The structure of intelligence is not static. Differentiation hypotheses suggest that the contribution of g and specific abilities on tasks performance varies with age, but most research focused solely on g. Here we sought to test intelligence differentiation in both its general and specific components using a bifactor modelling approach that – previously ignored in literature – should be better suited to this aim. In parallel, the possible differences in intelligence differentiation between males and females are explored for the first time. A population of 8866 youths (8–19 years old) of the Philadelphia Neurodevelopment Cohort completed 12 tasks measuring four components of cognition (complex cognition, executive control, episodic memory, and social cognition) and the WRAT-4. Using bootstrap and meta-regression analysis, the bifactor-(S·I - 1) model shows that the variance explained differently changes over time according to peculiar development patterns of specific components. Executive functions lose specificity with age and their variance is explained only by g at the end of adolescence, episodic memory is increasingly explained by g, and complex cognition is explained more by g in males – and less in females – when age increases, also suggesting a possible role of gender in intelligence differentiation. We conclude discussing the importance of using adequate statistical models and we proposed studying differentiation also at the level of specific abilities to account for the complexity of the developmental phenomenon that could be better described by development priority theories.

智力的结构不是一成不变的。差异化假设表明，g和特定能力对任务绩效的贡献随年龄而变化，但大多数研究仅关注g. 在这里，我们试图使用以前在文献中忽略的双因子建模方法来测试其一般和特定组成部分的智力差异，这种方法应该更适合这一目标。同时，首次探讨了男性和女性在智力差异上可能存在的差异。费城神经发育队列的 8866 名青年（8-19 岁）完成了 12 项任务，测量认知的四个组成部分（复杂认知、执行控制、情景记忆和社会认知）和 WRAT-4。使用自举和元回归分析，双因子-(S·I - 1) 模型表明，解释的方差根据特定组件的特殊发展模式随时间而变化。g在青春期末期，情景记忆越来越多地被g解释，而复杂认知在男性中更多地被g解释——而在女性中更少——当年龄增加时，这也表明性别在智力分化中可能发挥作用。最后，我们讨论了使用适当的统计模型的重要性，并建议在特定能力的水平上研究差异化，以解释发展优先理论可以更好地描述的发展现象的复杂性。