A growing body of evidence indicates that in many cultural settings fathers have lower T than non-fathers 
, including recent evidence that first-time fatherhood causes men's T to decline over a multi-year period 
. There are also cross-cultural data showing that fathers' involvement in the daytime care of their children may be a principal determining factor in whether and the degree to which fathers have lower T than other men 
. However, nothing is known about the role that nighttime father-child proximity may play in influencing paternal T. Thus, we used data from a large sample of 25–26 year old fathers in Metro Cebu, Philippines to test whether fathers' T differed if they coslept with their children at night or slept separately from them. A very high percentage (92%) of fathers in this sample reported practicing same surface cosleeping. We found that fathers who slept on the same surface as their children had significantly lower PM T and a significantly greater diurnal decline in T from waking (AM) to bedtime (PM) compared to men who slept solitarily. Drawing on men who became first-time fathers during the 4.5-year study period, we also showed that men's baseline T did not predict whether they practiced solitary sleeping or cosleeping at follow-up. However, among these same men, those that coslept with their children had a significantly greater longitudinal (over 4.5 years) decline in PM T compared to new fathers who slept separate from their children. These results are the first to show that T is comparatively lower among fathers practicing same surface cosleeping compared to solitary sleeping fathers and suggest that cosleeping may cause T to decline and/or remain relatively low.
In combination with limited ethnographic observations 
, a recent cross-cultural internet-based survey that sampled parents in the Philippines 
, and a previously published report that 99% of infants slept with their mothers in Cebu 
, our results indicate that familial cosleeping is culturally normative in the Philippines. Although in many popular and scholarly publications on familial sleeping practices same surface cosleeping is reduced simply to the label “bedsharing” 
the practice of parents and children sleeping together takes hundreds of diverse forms across cultures, most of which are not ordinarily characterized by the use of a “bed,” in the sense of the elevated mattress and box spring commonly used in the United States and much of Europe, Australia, and Canada. In this study, our respondents replied to a question using the Cebuano term “higdaanan
,” which literally translates as “something you sleep on, whether a bed, a mat, or a mattress on the floor.” Thus, though we cannot distinguish the specific forms of same surface cosleeping in this sample, it most likely involved fathers sleeping with their partners and child(ren) on various kinds of mats, thin mattresses, or blankets on the floor of their homes.
In American and many European societies, same surface cosleeping is a controversial practice. Certain medical and public health organizations argue that it poses a risk to child health under all circumstances 
. Elsewhere it has been levied that familial cosleeping may also negatively affect marital quality and parental sleep 
, though evidence from scientific studies to-date suggest that families that routinely sleep with their children generally avoid such relationship discord and sleep quality problems 
. In the health-related parameters we assessed, we found no significant differences between fathers based on their familial sleep practices for self-reported psychosocial stress, sleep duration, or sleep quality, all of which could potentially affect T, but we did observe a trend indicating that same surface cosleepers had a tendency to have been ill at a higher rate between 2007–2009. We also found a statistical trend for cosleeping fathers to be less likely to have sex with their partners on a weekly basis compared to solitary sleeping fathers. The extent to which this might be disruptive to marital cohesion in this cultural setting is presently unclear but warrants exploration in future studies.
Previously, in a separate study of men from this same sample, it was shown that men's T as single non-fathers did not predict their caregiving levels 4.5 years later at follow-up and that fathers who were the most involved with childcare at follow-up had the lowest T 
. Here we document similar findings in relationship to familial sleeping practices. Among men who transitioned from being non-fathers at baseline to being new fathers at follow-up, baseline T did not predict whether they coslept or slept separately from their children at follow-up. We also found that cosleeping fathers had a significantly greater longitudinal decline in PM T compared to solitary sleeping fathers, whose PM T increased, on average, between baseline and follow-up. Together these findings are suggestive that the lower evening T among same surface cosleeping fathers resulted from these fathers sleeping in close proximity to their children at night. Limited results from other species show that high T interferes with paternal investment, leading to lower offspring growth and reduced survival 
, and preliminary evidence from studies of human males suggests that lower T men are more sensitive to child needs 
. Thus, it is possible that decreases in paternal T associated with cosleeping could have beneficial implications for children. There is evidence from industrialized societies in many parts of the world that children of highly invested fathers fare better in many developmental domains, including, for example, greater self-esteem and socialization skills, higher academic performance, and lower delinquency 
. Thus, though it remains speculative at this juncture, lower T could amplify the beneficial effects of daytime paternal care and nighttime cosleeping, facilitating and/or enhancing fathers' responses to their children in those contexts, thereby contributing to better child health and development outcomes.
It is also plausible that same surface cosleeping fathers have lower T as a result of sleep disruption that is not experienced by solitary sleeping fathers. Notably, we found no differences in self-reported sleep duration or sleep quality based on familial sleeping arrangement in our study, suggesting that these factors are unlikely to account for the documented differences in T, although issues of self-report reliability in these domains have been raised 
. In addition, though no polysomnographic studies have been done on cosleeping fathers, prior research comparing polysomnography data from routinely bedsharing and solitary sleeping mothers revealed that bedsharing mothers had more transient, microarousals than mothers sleeping alone, but the two groups did not differ in time awake after sleep onset 
. While it remains to be seen whether similar polysomnographic-observed arousal patterns translate to fathers, laboratory studies have shown that extreme methods of sleep fragmentation, i.e. waking men up every 20 minutes throughout the night, lead to reduced T production 
. Further research is needed, generally, to assess causal relationships between naturalistic arousal patterns and men's T and, specifically, in the context of familial sleeping arrangements.
Fathers who practiced same surface cosleeping also showed a significantly greater diurnal decline in T from waking (AM) to evening (PM) compared to the decline seen in solitary sleeping fathers. It is noteworthy that while AM T did not differ by sleeping arrangement both the diurnal decline and PM T did. Recent studies have shown that short-term (~20–30 min) periods of father-child interaction have almost no immediate effect on paternal T 
. Thus, it is possible that the effects of father-child interaction in reducing fathers' T may take hours to come to fruition, which could help explain why same surface cosleeping fathers show a steeper diurnal decline in T and lower PM T overall, before going to bed, rather than an immediate effect on their AM T.
The physiological pathways through which this delayed effect might be possible are not well understood. It seems likely the process would be mitigated via the hypothalamic-pituitary-gonadal (HPG) axis, rather than through alternative physiological pathways that might rapidly affect circulating unbound T 
. If father-child sleep proximity causes changes in neurobiological function that reduce hypothalamic production of gonadotropin-releasing hormone (GnRH) and/or pituitary production of luteinizing hormone (LH), perhaps through the downstream effects of neurotransmitters/neurohormones such as dopamine, serotonin, norepinephrine, and/or endogenous opiates, reduced T might not be observed until later sampling, such as our PM saliva collection. This would be generally consistent with the previous proposal that AM T levels reflect circadian-sleep biology and are more impervious to social stimuli in humans and other hominoids whereas PM T is more responsive to social and behavioral context 
. That said, there is generally thought to be a 40–50 minute delay between changes in LH production and output of T from the testicular leydig cells 
, so the neuroendocrine mechanisms by which close nighttime father-child sleep proximity might cause sustained lower production of T over the course of the day remain to be elucidated. Ideally, future studies will be able to integrate both nighttime laboratory-based hormonal analysis and daytime sampling in order to track men's T (and other biomarker) changes overnight as they sleep near their children (or not) as well as how their hormones then shift over the course of the day after waking.
We also found preliminary evidence that the effects of having young, infant-aged (1 year old or less) offspring may affect men differently based on their familial sleeping arrangements. Men who practiced same surface cosleeping had lower T regardless of whether their youngest child was an infant or older than a year when compared to solitary sleeping fathers whose youngest child was older than 1 year. Solitary sleeping fathers of infants also had lower T than their solitary sleeping counterparts without infant-aged children, though there were few men in this category (n
6). Although these analyses are somewhat limited by small sample sizes in the solitary sleeping categories, our findings tentatively suggest that same surface cosleeping fathers may maintain lower T regardless of whether they have especially young children whereas solitary sleeping fathers' T may increase once offspring move out of infancy and become toddlers and beyond. These possibilities merit exploration in future longitudinal research.
This analysis has limitations that warrant mentioning. First, we asked fathers only about their sleeping arrangements on the night before salivary sampling, not their habitual activities. To our knowledge, families practicing cosleeping in Cebu generally do so regularly and for the entirety of the night, at least in part because of household space constraints. This differs from a common practice in more affluent societies in which infants and young children have their own bedrooms and are brought to the parents' room for portions of the night before being returned to their own sleeping quarters 
. Thus, though presently there are no relevant scientific data from the Philippines available on the subject, it seems likely that men's self-reports of their sleeping arrangements from the prior night are generally reflective of their more routine sleeping arrangements. However, there is also a chance that fathers who normally slept separately from their children might have been near them the night prior to sampling because of some extenuating circumstances, e.g. a distressed child or maternal absence. Thus, it is possible that the “solitary sleeping” category is mildly underrepresented, though we think it unlikely.
Second, as we noted in the Methods, we collected the PM sample the night before the collection of the AM sample in order to reduce participant burden and minimize interviewer field logistics. This study design allowed us to schedule single interviewer follow-up visits to each participant to collect both tubes. Prior research suggests that salivary T measures represent relatively stable baselines for subjects 
, which we take to be true here, particularly given our study's sample size. We also relate the PM and AM T values to a behavioral measure (whether men cosleep with their offspring) that is believed to be habitual in this cultural context. Consequently, we think that this approach allows us to capture relatively stable between-individual hormonal variability based on familial sleeping patterns. However, we also realize this design differs from studies that conduct AM and PM sampling on the same day. Our approach renders the interpretation of the diurnal change in T potentially more difficult, as our calculation is not change over a single day (see Methods), diverging from other studies [e.g. 64]
. However, if our T measurements are representative of relatively stable, day-to-day hormone levels, this issue is largely ameliorated. In total, we think it unlikely that an alternative research design, with both samples collected on the same day, would have substantially altered the results of the study or our interpretations thereof.
Finally, we collected single measurements of T at waking and in the evening for both the baseline and follow-up surveys 
. However, our single measurements of saliva do not introduce bias, but merely reduce the reliability of our biomarker measures and thereby limit our ability to detect relationships between hormones as well as with other variables. The relatively modest R-squared values of our regression models would also be expected to increase with greater measurement reliability. That said, the effect sizes (β coefficients) for our significant regression models are sufficiently large to suggest that our results are biologically meaningful 
. For example, in our study, bedsharing fathers had 19% lower PM T, on average, compared to solitary sleepers (Cohen's d
0.55, based on group means and pooled SD), which is similar to the percentage differences in T observed among men before and after engaging in competition 
or being exposed to visual sexual stimuli 
and is comparable in magnitude to the differences in T between men who engage in high versus low risk taking 
. The impact of low measurement reliability in our study was partially compensated for by the fact that we collected saliva samples at standardized times in a sample of men that exceeds the size of most prior studies of human male socioendocrinology.
In summary, our study is the first to test for relationships between cosleeping and paternal physiology, showing that fathers who slept near their children on the same surface had lower evening T and greater diurnal declines in T compared to fathers who slept separately from their children. In addition, we showed that fathers' T might respond differently as children age based on how families sleep, as same surface cosleeping fathers maintained lower T regardless of whether they were fathers to infants whereas solitary sleeping fathers with older children had higher T. These results are generally consistent with the idea that human paternal physiology has an evolved capacity to respond to childcare and direct contact with children 
. Future studies are needed to clarify whether differential effects of cosleeping versus solitary sleeping on men's T might influence their effectiveness as caregivers and potentially affect their children's development.