People with higher interoceptive sensitivity are more altruistic, but improving interoception does not increase altruism

People consistently show preferences and behaviors that benefit others at a cost to themselves, a phenomenon termed altruism. We investigated if perception of one’s body signals – interoception - may be underlying such behaviors. We tested if participants’ sensitivity to their own heartbeat predicted their decision on a choice between self-interest and altruism, and if improving this sensitivity through training would make participants more altruistic. Across these two experiments, interoceptive sensitivity predicted altruism measured through monetary generosity. Improving interoceptive sensitivity did, however, not lead to more altruistic behaviour. We conclude that there is a unique link between interoception and altruistic behaviour, likely established over an individual’s history of altruistic acts, and the body responses they elicit. The findings suggest that humans might literally ‘listen to their heart’ to guide their altruistic behavior.

3 performed the heartbeat detection task. Experiment 3 sequence: dictator game; other tasks (not reported here); smell sensitivity test; questionnaires (not reported here).
Thus, the larger amount is on some trials assigned to the decider ('favourable'), on others to the other person ('unfavourable', 15; our main text fig 1D). There are equal numbers of these two types of trials (favourable and unfavourable). The sum of money allocated to 'other' corresponds to monetary generosity and is an index of altruistic behavior. Unbeknownst to the participant, the 'other' person is the next participant in the study. As payment for the game, each participant receives the amount from two randomly selected trials from playing their own game, according to their own choice. They also receive the amount from 2 randomly selected trials from the previous participant's game, according to that participant's choice. The task consisted of 12 practice trials and 60 experimental trials.

Heartbeat detection task and training
During the heartbeat detection (discrimination) task (2), on each trial, participants listened, via laptop speakers, to a series of 20 auditory beeps simultaneous with their own heartbeat, recorded with electrocardiogram (ECG) electrodes on the chest. ECG 4 provided the input for in-house developed ExpyVR software (http://lnco.epfl.ch/expyvr), which produced the brief auditory beeps, triggered by the R-wave of the ECG in the synchronous condition, or it produced beeps at a speed of either 80% or 120% of the frequency of the participant's preceding two R-waves to create the asynchronous condition. Participants had to indicate on each trial if the beeps were in time with their own heartbeat or not. There were 16 trials. The proportion of correct answers indicates how good a person is at sensing their own heartbeat, and is an index of interoceptive sensitivity.

Training and control condition
In the interoceptive training experiment (experiment 2, N=57), one group of participants received interoceptive training: after the initial heartbeat detection task, they completed it a second time, this time with feedback (they were told, after their response to each trial, if their answer was correct or not). A third heartbeat detection task was then completed (without feedback) to evaluate the training effect. The control group were trained on a task assessing the detection of audio-visual synchrony as a control procedure. They saw geometric shapes appear on the computer screen. At the same time, they listened to sounds (beeps, as in the experimental condition) via the laptop speakers. Their task was to report if the visual shapes on the screen appeared at the same time as the auditory signals. They were given correct feedback on the answers in the training block, and all parameters were kept the same as for the experimental condition. The control group did not perform a second heartbeat detection task.

Questionnaires 5
Participants responded to questions from the Aspiration Index (3), a measure of materialism (4). It classifies a person's goals and values into 'intrinsic' and 'extrinsic' ones and indicates how much importance is given to each. (Intrinsic goals: Self-Acceptance, Affiliation, Community Feeling, Physical Fitness; extrinsic goals: Financial Success, Attractive Appearance, Social Recognition.) A relatively high importance assigned to intrinsic goals corresponds to low materialism. Participants also responded to questions from the Empathy Quotient for adults (5).

Help task
After finishing all other components of the study, receiving payment, and being told that they are 'free to go', participants were asked by the experimenter for help with a different study. The help takes the form of filling in as much as possible of a questionnaire, which consists of 25 pre-university level mathematics test questions.

Smell sensitivity test (Experiment 3)
We measured the threshold at which participants were able to detect weak odors (6).
The odor was n-butanol, applied at 16 concentrations (1/16 concentration to 1/1 concentration). Smell sensitivity was operationalized as the lowest concentration that could be reliably detected. Participants where required to wear a blindfold; they were 6 given sets of three pens (one at a time, pens were designed to dispense odors) to smell (held by the experimenter at a constant distance under the nose). Participants were asked to identify which of the three pens contained the odor (the other two were odorless). The test began with the weakest concentration. Every time the participant failed to identify the correct sniffing stick, they moved to the next odor concentration set. If they identified the odor correctly, the test was repeated at the same concentration level. The odor was considered reliably detected when a participant identified it correctly four times in a row. Participants were given a strong (1/1) and medium (1/8) strength test smell of n-butanol as reference before the experiment.

Heartbeat detection task and training
In the 16-trial task, identifying one's heartbeat correctly on all trials corresponds to the perfect score of 1. A score of around 0.5 indicates chance performance. We repeated a multiple regression as in experiment 1, including participants from both groups, to predict monetary generosity. We used the monetary generosity score from the first dictator game (before treatment) and heartbeat sensitivity from the first measurement (before treatment), to confirm findings from experiment 1. The multiple regression tested if monetary generosity is predicted by: interoceptive sensitivity, low materialism, empathy level. Using the enter method we found that the model misses

Help task
We repeated a multiple regression as in experiment 1, including participants from both groups, to predict helping time. We used the interoceptive sensitivity from the first measurement (before treatment), to confirm findings from experiment 1.

Smell sensitivity experiment
To test whether the association between altruism and interoception generalized to the detection of other faint sensory signals, we investigated olfactory detection thresholds. Olfaction is intimately linked to internal states, as it is well-established that olfactory sensitivity and perception are modulated by hunger/satiety states (7).
We tested 21 participants' performance detecting smells according to a wellestablished protocol (6), as well as their behaviour in the dictator game. The correlation for smell sensitivity was much weaker than for interoceptive sensitivity in experiments 1 and 2, and statistically not significant [r(21) = .16, p = .488, 95% CI [-.26, .54]]. This shows that link between altruistic behaviour and interoception shown in experiments 1 and 2 is unlikely to generalize to other perception modalities, and we therefore propose the link to altruistic behaviour might be specific to interoception.

Olfactory sensitivity and interoceptive sensitivity
For technical reasons, the HB detection task used together with smell sensitivity was different from the main experiments (1 and 2), the 'counting task' by Schandry (8).

Discussion and Limitations
In our experiments, we used heartbeat detection performance as an index of interoception. While heartbeat detection is an interoceptive process, and different interoceptive processes share neuroanatomical features (9), we don't currently have a sufficient amount of evidence to confirm that in humans, performance in this task is indicative of interoceptive ability in general, which encompasses other interoceptive processes (10). Future studies should address the generalizability of interoceptive performance, and its potential role in altruism.
It is conceivable that compliance of participants contributed to our initial results, in addition to the effect of interoceptive sensitivity. Non-compliant participants may expend less effort on the heartbeat detection task, and subsequently expend less effort on being altruistic. While this is possible, one would expect such a compliance effect foremost in the help task, and to a lesser degree in the dictator game. Data do not support this, as the association was stronger for the dictator game. Further, only 4 out of 29 participants in the training group in experiment 2 did not improve or did not retain a high interoceptive sensitivity level -suggesting that low compliance was not a problem.
In contrast to our findings for monetary generosity, performance in the 'help' task was not predicted by interoceptive sensitivity, empathy or materialism levels. One reason for this may be that the help task is a cruder measure of altruism and subject to more confounds than the dictator game performance: time spent helping was likely influenced by how much participants liked doing maths, maths ability, whether they had a subsequent appointment, their opinion of the experimenter, etc. The dictator game and the help task also differ in many important ways -in the former, the beneficiary is unseen and unknown and money rather than time is at stake -and may have different brain bases.
That less materialistic people may also act more generously is a promising thesis in need of further investigation. Because materialism has malleable components even in highly materialistic individuals (11), this presents an opportunity for interventions.
Such interventions could stimulate people to address 'bigger than self' problems, e.g. 18 regarding the climate or poverty, which are difficult to solve with appeals to selfserving motives (12).