Source-linked AI summary
Sex differences in intimate relationships
Vasyl Palchykov, Kimmo Kaski, János Kertész, Albert-László Barabási, Robin I. M. Dunbar
TL;DR
The paper examines how relationship preferences vary across the lifespan, using mobile-phone contact patterns to study sex differences. It reports stronger shifts in women’s relationship investment, including patterns the authors connect to pairbonds and parental investment.
Problem
The authors report previously unidentified sex differences in social and reproductive strategies and relationship investment across the lifespan.
Source: Discussion · p. 3Method
The study analyzes mobile-phone contact patterns, filtering for reciprocal contacts and applying additional filtering to same-age friend pairs.
Source: Discussion · p. 3, Initial data filtering · p. 4, Additional data filtering · p. 4Results
Women show stronger shifts in relationship preferences across reproductive and parental stages, while men maintain steadier patterns over time.
Source: Discussion · p. 3Takeaways & Limitations
The findings support the importance of female matrilineal relationships in human social organization.
Source: Discussion · p. 3Pooling individuals loses some of the richness of the original data.
Source: Discussion · p. 3
Abstract
from arXiv · showhide
Social networks have turned out to be of fundamental importance both for our understanding human sociality and for the design of digital communication technology. However, social networks are themselves based on dyadic relationships and we have little understanding of the dynamics of close relationships and how these change over time. Evolutionary theory suggests that, even in monogamous mating systems, the pattern of investment in close relationships should vary across the lifespan when post-weaning investment plays an important role in maximising fitness. Mobile phone data sets provide us with a unique window into the structure of relationships and the way these change across the lifespan. We here use data from a large national mobile phone dataset to demonstrate striking sex differences in the pattern in the gender-bias of preferred relationships that reflect the way the reproductive investment strategies of the two sexes change across the lifespan: these differences mainly reflect women's shifting patterns of investment in reproduction and parental care. These results suggest that human social strategies may have more complex dynamics than we have tended to assume and a life-history perspective may be crucial for understanding them.
Discussion
The authors interpret mobile-phone contact patterns as evidence that relationship investment varies by sex and age, with women showing stronger shifts across reproductive and parental stages. They also identify female matrilineal relationships as important to human social organization.
Discussion
- Women are more focused than men on opposite-sex relationships during reproductive years, which the authors interpret as heavier investment in pairbonds.This interpretation assumes mobile communication captures subscribers’ most important relationships and that contact strength reflects emotional closeness.Source: Discussion · p. 3
- Women’s attention appears to shift from spouses toward younger females, assumed to be daughters, gradually reaching a new asymptote around age 60.The authors suggest this gradual transition may reflect the arrival of grandchildren.Source: Discussion · p. 3
- Women switch people between preference ranks more than men, consistent with more pronounced shifts from mate choice toward personal and grandparental investment.The authors contrast women’s changing pattern with men’s steadier spouse preference and more even second-position gender balance.Source: Discussion · p. 3
- The findings support a particularly important role for mother-daughter relationships in human social organization, irrespective of dispersal pattern.The authors contrast this interpretation with emphasis on male-male relationships in patrilineal accounts.Source: Discussion · p. 3
- Pooling many individuals reveals lifespan patterns but loses some richness; the authors suggest analyzing contact directionality and temporal network motifs for further insight.The analyses focus on contact presence or absence.Source: Discussion · p. 3
Methods
The sample was concentrated among subscribers aged 20–60, but rarer age groups remained represented in substantial numbers.
Methods
- Although most recorded subscribers were aged 20–60, the saturation sample included about 5,000 people aged 60–80.The authors cite sample size and saturation as reasons these relatively rare age classes were still represented.Source: Methods · p. 3
Initial data filtering
The initial filtering restricted the analysis to subscribers with reciprocal contact and known demographic and subscription characteristics.
Initial data filtering
- Only calls and texts between people with at least one contact in each direction were retained to filter out spurious and professional calls.The passage gives accidental or wrong-number events and call-centre calls as examples.Source: Initial data filtering · p. 4
- Included subscribers had known age and gender and only one registered subscription.Source: Initial data filtering · p. 4
- The analysis treated the subscriber as the phone’s real user when no obvious deviation was observed.Source: Initial data filtering · p. 4
Additional data filtering
Additional filtering addressed an anomalous peak in same-age, same-gender ego-friend pairs that could reflect multiple subscriptions used by other people.
Additional data filtering
- A peak in same-age, same-gender pairs contradicted the expected similarity of same-age and one-year-different-age friend counts and could reflect multiple subscriptions.The suspected subscriptions might be registered to one person but used by children or partners who contact the ego.Source: Additional data filtering · p. 4
- For each ego age and gender, the procedure estimated real same-age, same-gender pairs using the larger count of pairs with friends one year older or younger.It then randomly retained that estimated number of same-age pairs to estimate the average gender of the best friend reliably.Source: Additional data filtering · p. 4
- The same filtering approach was repeated for each age-gender group and applied to second-best-friend pairs as well.Its effect is shown in Supplementary Fig. S2.Source: Additional data filtering · p. 4
Error estimation
The section estimates sampling uncertainty in best-friend gender by age and ego gender, using Bayesian credible intervals for the male-friend fraction and translating them to average gender.
Error estimation
- Finite-sample errors appear as error bars, with analysis focused on best friends of egos grouped by age and gender.The calculation is repeated for each age and gender group.Source: Error estimation · p. 4
- The male-friend fraction is estimated from the number of male friends among the remaining ego/friend pairs, with its likelihood-based estimate at m/n.The passage describes a Bayesian inference for the fraction of male best friends.Source: Error estimation · p. 4
- Credible bounds for the male-friend fraction are obtained using the incomplete regularized beta function and its inverse.The passage introduces an equation for the lower bound and obtains the upper bound similarly.Source: Error estimation · p. 4
- The credible interval for the fraction of male best friends is converted to an interval for average gender using f = (2x_min - 1, 2x_max - 1).The bounds are calculated for egos in each age and gender group.Source: Error estimation · p. 4
Additional information
The supplementary material extends the paper’s relationship analyses with population, filtering, contact-measure, friend-rank, and age-cohort views.
Additional information
- Subscriber counts favor males at every age (g = 0.13), while the set of best friends is very slightly female-biased.Supplementary Figure S1 contrasts subscriber age profiles with best-friend age profiles.Source: Figure S1 · p. 6, Additional information · p. 6
- For 50-year-old female egos, the original best-friend age distribution has a same-age, same-gender peak that filtering may reduce.The peak may arise from multiple subscriptions used by family members.Source: Figure S2 · p. 6
- Best- and second-best-friend gender patterns show generic features across definitions based on call counts, call duration, or text-message counts.Supplementary Figure S3 places best friends in the left column and second best friends in the right.Source: Figure S3 · p. 7
- Supplementary Figure S4 tracks average gender for third-, fourth-, and fifth-ranked friends by age and gender.Males are blue squares and females are red circles.Source: Figure S4 · p. 8
- Among egos aged 20, 30, and 40, best-friend age-and-gender distributions show multiple maxima and asymmetry in opposite-gender ages around ego age.Each data point gives the probability of a best friend of a specified age and gender.Source: Figure S5 · p. 8
- For egos aged 50, 60, and 70, best-friend distributions show increasing communication with the younger generation.Each data point gives the probability of a best friend of a specified age and gender.Source: Figure S6 · p. 9
- For second-best friends of similar age, younger egos more often select someone of the same gender, while older egos more often select someone of the opposite gender.The profiles resemble those for best friends, and each data point represents a probability for a specified friend age and gender.Source: Figure S7 · p. 10, Figure S8 · p. 11