Mandrill mothers associate with infants who look like their own offspring using phenotype matching
- ISEM, Univ Montpellier, CNRS, IRD, EPHE, France
- Behavioral Ecology and Sociobiology Unit, German Primate Center, Leibniz Institute of Primate Research, Germany
- Projet Mandrillus, SODEPAL, Gabon
- SODEPAL-COMILOG, Gabon
- CEFE, Univ Montpellier, CNRS, EPHE, IRD, France
Figures
Figure 1
Summary of fixed effects' parameters included in the model analyzing average facial distance among infants.
For each estimate, the 50% (inner), 70% (middle), and 95% (outer) Wald confidence intervals are shown. Pink shades highlight variables of interest while blue shades correspond to control variables. The following variables are displayed: father identity (top three rows: infants born to different fathers -’Father:different’-; infants conceived during the same alpha male tenure -’Alpha:same’-; infants born to the same father -’Father:same’-; reference category: Alpha:different); mother identity (infants born to the same mother -’Mother:same’-; reference category: Mother:different); infants’ difference in age (‘Age difference’); infants’ difference in sexes (‘Sex:FM’ and ‘Sex:FF’; reference category: MM). Bold y-axis labels highlight variables with significant effects (p<0.05).
Figure 2
Summary of fixed effects' parameters included in the three models analyzing spatial association across dyads of (a) mothers and other infants; (b) infants; and (c) mothers.
For each estimate, the 50% (inner), 70% (middle) and 95% (outer) Wald confidence intervals are shown. Pink shades highlight the variable of interest (‘facial distance’) while blue shades correspond to control variables. Note that a negative estimate (as for ‘facial distance’) indicates a negative correlation between spatial association and that variable (‘facial distance’): individuals associate more with low values of ‘facial distance’ (high resemblance). The following variables are displayed: infants’ residuals of facial distance (‘Facial distance’); different vs. same mothers’ matrilines (‘Matriline:same’; reference category: different); mothers’ relatedness (‘Relatedness’); mothers’ difference in rank (‘Rank difference’); infants’ difference in age (‘Age difference’); infants’ difference in sexes (‘Sex:FM’ and ‘Sex:FF’; reference category: MM). Bold y-axis labels indicate variables with significant effects (p<0.05). Pictures depict three male infants with their average facial distances: B and C resemble each other most, in contrast to A.
Figure 3
Graphical representation of the ‘second-order kin selection’ process.
The actor (here a focal mother, face on the top) has the control over her behaviors with social partners (all three faces on the bottom including actor’s offspring, other kin and non-kin in red). Plain black arrows represent the inclusive fitness framework of the kin selection theory. Dashed red arrows represent the second-order kin selection process where an actor’s social behaviors towards a non-kin recipient (e.g. an offspring’s paternal half-sib) favor the latter’s social behaviors towards the actor’s kin (offspring or other kin). We provide a few examples (A: the mechanism explored in this study; B: a generalization of the mechanism to other actor’s kin) where this process may occur (and see discussion). Importantly, the second-order kin selection necessitates that non-kin recipient in red shares genetic or reproductive interests with the actor’s kin (double-arrows).
Appendix 1—figure 1
Fathers do not resemble their own offspring.
We used deep neural network analyses to estimate father-offspring resemblance (using the same methodology as for infant-infant resemblance; see main text) and did not find evidence that infants resemble their father. We considered the 32 infants for which the father was known (sired by 12 different fathers) and compared their facial distances across dyads of father-offspring and non-father-infant (N=47,724 pairs of pictures from a total of 590 male and infant portraits) using a LMM (response variable: facial distance; explanatory variables: infant’s sex and whether the male was the father or not). We considered the identity of the adult male and the identity of the infant as two random variables. We fitted a heteroscedastic residual variance with prior weights defined as the total number of pairs of pictures collected on each dyad of infants (giving more weight to those pairs with more numerous pictures). We found that infants do not resemble their father (‘dad’; figure above) more than any random male (‘no dad’; N=352 pairs; p=0.39 by likelihood ratio test; there is also no effect of the sex of the infant: p=0.74). The highly pronounced sexual dimorphism and morphological differences between immatures and adults in this species likely explain this result.
Appendix 1—figure 2
Mother-offspring spatial association across offspring’s age.
The figure above represents the percentage of scans during which the mother was associated (0–5 m) to her offspring averaged across different offspring’s age classes (in days). The figure was obtained from a total of 11,926 scans performed on 217 different infants aged 0–1 yr (for 11,003 scans, the mother was 0–5 m away from her offspring). Sample sizes are as follow: 0–30 days: 2,255 scans; 31–90 days: 3,549 scans; 91–180 days: 3,148 scans; 181–300 days: 1,983 scans; 301–365 days: 991 scans. Although, this pervasive mother-offspring association should, intuitively, translate into high mother-mother association (that we do not observe with respect to infant-infant facial resemblance), the variance observed is high, suggesting strong variation in association patterns across mother-offspring dyads. In addition, mothers may be located 5 m away or less from their offspring and from other similar-looking infants without being less than 5 m away from these infants’ mothers.
Appendix 1—figure 3
Correlation between infant-infant association rate and mother-other infant association rate.
For the 282 study pairs of infants, we calculated the raw rate of association per dyad of infants (total number of scans where both infants were in proximity divided by the total number of scans performed on the two members of the dyad) and plotted it against the average raw rate of association of each mother with the other infant. We found that both association rates are highly positively correlated (Pearson correlation test: N=282, r=0.85, p<0.0001).
Appendix 1—figure 4
Infant-infant association rates a year later.
We split our initial data set on infant-infant associations between those dyads that were associated early in life (0–1 year) and those that were not. We then retrieved the association rates (number of scans where both individuals were recorded in proximity divided by the total number of scans performed on them both) of these infants (those that survived and for whom we had detailed data on spatial association), a year later (when they were aged 1–2 years). Infants that did associate at 0–1 year also associate, on average, three times more at 1–2 years than those that did not. Sample sizes are provided within bars and represent the number of dyads of juveniles aged 1–2 years. This result, although preliminary, suggests that early association during infancy may pervade later in life.
Appendix 1—figure 5
Learning curve for the face identification task.
Evolution of accuracy for the training set (in blue) and validation set (in orange) during the run that yielded the highest accuracy (93.42%). The small difference between the training accuracy and the validation accuracy indicates limited or no effect of overfitting.
Tables
Table 1
Predictors for facial distance among infants.
Significant predictors (p<0.05) are in bold (LMM with a Gaussian response).
| Infants’ facial distance N=2556 dyads | |||||
|---|---|---|---|---|---|
| χ2 | p > χ2 | Estimate | Standard error | ||
| Predictors | Father/alpha male at infants’ conception* Same alpha vs. same father Different alphas vs. different fathers Same alpha vs. different alphas | 28.07 5.87 1.06 8.16 | <0.001 0.015 0.30 0.004 | Same alpha: –0.24 Different father: 0.10 Same father: –0.67 | 0.08 0.09 0.17 |
| Infants’ mother(s)† | 0.32 | 0.57 | 0.12 | 0.19 | |
| Infants’ difference in age | 1231.8 | <0.001 | 1.18 | 0.03 | |
| Infants’ sex‡ | 0.84 | 0.66 | FF: –0.36 | 0.41 | |
| FM: –0.17 | 0.21 | ||||
-
*
Reference: different alpha males.
-
†
Reference: different mothers.
-
‡
Reference: MM (male-male).
Table 2
Predictors of the spatial associations recorded across mother-infant, infant-infant and mother-mother dyads.
Significant predictors (p<0.05) are in bold (GLMM with negative binomial response family and log link). The reported dispersion parameter is the so-called shape parameter of the negative binomial distribution.
| Mother-infant associationN=580 dyadsDispersion parameter = 2.43 | Infant-infant associationN=282 dyadsDispersion parameter = 2.635 | Mother-mother associationN=325 dyadsDispersion parameter = 1.281 | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| χ2 | p > χ2 | Estimate | Standard error | χ2 | p > χ2 | Estimate | Standard error | χ2 | p > χ2 | Estimate | Standard error | ||
| Predictors | Infants’ facial distance (residuals) | 7.72 | 0.005 | –0.14 | 0.05 | 5.30 | 0.021 | –0.16 | 0.07 | 1.33 | 0.248 | –0.08 | 0.07 |
| Infants’ matriline* | 18.11 | <0.001 | 0.70 | 0.15 | 7.85 | 0.005 | 0.61 | 0.21 | 3.45 | 0.063 | 0.44 | 0.23 | |
| Mothers’ relatedness | 31.17 | <0.001 | 0.27 | 0.05 | 24.04 | <0.001 | 0.32 | 0.06 | 17.01 | <0.001 | 0.31 | 0.07 | |
| Mothers’ rank difference | 37.47 | <0.001 | –0.37 | 0.06 | 10.74 | 0.001 | –0.28 | 0.08 | 28.49 | <0.001 | –0.44 | 0.08 | |
| Infants’ age difference† | 13.77 | <0.001 | –0.23 | 0.06 | 10.49 | 0.001 | –0.27 | 0.09 | 6.20 | 0.013 | –0.20 | 0.08 | |
| Infants’sex† | 1.39 | 0.499 | FF: –0.20 | FF: 0.21 | 0.66 | 0.719 | FF: –0.21 | FF: 0.27 | 8.94 | 0.011 | FF: –0.72 | FF: 0.25 | |
| FM: –0.18 | FM: 0.15 | FM: –0.11 | FM 0.21 | FM: –0.50 | FM: 0.19 | ||||||||
-
*
Reference: different matrilines.
-
†
Reference: MM (male-male).
Appendix 1—table 1
The first table (a) presents possible mechanisms of paternal kin recognition to explain increased nepotism among paternal half-sibs in mandrills (Charpentier et al., 2007; Charpentier et al., 2020), and their associated empirical evidence (or lack thereof) with respect to an increased facial resemblance among paternal half-sibs; the second table (b) lists the empirical evidence (or lack thereof) for different possible evolutionary functions.
| a. | ||
|---|---|---|
| Possible mechanism(s) | Empirical data in mandrills | Likelihood in paternal half-sib mandrills |
| Self-phenotype matching | Absence of any physical substrate (mirror) to allow facial self-recognition | Low. In addition, facial similarity as a by-product of selection on other self-evaluable phenotypic traits (e.g. odours) would not explain increased facial resemblance among paternal half-sibs |
| Phenotype matching with father as a template | Adult male mandrills do not resemble their offspring at any age because of pronounced sexual dimorphism (see Appendix 1—figure 1) | Low. |
| Familiarity mediated by fathers: Males know with whom they reproduced and associate either with the potential mothers or their offspring (or both) following births favoring secondary association among paternal half-sibs | Male mandrills are not responsible for establishing spatial associations with either females or infants (MJEC and BRT, Pers. Obs.) -Half of the male mandrills present during a reproductive season are absent from the group the next birth season (this study) | Low. In addition, these two mechanisms alone fail to explain increased facial resemblance among paternal half-sibs that necessarily involves some forms of phenotype matching |
| Familiarity mediated by mothers: Females know with whom they reproduced and associate either with the father, the other females that also reproduced with the same father or their offspring (or a combination of all three) following births, favoring secondary association among paternal half-sibs | -Mother-mother association and grooming relationships do not depend on facial similarity among offspring (this study) -Possible memory and field constraints as females would need to remember all copulation events that occurred with other females in a dense habitat | |
| A mix between familiarity and phenotype-matching+father mediation: Males know with certainty at least one paternity (e.g. thanks to patterns of copulations) and associate either with the mothers or their offspring (or both) following births when these offspring resemble more to their own (certain) offspring, favoring secondary association among paternal half-sibs | -Male mandrills are not responsible for establishing spatial associations with either females or infants (MJEC and BRT, Pers. Obs.) -Half of the male mandrills present during a reproductive season are absent from the group the next birth season (this study) | Low. This mechanism may explain increased facial resemblance among paternal half-sibs but is not parsimonious given males’ patterns of residency and the fact that males are not responsible for establishing proximities with infants or their mothers. In addition, it requires at least one event of paternity certainty. This mechanism also fails to explain why mothers associate more with strange but resembling infants but not with their mothers if fathers mediate these associations |
| A mix between familiarity and phenotype-matching+mother mediation: Mothers associate either with the mothers or their offspring (or both) following births when these offspring resemble their own (certain) offspring more, favoring secondary association among paternal half-sibs | Mothers associate more with strange infants that resemble their own offspring more, possibly favoring secondary association among paternal half-sibs (this study) | High. This mechanism may explain increased facial resemblance among paternal half-sibs and their long-term nepotism. Mother mediation could have also been selected to favor paternal care and offspring protection against infanticide (patterns of mother-infant or infant-infant association would be by-product of association to a father), but it would not explain alone the increased facial resemblance among paternal half-sibs |
| b. | ||
| Evolutionary functions | Empirical data in mandrills | Likelihood in mandrills |
| Nepotism | -Paternal half-sibs are numerous in the group (>2 times more numerous than maternal half-sibs) because of high male reproductive skew (Charpentier et al., 2020; Charpentier et al., 2005) -Nepotism occur among paternal half-sibs in juvenile and adult female mandrills (Charpentier et al., 2020; Charpentier et al., 2007) | High. Potentially elevated fitness benefits to recognize and favor paternal half-sibs and strong empirical support |
| Inbreeding avoidance | -Alpha males’ tenure (<15 months) and males’ length of stay in the group (<23 months for males aged ≥10 yrs) are restricted -Sex-biased dispersal: most males emigrate before being reproductive (<7 yrs; males start reproducing around 10 yrs) | Low. Father-daughter reproduction is highly unlikely; inbreeding among paternal half-sibs may occasionally occur if males reproduce with their sisters before emigrating; mothers should avoid rather than favor association with highly resembling infants; other mechanisms have probably evolved, such as sex-biased dispersal |
| Paternal care and offspring protection against infanticide | -Social relationships between adult males and infants are highly limited (e.g. absence of affiliation) although, in captivity, fathers are spatially closer to their own offspring than to unrelated juveniles (Charpentier et al., 2007) -There is indirect evidence of infanticide in mandrills -Only 54.5% of alpha males and 44.7% of subordinates are present during the birth season following the reproductive season they experienced (based on 69 male.years) -When present, males stay less than a year in their offspring’s group (9.5 months on average for alpha males; 5.5 months for subordinates; based on 33 male.years) | Medium. Patterns of male residency do not offer strong support for this hypothesis, but paternal care may occur early in life in the form of increased spatial association or support during agonistic interactions; offspring protection against infanticide may also occur |
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Mandrill mothers associate with infants who look like their own offspring using phenotype matching
eLife 11:e79417.
https://doi.org/10.7554/eLife.79417
