The Cattell-Horn-Carroll theory is the principal model at present to define and assess intelligence. However, several authors have suggested that the structure of intelligence is not stable throughout life. For example, children tend to present a general ability, which is invested in the development of specific and differentiated abilities during adolescence, reflecting the effect of schooling, experiences, and interests. In this study, an intelligence nine subtests battery was applied to a sample of 470 5-, 7-, and 9-year-old Portuguese children, from public and private basic schools. Two structure models have been tested, one assuming a global general factor loading by the nine subtests, and a second one formed by two factors integrating the verbal and non-verbal subtests. The results provide evidence of better fit of the second model where intelligence performance is organized in traditional verbal and non-verbal factors. This structure model is confirmed taking 5-, 7-, and 9-year-old children subgroups. A posterior path analysis verifies the impact of socio-familial variables in both factors. More than parents’ educational levels and professional status, the results suggest the sibship size appears as more relevant on explanation of children cognitive performance.