Body odours as a chemosignal in the mother–child relationship: new insights based on an human leucocyte antigen-genotyped family cohort

Mothers are able to identify the body odour (BO) of their own child and prefer this smell above other BOs. It has hence been assumed that the infantile BO functions as a chemosignal promoting targeted parental care. We tested this hypothesis and examined whether children's BOs signal genetic similarity and developmental status to mothers. In addition, we assessed whether BOs facilitate inbreeding avoidance (Westermarck effect). In a cross-sectional design, N = 164 mothers participated with their biological children (N = 226 children, aged 0–18 years) and evaluated BO probes of their own and four other, sex-matched children. Those varied in age and in genetic similarity, which was assessed by human leucocyte antigen profiling. The study showed not only that mothers identified and preferred their own child's BO, but also that genetic similarity and developmental status are transcribed in BOs. Accordingly, maternal preference of their own child's odour changes throughout development. Our data partly supported the Westermarck effect: mothers' preference of pubertal boys' BOs was negatively related to testosterone for the own son, but not for unfamiliar children. This article is part of the Theo Murphy meeting issue ‘Olfactory communication in humans’.


Supplementary information on body odour matching (compare manuscript, method: BO matching):
For the own children, the mean similarity of shared HLA-B and -C alleles between mother and child across the whole sample was 2.2 alleles (SD = 0.53, ranging from 2 -4 alleles, with 87% sharing 2 alleles, 8 % sharing 3 alleles, and 5 % sharing 4 alleles). For shared HLA-B and -C alleles between mothers and matched body odour (BO) samples across all HLA-similar unfamiliar children, the mean similarity was 1.75 alleles (SD = 0.62; for same aged children: M = 1.8, SD = 0.59, for the different aged children: M = 1.65, SD = 0.64; ranging from 1 to 4 alleles, with 36 % sharing 1 allele, 59 % sharing 2 alleles; 3 % sharing 3 and 2% sharing 4 alleles). Between mothers and matched BO samples across all HLA-dissimilar unfamiliar children, similarity was 0, as previously defined.

Supplementary tables
Supplementary

Bayesian analyses (compare manuscript, method: statistical analyses)
In order to additionally explore the likelihood of our hypotheses, we used Bayesian statistics [1] using JASP software [2]. We included only H1 and the first assumption of H2 for that analysis, as for these hypotheses basic assumptions about the effect sizes are available based on a previous publication [3].

H1: Mothers are able to identify their own child above chance
We performed a Bayesian Binomial test using the identification of own child (yes/no) as the dependent variable: Based on a previous study [3] we assumed that the identification ability for the choice between three BO samples is at 75%; and rescaled this to our design with a choice between six BO samples, = 37.5%. We, therefore, did not follow the suggestions from Dienes & Mclatchie (2018, [1]) to divide the value by 2, because the result would then equal the chance level (null hypothesis = correct identification of child does not differ from chance level, chance level = 16.67 %). We hence tested evidence for the alternative hypothesis (correct identification of own child above chance) against 0.375. Additionally, we tested evidence for the null hypothesis against 0.167. We did this for each age group, as listed in the

H2: Mothers prefer the BO of their own child over other children
We performed the Bayesian Factor for a one sample t-test using the preference of the own vs. unfamiliar child as the dependent variable. Based on a previous study [3] we assumed that the difference in pleasantness ratings between the own and unfamiliar children´s BO is 1.3 points (pleasantness ratings on a scale ranging from 1 to 10); and rescaled this to our design (pleasantness ratings on a scale ranging from 1 to 100) thereby expecting a difference of 13 points.
According to suggestions from Dienes & Mclatchie (2018), we divided this value by 2; thus, we tested against 6.5.
Again, we tested evidence for the alternative hypotheses (significant difference between pleasantness ratings for own vs. unfamiliar children) against 6.5, and additionally, we tested null hypothesis (no difference between own and unfamiliar children) against 0. We performed the analyses for each age group. Results are presented in the The Table shows a BF below 3 for all groups except of age group 9-13 in comparison to the null hypothesis, indicating that the null hypothesis is not supported for three groups, but supported for 9-13 year old children. In addition, the BF is higher 3 for all groups except for age group 9-13, indicating that the alternative hypothesis is supported for those groups. For age group 9-13 the null hypothesis is supported, meaning that there is no pleasantness preference of the own child´s BO compared to unfamiliar children.