Darwin experienced them while hearing the anthem in King’s College Chapel, which gave him “intense pleasure so that [his] backbone would sometimes shiver”([1], p. 61). Nabokov elevated them as “the telltale tingle” needed to bask in the book of genius [2]. When people encounter art, they may experience intense physiological responses that coincide with peak subjective pleasure and emotions, described as “chills” [3]. Since chills are clear, measurable events [4] that link subjective responses with physiological manifestations [4–7], they serve as a model for studying responses to art [3,6–11]. Investigations of chills have revealed that the pleasure evoked by music and poetry recruits neural systems similar to those involved in the evaluation of other biologically relevant non-abstract stimuli [3,7] and have linked stable subjective differences to inter-individual variability at both physiological and neurobiological levels [8,10,12].

Here, we investigate the genetics behind individual differences in proneness to chills, which we suggest can offer an additional approach for empirically studying how humans respond to art. As with many other complex human traits [13], proneness to chills is partly heritable [14]. For example, across diverse cultures, proneness to chills from visual art and poetry (referred to as aesthetic chills) has moderate heritability (h²). Genetic effects have been estimated to account for 36–43% of the variance in large twin-based samples—including more than 10,000 individuals from the Netherlands [14], and thousands from Germany, Denmark, North America, Australia, and Japan [15]—with the remaining variance attributed to residual effects, such as environmental influences and measurement error, with no evidence for environmental effects shared by family members [14]. However, key questions about the genetics of inter-individual differences in proneness to chills remain unanswered. First, to date, h² estimates have been derived from family studies only, mainly from those relying on the Classical Twin Design (CTD). As such, it is unclear whether previous results reflect molecular genetic effects or biases introduced by unmet assumptions inherent to the CTD. Second, to the best of our knowledge, there is a lack of evidence regarding whether genetic effects on aesthetic chills may extend to other artistic domains, such as music, since the only direct estimates derived so far are specific to chills from visual art and poetry. This gap raises additional questions of whether sensitivity to different art forms is shaped by entirely different processes or is tied to similar underlying biological mechanisms. Finally, it is still unclear whether the genetics of proneness to chills is indeed an index of broader predispositions towards art, or if it reflects an independent phenomenon.

To resolve these uncertainties, we used a large sample of more than fifteen thousand genotyped individuals from Lifelines, a large multi-generational cohort study of individuals living in the North of the Netherlands [16], from whom we gathered self-reports on their proneness to aesthetic and music chills. We applied a molecular approach to quantify h² that contrasts patterns of measured genotypic resemblance with phenotypic resemblance in unrelated and related individuals [17,18]. The strength of this approach relies on the possibility of partitioning family-based (also called pedigree-based) estimates for h² () into two types of h². The first, obtained from unrelated and distantly related individuals and referred to as the Single-Nucleotide Polymorphism heritability (), estimates the additive effect of genetic variants common in the population under study, captured using DNA genotyping chips. The second, obtained instead from related individuals, assesses additional additive effects of genetic variants untagged by genotyping chips, as well as biases due to, for instance, non-additive genetic and environmental effects shared between members of families [19,20]. The gap between such and estimates is sometimes referred to as the (still) missing heritability [21]. Moreover, by collecting information on people’s proneness to chills evoked by different art modalities, we could investigate similarities and differences between responses to diverse forms of artistic, poetic and musical expression. We accomplished this by estimating genetic correlations () between aesthetic and music chills. Finally, we investigated the degree to which proneness to chills might reflect a broader genetic predisposition towards art. To do so, we leveraged the largest published genome-wide association study (GWAS) on personality to date [22] to construct an openness to experience polygenic index (PGI), including artistic interests and active imagination, testing their association with proneness to chills in our target sample.

Working with Lifelines [16], we gathered self-report data on proneness to chills in response to diverse art forms from 35,114 adults (21,249 women). Some 15,615 of these individuals (9,907 women) had genotype data available, and their ages covered the entire adult lifespan, ranging from 18 to 96 years old. Proneness to aesthetic and music chills was assessed using the self-reported items “Sometimes when I am reading poetry or looking at a work of art, I feel a chill or wave of excitement” [23] and “I sometimes feel chills when I hear a melody that I like” [24], respectively. Participants were asked to report how much they agreed with these statements by choosing one out of five response options ranging from “strongly disagree” to “strongly agree”. Within-trait descriptive statistics and between-trait phenotypic similarities and differences are reported in the S1 Text.

To assess the extent to which genetic influences account for variation in proneness to aesthetic and music chills, we applied a linear mixed model with a Genome-based Restricted Maximum Likelihood (GREML) estimator [18]. This method estimates the amount of the observed phenotypic variance that can be explained by genomic relatedness between individuals, as indexed by SNP-derived Genetic Relatedness Matrices (GRM). We focused our analysis on a total sample of 15,606 individuals, removing a total of 9 individuals from pairs with estimated genomic relatedness () larger than .90 (i.e., identical twins, see Table A in S1 Text). Taking advantage of the family relatedness structure of the participants included in the Lifelines cohort, we constructed two complementary SNP-derived GRM, one representative of across all 15,606 individuals and another for which below a set threshold (i.e., <  .05, the recommended genetic relatedness threshold for GREML-based analysis in family data [19], see also S1 Text) was adjusted to zero [17]. Using this approach, also known as threshold GREML [25], we could partition into and , the latter assessing effects beyond common SNP variation tagged by genotyping array (see Fig A in S1 Text). As shown in Table 1, GREML-based analysis of adjusted two-stage rank normalised and residualised data for age, sex, genotyping array, and ten genomic principal components (PC) indicates significant for aesthetic (χ²(1) = 6.30, p = .006) and music (χ²(1) = 9.25, p = .001) chills, with explaining up to 29% of the total variance in proneness to chills. These findings indicate that the proneness to chills from art, poetry and music is influenced by genetic variation. Results obtained from raw, transformed data, and by using a different threshold ( < .02), were similar (see Tables B and C in S1 Text).

Finally, we assessed whether the genetics of proneness to chills could in part be reflected by broader genetic associations with art interest. We derived individual PGI from the largest published genome-wide association study of openness to experience, which was previously conducted in a fully independent sample of 220,015 individuals [22]. To construct the PGI, we used SNP associations with openness to experience, assessing artistic interests and active imagination (see Methods). We found that PGI were significantly associated with proneness to aesthetic (β = .05, p = 4.79 × 10⁻¹⁰) and music chills (β = .03, p = 1.97 × 10⁻⁶), explaining 0.3% ( 95% CI [0.01%, 0.4%]) and 0.1% ( 95% CI [0.0%, 0.3%]) of the overall variance, respectively. Similar results were obtained in a subset of 10,703 distantly related and unrelated individuals ( < .05; See S1 Text). Further analyses provided no evidence for heterogeneous PGI associations with either aesthetic or music chills (Q_PGI = 1.95, p = .10), suggesting that PGI explained about 0.5% (95% CI [0.2%, 0.8%]) of the overall covariance between the two traits (see Methods, S1 Text, and Fig B in S1 Text). Thus, genetic variation associated with individual differences in openness to experience, including art interest, contributes to proneness to chills across art domains. Table 1 provides a comprehensive overview of the results of these different investigations.

We note that assortative and cross-assortative mating can lead to bias when estimating and . Assortative and cross-assortative mating refer to phenomena where two individuals are more likely to mate when they resemble each other in one or more (e.g., cross) traits. These phenomena have been shown to confound and by upwardly biasing estimates [26,27]. To provide evidence of minimal upward bias on , and help solidify the interpretation of as an estimate for widespread pleiotropy unconfounded by both types of assortment, we estimated trait and cross-trait partner correlations in proneness to chills. To do so, we studied 3408 partners for whom phenotypic data were available. We observed small yet significant partner correlations for aesthetic chills, = .13 (95% CI [.10, .16]), and music chills, = .11 (95% CI [.08, .15]). There was no evidence of cross-trait partner correlation, with a non-significant correlation of only .03 (95% CI [-.01, .06]). These findings indicate that assortative and cross-assortative mating do not largely bias our estimates for and (for details, see S1 Text and reference [26]).

Our results provide converging evidence for the existence of a genetic predisposition to the ability to experience chills from cultural products, such as artistic, poetic, and musical expression. By investigating the similarity between genotypic and phenotypic resemblance in related and unrelated individuals, we find significant SNP-based heritability for aesthetic and music chills, aligning with the claim that proneness to chills may represent a “vector of biological variation” [28]. Consistent with other studies [14,15], the substantial proportion of variance attributable to residual effects—including environmental influences, measurement error, but also stochastic developmental processes unique to each individual—indicates that factors beyond genetic variation also contribute to individual differences in proneness to chills. Furthermore, we reveal a substantial genetic correlation between the traits, which indicates that proneness to chills from different art modalities is partly tied to shared genetic variation. In line with these findings, we also find that PGI for openness to experience, including artistic interests, are significantly associated with proneness to aesthetic and music chills. At the same time, the genetic correlation between propensity to aesthetic and music chills is significantly lower than 1, indicating less than full overlap in the underlying genetic architecture for responses to different forms of artistic expression.

Notwithstanding the significant SNP-based heritability, this explained only around a fourth of the total pedigree heritability and a fifth of the previously reported twin-based heritability for aesthetic chills (obtained from another sample of the Dutch population [14]). While in line with well-established concepts of “missing heritability” [29], as SNP-based estimates tend to explain only one-third to one-half of family-based estimates of complex human traits [20,21,30], the gap observed here for proneness to chills seems relatively large. (Note, however, that this is still smaller than the gap observed for some other behavioural characteristics, for which is only about one-tenth of the ; see reference [30] for an example). The observed gap between SNP- and family-based heritability estimates likely involves a complex mixture of factors. These include contributions of imperfect tagging of causal variants from genotyping arrays, such as rare variants, non-additive genetic effects, such as dominance deviations or higher-order epistasis (gene-by-gene interaction at different genetic loci), and common environmental effects shared across members of families [17]. Additional contributors may include unequal environmental sharing among relatives and interactions between genes and shared environmental factors [31].

It is worth noting here that while estimates can be considered the upper bound of the additive genetic influences that genome-wide association studies (GWAS) could capture on the same genotyping platform, they are generally expected to underestimate the true heritability of the trait in the population under study. This discrepancy arises because the GRM imperfectly captures the contribution of ungenotyped variants among unrelated or distantly related individuals. By contrast, when the same GRM is obtained from pairs of individuals above a certain threshold of relatedness (e.g., ≥ .05), the estimates tend to be closer to the true heritability. This is because, at higher levels of relatedness (larger values of ), the GRM better approximate the genetic covariance across the genome. Therefore, a gap between and is expected [17]. At the same time, the estimates derived from related individuals can be upwardly biased if shared environmental effects, non-additive genetic effects, or effects of rare variants are present [32]. Previous CTD studies on proneness to aesthetic chills found no evidence of shared environmental effects [14,15], with correlations among monozygotic twins exceeding twice those of dizygotic twins [14]. These findings, under the assumptions of identical shared environmental influences in monozygotic and dizygotic twins and no gene-by-shared environment interaction, suggest that non-additive genetic effects, rather than shared environmental effects, may partly contribute to the gap between and as found in this study. However, an alternative explanation, which cannot be ruled out in the current study, is also possible: gene-by-shared environment interactions. If gene-by-shared environment interactions were to influence phenotypic variation, then both the current and the previously reported may have yielded upwardly biased estimates of h² [33]. Furthermore, although we noted that the observed weak correlation between partners was not sufficient to largely bias estimates in this study, it is still possible that assortment may have previously upwardly biased estimates (as these were obtained from a constrained ADE model, which is expected to provide inflated estimates under assortative mating [33]). Therefore, assortative mating may still play a role in the gap observed between the estimates provided in this study and the ones reported from previous CTD studies. Future studies with different research paradigms will help to better understand this gap. In particular, more sophisticated twin-based approaches, such as extended family twin designs (in which parents, partners, and siblings of twins are included [33,34]), may be particularly suited to partly explain the gap observed in this study.

In addition, our study reveals the association between proneness to chills and PGI for openness to experience—assessed as individuals’ interest towards the arts and self-reported active imagination (see below and Methods). We note that the percentage of variance in proneness to aesthetic and music chills explained by the PGI is less than 1%. However, the modest proportion of variance accounted for here is in line with theory-driven expectations on PGI-based predictions. In particular, it is known that the expected percentage of phenotypic variance that can be explained by a PGI is a function of the heritability of the target phenotype being predicted, the one from which the PGI is derived, and their genetic correlation, as well as the sample size of the original GWAS, along with the effective number of SNPs included [35,36]. Considering the observed of proneness to chills and the previously reported heritability of openness to experience [22] ( = .05), then the maximum expected percentage of variance that we could have explained is ~ 1% (see S1 Text) for aesthetic and music chills, even assuming an unlikely perfect genetic correlation of 1. These expectations are not far from our observed , thus contextualising our estimates for the absolute percentage of variance explained by the PGI as being relatively close to theoretical expectations.

This study comes with a few limitations. First, our PGI analyses made use of summary statistics from GWAS efforts conducted with self-reports that included art interest and active imagination [22]. As such, individual PGI assess also other indicators of openness to experience beyond art interest alone. Second, the results in the present study were obtained from a population that was representative of the Netherlands and focused on individuals of European descent. Further studies should assess the generalisability of our findings beyond ancestries of European descent and culture. Indeed, the cross-cultural prevalence of experiencing chills [37] and proneness towards these [23], and the observation that such proneness may rely less on cultural bounds and influences from Western humanistic traditions than other traits [23,28], leads to intriguing questions for future research. In particular, it will be interesting to ask whether proneness to chills may be used to investigate the genetics of human attitudes towards art across cultures. Moderating effects introduced by cultural or generational differences in genetic influences could be an especially promising avenue of study [38]. Third, our results do not provide insights into how molecular associations with proneness to chills may extend to biological mechanisms underlying related physiological processes, such as piloerection (i.e., hairs standing on end). Previous research suggests that piloerection tends to co-occur with the experience of chills [39], although this is not always the case [40]. At the same time, the findings of genetic associations with proneness to chills may allow future research to better disambiguate between these two seemingly tied phenomena. Fourth, sampling bias in the Lifelines cohort may affect heritability and genetic correlation estimates. While Lifelines represents the northern Netherlands well, women, middle-aged individuals, and individuals with higher educational attainment tend to be slightly overrepresented [41]. Since proneness to chills is linked to both demographics and education [42], this may introduce bias in genetic estimates. However, evidence from other biobanks suggests that heritability is typically only modestly underestimated under such bias [43], with downstream analyses being more strongly affected [44]. Therefore, genome-wide association studies may be more susceptible to this type of bias than the present study. Future work should address this issue, for example, by applying weighting to genetic associations based on census data on sex and educational attainment, to better align the sample estimates to the effects in the overall population [45]. Fifth, the items used to assess proneness to chills are likely subject to measurement error. For instance, the test-retest over a 5-year timespan for aesthetic chills in adults has been previously estimated to range between .58 and .61, in a comparable but not overlapping Dutch sample [14]. As such, our h² estimates are likely downwardly biased by intra-individual variance (including measurement error), and therefore should be considered conservative lower bounds for the additive effects of SNP variation.

In conclusion, our research highlights shared molecular heritability for proneness to chills from art, poetry, and music. Furthermore, it reveals that proneness to chills is associated with a PGI for openness to experience, partly indexing genetic associations with art interest. Yet it also reveals larger estimates for genetic effects on proneness to chills in related individuals, hinting at deviations from pure additive genetic transmission of sensitivity to art and underscoring the need for extended intergenerational designs to relax classic modelling assumptions. We anticipate that larger-scale GWAS of proneness to chills, combined with extended family designs, could further illuminate the origins and biological bases of affective responses to art. Such studies would enable the exploration of both molecular and cultural processes and may even clarify whether the aetiology of these subjective experiences differs from that of other sensory experiences [46].