulator, the participants spent 10 min alone while filling in two questionnaires: one measuring selfreported empathy (Interpersonal Reactivity Index; Davis, 1983) and F I G U R E 1 Timeline of the study procedure (below) and timing of infant simulator sounds (above) the PANAS questionnaire for the second time. After 10 min, the second saliva sample (T2) was collected in the same way as the first sample. The whole procedure is illustrated in Figure 1.
Within days after the laboratory visit, the participant received a link to an online questionnaire, which included questions about background information such as education, income, and relationship status and length. The questionnaire also included items assessing depressive symptoms (CES-D; Radloff, 1977), anxiety (STAI; Bieling et al., 1998)  were replaced with the mean of the participant's other values.

Covariates
Age of the participant, time of day, cycle day, relationship duration, and years of education were initially considered as covariates based on earlier research. For mothers, the age of the infant and the time from the last breastfeeding were also included.

Statistical analyses
Video recordings were missing from three participants due to technical problems, but these participants' hormonal data were included in the analysis. Associations between the hormonal, behavioral, and background data were further investigated with Pearson correlation coefficients separately within the two groups (Table 3). In addition, to explore the influence of the categorically coded hormonal birth control use, we compared hormonal levels between women who did and did not use hormonal birth control separately in nulliparous and primiparous women using independent samples t-tests. Two separate repeatedmeasures ANOVAs were used to examine the change in participants' positive and negative affects during the interaction task. Time (before and after the interaction) was the within-subjects variable and parity was the grouping variable.
As can be observed from Table 1, the groups differed in some background variables. Age and the use of hormonal birth control were included as covariates in the main analysis. To answer the three main research questions, linear mixed models were conducted. One model was built for each hormone making it in total of four models. The −2-log likelihood ratio scale was examined as a determinant of model fit. First, the repeated factor time (T1, T2) was added to the model, followed by within-subjects covariates motherese, affectionate touch, age, and a dummy variable representing the use of hormonal birth control. Next, the between-subjects factor parity was added. For the testosterone model, cortisol was also added as a covariate, to test whether the association between testosterone and maternal behavior was dependent on baseline cortisol level. Main effects of within-and between-subjects variables were interpreted before adding any interaction terms to the models. Next, the interaction terms (parity × time, parity × motherese, parity × affectionate touch, time × motherese, time × affectionate touch, and the interaction terms of cortisol and the behavioral variables) were added. The final models are presented in the results.
Finally, to explore the relation of fertility motivation to the behavioral variables and hormonal reactivity within the nulliparous group, Pearson correlation coefficients between the ABS questionnaire scores and the behavioral and hormonal variables were calculated.

Descriptive statistics
As reported in Table 1 Table 2.
Nulliparous women not using hormonal birth control had higher baseline testosterone levels than nulliparous women who did use hor- In nulliparous women, the covariates (age, educational years, relationship length, menstrual phase, duration of the simulator crying, or time of day) did not correlate with any hormonal baselines or reactivity, or with behavior toward the infant simulator. The correlation coefficients between the study variables for nulliparous women are presented in Table 3 (below the diagonal).
In primiparous women, time from last breastfeeding was positively correlated with baseline oxytocin (T1: r = .39, p = .01) and testosterone reactivity (r = .46, p = .002). Time of day was inversely associated with testosterone reactivity (r = -.31, p = .041). Educational years were associated with the use of motherese (r = .34, p = .017). Other covariates (relationship duration, menstrual phase, age, duration of simulator crying, or own infant's age) were not associated with any of the hormonal baseline levels, hormonal reactivity, or behavior toward the infant simulator within the primiparous group. Correlation coefficients for primiparous women are presented in Table 3 (above the diagonal).

Hormonal reactivity and behavior towards the infant simulator
The linear mixed models for all four hormones are presented in Table 4.

Oxytocin
Although mean values of the original data ( Abbreviation: CI, confidence interval; ICC, intraclass correlation coefficient. a Dummy variable (0 = not using hormonal birth control, 1 = using hormonal birth control).
affectionate touch showed significant main effects or interaction with parity or time.

Estradiol
For estradiol, there were no significant main effects of time or parity nor were there interactions between parity and time. Neither motherese nor affectionate touch had significant main effects or interaction with parity or time.

Testosterone
For testosterone, there was a main effect of affectionate touch (p = .032). More affectionate touch was associated with lower overall testosterone levels across the two measurement points. There were no significant main effects of parity or motherese. Time showed a significant interaction with motherese (p = .043). To illustrate this interaction, standardized simple slopes were plotted for one standard deviation below and above the mean for motherese (Figure 2). The coefficients were in different directions signaling interaction. At low levels of motherese, testosterone levels were decreasing (β = −0.28, p = .346), whereas in high levels of motherese testosterone levels were slightly increasing (β = 0.153, p = .598).
Other interactions were not significant. To test for effects related to the dual hormone hypothesis, baseline cortisol was also included in the testosterone model. However, no interactions with cortisol emerged.

Cortisol
Time had a significant effect on cortisol (p = .002), with cortisol levels showing an overall decrease from T1 to T2 (t(114) = 2.84, p = .005, d = 0.28) as can be observed from Figure 3. There was also a significant interaction between parity and affectionate touch (p = .021).
We examined this interaction with additional linear mixed models conducted separately for both groups. For primiparous women, we found a significant main effect of affectionate touch on cortisol (p = .035), as more affectionate touch was associated with lower overall cortisol levels across the two measurement points (Figure 4). For nulliparous women, there was no significant main effect of affectionate touch on cortisol levels (p = .312).

Fertility motivation and its associations with behavior and hormones in nulliparous women
In nulliparous women, fertility motivation scores were positively associated with the amount of motherese used with the infant simulator (r = .47, p = .001) but not with affectionate touch. Fertility motivation was inversely correlated with baseline testosterone (T1: r = -.40, p = .001). The correlation coefficients are presented in Table 3 (below the diagonal).

DISCUSSION
The primary aims of the present study were to explore potential dif- Decreasing cortisol levels have also been observed in mothers when they take care of their own infant (Bos et al., 2018). Thus, the infant simulator appears to produce similar cortisol reactivity compared to interaction with a real infant. The decreasing cortisol levels in this study were opposite to earlier results with pregnant women in Bos et al. (2018), which suggests that during pregnancy cortisol reactivity may differ from other life situations.
We also observed an interaction between the amount of affectionate touch and parity in overall cortisol levels. In primiparous women, lower cortisol levels across the measurement points were associated with more affectionate touch, whereas in nulliparous women no association between cortisol levels and affectionate touch was found. The same difference was evident from the correlation coefficients (Table 3).

F I G U R E 4 Correlation between overall cortisol levels and affectionate touch in primiparous women
This finding is novel and suggests that the association between cortisol and affectionate touch might be different in primiparous and nulliparous women. Earlier studies have not compared primiparous and nulliparous women in this regard, but our finding of an inverse relation between cortisol levels and affectionate touch in mothers is in line with earlier research suggesting that lower baseline cortisol levels are associated with higher maternal sensitivity (Finegood et al., 2016;Gonzalez et al., 2012) and less intrusive parenting behavior (Mills-Koonce et al., 2009). In addition, mothers' sensitivity toward an infant simulator has been observed to correlate with their sensitivity toward their own infant (Bakermans-Kranenburg et al., 2015).
Together with the current study, those findings support the use of the infant simulator as a valid method for investigating individual differences in caretaking behavior and suggest that mothers differ from nonmothers regarding the association between cortisol and caretaking behavior also during naturalistic caretaking situations. In the current study, mothers also used more motherese with the simulator than nonmothers, suggesting that they were behaving with the simulator as with their own infant. However, we did not control for the reported seriousness or reality value (see Bos et al., 2018;Voorthuis et al., 2013) of the interaction with the simulator, which is necessary in the future studies.
There was a negative association between testosterone levels and the amount of affectionate touch when taking care of the simulator.
This is in line with the earlier research with male samples (Bos et al., 2018;Meijer et al., 2019;Weisman et al., 2014)  reactivity. Furthermore, we did not find cortisol to moderate the association between testosterone and behavior. One reason for this might be due to the all-female sample in our study. Earlier human and animal studies have had either mixed-sex or all-male samples and it is possible that the moderating effect of cortisol on testosterone levels is more evident in males than females.
The two groups did not show oxytocin reactivity toward the infant simulator, nor did they differ in their oxytocin reactivity toward the simulator. This was contradictory to our expectations and results from studies investigating parental oxytocin reactivity in response to interaction with their infant (e.g., Feldman et al., 2010). However, in our study, the amount of oxytocin was below the manufacturer's limit in 48% of the sample and 36% of the saliva samples were unanalyzable due to the low levels of oxytocin or insufficient amount of saliva, which means that oxytocin data were based on imputed values to a greater extent than was the case for other hormones. Therefore, the oxytocin results should be evaluated with caution. The wide variability of reported oxytocin levels across the literature has resulted in criticism toward measuring oxytocin from saliva (Horvat- Gordon, et al., 2005;McCullough et al., 2013). Many of the previous studies that reported highly variable oxytocin levels were missing the extraction step of the analysis. In this study, the oxytocin samples were extracted before the assay, which, as a downside, partially explains the attrition in the oxytocin data. Furthermore, similar to earlier studies, in our sample baseline oxytocin levels were correlated with the time since the last breastfeeding in mothers. This indicates that oxytocin levels as measured in the present study reflect true variation as in earlier studies oxytocin levels have been found to start to increase after breastfeeding reaching their peak just before the next feeding (Carter et al., 2007;de Jong et al., 2015;White-Traut et al., 2009). It is noteworthy that a considerable proportion of the literature documenting associations between peripheral oxytocin and parenting behaviors is based on data from a single laboratory (see Grumi et al., 2021). Therefore, it is vital to replicate and extend such associations with greater diversity of samples and methods.
Compared to other hormones in this study, there is very little research on the relation of estradiol to caretaking in humans. In our study, estradiol levels did not change in response to interaction with the simulator, nor were they associated with caretaking behavior. The few earlier studies linking estradiol levels to parenting or relationship outcomes (Edelstein et al., 2017;Glynn et al., 2016) have studied estradiol levels during pregnancy. As estradiol levels increase during pregnancy and decrease rapidly after giving birth (Fleming, Ruble, et al., 1997), associations between estradiol and caretaking behavior may be more prominent during the prenatal period. In addition, associations between estradiol and behavior might depend on individual progesterone or testosterone levels, which have been found to be important in earlier studies. For example, smaller decline in the estradiol to progesterone ratio during pregnancy has been associated with higher postpartum feelings of attachment toward the infant (Fleming, Ruble, et al., 1997), and in fathers high testosterone levels combined with high, but not low, estradiol levels have been associated with lower sensitivity . In the future, it is relevant to also measure progesterone, which might have a moderating effect on estradiol reactivity.
In nulliparous women, higher fertility motivation was associated with more motherese directed to the simulator. This finding is novel and indicates that women's positive feelings toward infants and motivation to become a mother affect their caretaking behavior in a positive way. In addition to behavior, fertility motivation was negatively associated with testosterone levels. In line with the Challenge hypothesis (Archer, 2006), the inverse relation between fertility motivation and testosterone may suggest that, similarly to men, preparation to parenthood is associated with declining testosterone levels also in women.
Another possible explanation for the negative association between testosterone and fertility motivation could be that women who have lower testosterone levels in general have more positive views on babies. This would also be in line with earlier results showing higher testosterone levels to be associated with lower self-rated reproductive ambition (Deady et al., 2006). Fertility motivation or "baby fever," although being a popular subject on the media, has not yet been studied extensively. It is unclear whether fertility motivation predicts future pregnancy in women. In addition, the potential associations between fertility motivation and hormones and caretaking behavior in men are an important target for future studies.
Together with earlier studies using the infant simulator (Bos et al., 2018;Voorthuis et al., 2019), this study supports the use of the infant simulator in comparing mothers and nonmothers: the infant simulator elicited hormonal reactivity in both mothers and nonmothers, and the two groups showed similar and partially different patterns of associations between hormonal levels and caregiving behavior. However, our results are preliminary at best and require replication in independent samples and greater variability of caregiving behaviors in the future. In addition, it remains unclear when during the transition to parenthood the potential differences in hormonal reactivity and caretaking behavior begin to emerge. Longitudinal research designs would be important to determine whether the differential responses emerge in mothers due to the biological changes associated with pregnancy or whether they are induced by caretaking experiences with their infants. Comparing biological and adoptive parents in a simulated caretaking situation might reveal an answer to this question. In addition, the impact of hormones on different aspects of caretaking behavior may be indirect and operate through motivational processes that may affect parental behavior. For example, oxytocin has been linked to approach motivation (MacDonald & MacDonald, 2010;Soriano et al., 2020) and activity of the reward system of the brain (MacDonald & MacDonald, 2010).
The heightened motivation toward babies could thus promote infantoriented behavior such as the use of motherese. In future studies, it will be important to investigate the role of motivation toward babies for parental behavior in mothers and nulliparous women in greater detail.

ACKNOWLEDGMENTS
The authors would like to thank Prof. Anna Rotkirch for providing the Finnish version of the fertility motivation questionnaire. This research was supported by grants from the Academy of Finland (#307657 and #321424).