Data from: What caused over a century of decline in general intelligence? Testing predictions from the genetic selection and neurotoxin hypotheses

Several converging lines of evidence indicate that general intelligence (g) has declined in Western populations. The causes of these declines are debated. Here, two hypotheses are tested: (1) selection acting against genetic variants that promote g causes the decline and (2) the presence of neurotoxic pollution in the environment causes the decline. A linear mixed model was devised to test (1) and (2), in which the secular decline in a “heritable g” (g.h) chronometric factor (comprised of convergent indicators of simple reaction time, working memory, utilization frequencies of high difficulty and also social-intelligence-indicating vocabulary items and per capita macro-innovation rates) was predicted using a neurotoxin chronometric factor (comprised of convergent secular trends among measures of lead, mercury and dioxin + furan pollution, in addition to alcohol consumption) and a polygenic score chronometric factor (comprised of polygenic score means for genetic variants predictive of g, sourced from US and Icelandic age-stratified cohorts). Bivariate correlations revealed that (other than time) only the polygenic score factor was significantly associated with declining g.h (r = .393, p < .05 vs. .033, ns for the neurotoxin factor). Using a hierarchical linear mixed model approach incorporating 25 year lags between the predictors and g.h, time period, operationalized categorically as fifths of a century, accounted for the majority of the variance in the decline in g.h (partial η^2 = .584, p < .05). Net of time period and neurotoxins, changing levels of polygenic scores also significantly predicted variance in the decline in g.h (partial η^2 = .253, p < .05); however, changing levels of neurotoxins did not significantly predict variance in g.h net of time (partial η^2 = .027 ns). Within-period analysis indicates that the independent significant positive effect of the polygenic score factor on g.h was restricted to the third fifth of a century period (β = .202, p < .05).

多项汇聚的证据表明，西方人群的通用智力（general intelligence，以下简称g）已出现长期下降。此类下降的成因尚存争议。本文对两项假说展开检验：其一，针对促进g的遗传变异的选择作用导致了该下降；其二，环境中存在的神经毒性污染物引发了该下降。我们设计了线性混合模型以检验上述两项假说，模型以可遗传通用智力（heritable g，下文简称g.h）计时因子（chronometric factor）的长期下降趋势作为预测目标——该计时因子由多项汇聚式指标构成：简单反应时、工作记忆、高难度词汇与体现社会智力的词汇的使用频率，以及人均宏观创新率。模型的预测因子包含两项计时因子：其一为神经毒素计时因子，由铅、汞、二噁英与呋喃类污染物的浓度变化趋势，以及酒精消费情况的汇聚趋势构成；其二为多基因得分（polygenic score）计时因子，其数据取自美国与冰岛按年龄分层的队列（age-stratified cohorts），由预测g的遗传变异对应的多基因得分均值构成。双变量相关分析显示，除时间变量外，仅多基因得分因子与g.h的长期下降存在显著关联（r=0.393，p<0.05；神经毒素因子的相关系数为0.033，无统计学意义）。采用分层线性混合模型方法，并将预测因子与g.h之间设置25年的滞后效应，结果显示，以每五分之一世纪分类的时间周期变量，解释了g.h下降的大部分方差（偏η²=0.584，p<0.05）。在控制时间周期与神经毒素变量后，多基因得分的变化仍能显著预测g.h下降的方差（偏η²=0.253，p<0.05）；但在控制时间变量后，神经毒素水平的变化并未显著预测g.h的方差变化（偏η²=0.027，无统计学意义）。同期分析表明，多基因得分因子对g.h的独立显著正向效应仅存在于第三个五分之一世纪区间（β=0.202，p<0.05）。