Modular morals: Mapping the organization of the moral brain

of


Introduction
Recent research suggests that morality is a collection of biological and cultural traits that promote cooperation (Curry, 2016;Greene, 2015;Haidt & Kesebir, 2010;Rai & Fiske, 2011;Sterelny & Fraser, 2017;Tomasello & Vaish, 2013).The basic argument is that humans are social creatures who have lived together in social groups for millions of years.During this time, humans have faced a range of different nonzero sum problems of cooperation, and have evolved and invented a range of different solutions to them.These cooperative solutionsin the form of traits, strategies, dispositions, behaviours, rules, norms, institutions, and technologiesmotivate cooperative behaviour, and provide the criteria by which we judge the behaviour of ourselves and others.And it is these cooperative solutions that philosophers and others have called 'morality'.
This approach predicts that, because there are many different types of cooperation, there will be many different types of morality; and it tells us what they will be.For example, drawing on the theory of nonzero sum games, the theory of Morality as Cooperation (MAC) has argued that there are (at least) seven distinct types of cooperation: (1) the allocation of resources to kin; (2) coordination to mutual advantage; (3) social exchange; and conflict resolution through contests featuring (4) hawkish displays of dominance and (5) dove-ish displays of submission; (6) division of disputed resources; and (7) recognition of prior possession.And each of these types of cooperation gives rise to a corresponding type of morality: (1) family values, (2) group loyalty, (3) reciprocity, (4) heroism, (5) deference, (6) fairness and (7) property rights (Curry, 2016;Curry et al., 2021).These seven different types of morality appear to be evolutionarily ancient (Curry, 2016), genetically (Zakharin et al., in prep) and psychometrically distinct (Curry et al., 2019), and crossculturally universal (Alfano, Cheong, & Curry, 2024;Curry, Mullins, & Whitehouse, 2019). 1 However, it remains unclear how these seven types of morality are implemented at a neuroanatomical level.
One possibility is that natural selection has equipped the human mind with a range of domain-specific adaptations for cooperation, each of which is responsible for a distinct domain of moral decision-making (Barrett & Kurzban, 2006;Tooby & Cosmides, 1992).If there are multiple special-purpose moral 'modules', and if these modules were neuroanatomically localised, then we would expect to see relationships between the seven types of morality and seven different brain regions (a 'seven-to-seven' model).
A second possibility is that the human mind possesses a single more domain-general mechanismperhaps a 'harm avoidance' mechanism, or a 'norm acquisition device' -which learns a range of cooperative moral rules (Sterelny, 2010).If there is one general-purpose moral module, and if this module was neuroanatomically localised, then we would expect to see relationships between the seven types of morality and one brain region (a 'seven-to-one' model).
A third possibility is that morality is the product of a combination of domain-specific and domain-general mechanisms.Perhaps some types of morality are the product of domain-specific mechanisms, and some are learned by a domain-general mechanism; perhaps one mechanism accounts for more than one type of morality, such that there are fewer than seven mechanism in total; perhaps multiple types of morality are the product of multiple, overlapping mechanisms; and so on.If so, and if the mechanisms are neuroanatomically localised, then we would expect to see relationships between the seven types of morality and multiple, disparate or overlapping groups of brain regions ('seven-to-some' model).
Recent work on the genetics of morality has provided evidence for the combined 'seven-to-some' model, according to which morality is the product of the seven predicted domain-specific mechanism as well as a more domain-general mechanism (Zakharin et al., in prep).However, the neuroanatomical implementation of these mechanisms remains unclear.
And so, in an initial attempt to distinguish between these possibilities, here we investigate the relationship between these seven moral traits and neuroanatomyspecifically, regional grey matter volume (GMV) using magnetic resonance imaging (MRI) 2 data.Previous research has found GMV correlates of individual differences in psychological traits such as intelligence, decision making, personality and social cognition (Kanai & Rees, 2011).More specifically, MRI data has shown that empathy is negatively correlated with GMV in the insula (Li et al., 2020;Raschle et al., 2018).Further, psychopathic traits have been associated with numerous GMV variations (Johanson et al., 2019).And, in terms of methodology, self-report measures of moral, social and political values have been shown to correlate with GMV (Patil et al., 2017;Zacharopoulos et al., 2017;Takeuchi et al., 2014).There has also been research conducted on the relationship between the previous types of morality outlined and GMV (see systematic review and meta-analysis below).
However, no previous review or study has looked at the relationship between grey matter volume and all seven of the types of morality identified by MAC.So, the present study had two primary aims.First, to conduct an anatomical likelihood estimation (ALE) meta-analysis (Eickhoff et al., 2009).This involved a systematic review of the literature on previous studies that explored the relationship between these seven moral domains and grey matter volume.We chose grey matter volume because most of the previous morality literature has used this as a measure of neuroanatomy, as opposed to cortical thickness and surface area.Second, we analyze the relationship between seven types of morality (measured by the Morality as Cooperation Questionnaire), and GMV (measured by MRI scans) in a large twin sample (n = 607).

Literature search and selection
The review involved a keyword search (see Table 1 for search terms) of Web of Science and the BrainMap database (http://brainmap.org),as well as a review of work cited in review papers, direct searches on the names of frequently occurring authors, emails to those authors, and an open call for papers on social media (literature search stopped in May 2020).Studies were considered for inclusion if they met the following criteria: MRI was the imaging method; participants either underwent a behavioural or self-report measure related to one of MAC's seven domains of morality (Family, Group, Reciprocity, Heroism, Deference, Fairness or Property); whole-brain analyses were applied; participants were free from psychiatric or neurological diagnoses; participants were adults, and no pharmacological methodologies were used.

Anatomical likelihood estimation (ALE)
To determine the brain regions that were meta-analytically significant across the identified studies, we employed a coordinate-based meta-analytic approach using the ALE algorithm with GingerALE software (https://www.brainmap.org/ale)(Eickhoff et al., 2009).ALE determines the convergence of foci reported from different neuroimaging studies with coordinates of peak activations in standardized space (Laird et al., 2005;Turkeltaub et al., 2002).Foci are interpreted as spatial probability distributions.Their widths are based on empirical estimates of the spatial uncertainty based on between-subject and betweentemplate variability of the neuroimaging data (Eickhoff et al., 2009).The sample sizes of the studies reporting the foci are used to weight the between-subject variability.Effects are sample-size weighted and therefore, the ALE algorithm assumes a more reliable approximation of anatomical relevance for larger sample sizes, which are modelled with smaller Gaussian distributions (Eickhoff et al., 2009).

Table 1
Search terms for the literature review.
Family "kin altruism" OR "kin selection" OR "inclusive fitness" OR "relatives" OR "hamilton" OR "incest aversion" OR "paternal investment" OR "kin recognition" OR "parental care" Group "coordination problem" OR "coordination game" OR "stag hunt" OR "assurance game" OR "loyalty" OR "solidarity" OR "mutualism" OR "coalition" OR "group membership" Reciprocity "prisoners dilemma" OR "prisoner's dilemma" OR "prisoners' dilemma" OR "tragedy of the commons" OR "public goods game" OR "free-rider" OR "reciprocity" OR "reciprocal altruism" OR "trust*" OR forgiv*" Heroism "hawk" OR "bravery" OR "prestige" OR "prowess" OR "costly signal" Deference "submis*" OR "defer*" Division "fair*" OR "equity" OR "equality" Property "possession" OR "endowment effect" OR "prior ownership" OR "territory" Grey Matter "grey matter volume" OR "gray matter volume" OR "voxel-based morphometry" OR "VBM" OR "grey matter" OR "gray matter" 1 MAC's systematic taxonomy of moral domains provides a principled and more comprehensive alternative to earlier ad hoc taxonomies, such as Moral Foundations Theory (MFT), which omits important domains such as family values, reciprocity, heroism and property, and includes domains, such as 'purity', which are not associated with a particular forms of cooperation, and which subsequent research has suggested are not coherent domains (Curry et al., 2019;Fitouchi et al., 2023;Gray et al., 2022).
2 For a comparable review of the fMRI literature, see: (Wilkinson & Curry, in prep).
An ALE map is obtained by calculating the union of the individual anatomical maps created from the maximum probability associated with each voxel (Turkeltaub et al., 2012).This ALE map is compared to a nulldistribution of random spatial association between studies employing a non-linear histogram integration algorithm (Eickhoff et al., 2012;Turkeltaub et al., 2012).Significant results were assessed at a cluster-level corrected threshold of p < 0.05 with a cluster defining threshold of p < 0.001 and 1,000 permutations, this is expected to generate 1 false positive (Eickhoff et al., 2012).

Results
The literature review identified a total of nine previous studies with a total of 35 foci for 728 participants (Table 2).The average number of participants across all studies was N = 57.Within these studies there were twelve different behavioural or self-reported measures that were tested for correlation with grey matter volume.There were no studies in the Family domain and there was an overlap in the studies found for the Heroism and Deference domain.Group and Reciprocity were found to be positively correlated with grey matter volume in the dorsomedial prefrontal cortex (Baumgartner et al., 2013;Haas et al., 2015).Reciprocity and Possession were found to be positively correlated with grey matter volume in the inferior frontal gyrus (Haas et al., 2015;Votinov et al., 2010).Group, Reciprocity, Deference and Property all correlated with unique regions (Baumgartner et al., 2013;Getov et al., 2015;Haas et al., 2015;Kumaran et al., 2012;Lewis et al., 2012;Li et al., 2017;Morishima et al., 2012;Votinov et al., 2010;Watanabe et al., 2014).

Anatomical likelihood estimation
The ALE meta-analysis suggests that there are relationships between specific types of morality and distinct as well as overlapping brain regions (Table 3; Fig. 1).The left uncus, right superior frontal gyrus and right insula are all related to Heroism and Deference indicating that these areas may be related to status and hierarchy (Fig. 2a).Grey matter volume in the right medial frontal gyrus is negatively correlated with Group and grey matter volume in the left middle frontal gyrus is positively correlated with Reciprocity.These areas also show relationships to the Property domain but with opposite correlations and hemispheric differences (Fig. 2b).Given the opposite correlations, it could be the case that grey matter volume in these areas relate to conflicting values in these domains, although the hemispheric differences may reflect distinct brain regions.

Discussion
The meta-analysis provides tentative support for the hypothesis that morality is the product of a combination of domain-specific and domaingeneral mechanisms (the third, 'some to some' model).All the moral domains relate to unique sets of brain regions, but these sets are overlapping.
One limitation of the meta-analysis is that it included a relatively small number of studies in the analysis; and each had different Hem: hemisphere; L: left; R: right; x,y,z: MNI 3-dimensional coordinate system.*significant at p < 0.001.methodologies, and relatively small sample sizes.Many of the measures used in each domain were indirect measures of the corresponding moral values in each domain.For example, trust and forgiveness are subcomponents of reciprocity, and it is unclear whether they predict reciprocity as a whole.To overcome these limitations, we proceeded to investigate the relationship between the seven types of moralityas measured by the Morality as Cooperation Questionnaire (Relevance)and grey matter volume in a large neuroimaging (n = 607) sample.

Participants
Participants were Australian twins and siblings from the ongoing Brisbane Longitudinal Twin Study (BLTS) at QIMR Berghofer Medical Research Institute.Over the past 30 years, these twins have participated in several different studies that collected data on a host psychological variables, including grey matter volume (MRI), and moral values.
Data on neuroanatomical structure (MRI) was collected by the Queensland Twin Imaging Study (QTIM) (Couvy-Duchesne et al., 2018;Mitchell et al., 2019;Wright & Martin, 2004).Data on moral values was collected by the Genetics of Human Agency Study (Zakharin et al., in prep).We had MRI and moral data for a total of 607 healthy participants (429 females; mean age at MRI = 22.54; mean age at MAC-Q-R = 31.25;237 MZ twins, 271 DZ twins, remainder siblings).
A power analysis (two tails, alpha = 0.05, power = 0.8, 7 predictors) suggests that a study with a sample of 607 can detect effect sizes of d ≥ 0.23, which is sufficiently sensitive given that the smallest effect size found in our meta-analysis was d = 1.47 (Faul et al., 2007).
Surface area (SA) was measured using FreeSurfer (v5.3; https ://surfer.nmr.mgh.harvard.edu)as previously described and contributed to the ENIGMA consortium GWAS meta-analysis of brain surface area and thickness (Grasby et al., 2020).Grey matter volume and SA were extracted for 34 regions of interest per hemisphere from the Desikan-Killiany atlas (Desikan et al., 2006) contained within Free-Surfer.Further, grey matter volume was extracted for 20 subcortical regions.Two global measures, whole-brain grey matter volume and total surface area were also extracted.Cortical reconstructions and ROI labelling were checked using the standardised procedures of the ENIGMA consortium (enigma.ini.usc.edu)(Thompson et al., 2014), with any incorrectly delineated cortical structures also being excluded from the analysis.

Moral measures
Moral values were assessed using the 21-item Relevance subscale of the Morality-as-Cooperation Questionnaire (MAC-Q-R) ( Curry, Jones Chesters, & Van Lissa, 2019; see also: van den Berg, Kroesen, & Chorus, 2022; Yilmaz, Harma, & Dogruyol, 2021) measuring the relevance of family, group, reciprocity, heroism, deference, fairness and property when deciding whether something is morally right or wrong.Participants answered on a 6 response rating scale from 'Not at all relevant' to 'Extremely relevant'.For further information on this dataset, see: Zakharin et al., in prep

Regression analysis
For our primary analysis, each regional grey matter volume was regressed on the measures of moral relevance using a linear regression, where left and right brain regions were considered separately.The regression analysis was run in R (Team, 2017).We used a linear-mixed effects model.Age, sex and grey matter volume were included as fixed effects.We accounted for genetic relatedness by including random intercepts and slopes for zygosity within each family.To correct for multiple testing error across all brain regions, the effective number of independent observations was estimated using Matrix Spectral Decomposition (Nyholt, 2004) in each lobe before undergoing Bonferroni correction (Cingulate: p < 0.05/4, Frontal: p < 0.05/13, Occipital: p < 0.05/4, Parietal: p < 0.05/5, Temporal: p < 0.05/5).Regressions were also run independently for total grey matter volume per lobe.As a Fig. 2b.The horizontal slice (left) shows the locations of the medial frontal gyrus, grey matter volume in this area positively correlates with property (blue) and negatively correlates with group (green).The coronal slice (right) shows the locations of the middle frontal gyrus, grey matter volume in this area positively correlates with property (blue) and negatively correlates with reciprocity (black/orange).sensitivity analysis, regression analyses using the surface area of the 34 regions of interest per hemisphere were also conducted (Supplementary Material).

Results
Descriptives (mean, standard deviation) and Cronbach alphas for the seven MAC-Q-R scores are shown in Table 4. Exploratory whole brain analysis showed relationships between moral values and grey matter volume (Table 5).There were positive associations between Group and the left precuneus (p = 0.011) and Division and the left middle temporal (p = 0.037).There were negative associations between Reciprocity and the left cuneus (p = 0.017), Deference and the left cuneus (p = 0.046), Heroism and the left pars triangularis (p = 0.039), Division and the left rostral middle frontal (p = 0.012) and Property and the right lateral occipital (p = 0.021) and right medial orbito frontal (p = 0.013).In the subcortical regions there were positive associations between Family and the lateral ventricle (p = 0.023), Deference and the hippocampus (p = 0.011) and Heroism and the hippocampus (p = 0.016).There was a negative association between Reciprocity and the nucleus accumbens (p = 0.033).However, none of these findings were significant after controlling for multiple comparisons (Bonferroni correction).Similar results were observed when examining the relationship with grey matter volume per lobe (Table S1) and surface area (Table S2).

Discussion
The neuroanatomical study found no association between grey matter volume and moral values.More conservatively, given the power of the study, we can conclude that there are no associations between grey matter volume and moral values greater that d = 0.23, and hence any such associations would be small.

General discussion
The two studies reported here did not find any consistent relationship between moral values and grey matter volume.The meta-analysis suggested that morality could be the product of a combination of domain-specific and domain-general mechanisms.However, the regression analysis did not identify any significant correlations between moral values and grey matter volume after controlling for multiple comparisons.Hence the current study does not provide sufficient evidence to support a domain-specific, domain-general, or combination neuroanatomical model of moral psychology.This suggests that, whatever mechanisms are responsible for morality, either they are (i) not neuroanatomically localised, (ii) their localisation is not manifest in grey matter volume, or (iii) any manifestations in grey matter volume are small (d < 0.23).
The general view that brain functions are neuroanatomically localised, despite some early support, has more recently been called into question.The relationships between structural brain properties and behaviour and personality measures in general have been notoriously difficult to replicate (Boekel et al., 2015;Melonakos et al., 2011).Hence morality may be the product of mechanisms and processes that are distributed across the brain (Uttal, 2004).Alternatively, the functional localisation of the mechanism(s) responsible for morality, may be expressed not in grey matter volume, but in some other aspect of the braindetectable not by MRI but by, for example, multivariate pattern recognition (Calhoun et al., 2017) and network analysis techniques (Sporns, 2014).These methodologies relate to the functional activation within brain scans, and there is a plethora of morality and cooperation studies related to activation (Bitsch et al., 2018;Bortolini et al., 2017;Duarte et al., 2017;Hu et al., 2018;Niemi et al., 2018;Platek & Kemp, 2009).Further investigations of the location of these moral mechanisms would require carefully designed functional neuroimaging studies.
Of course the current studies have a number of limitations, that could be overcome by further research.
First, the MRI data was parsed into 88 areas (68 cortical and 20 subcortical) identified by the Desikan-Killarney brain atlas as adopted by the ENIGMA consortium (Grasby et al., 2020;Mitchell et al., 2020).Parsing the data into the 1,105 smaller areas identified by the MNI and Tailarach brain atlases may have led to results that are more directly comparable to previous findings.
Second, the absence of prior predictions about the neuroanatomical locations of morals meant that we had to conduct an exploratory whole brain analysis, which increased our multiple testing burden when defining significance.Future theory-driven, confirmatory research, perhaps drawing on phylogenetic findings, (Cisek, 2019) might lead to different conclusions.
Third, we used MAC-Q-R, a self-report measure of seven moral values.Although the MAC-Q-R has been well-validated (Curry et al., in prep), including a high test-retest reliability of 0.85 (Curry et al., 2019), all self-report measures have their limitations.Other measurespositing different moral factors (Fiske, 1992;Haidt and Kasebir, 2010;Schwartz, 1992), or measuring MAC's seven factors in different ways (such as vignettes, or behavioural tasks) -may yield different results.
Fourth, the present study's cohort was in an English-speaking Western, Educated, Industrialised, Rich, Democratic cultures (W.E.I.R. D) (Henrich et al., 2010) therefore results may not be applicable to samples that are not W.E.I.RD.Future research should test the relationship between morality and brain structure in a wider range of languages and cultures.
In addition, future work using more genetically informative designs may help to resolve the question of how morals are neuroanatomicaly implemented.One recent study showed that relatively little of the difference in morals are due to genes or shared environment and most is due to non-shared environment (Zakharin et al., in prep).Looking further into the relationship and therefore causality of morality and structural differences would be useful, as has been done in studies of brain volume and intelligence and educational attainment (Mitchell et al., 2020;Toga & Thompson, 2005).Further, longitudinal studies of brain development and changes will also be valuable to further uncover this direction of causality (Hedman et al., 2012).In addition, it would be informative to conduct similar studies using other measures of neuroanatomysuch as functional magnetic resonance imaging (fMRI) to measure blood-oxygen-level-dependent signals, transcranial magnetic stimulation (TMS) to inhibit or suppress cortical excitability, electroencephalography (EEG) to measure electrical signals from neurons and brain lesion studies.
In conclusion, the present studies found no support for the theory that variation in moral values is related to variation in regional grey matter volume.Future work should make use of other measures of morality, and other measures of brain function, as well as phylogenetically informed a priori predictions about the localisation of brain functions.Such research would shed further light on the nature of human morality.

Funding information
Many thanks to Alexander Vostroknutov for the invaluable feedback on the project and the paper.The Genetics of Human Agency project was a data collection conducted within the Brisbane Longitudinal Adolescent Twin Study.
The QTIM study was supported by the National Institute of Child Health and Human Development (R01 HD050735), and the National Health and Medical Research Council (NHMRC 486682, 1009064), Australia.
This study was funded by a one-off grant from the John Templeton Foundation.More info about the grant (broader objectives), can be found here: https://www.geneticshumanagency.org/about/.
Oliver Scott Curry's time on the project was also supported by kindn ess.org.

Open practices statements
None of the data or materials for the experiments reported here is openly available but access can be requested, and none of the experiments were preregistered.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Relationship between moral domains and grey matter volumes in brain regions.

Table 2
Subject table of previous studies analyzing the relationship between grey matter volume and moral domains.
x,y,z: Montreal Neurological Institute and Hospital (MNI) 3-dimensional coordinate system.Behavioural/self-report variables were tested for correlation with grey matter volume.Statistical Parametric Image (SPI).

Table 3
Anatomical likelihood estimates for brain regions across moral domains.

Table 4
Moral Relevance Results.

Table 5
P-values of effect (Beta)of association between grey matter volume and moral domains assessed using linear regression analysis.Significant (p < 0.05) prior to (*) and after (**) multiple testing correction.