Great thinkers of the past from Aristotle to Shakespeare have remarked that creative genius and insanity are often characterized by the same unleashing of thoughts and emotions. This is supported by epidemiological studies demonstrating overlap between psychiatric disorders and creativity1,2,3,4,5,6,7. In a large Swedish study, Kyaga et al. found that individuals with bipolar disorder and healthy siblings of people with schizophrenia and bipolar disorder are over-represented in creative professions but those with major depression and their relatives are not. This overlap appears to be independent of IQ8. By contrast, members of the creative professions are not more likely to be diagnosed with psychiatric diseases with two exceptions: writers are more likely to be diagnosed with psychiatric disorders in general and members of all of the creative professions are more likely to be diagnosed with bipolar disorder9.

Psychological theories propose that the schizophrenic spectrum is accompanied by a decrease in practical reasoning, as schizophrenia patients outperform controls in logical deduction that is in conflict with practical reasoning8,10. Furthermore, it has been suggested that those less restrained by practical cognitive styles may have an advantage in artistic occupations8. These results provide support for the notion that creativity and psychiatric disorders, particularly schizophrenia and bipolar disorder, share psychological attributes. However, whether and to what degree this is due to shared environment or genetics has not been assessed with modern genomic tools.

Creativity can be viewed in various ways11,12, and, although it is a difficult concept to define for scientific purposes, the creative person is most often considered one who takes novel approaches requiring cognitive processes that are different from prevailing modes of thought or expression11. Thinking differently from others is therefore a prerequisite for creativity11. Schizophrenia and bipolar disorder are disorders of thoughts and emotions, which means that those affected show alterations in cognitive and affective processing. Yet it is unclear whether the cognitive deviations of psychiatric patients and of creative individuals can be explained in part by shared genetic variation. We have previously shown that control carriers of copy number variants (CNVs) conferring risk of schizophrenia have cognitive abnormalities that are akin to those encountered in schizophrenia although not as severe13. The population controls carrying the neuropsychiatric CNVs did not have schizophrenia, autism, bipolar disorder or a diagnosis of intellectual disability. Hence, variants in the genome that confer modest to high risk of psychiatric diseases affect cognition in controls.

Here we investigated whether common variants that affect the risk of schizophrenia and bipolar disorder, and thus are potentially detrimental to the individuals who carry them, may also underlie cognitive traits that can be advantageous to society. Polygenic risk scores, or cumulative genetic risk profiles from across the genome, were used to explore this question. Based on results from two mega-analyses14,15, neither of which included Icelandic data, we generated separate polygenic risk scores for schizophrenia and bipolar disorder. These scores were then examined in a sample of 86,292 individuals from the general population of Iceland.

We first tested the ability of these polygenic risk scores to predict their corresponding disorders. Prediction accuracy (variance explained) was assessed throughout the study using Nagelkerke's pseudo-R2, the likelihood-based measure used in previous work14,15. We found that both scores did associate with their matching disorder in Iceland (n = 583 and 500 for schizophrenia and bipolar disorder, respectively), with prediction accuracy plateauing at a P-value threshold of around 0.1 to 0.2 (Fig. 1a). The maximum variance explained was 5.5% for schizophrenia and 1.2% for bipolar disorder. To consider all polygenic risk scores on the same scale, scores were centered and scaled to have a mean of 0 and a s.d. of 1. Odds ratios (ORs) computed using these scores were 2.22 for schizophrenia and 1.46 for bipolar disorder based on a P value threshold of 0.2.

Figure 1: Polygenic risk scores for schizophrenia and bipolar disorder predict their corresponding disorder and creativity. (a) Prediction of schizophrenia (SCZ) and bipolar disorder (BD) in Iceland using polygenic risk scores derived from independent GWASs of these disorders14,15. Nagelkerke's pseudo-R2 is shown for scores derived using ten significance thresholds. Variance explained on the liability scale, which is adjusted for case-control ascertainment, is around 8% for schizophrenia and 2% for bipolar disorder, in keeping with previous estimates14. (b) Prediction of creativity in Iceland using scores derived from independent GWASs of schizophrenia and bipolar disorder14,15. Nagelkerke's pseudo-R2 is shown for scores derived using ten significance thresholds. Full size image Download Excel source data

Next, we tested for an association between the polygenic risk scores and creativity. Creative individuals were defined as those belonging to the national artistic societies of actors, dancers, musicians, visual artists and writers (n = 1,024; Supplementary Table 1). We found that both schizophrenia and bipolar disorder polygenic risk scores were associated with creativity, with the schizophrenia and bipolar disorder scores explaining a maximum of 0.24% and 0.26% of the variance of creativity, respectively (Figs. 1b and 2). At a P-value threshold of 0.2, ORs for creativity were 1.17 for both scores (P = 5.2 × 10−6 and 3.8 × 10−6 for the schizophrenia and the bipolar scores, respectively).

Figure 2: Nagelkerke's pseudo-R2 in Iceland for schizophrenia (SCZ), bipolar disorder (BD), artist (Art) and university degree (Univ) based on schizophrenia and bipolar disorder polygenic risk scores derived at the significance threshold of P < 0.2. Full size image Download Excel source data

In contrast to these results, none of 20 non-psychiatric diseases tested, including three phenotypes affecting cognition, were significantly associated with the psychosis polygenic risk scores (Supplementary Table 2). We also found no significant association with the psychosis scores for five other types of profession (Supplementary Table 3).

In an analysis including both scores as predictors of creativity, both scores were significantly associated with artist status, together accounting for 0.39% of the variance in creativity (P = 6.8 × 10−8), somewhat less than the sum of their independent scores owing to the positive correlation between them (r = 0.23). ORs were similar for four different classes of artist (actor or dancer, musician, visual artist and writer) using either schizophrenia or bipolar disorder scores (Supplementary Table 4).

To determine whether the association with creativity is driven by general cognitive function or educational attainment, we also tested for an association of the polygenic risk scores with the number of years in school or having a university degree. We found that higher bipolar disorder polygenic risk scores were associated with greater educational attainment (β = 0.15, P = 4.8 × 10−9 and OR = 1.09, P = 5.2 × 10−7 for number of years in school and university degree, respectively) and that higher schizophrenia scores also predicted greater educational attainment, although less significantly (β = 0.096, P = 1.2 × 10−4 and OR = 1.04, P = 0.014 for number of years in school and university education, respectively). In the subset of the data with education information, both bipolar disorder and schizophrenia scores remained significant predictors of creativity when educational attainment was included in the model, with little or no attenuation of the effect (OR = 1.18 and 1.10 for the schizophrenia and bipolar scores, respectively; for schizophrenia scores, P = 1.2 × 10−4 and P = 3.6 × 10−4 without and with adjustment for education, respectively, and for bipolar disorder scores, P = 0.0078 and P = 0.038 without and with adjustment for education, respectively), suggesting that the association of the polygenic risk score with creativity cannot be accounted for by differences in educational attainment.

Schizophrenia polygenic risk scores are elevated in relatives of schizophrenia patients, such that scores of first degree relatives are expected to be elevated by half the elevation of schizophrenia patients and those of second degree relatives are expected to be elevated by one quarter. The elevation in the group of creative individuals was 20% of that seen in patients with schizophrenia. It is conceivable that the association we observe between the polygenic schizophrenia risk score and creativity could be driven by having a relative with schizophrenia. In this case the observed elevation in creative people would demand that, on average, they would have a second degree relative with schizophrenia. The actual average meiotic distance between creative individuals and their closest relative with schizophrenia was 6.2 meioses. That would only account for an elevation of the polygenic schizophrenia risk score of 5.03%, which does not explain the observed association with creativity (0.21, compared to 0.045 under the null hypothesis; P = 3.3 × 10−4).

We also examined the association of creativity and psychosis polygenic risk scores conditional on the fraction of first, second and third degree relatives who were schizophrenia or bipolar disorder patients. This resulted in little change to the significance of the association between creativity and psychosis polygenic score (OR = 1.17, P = 7.1 × 10−6 and OR = 1.17, P = 4.3 × 10−6 for schizophrenia and bipolar disorder scores, respectively), demonstrating that the association is not solely the result of a closer relationship between creative individuals and psychosis patients.

Finally, we investigated the association between creativity and psychosis polygenic risk scores in four longitudinal studies from the Netherlands (n = 18,452) and Sweden (n = 8,893). Two creativity phenotypes were used: artistic profession, which was available for all cohorts, and a quantitative measure of creativity derived from the Creative Achievement Questionnaire (CAQ) and assessing an individual's activities in the fields of visual arts, music, dance, writing and theater (CAQ-Arts) that was available for a subset of one study. For artistic profession, similar ORs as those for artistic society membership in Iceland were observed (in the combined cohorts, OR = 1.23, P = 0.0021 and OR = 1.23, P = 8.6 × 10−4 for schizophrenia and bipolar disorder scores, respectively; Table 1). The CAQ-Arts measure was associated with both scores in the expected direction (Table 1). The power of the replication cohorts to detect an effect of similar magnitude to that in the discovery sample was around 70–80%.

Table 1: Prediction of artist status by schizophrenia and bipolar disorder polygenic risk scores in four replication cohorts Full size table

It has been suggested that alleles conferring risk of disorders reducing fecundity, such as schizophrenia, may persist through balancing selection if their negative fitness effects are offset by benefits16,17. However, creative individuals in our sample had fewer children than population controls (effect derived from a Poisson log-linear model = 0.92, P = 6.8 × 10−5), making an offset manifested through creativity unlikely.

The study of the association between creativity and psychiatric disorders has in the past primarily been addressed in two types of studies: studies assessing psychiatric disorders in eminent creative individuals and studies assessing creativity in psychiatric patients and their relatives. Our study lends support to direct influences of genetic factors on creativity as opposed to sharing an environment with individuals with psychosis influencing creative aptitude. Thus, the main finding presented here is that creativity, conferred, at least in part, by common genetic variants, comes with an increased risk of psychiatric disorders conferred by the same genetic variants. How this genetic overlap fits into evolutionary models of disease persistence remains to be determined.