| Literature DB >> 25143430 |
Rosalind Arden1, Maciej Trzaskowski2, Victoria Garfield3, Robert Plomin2.
Abstract
Drawing is ancient; it is the only childhood cognitive behavior for which there is any direct evidence from the Upper Paleolithic. Do genes influence individual differences in this species-typical behavior, and is drawing related to intelligence (g) in modern children? We report on the first genetically informative study of children's figure drawing. In a study of 7,752 pairs of twins, we found that genetic differences exert a greater influence on children's figure drawing at age 4 than do between-family environmental differences. Figure drawing was as heritable as g at age 4 (heritability of .29 for both). Drawing scores at age 4 correlated significantly with g at age 4 (r = .33, p < .001, n = 14,050) and with g at age 14 (r = .20, p < .001, n = 4,622). The genetic correlation between drawing at age 4 and g at age 14 was .52, 95% confidence interval = [.31, .75]. Individual differences in this widespread behavior have an important genetic component and a significant genetic link with g.Entities:
Keywords: cognition(s); cognitive ability; creativity
Mesh:
Year: 2014 PMID: 25143430 PMCID: PMC4232264 DOI: 10.1177/0956797614540686
Source DB: PubMed Journal: Psychol Sci ISSN: 0956-7976
Fig. 1.Sample drawings of one pair of monozygotic twins (left) and one pair of dizygotic twins (right), with the scores the drawings received.
Zero-Order Phenotypic Correlations Among the Key Measures (N = 7,752 pairs)
| Variable | Drawing at 4 | |
|---|---|---|
| .33 [.32, .35] | ||
| .20 [.17, .22] | .24 [.21, .27] |
Note: The correlations and 95% confidence intervals, in square brackets, are maximum-likelihood estimates. All three correlations are significant, p < .001 (two-tailed).
Heritability of the Key Measures (Estimated From Univariate Models)
| Measure | Heritability |
|---|---|
| Drawing at age 4 ( | .29 [.22, .35] |
| .29 [.27, .32] | |
| .50 [.38, .61] |
Note: The values in square brackets are 95% confidence intervals.
Fig. 2.Trivariate Cholesky decomposition showing estimates (with 95% confidence intervals in brackets) of squared standardized path coefficients (proportion of variance accounted for) for additive-genetic (A), shared-environment (C), and stochastic or person-specific-environment (E) effects. The subscripts 1, 2, and 3 refer to g at age 4, drawing at age 4, and g at age 14, respectively.
Fit Statistics Comparing the Saturated Full Model With the Trivariate Cholesky Model
| Model | –2LL | AIC | BIC | |
|---|---|---|---|---|
| Saturated | 85,530.89 | 33,976 | 17,578.89 | –218,748.18 |
| Trivariate | 85,604.07 | 34,009 | 17,586.07 | –218,970.54 |
Note: A chi-square test indicated that the fit of the trivariate Cholesky model was significantly better than the fit of the saturated model, χ2 = 73.18, p = .0000711. LL = log likelihood; AIC = Akaike’s information criterion; BIC = Bayesian information criterion. Smaller (more negative) BIC values indicate better fit.