Figures and data in Oxytocin restores context-specific hyperaltruistic preference

Figures
Tables
Additional files

11 figures, 3 tables and 1 additional file

Figures

Figure 1 with 1 supplement

Download asset Open asset

Experimental design and task.

(A) Experimental task. Subjects performed a money-pain trade-off task in which they were designated as deciders. Four conditions (gain-self, gain-other, loss-self, loss-other) were introduced across decision contexts (gain vs. loss) and shock-recipients (self vs. other). Prior to the task, subjects received a neutral description of monetary values and potential shocks associated with choices options (see Figure 1—figure supplement 1 for details). In each trial, subjects were asked to choose between two options with various amounts of monetary and harm consequences. The chosen option was highlighted for 1 s after subjects’ decisions. (B) Procedures of the oxytocin study (study 2). Before the task, a pain calibration procedure was performed on each subject to determine their pain thresholds for electrical shock stimuli. Subjects were then administered with 24 IU oxytocin nasal spray or placebo (saline). Then, 35 min later, subjects commenced the money-pain trade-off task. Finally, they filled out questionnaires including post-task surveys and assessments of personality traits.

Figure 1—figure supplement 1

Download asset Open asset

Experimental instructions provided to subjects.

Choice options were provided to subjects in a neutral and descriptive manner, with the focus on the relevant components (shocks and money) across four conditions.

Figure 2 with 2 supplements

Download asset Open asset

Context-specific hyperaltruistic preferences.

(A) Subjects chose less painful options more frequently for others than for themselves in the gain context, demonstrating a hyperaltruistic preference. However, this tendency was absent in the loss context. (B) The harm aversion parameter κ for others was significantly greater than that of self in the gain context but not in the loss context (also see the ‘Materials and methods’ section and Figure 2—figure supplement 2 for the harm aversion model analysis and model comparison results). (C) Furthermore, a mixed-effect logistic regression analysis showed that the relative harm sensitivity, calculated as the difference of regression coefficients of $Δ s$ in the other- and self-conditions ( $o t h e r β_{Δ s} - s e l f β_{Δ s}$ ), was significant in the gain context, but not in the loss context (also see Figure 2—figure supplement 1 for details). (D) However, the relative money sensitivity, the difference of regression coefficients of $Δ m$ in the other- and self-conditions ( $o t h e r β_{Δ m} - s e l f β_{Δ m}$ ), did not show contextual specificity. Error bars represent SE across subjects (n=80). NS, not significant; ** $P < 0.01$ and *** $P < 0.001$ .

Figure 2—figure supplement 1

Download asset Open asset

Mixed-effect logistic regression analysis results for study 1 (regression model 1).

∆m and ∆s represent the objective differences in money and electric shocks between the more and less painful options. $∆ s$ and $∆ m$ are numerical variables, while the context (gain vs. loss) and recipient (other vs. self) are categorical variables. The $Δ s \times context \times r e c i p i e n t$ interaction was further elucidated in Figure 2C and D. Con: context; Rec: recipient. Error bars represent 95% confidence interval (CI). NS, not significant, * $P < 0.05$ , ** $P < 0.01$ , and *** $P < 0.001$ .

Figure 2—figure supplement 2

Download asset Open asset

Model comparison results.

To investigate whether the choice consistency parameter (γ) was condition specific, four candidate models were compared: (M1) γ was constant across all conditions; (M2) γ varied based on the pain recipient (self vs. other); (M3) γ varied between decision contexts (gain vs. loss); and (M4) γ varied across all four conditions (self-gain, other-gain, self-loss, and other-loss). Model 1 (M1) performed the best among all the four candidate models both in Study 1 (A) and Study 2 (B).

Figure 3 with 2 supplements

Download asset Open asset

Individual difference in moral preferences.

Hyperaltruism ( $κ_{o t h e r} - κ_{s e l f}$ ) was negatively associated with instrumental harm (IH) attitudes (A) but not with impartial beneficence (IB) attitudes (B). (C) The correlation between IH and subjects’ relative harm sensitivities ( $o t h e r β_{Δ s} - s e l f β_{Δ s}$ ) was marginally different between the gain and loss contexts (also see Figure 3—figure supplement 2A). (D) However, no significant associations were observed between subjects’ relative monetary sensitivities ( $o t h e r β_{Δ m} - s e l f β_{Δ m}$ ) and IH or IB (see Figure 3—figure supplement 2B). Please note that since both IH and IB were correlated with empathic concern (EC) scores (see Figure 3—figure supplement 1), the multiple regression analysis was conducted with EC included as a covariate of no interest. The regression coefficients in (A, C, D) show the relationship between IH and moral behavior, controlling for EC and IB, while the coefficient in (B) shows the relationship between IB and moral preference, controlling for EC and IH. NS, not significant.

Figure 3—figure supplement 1

Download asset Open asset

The relationship between empathic concern (EC) and utilitarian moral personality traits.

Correlation between EC and hyperaltruism ( $κ_{o t h e r} - κ_{s e l f}$ ), instrumental harm (IH), impartial beneficence (IB) both in study 1 (**A–C**) and study 2 (D-F). All the correlations were performed using Pearson correlations.

Figure 3—figure supplement 2

Download asset Open asset

Association between impartial beneficence (IB) and subjects’ relative harm/money sensitivities in Study 1.

IB showed no significant correlation with subjects’ relative harm sensitivities ( $o t h e r β_{Δ s} - s e l f β_{Δ s}$ ) (A) or relative monetary sensitivities ( $o t h e r β_{Δ m} - s e l f β_{Δ m}$ ) (B). The regression coefficients showed the association between IB and relative harm/monetary sensitivities, controlling for empathic concern (EC) and instrumental harm (IH).

Figure 4 with 2 supplements

Download asset Open asset

Oxytocin significantly promoted hyperaltruistic preference in the loss context.

(A) In contrast to the placebo condition, oxytocin administration restored hyperaltruistic behavior in the loss decision context. (B) Model-based hyperaltruistic parameter ( $κ_{o t h e r} - κ_{s e l f}$ ) showed similar patterns: hyperaltruistic tendency was reduced in the loss context of the placebo session but restored in the oxytocin session (also see Figure 4—figure supplements 1 and 2 for the influence of oxytocin on the persoanlity triats and the relationship between hyperaltruistic preference and these moral personality triats). Error bars represent SE across subjects (n=46). NS, not significant; * $P < 0.05$ , ** $P < 0.01$ , and *** $P < 0.001$ .

Figure 4—figure supplement 1

Download asset Open asset

Oxytocin did not influence personality traits in study 2.

(**A–C**) Oxytocin did not affect subjects’ ratings on empathic concern (EC), instrumental harm (IH), or impartial beneficence (IB). Error bars represent SE. NS, not significant.

Figure 4—figure supplement 2

Download asset Open asset

Relationships between hyperaltruism and utilitarian moral personality traits in study 2.

(**A–B**) In the placebo session, hyperaltruistic preferences ( $κ_{o t h e r} - κ_{s e l f}$ ) showed a marginally negative correlation with instrumental harm (IH) but no significant correlation with impartial beneficence (IB). (**C, D**) In the oxytocin session, hyperaltruistic preferences exhibited significantly negative correlation with IH and IB. The regression coefficients in (A, C) show the relationship between IH and moral preference, controlling for empathic concern (EC) and IB, while the coefficient in (B, D) shows the relationship between IB and moral preference, controlling for EC and IH.

Figure 5 with 3 supplements

Download asset Open asset

Oxytocin effect on relative harm/ money sensitivities.

(**A, B**) The effect of oxytocin on relative harm/ money sensitivity was assessed via the mixed-effect logistic regression analysis (regression model 2). Oxytocin significantly modulated the context-specificity of relative harm sensitivity ( $o t h e r β_{Δ s} - s e l f β_{Δ s}$ ), while having no effect on the contextual differences of the relative monetary sensitivities ( $o t h e r β_{Δ m} - s e l f β_{Δ m}$ ) (also see Figure 5—figure supplements 1 and 2 for detailed results of the regression analysis). (C) The decision context modulated the correlation between instrumental harm (IH) and subjects’ harm sensitivities in the placebo session. (D) The modulation effect of decision context was absent in the oxytocin session (also see Figure 5—figure supplement 3 for the oxytocin’s effect on the association between relative monetary sensitivities and IH). Error bars represent SE across subjects. The regression coefficients in (C, D) showed the association between IH and relative harm sensitivities, controlling for empathic concern and impartial beneficence. NS, not significant; * $P < 0.05$ , ** $P < 0.01$ .

Figure 5—figure supplement 1

Download asset Open asset

The main results of mixed-effect logistic regression analysis in study 2 (regression model 2).

∆m and ∆s represent the objective differences in money and electric shocks between the more and less painful options. The $Δ s \times Treat. \times Con. \times Rec.$ interaction was further explained in Figure 5, Figure 5—figure supplement 2. The $Δ m \times Treat. \times Con. \times Rec.$ interaction was further explained in Figure 5B. Treat.: treatment (placebo vs. oxytocin); Con.: decision contexts (gain vs. loss), and Rec.: recipient (self vs. other) are all categorical variables. Error bars represent 95% confidence interval (CI). NS, not significant, * $P < 0.05$ and *** $P < 0.001$ .

Figure 5—figure supplement 2

Download asset Open asset

Pain sensitivity across experimental conditions.

The administration of oxytocin significantly reduced participants’ pain sensitivities yet also restored the pain sensitivity patterns in both the gain and loss conditions relative to the placebo session. Error bars represent SE across subjects. NS, not significant, * $P < 0.05$ and *** $P < 0.001$ .

Figure 5—figure supplement 3

Download asset Open asset

Associations between relative harm/money sensitivities and utilitarian moral personality traits in study 2.

(**A, D**) The relative monetary sensitivities ( $o t h e r β_{Δ m} - s e l f β_{Δ m}$ ) were not significantly related with instrumental harm (IH) in either the placebo or oxytocin session. (**B, E**) In both the placebo and oxytocin sessions, there was no significant correlation between relative harm sensitivity ( $o t h e r β_{Δ s} - s e l f β_{Δ s}$ ) and impartial beneficence (IB). (**C, F**) No significant correlation was found between relative monetary sensitivities and IB in either the placebo or oxytocin sessions. (A, D) illustrates the relationship between IH and relative monetary sensitivities, with empathic concern (EC) and IB controlled. (B, C, E, F) described the relationship between IB and relative harm/monetary sensitivities, controlling for EC and IH.

Figure 6

Download asset Open asset

Oxytocin modulated the contextual influence on hyperaltruistic behaviors.

(A) The non-parametric Friedman tests showed that monetary loss (relative to gain) significantly reduced subjects’ perception of harm framing in the task. Oxytocin augmented harm framing perception, particularly in the loss context, effectively removing the contextual specificity of harm framing perception. (B) The conceptual diagram of the moderated mediation model. We assume that the perceived harm framing mediates the relationship between instrumental harm and relative harm sensitivity, with decision context moderating the mediation effect. (C) The moderating effect of decision context was significant under placebo condition. However, oxytocin obliterated the contextual moderation effect by reinstating the mediating role of harm perception in the loss context. Error bars represent SE across subjects. NS, not significant; * $P < 0.05$ , ** $P < 0.01$ .

Author response image 1

Download asset Open asset

Figs.

2B & 4B in the original manuscript in both studies.

Author response image 2

Download asset Open asset

Author response image 3

Download asset Open asset

Author response image 4

Download asset Open asset

Author response image 5

Download asset Open asset

Tables

Table 1

Contribution of empathic concern (EC), instrumental harm (IH), and impartial beneficence (IB) scores on subjects’ behavior in study 1.

The table presents results from three multiple regression analyses. The dependent variables in the three analyses were hyperaltruism ( $κ_{o t h e r} - κ_{s e l f}$ ), relative harm sensitivity ( $o t h e r β_{Δ s} - s e l f β_{Δ s}$ ), and relative monetary sensitivity ( $o t h e r β_{Δ m} - s e l f β_{Δ m}$ ). Results in bold are further illustrated in Figure 3C.

Predictors	Hyperaltruism				Relative harm sensitivity				Relative monetary sensitivity
Predictors	β	SE	t	P	β	SE	t	P	β	SE	t	P
Intercept	–0.086	0.153	–0.563	0.574	–0.086	0.162	–0.532	0.596	0.121	0.147	0.825	0.411
EC	0.011	0.006	2.080	0.039	0.012	0.006	2.063	0.041	0.004	0.005	0.738	0.461
IH	–0.039	0.015	–2.501	0.013	–0.053	0.016	–3.224	0.002	–0.022	0.015	–1.476	0.142
IB	0.009	0.022	0.421	0.675	0.019	0.023	0.804	0.423	–0.008	0.021	–0.369	0.713
Context (gain vs. loss)	–0.163	0.217	–0.754	0.452	–0.298	0.229	–1.302	0.195	0.058	0.208	0.280	0.780
EC × context	0.002	0.008	0.216	0.829	0.003	0.008	0.383	0.702	–0.004	0.007	–0.484	0.629
IH × context	0.016	0.022	0.726	0.469	0.045	0.023	1.959	0.052	–0.001	0.021	–0.034	0.973
IB × context	–0.004	0.031	–0.115	0.908	–0.009	0.033	–0.268	0.789	–0.001	0.030	–0.032	0.974

Table 2

Contribution of empathic concern (EC), instrumental harm (IH), and impartial beneficence (IB) scores on participants’ behavior in study 2.

Tables 2a (placebo session) and 2b (oxytocin session) show the results of three multiple regression analyses. The dependent variables in the three analyses were hyperaltruism (κ_other – κ_self), relative harm sensitivity (otherβ_Δs – selfβ_Δs), and relative monetary sensitivity (otherβ_Δm – selfβ_Δm). Results in bold are further illustrated in Figure 5C and D.

a. Placebo condition
Predictors	Hyperaltruism				Relative harm sensitivity				Relative monetary sensitivity
Predictors	β	SE	t	P	β	SE	t	P	β	SE	t	P
Intercept	–0.085	0.123	–0.688	0.493	0.065	0.17	0.382	0.704	–0.205	0.266	–0.771	0.443
EC	0.005	0.004	1.162	0.248	0.006	0.006	1.086	0.281	–0.004	0.009	–0.444	0.658
IH	–0.032	0.017	–1.904	0.06	–0.071	0.024	–3.005	0.003	0.035	0.037	0.967	0.336
IB	0.047	0.021	2.2	0.031	0.029	0.029	1	0.32	0.086	0.046	1.887	0.063
Context (gain vs. loss)	–0.034	0.174	–0.197	0.844	–0.34	0.241	–1.412	0.162	0.323	0.375	0.862	0.391
EC × context	0.002	0.006	0.332	0.74	0.005	0.008	0.601	0.549	0.003	0.012	0.264	0.792
IH × context	0.018	0.024	0.737	0.463	0.077	0.033	2.297	0.024	–0.062	0.052	–1.187	0.238
IB × context	–0.051	0.03	–1.711	0.091	–0.047	0.042	–1.14	0.257	–0.084	0.065	–1.295	0.199
b. Oxytocin condition
Intercept	0.138	0.125	1.1	0.275	0.056	0.109	0.511	0.611	0.049	0.253	0.195	0.846
EC	0.01	0.004	2.397	0.019	0.006	0.004	1.675	0.098	0.001	0.009	0.157	0.876
IH	–0.047	0.017	–2.721	0.008	–0.037	0.015	–2.458	0.016	0.053	0.035	1.541	0.126
IB	–0.034	0.019	–1.813	0.073	–0.009	0.016	–0.582	0.562	–0.041	0.038	–1.091	0.279
Context (gain vs. loss)	–0.071	0.177	–0.401	0.69	0.092	0.154	0.596	0.553	–0.372	0.358	–1.04	0.301
EC × context	0.002	0.006	0.385	0.701	–0.002	0.005	–0.336	0.737	0.012	0.012	0.985	0.327
IH × context	0.008	0.024	0.323	0.747	–0.006	0.021	–0.272	0.786	–0.045	0.049	–0.918	0.361
IB × context	–0.01	0.026	–0.384	0.702	–0.012	0.023	–0.526	0.6	0.024	0.053	0.457	0.649

Table 3

Moderated mediation analysis results for study 2.

In the moderated mediation model, IH serves as the independent variable, harm framing report as the mediator, relative harm sensitivity as the dependent variable, and decision context (gain vs. loss) as the moderator. Coefficient a represents the influence of the independent variable on the mediator, while coefficient b represents the effect of the mediator on the dependent variable, and coefficient c’ represents the direct effect (insignificant c’ suggests a full mediation effect). The coefficients in bold indicate the moderation effect of the context (gain vs. loss) in both the placebo and oxytocin sessions ^∗P<0.05, ^∗∗P<0.01, and ^∗∗∗P<0.001.

Predictors	Placebo		Oxytocin
	Mediator	Dependent variable	Mediator	Dependent variable
	Harm framing report β (SE)	Relative harm sensitivity β (SE)	Harm framing report β (SE)	Relative harm sensitivity β (SE)
Intercept	1.196 (0.279) ***	0.069 (0.023) **	1.609 (0.239) ***	0.049 (0.014) ***
Context (gain vs. loss)	–0.739 (0.396)	–0108 (0.032) **	–0.196 (0.338)	–0.002 (0.019)
Empathic concern (EC)	0.020 (0.046)	0.008 (0.004) *	0.023 (0.040)	0.005 (0.002) *
Impartial beneficence (IB)	–0.116 (0.239)	0.010 (0.019)	–0.270 (0.179)	–0.008 (0.010)
Instrumental harm (IH)	–0.758 (0.257) *** a	–0.018 (0.016) c’	–0.753 (0.220) *** a	–0.017 (0.011) c’
Context × IH	0.784 (0.343) *	–	–0.137 (0.291)	–
Harm framing report	–	0.039 (0.008) *** b	–	0.028 (0.006) *** b

Additional files

MDAR checklist: https://cdn.elifesciences.org/articles/102756/elife-102756-mdarchecklist1-v1.docx
Download elife-102756-mdarchecklist1-v1.docx

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)

Mendeley

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

Hong Zhang
Yinmei Ni
Jian Li

(2025)

Oxytocin restores context-specific hyperaltruistic preference

eLife 13:RP102756.

https://doi.org/10.7554/eLife.102756.3

Share this article

Cite this article

Experimental design and task.

Experimental instructions provided to subjects.

Context-specific hyperaltruistic preferences.

Mixed-effect logistic regression analysis results for study 1 (regression model 1).

Model comparison results.

Individual difference in moral preferences.

The relationship between empathic concern (EC) and utilitarian moral personality traits.

Association between impartial beneficence (IB) and subjects’ relative harm/money sensitivities in Study 1.

Oxytocin significantly promoted hyperaltruistic preference in the loss context.

Oxytocin did not influence personality traits in study 2.

Relationships between hyperaltruism and utilitarian moral personality traits in study 2.

Oxytocin effect on relative harm/ money sensitivities.

The main results of mixed-effect logistic regression analysis in study 2 (regression model 2).

Pain sensitivity across experimental conditions.

Associations between relative harm/money sensitivities and utilitarian moral personality traits in study 2.

Oxytocin modulated the contextual influence on hyperaltruistic behaviors.

Figs.

Contribution of empathic concern (EC), instrumental harm (IH), and impartial beneficence (IB) scores on subjects’ behavior in study 1.

Contribution of empathic concern (EC), instrumental harm (IH), and impartial beneficence (IB) scores on participants’ behavior in study 2.

Moderated mediation analysis results for study 2.

MDAR checklist

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)