Adult male and female laboratory mice (outbred strain NMRI, 2–3 months old) with no previous sexual or parental experience were housed at the University of Ulm. As previously (Ehret et al., 1993), mice were randomly assigned to one of three treatment groups consisting of (1) naïve males without any experience with pups (N), (2) males co-caring with their female mate for their first litter for 5 days (5D), or (3) males co-caring with their female mate for their first litter until weaning and their second litter for 5 days. The total co-caring time of the latter males (first and second litter) was 26–27 days (27D) since after delivery of the first litter, females in post-partum estrus paired immediately, and, while nursing the first litter up to weaning at 21 days, delivered the second litter after 21–22 days of gestation. Mice were housed at 22°C and a 12 h light–dark cycle (starting at 7.00 h) in plastic cages (26.5 × 20 × 14 cm) either with other naïve males (group N; 4 brothers per cage) or with one female (groups 5D and 27D). Water and food were available ad libitum. The experiments were carried out in accordance with the European Communities Council Directive (86/609/EEC) and were approved by the appropriate authority (Regierungspräsidium Tübingen, Germany).

Pup retrieval behavior was tested as previously described (Ehret and Koch, 1989; Ehret et al., 1993). Male mice (alone or with litter mate and pups) were placed overnight on a running board (length 110 cm, width 8 cm) with a central nest depression (Figure 1A) suspended in a soundproof and anechoic room. The running board was supplied with mouse chow and nest material from the cage and covered with a plexiglass hood with an opening for the tip of a suspended water bottle. The following day, retrieving tests were conducted under dim red lighting (<1 lx), ensuring that males relied on the sense of smell, audition, and touch to respond and retrieve pups. Before the start of testing, the hood and the female with the pups (groups 5D and 27D) were removed and placed back in their home cage outside the soundproof room. Seven pups were kept in the room to serve as targets for retrieving. The males were habituated for at least 30 min to the new situation after the hood (group N) or the hood and female with pups (groups 5D and 27D) were removed. Prior to testing males in the N group, 5-day-old foster pups were rubbed with the nest material of the respective males to provide a familiar scent. During the test, the seven pups (always 1–3 pups at a time) were randomly placed on the running board, at least 30 cm away from the nest. A male was scored as a ‘retriever’ if he retrieved all seven pups to the nest within 10 min. A male was scored as ‘non-retriever’ if he had not retrieved any pups after 10 min. The distinction between retriever and non-retriever was clear-cut because once a male had retrieved the first pup, he retrieved all. Retrieving tests were immediately stopped, and the pups were removed if a male attacked any of the pups (‘aggressor’).

Immediately following the pup retrieval test, males were euthanized by cervical dislocation, and their brains were quickly dissected and frozen over liquid nitrogen. Brains were mounted using O.C.T. compound, and frontal sections were serially cut (30 μm) using a cryostat (Thermo Scientific Microm HM560) and immediately mounted on slides to keep track of the left and right sides of the brain. Sections were fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (PB; pH = 7.4) for 60 min. Sections were rinsed thrice for 5 min with 0.1 M PB between subsequent processing steps. Slides were incubated with 0.2% Triton X-100 (Sigma) in 0.1 M PB for 45 min to permeabilize cells and with 1% H₂O₂ (Merck) in 0.1 M PB for 20 min to block endogenous peroxidase. In a humidified chamber, the sections were treated with 2% normal goat serum in 0.1 M PB for 60 min to block non-specific staining, then with the polyclonal rabbit anti-aromatase antibody (Abcam Cat# ab35604, RRID: AB_867729; 0.002 mg/mL in 2% normal goat serum) at 4°C for 72 h. Sections were then incubated with the secondary antibody from the goat against rabbit horseradish peroxidase (Dako, Cat# PO448, 0.00125 mg/mL in 0.1 M phosphate buffer) at room temperature for 60 min. Immunoreactivity was revealed by incubating the sections in 0.015% diaminobenzidine tetrachloride (DAB; Sigma), 0.023% NiCl, and 0.013% H₂O₂ in 0.1 M PB for 7 min. Sections were rinsed in PB, dehydrated in a series of alcohols and xylene, and coverslipped with Entellan (Merck). At the time of purchase, the primary antibody was recommended for aromatase immunohistochemistry on mouse tissue, and the supplier provided positive quality controls for mouse ovarian tissue. Dutta et al. (2014) also performed a Western blot with mouse tissue that confirmed the specificity of the antibody. Previous studies have used this antibody to detect aromatase immunoreactivity in mouse ovaries (Laws et al., 2014), brain stem (Ji et al., 2017), and auditory cortex (Charitidi and Canlon, 2010). We tested the specificity of the secondary antibody labeling by incubating sections with PB instead of the primary antibody. Immunoreactivity was completely absent in all these endogenous tissue background control sections (Figure 2A).

Sections were independently analyzed by two experimenters who were blinded to the experimental conditions. To accommodate the variable intensity of the immunolabeling, they initially reviewed multiple sections from all animal groups to establish consensus on criteria for aromatase-specific labeling. Brain areas were identified using a 10x objective, while aromatase-positive cells were located with a 40x objective. Cells were classified as expressing aromatase if they (1) displayed more intense gray staining than the background and (2) contained black granulae typically clustered around the cell nucleus (Figure 2B). Structures presenting single or multiple black granulae without the associated gray cellular structure were not counted as aromatase-positive (Figure 2B). This procedure led to (a) a high agreement between the experimenters’ counts, (b) the identification of unequivocally aromatase-positive cells, and (c) relatively low counts of aromatase-positive cells per evaluated brain area in each section.

Sections were analyzed using a light microscope (Axiophot, Zeiss, Germany). The brain areas in which aromatase-positive cells were counted are listed in Table 1. Areas were matched to the mouse brain atlas plates and associated Bregma coordinates (Paxinos and Franklin, 2001) and anatomically matched across all experimental animals. These areas largely overlap with those showing ERα-positive cell dynamics during the acquisition of pup-caring experience in male (Ehret et al., 1993) and female (Ehret and Buckenmaier, 1994) mice. Aromatase-positive cells were counted separately in the right and left hemispheres of the brain. Counts per section for a given animal, brain area, and hemisphere were averaged between the maximum number of consecutive sections that could be matched across all animals, i.e., two sections for LS, BNST, VMH, and nucleus accumbens (NAC), or three sections for PIR, MPOA, cortical amygdala (CAM), and medial amygdala (MAM). These average counts from the individual animals were used in statistical analyses.

All statistical analyses were carried out using R v4.2.0 (R Core Team, 2022).

Only males with paternal experience retrieved pups (Figure 1A). Figure 1B shows that naïve males with no previous sexual experience and pup contact mostly ignored the pups (n = 5/8) or were aggressive (n = 3/8). The majority of fathers with 5 days (5D) or 26–27 days (27D) of pup experience retrieved pups (n = 10/12 and 11/12, respectively), and only a few ignored pups (n = 2/12 and 1/12, respectively). No father was aggressive.

The probability of observing aggressive behavior toward pups was significantly higher in naïve males compared to fathers regardless of experience (5D and 27D, Figure 1B). The probability of other non-paternal behavior like ignoring pups was higher in naïve males compared to 27D fathers. Parental behavior measured as pup retrieval was significantly higher in fathers (5D and 27D) compared to naïve males. There were no significant differences in the probability of pup-retrieving between the groups of fathers, 5D and 27D (Figure 1B).

We found aromatase-positive cells in each examined area and with each experience level. In absolute terms, the largest numbers of aromatase-positive cells per section were seen in PIR, LSI, MAM, and CAM (see unstandardized data in Supplementary Figure S2). Aromatase labeling in the LSV, PIR, VMH, and MAM of the N and 27D groups showed qualitative differences between experimental groups and brain hemispheres (Figure 2). Standardized cell counts together with the group means are shown in Figure 3A along with our estimates of the posterior differences in aromatase counts by hemisphere (Left–Right, Figure 3B) and between levels of experience (Figure 3C).

We found evidence for significantly greater numbers of aromatase-positive cells in the left hemisphere of the MPOA and BNST in all three experimental groups (Left – Right >0, Figure 3B). The NAC exhibited a similar trend to MPOA and BNST, but with a just insignificant contrast in the 27D group. We did not detect significant lateralization of aromatase counts in any of the experimental groups in the amygdala (CAM and MAM), VMH, or posterior PIR (Bregma −1.34). In the other brain areas, the appearance of lateralization strongly depended on the level of pup experience. Long-term experience (27D) led to greater numbers of aromatase-positive cells in the right hemispheres of anterior PIR (Bregma 0.98), the LSV, and the anterior LSI (Bregma 0.14). Posterior LSI (Bregma 0.98) exhibited a similar trend that was just non-significant. In the LSD, pup experience had a short-term effect (5D) during which counts in the left hemisphere were greater. This effect was not present in the 27D group (Figure 3B).

In all cases in which we found significant changes in aromatase counts with pup experience, the number of aromatase-positive cells increased (Figure 3C). This experience-dependent increase was pervasive for the amygdala (CAM and MAM), the VMH, and the posterior PIR (Bregma −1.34). For these areas, the increase in aromatase positive cells was present in both hemispheres and with increasing experience, i.e., experience contrasts had significant positive values for both brain hemispheres and all experimental groups. In the LSD, LSI, and LSV (both Bregma coordinates) and the anterior PIR (Bregma 0.98), aromatase-positive cells increased significantly only in the right hemisphere of 27D males. Experience-dependent effects were absent in the NAC, the BNST, and the MPOA, except for a significant effect in the right MPOA of 27D males, i.e., 27D males had significantly more aromatase-positive cells than 5D males (Figure 3C).

We examined the unexplained (residual) variation in aromatase cell numbers among brain areas, as shown in Figure 4. We found positive residual correlations between all brain areas, with the sole exception of CAM (−1.34) and LSD (0.98), which were negatively correlated. Three subgroups of covariation are obvious. First, the LS is highly connected, and each subregion (D, I, V) is correlated across Bregma coordinates and with at least one other subregion. The LS appears to form the backbone of variation, to which other nodes are connected. This includes subgroups comprised of MPOA (connected to LSV), NAC and BNST (connected to LSI), and CAM and VMH (connected to LSD). Second, the subgroup of MAM and anterior PIR appears to be weakly connected to other parts of the graph (MPOA and VMH) because the edges are significant only with p < 0.10. Third, posterior PIR did not appear to be weakly correlated with any other analyzed brain area despite the increase in aromatase expression with experience similar to CAM, MAM, and VMH (see Figure 3).

The transition to fatherhood is distinctly marked by a reduction in aggressiveness toward pups and an increase in pup retrieval behavior (Figure 1). However, some males continue to ignore pups, and notably, one second-time father failed to retrieve pups during the pup retrieval test. To explore this further, we investigated whether variations in aromatase cell counts in any specific area could predict a male’s reaction to pups (contrast of reacting – ignoring mean cell counts, Figure 5). Our analysis revealed that males who reacted to pups had significantly higher numbers of aromatase cells in the MAM (p = 0.011) and exhibited a mild tendency toward higher cell counts in the CAM (p = 0.06).

Aromatase expression in the male mouse brain has been shown via immunostaining with several protocols using different antibodies (Akther et al., 2015; Foidart et al., 1995), partly with enhancement through transgenic techniques (Stanić et al., 2014). In agreement with the published data, we found aromatase expression in all studied brain areas of our three experimental groups of male mice (Figure 3). Differences in the levels of aromatase expression between studies relate to differences in the staining methods, data analysis, and the social environment of the animals prior to and during the behavioral tests. For example, in our study, dynamics of paternal behavior and associated changes in aromatase levels have been examined while the fathers were continually housed with their female mate except during the pup-retrieval tests. This is different from the study of Akther et al. (2015), in which the relationship between pup-retrieving behavior in males and the dynamics of aromatase levels was tested throughout different ethological contexts (presence, absence, and reintroduction of their female mates and pups). In that previous study, brain aromatase expression was highest when the fathers were housed with their female partner but without their pups, followed by fathers housed with the whole family (Akther et al., 2015). Therefore, the observed changes in aromatase levels in that study can be related to male–female interaction. Nevertheless, the strong reduction of pup retrieval in fathers after injection with an aromatase inhibitor (letrozole; Akther et al., 2015) shows, along with the results of our present study, that proper levels of aromatase expression are integral to the display of paternal behavior in mice.

Interestingly, circulating testosterone is negatively correlated with paternal behavior in various rodent species (reviewed in Bales and Saltzman, 2016; Numan and Insel, 2003), and testosterone can inhibit paternal care in mice (Okabe et al., 2013). Our study did not examine the source of testosterone, the substrate of aromatase in the brain. However, previous studies have shown that changes in circulating testosterone may not be reflected locally in the brain. Neurosteroids can be locally produced, and levels in the brain may be independent of circulation or gonadal status (Giatti et al., 2019; Hojo et al., 2008, 2004).

In the following, we discuss the observed presence and changes of aromatase levels in the studied brain areas of the limbic system in order to see how these areas may contribute to changing male behavior towards pups from aggressive or ignoring in naïve males to caring in experienced fathers (Figure 1). Furthermore, we compare the dynamics of aromatase-positive cells with the dynamics of ERα-positive cells in the given brain areas to reveal the extent to which these align to support the hypothesis about aromatase influencing indirectly (via estrogen synthesis and ER binding) paternal behavior.

Three nuclei, the MAM, CAM, and VMH, were most affected by pup experience with increases in the number of aromatase-positive cells from N to 5D to 27D animals (Figures 3A,C; Supplementary Figure S2). These areas are associated with the regulation of aggression compared to sexual and parental behavior (e.g., Chen et al., 2019; Dwyer et al., 2022; Hashikawa et al., 2016; Lee et al., 2014; Sheehan et al., 2001; Unger et al., 2015; Yang et al., 2017). Important changes in aggression in the course of becoming a caring father are mediated by olfactory signals from the vomeronasal organ via the accessory olfactory bulb to the CAM and MAM (e.g., Hashikawa et al., 2016) and concern the switch from aggressive to caring behavior toward pups (Tachikawa et al., 2013). Concomitantly, aggressive motivation in fathers increases against male intruders to protect their own pups from potential pup killers, as shown in house mice (Shabalova et al., 2020), Californian mice (Trainor et al., 2008) and prairie voles (Getz et al., 1981). This increased aggression toward intruders was not mediated by changes in ERα or ERß expression in limbic centers but reflected in increased neuronal activation (FOS expression) in the MPOA, MAM, and, potentially, the VMH in California mice (Trainor et al., 2008). In house mice fathers, aggressive behavior toward intruders is regulated by oxytocin and activation of neurons in areas of the hypothalamus (periventricular and supraoptic nuclei; Shabalova et al., 2020). Further research could examine whether changes in aggression in naïve vs. fathers are also related to aromatase levels in the amygdala and VMH.

In contrast to aromatase expression (present study), the number of estrogen receptors (ERα) in MAM, CAM, and VMH did not change with parental experience in male mice (5D or 27D) (Ehret et al., 1993). Together, this indicates that the increased presence of local estrogen synthesis does not facilitate the expression of ERα, as has been shown in the VMH (Devidze et al., 2005; Malikov and Madeira, 2013), especially in combination with parental experience (Koch, 1990). Few studies have explored the role of other estrogen receptors in paternal care in rodents (Hyer et al., 2017; Torres et al., 2024). Other subtypes (ERβ and membrane receptor GPER) may be involved in the expression of parental behavior. Our study indicates that parental experience is associated with changes in aromatase expression, which can potentially affect the balance of androgen to estrogen levels in the brain. Further studies should also examine whether androgen receptors, expressed in the same brain areas as aromatase (Dart et al., 2024), are implicated in parental behavior (e.g., see research in Mongolian gerbils; Martínez et al., 2019). We suggest that the increase in aromatase expression in the MAM, CAM, and VMH in new (5D) and highly experienced (27D) fathers (Figure 6) supports the control of instinctive behavior (Honda et al., 2011), including inhibition of infanticide, via increased local synthesis of estrogens on neurons (e.g., Cornil et al., 2006, 2013).

In house mouse mothers, pup-retrieving performance does not depend on olfactory signals from the main olfactory bulb and its projections to the PIR (Koch and Ehret, 1991). Pup-naïve females, however, needed this olfactory information to learn to become as maternal as the mothers (Ehret and Buckenmaier, 1994; Koch and Ehret, 1991). While learning pup cues for becoming maternal, virgin females with maternal experience have an increase in ERα-positive cells in the posterior PIR compared to females without experience (also in the entorhinal cortex; Koch, 1990). Similarly, as males became caring fathers, ERα-positive cells became visible in the posterior PIR (also in the entorhinal cortex; Ehret et al., 1993). Given the significant increase of aromatase expression in the posterior PIR with increasing paternal experience (Figure 3C), it is possible that in PIR, unlike MAM, CAM, and VMH, estrogen may have induced its own receptors (tested for ERα) and enhanced olfactory learning of pup cues for increased paternal performance.

The significant increase of aromatase-positive cells only in the right hemisphere in the anterior PIR of 27D fathers does not directly correlate with any evidence. In studies on ERα expression in the limbic areas of female and male mice under various conditions of the reproductive cycle, experience with pups, and intact or lesioned olfaction, no hemisphere differences have been reported in any area (Ehret et al., 1993; Ehret and Buckenmaier, 1994; Koch, 1990; Koch and Ehret, 1991). The splitting of PIR in regions with experience-dependent lateralized (anterior PIR) and non-lateralized (posterior PIR) aromatase change shows compartmentalization of the PIR. Functional relationships need to be elucidated in further studies.

The main effects of pup-experience on aromatase expression in the LS were changes in the left-right balance (Figure 3B) and significant increase of aromatase-positive cells only in the right LS of highly experienced (27D) fathers (Figure 3C). As in the PIR, ERα-positive cells in the LS become visible in caring fathers compared to pup-naïve males (Ehret et al., 1993). Since hemisphere differences in ERα-positive cells have not been found, we cannot directly relate increased right-hemisphere aromatase expression with a non-lateralized increase of ERα expression in the LS of both hemispheres. Lateralized input to the rostral parts of the LSD and LSI comes from the hippocampal CA1 and subiculum region (work in mice and rats; Rizzi-Wise and Wang, 2021; Risold and Swanson, 1997). In rodents, the CA1 shows lateralization in synapse morphology and receptor distribution (Shinohara et al., 2008; Xiao and Jordan, 2002) and left–right dissociation of memory processes (Shipton et al., 2014) with right-hemisphere dominance of spatial memory in mice (Shinohara et al., 2012). Also, androgen receptor levels are higher in the left hippocampus in male rats (Xiao and Jordan, 2002). It is unclear, however, whether and how lateralized hippocampal input to the LS could have caused lateralized aromatase expression in the LS of 27D fathers (Figure 3C). Right hemisphere lateralization is associated with the regulation of mood, affect, and stress, and with sustained and global arousal and attention (e.g., Ehret, 2006; Palomero-Gallagher and Amunts, 2022). Speculatively, our observed right-hemisphere LS bias in 27D fathers could be related to the stress of continuous arousal and steady involvement in a routine of pup care over several weeks.

The MPOA and the BNST and their output via the NAC are essential for controlling rodent parental behavior (e.g., Numan, 2006; Bridges, 2015; Wu et al., 2014). Significant increases have been observed in the MPOA in both the number of and area occupied by ERα-positive cells in 5D and 27D pup-experienced father mice (Ehret et al., 1993) and, although absent in pup-naïve males, ERα-positive cells are present in the BNST of 27D fathers (Ehret et al., 1993; Koch and Ehret, 1989a,b). The number of ERα-positive cells in females is also higher in the MPOA and BNST of one-day pup-experienced virgin females (Ehret and Buckenmaier, 1994). In the current study, we did not observe changes in the number of aromatase-positive cells in the BNST and NAC and found only a small increase in the MPOA in only the 27D males relative to 5D. This suggests that changes in sensory processing and behavior with experience in MPOA, BNST, and NAC may be more dependent on the regulation of estrogen signaling through binding to ERα than on local estrogen action after synthesis via aromatase. However, further research on other estrogen receptors is needed. Brain areas that regulate social behavior, including parental care, are conserved in vertebrates (O’Connell and Hofmann, 2012). However, neuroendocrine ligand spatial profiles are more variable than receptor distributions within the social decision-making network among vertebrates, although there are regional differences (O’Connell and Hofmann, 2012). Our study supports the existence of experience-dependent regional differences, specifically indicating that the effects of the MPOA, BNST, and NAC on parental behavior may depend on changes in ERα distribution. In contrast, the influence of the VMH and the amygdala may be driven by changes in aromatase expression. Overall, our results highlight the area-specific regulation of receptor and ligand synthesis based on parental experience.

The brain areas MPOA, BNST, and NAC exhibited substantial lateralization of the presence of aromatase-positive cells, biased toward the left hemisphere (Figure 3B). Left-hemisphere dominance has been noted in numerous contexts, including maternal perception and processing of pup-ultrasounds, in the latter case, through cFos expression (Ehret, 1987; Geissler et al., 2016). Maternal pup retrieval behavior likewise is associated with left- but not right-hemisphere oxytocin receptor expression and action (Marlin et al., 2015). To what extent can these sources of left-hemisphere bias in females (primarily auditory cortex) inform our findings? The left-hemisphere bias in females has primarily been observed in the auditory cortex (Geissler et al., 2016; Marlin et al., 2015). Lateralization of cFos expression has not been observed in the neural responses to pup presence in the limbic system of female mice, MPOA and BNST included (Geissler et al., 2013). However, neural activation characterized by cFos labeling concerns the input side of neurons (e.g., Sheng et al., 1990, 1993). Provided such absence of hemisphere differences in activation at the input side of the MPOA and BNST also existed in males/fathers, we can hypothesize that the greater left hemisphere aromatase expression in the MPOA and BNST may lead to a left-hemisphere bias of activation at the output of the MPOA and BNST after local estrogen action (enabled via aromatase) has taken place. Whether such a functional bias exists and how it may contribute to the preferred processing of pup ultrasounds to activate appetitive pup-searching behavior and possible left-hemisphere-dominant ultrasound perception in pup-experienced male mice similar to mothers remains to be shown.

The brain’s connectivity suggests that counts of active cells (e.g., c-Fos indicative of brain activation) within individuals may be correlated across regions. For example, aromatase could be uniformly lower or higher in some individuals than others or vary highly non-uniformly across brain areas. We expect that such variation will contribute to individual differences in behavior, such as some fathers continuing to ignore pups (Figure 1B). Our study found significant residual correlated variation among brain areas (Figure 4), generally not depicted in the data shown in Figure 3. Our analysis points to a central role of the LS in the variation of aromatase expression in the mouse brain. The LS has a central functional role in the brain, integrating cognitive and emotional information in the regulation of social relationships and instincts (Besnard and Leroy, 2022; Menon et al., 2022; Rizzi-Wise and Wang, 2021), including parental behavior (Fleischer and Slotnick, 1978; Slotnick and Nigrosh, 1975). Interestingly, these partial correlations (shown in Figure 4) reflect the main neural connections of the subregions LSD and LSV. For example, there is input from the amygdala to the LS supporting the LSD-CAM partial correlation, and there is an output from the LS to the MPOA supporting the LSV-MPOA partial correlation (Besnard and Leroy, 2022; Menon et al., 2022; Rizzi-Wise and Wang, 2021; Risold and Swanson, 1997). Thus, residual variation in aromatase expression (Figure 4) appears to mirror functional connections and relationships of the LS. Notably, the posterior PIR and, to a lesser degree, the anterior PIR and the MAM remained outside this connectivity. Indeed, Supplementary Figure S2 shows that posterior PIR (Bregma −1.34) was the sole case in which we observed substantial divergence in overall cell counts between the two cohorts of animals. Together, these observations on correlated individual variations indicate both coherent relationships between brain areas in a functional network and the presence of possibly unknown additional factors intrinsic to given experiments. In our present case, such factors modulated the numbers of aromatase-positive cells in PIR and MAM. Therefore, our study shows the usefulness of correlative modeling in assessing and estimating the relationships between the contributions of the members of a neuronal network and its overall functional properties. This is another example of correlative modeling supporting the adequate interpretation of absolute data.

Our results demonstrate that pup retrieval and aggression behaviors are closely related to the paternal experience groups (5D and 27D). Consequently, we aimed to investigate why some fathers do not appear to respond to their own pups, continuing to ignore them much like naïve males. Our analysis revealed that males who engaged with pups exhibited a significantly greater number of aromatase-expressing cells in the amygdala. Although tentative, these results (Figure 5) suggest that the amygdala, particularly the medial amygdala (MAM), may mediate robust emotional responses to pups. The longitudinal responses of males—both naïve and experienced fathers—to pups could clarify the repeatability of responses and whether the tendency to ignore is a transitional phase between aggression and retrieval or a distinct behavioral state. Such research would require a considerably larger sample size than was feasible in this study to effectively capture the natural variation in individual behavior.