In this manuscript, we examined paternal age effects on offspring fitness. As an index of fitness, we mainly focused on the offspring’s number of children. Our goal was to isolate the mutational aspect of paternal age effects. To do so, we needed to rule out many alternative pathways in which paternal age might be associated with offspring fitness. Most importantly, we compared full siblings in a multilevel
regression model and adjusted for the average paternal age within the family.
Because all children of a couple have the same random chance to inherit some of their genes, this approach allowed us to rule out that less fit fathers simply found partners later in life and their children inherited low fitness. However, we also had to
24 We examined four different populations. One was 20th-century Sweden, our data were based on governmental records. The other three populations were from pre-industrial times (1720-1850). Church records were digitalised and used to
reconstitute genealogies for the Saint-Lawrence valley, Québec (Canada), the Krummhörn (Germany) and four historical Swedish regions.
From our analyses of these genealogies, three main conclusions are relevant to this dissertation.
First, we found negative paternal age effects in all four populations that we
examined. They were small, as predicted, but remained after adjusting for a lot of potential confounds in our robustness checks.
Second, average paternal and maternal ages at birth rose in 20th-century Sweden from 1970 onwards. However, from 1930 until 1970 they dropped. In 2010, they were at similar levels as in 1930 (around 33). More interestingly though, average parental ages were still below historical averages of the three pre-industrial
populations that we also examined. This seems counter-intuitive only because most previous studies focused on maternal age at first birth. Compared to a historical baseline, the age at first birth is indeed delayed, but because people are also having far fewer children on average, the age at last birth and the average age at birth are earlier.
Third, differences between paternal age effects in the pre-industrial populations and 20th-century Sweden were substantial for infant mortality but not for the aggregate
25 clear differences between populations. Although paternal age predicted increased infant mortality in all four populations, infant mortality on average is so much lower in 20th-century Sweden that the effect was insubstantial in comparison to the other populations. However, when examining the paternal age effect on aggregate
offspring reproductive success (including low reproductive success caused by early mortality), two things became clear: 20th-century Sweden did not stand out as exhibiting the smallest effect size, the effect size in Québec was smaller. Moreover, across 26 different model specifications all of which had some degree of plausibility, the effect sizes varied more than across populations. Because we probably cannot identify one true, best model, we cannot clearly conclude that selection against mutations is relaxed.
We were interested in the question whether relaxed postnatal survival selection is compensated by sexual selection later in life. However, we only had data on marriage and divorces, which are poor indices of mating success in 20th-century Sweden, because marriage is no longer a social or legal prerequisite for being in a relationship, cohabiting, or having children. Further, because we had no data on abortions, we could not clearly conclude whether infant survival selection was truly relaxed or displaced to before birth. Approximately 20% of all pregnancies are aborted in the modern Western world, but only few of these abortions are
“therapeutic”, i.e. aim to end a pregnancy where a potential birth defect was detected. Still, because the paternal age effects we found were quite small, these abortions might explain (part of) the difference.
26 In this study, we collected daily online self-reports from a large sample of women. A final sample of 1043 women filled out a short survey every day until they had
contributed up to 40 days.
Our goal was to replicate and extend previous studies’ reports that women’s sexual interests change around ovulation. To this end, we asked our participants about their menstruation dates and contraceptive methods in the study. From the menstruation dates, we could then estimate the probability of being in the fertile window for each diary day. In multilevel models, we then tested whether being in the fertile window was associated with psychological changes. We also tested whether that such changes were absent among hormonal contraception users, who do not experience ovulation and the concurrent hormonal changes.
A key theoretical prediction was that women’s assessment of their partners’
attractiveness for a short-term sexual relationship should moderate the shifts in sexual desire in such a fashion, that women with more attractive partners experience increases in in-pair desire, but not extra-pair desire and vice versa for women with less attractive partners. The purported evolutionary function of this moderation pattern is to obtain good genes for the offspring.
Previous studies had supported this prediction in small studies, but many
methodological criticisms of the previous literature were raised. Namely, previous studies had often used small sample sizes, often gathered data from women only on one or two days and did not preregister their methodology. This combination of problems is now thought to lead to overestimation of effect sizes and false positives.
27 We found ovulatory changes, that is peaks in the fertile window restricted to non-hormonal contraceptive users, for several outcomes. Namely, we replicated changes in extra- and in-pair sexual desire and behaviour, and in self-perceived sexual
desirability.
However, we did not confirm the predicted moderation patterns. Even though our sample size of naturally cycling women was larger than the combined sample sizes of previous studies and about ten times larger than the average previous study on the subject, we found no significant moderation patterns.
Previous studies had mostly excluded women using hormonal contraception from participating to save costs. Our online approach allowed us to include them in our study and directly test whether the ovulatory changes observed among naturally cycling women were absent. They were. Hence, it seems possible that hormonal contraception would flatten cyclical changes in mate preferences. However, we found no evidence for such changes when we examined the moderation of sexual desire changes by the partner’s short-term attractiveness. We would have predicted hormonal contraception users to permanently have the sexual desire of naturally cycling women in the luteal phase.
Future studies should examine whether mate preferences change across the cycle at all, in studies that also include single women. If they do not, then hormonal contraception is unlikely to have any effect on mate preferences either, although randomised controlled trials are necessary to rule this out with finality.
28 menstrual cycle. Mean levels may also be affected, but experimental studies are necessary to test this because of confounding variables. In our study, women using hormonal contraception differed from non-users not only in contraceptive method but also in other ways, mainly in age. Unfortunately, existing randomised
placebo-controlled trials of hormonal contraceptives usually ignore the menstrual cycle.
Package leaflets for hormonal contraception currently point out changes in libido as potential side effects, but are very unspecific. The reason for this might be that there is large heterogeneity in how sexuality varies across the cycle when not taking hormonal contraception, and thus, response to it may also vary across individuals.