The role of oxytocin on self‐serving lying

Abstract Introduction The effects of intranasal administration of the neuropeptide oxytocin on social cognition and behavior are highly specific. Potentially situational and personal variables influence these effects. The aim of the present study was to investigate effects of oxytocin administration on self‐serving lying, including situational effects. Methods A total of 161 adult males participated in a randomized double‐blind placebo‐controlled between‐subject intranasal oxytocin administration (24 international units) study. Self‐serving lying was assessed using three subsequent rounds of the die‐in‐a‐cup paradigm, in which different degrees of lying can be implemented by the participants that can be determined on group level. Results Oxytocin administration seemed to promote self‐serving lying, particularly in the third (last) round and only to a certain degree (not to the maximum possible). Conclusions Our findings demonstrate that oxytocin administration can promote self‐serving lying when given repeated opportunities to lie. Moreover, exploratory results presented in the Supplementary Material indicate that the sensitivity to the effects of intranasal oxytocin in this domain might be moderated by individual differences in the oxytocin receptor gene.

Chi 2 = 0.03, p = .874 Note. The differences in the total N are due to not detected genotypes in some SNPs for some participants; alleles are derived from 5'-3' strand.
Regarding the rs237887-rs2268491-rs2254298 haplotype block, five individual haplotypes were detected with only three individual haplotypes of interest due to the number of people carrying the respective individual haplotypes: the GCG-, GTA-and the ACG-individual haplotypes. Within the rs53576-rs2268498 haplotype block, four individual haplotypes could be detected with also three individual haplotypes of interest: the AC-, AT-and GT-individual haplotypes. Compared to the GCG-, GTA-, ACG-(rs237887-rs2268491-rs2254298), AC-, AT-, and GT-(rs53576-rs2268498) individual haplotypes the other individual haplotypes were only rarely found (<5%). Therefore, statistical testing would not be meaningful, which is why they were not considered in further analyses. Haplotype frequencies are presented in Supplementary  Table 3 shows the number of participants in each (sub-)sample analyzed with regard to the behavior in the die-in-a-cup paradigm.
Of note, the genotypes of each individual SNP are not equally distributed for most of the SNPs (see Supplementary Table 1). Accordingly, when splitting the sample into genotype and treatment groups some of the resulting subsample sizes are extremely small (e. g. 6 participants carrying the AA genotype in the rs2254298 and receiving OXT treatment). Therefore, it was decided to not report results of interaction effects between treatment (PLC vs. OXT) and genotype of each single SNP. Instead, it was decided to focus on the results regarding the individual (OXTR) haplotypes.
Of final note: The experiment was carried out blinded for the genetic data.

Confounding variables
First, the groups of carriers versus non-carriers of all six individual haplotypes were compared regarding age and the Honesty-Humility (sub)scales of the HEXACO-PI-R (Lee & Ashton, 2016). The HEXACO-PI-R personality trait questionnaire was assessed during the CGBBP and was therefore not influenced by treatment; hence, examining genetic effects across treatment groups is justified. These analyses were implemented using Mann-Whitney U tests.
The groups split by PLC and OXT treatment were compared as explained in the main manuscript.
Finally, the groups split by PLC versus OXT treatment and individual haplotype carriers versus noncarriers (2×2 designs) were compared in light of all the previously mentioned possible confounding variables (always comparing 4 groups: PLC, non-carriers vs. PLC, carriers vs. OXT, non-carriers vs. OXT, carriers; for each of the six individual haplotypes) using Kruskal-Wallis tests.
The die-in-a-cup paradigm Based on the round-specific effects of OXT on lying behavior (see main manuscript), the effects of PLC versus OXT treatment and the six individual haplotypes were analyzed on each round separately.
Therefore, the distributions of reported numbers in each separate round split by treatment (PLC vs. OXT) and individual haplotypes (carriers vs. non-carriers; 2×2 designs) were investigated. To test statistically for significance of the deviations from the expected equal distribution, Chi 2 -tests were calculated. If Chi 2 -tests revealed significant (p < .05, two-tailed) deviations, the observed frequencies of each individual number were compared with the expected frequency (1/6 th ) using binomial tests (two-tailed) (see for example Wibral, Dohmen, Klingmüller, Weber, and Falk (2012) for a similar approach). Chi 2tests were also implemented to compare the distributions found in carriers of the respective individual haplotype in the OXT versus the PLC group and the distributions found in non-carriers of the respective individual haplotype in the OXT versus the PLC group.

Supplementary Material
Results:

Confounding Variables
No significant differences in age or the HEXACO-PI-R (Lee & Ashton, 2016) Honesty-Humility (sub)scale(s) were observed between the individual haplotype groups (for each of the six individual haplotypes comparing carriers vs. non-carriers), which would hold after correction for multiple testing (0.05 / 6 = 0.0083; divided by six because six individual haplotypes on the OXTR gene were investigated) (all p-values > .035).
No significant differences between OXT and PLC group were observed in the possible confounding variables (see paragraph 3.1 in the main manuscript).
In the groups split by treatment (PLC vs. OXT) as well as carriers and non-carriers of each of the six individual haplotypes (2×2 designs), Kruskal-Wallis tests showed no significant difference in the possible confounding variables, which would hold after correction for multiple testing (0.05/(6×2) = 0.05/12 = 0.0042; divided by 6×2 because six individual haplotypes on the OXTR gene were investigated in two groups each (PLC and OXT) -all p-values > .008). As a result of this, it was decided not to include these variables as confounding variables in further analyses.

Results of effects of treatment and individual haplotype on each round separately
In the PLC group, no meaningful effect of any individual haplotype (investigating carriers and noncarriers) was found on any round (only GT non-carriers under PLC showed a deviation from the equal distribution in the third round with a p = .049; the other respective Chi 2 -tests revealed p-values > .083).
In the OXT group significant results were observed, particularly in the third round (only one significant effect in the second round (p < .05), but none in the first round). In detail, a significant deviation of the distribution of numbers reported in the third round from the equal distribution could be detected in several groups. No significant differences were observed in the distributions between OXT group and PLC group (comparing the respective carriers (with carriers) and non-carriers (with non-carriers) of the OXT and PLC groups) in the first and second round. However, several significant differences were observed in the third round. These results are presented in Supplementary Table 4

Supplementary Figure 2. Distributions of numbers reported in the 3 rd round (in %) in GCG individual
haplotype (rs237887-rs2268491-rs2254298) carriers and non-carriers in the OXT group. Binomial tests were only calculated for the carriers group as only in this group the Chi 2 -test revealed a significant deviation from the equal distribution: *p < .05, **p < .01, ***p < .001 (two-tailed); n = number of participants in the respective group.

Supplementary Figure 3. Distributions of numbers reported in the 3 rd round (in %) in GT individual
haplotype (rs53576-rs2268498) carriers and non-carriers in the OXT group. Binomial tests were only calculated for the non-carriers group as only in this group the Chi 2 -test revealed a significant deviation from the equal distribution: *p < .05, **p < .01, ***p < .001 (two-tailed); n = number of participants in the respective group.

Additional Post-Hoc Analyses
Next to the GCG (rs237887-rs2268491-rs2254298) carriers and the GT (rs53576-rs2268498) noncarriers, also in the groups of GTA non-carriers and ACG carriers (rs237887-rs2268491-rs2254298) as well as in the groups of AC and AT carriers (rs53576-rs2268498) receiving OXT, highly significant results with regard to the deviation of the distributions of reported numbers in the third round from the equal distribution were found (see Supplementary Table 4). To examine these effects in more detail and to check whether the significant results in these groups would be driven by the participants who also belong to the carriers or non-carriers groups of the individual haplotypes, in which we found the strongest results in the respective block (namely GCG (rs237887-rs2268491-rs2254298) carriers or GT (rs53576-rs2268498) non-carriers), we implemented additional analyses. Therefore, we split each of the Additionally, the results also underline that the effects of increased lying behavior found in AC and AT (rs53576-rs2268498) carriers presented in Supplementary Table 4 are due to the significant results in the subgroup of participants (also) not carrying the GT (rs53576-rs2268498) individual haplotype.

Supplementary Material
Discussion: Our results demonstrate potential interactions between OXT treatment and OXTR genetics with the intranasal OXT effect on lying behavior potentially being specific to GCG (rs237887-rs2268491-rs2254298) carriers and GT (rs53576-rs2268498) non-carriers. In detail, OXT treatment seems to enhance lying behavior only in people carrying certain individual haplotypes in the OXTR gene (hence, in the PLC group no effect of individual haplotypes could be observed). The direct comparison of OXT and PLC sub-groups also revealed several significant results. However, none would hold after correction for multiple testing.
Additionally, except rs2268498 (Reuter et al., 2017), the OXTR SNPs under investigation are placed in an intronic region of the gene. Accordingly, their functionality on a biochemical level is unknown. Thus, it is not possible to directly conclude potential functional differences between the groups of certain individual haplotype carriers versus non-carriers. Nevertheless, findings from animal models throw possible light on this issue. For example, a first study in prairie voles indicates that variation in intronic regions of the OXTR, especially in / near so called cis-regulatory elements (regulatory elements in noncoding sequences of the DNA (e. g. intronic regions)), might contribute to differences in OXTR expression especially in brain regions associated with social attachment (King, Walum, Inoue, Eyrich, & Young, 2016). This might be also of importance for explaining the present interaction effects between individual OXTR haplotypes and OXT treatment on lying behavior. Thus, higher transcription and translation of the OXTR gene finally leading to higher expression of OXTRs in brain areas associated with social behaviors, such as (self-serving) lying, might lead to a more efficient processing of the Nevertheless, as can also be seen in the distributions, some seem rather "inconsistent". This might be due to the fact, that with such a small number of participants (times the die was thrown) actually no equal distribution can be expected. To validly expect an equal distribution (and test the actual distribution against the equal distribution) potentially more observations as the ones presented / analyzed in the present treatment by genetics interaction design would be necessary. Finally, it is likely that there are additional genes and polymorphisms, which could influence the effects of intranasally applied OXT (e. g. polymorphisms in the CD38 (Cluster of Differentiation 38) gene or in the AVPR1a (argininevasopressin receptor 1a) gene (see for example (Israel et al., 2008;Neumann & Landgraf, 2012;Sauer, Montag, Worner, Kirsch, & Reuter, 2012;Stoop, 2012)). For the present study, however, it was decided to investigate only well-established candidate gene polymorphisms of the oxytocinergic system with respect to lying, which have previously been associated with social cognition and relevant behaviors and/or have a known biochemical function (e. g. (Bakermans-Kranenburg & van Ijzendoorn, 2014;Reuter et al., 2017;Walter et al., 2012)).
Overall the present results support the assumption that OXT effects might depend on genetic predispositions of the OXTR gene. This underlines the importance of assessing genetic moderators in OXT administration studies (Bartz, Zaki, Bolger, & Ochsner, 2011;Chen et al., 2015;Feng et al., 2015;Montag et al., 2013;Shamay-Tsoory & Abu-Akel, 2016). Nevertheless, given shortcomings of the present study replication studies are urgently needed.