The most replicated finding in intelligence research is the positive manifold, which refers to the finding that intelligence tests tend to be positively correlated. According to Spearman, this positive manifold can be accounted for by a general factor of intelligence (g factor). The most common interpretation of g is that it represents a general mechanism that influences performance on all intelligence tests. g is considered as a reflective latent variable, a causal psychological attribute. However, recent models suggested that g is a formative latent variable; g is not the cause of the correlations between tests, but the consequence. It does not represent a causal psychological attribute. The psychometric network modeling can be used to test this formative hypothesis, because psychometric network analysis does not assume latent common causes.

The aim of this study was to estimate the underlying structure of the Wechsler Intelligence Scale for Children-Fifth Edition (WISC–V) with five standardization sample age groups (6-7, 8-9, 10-11, 12-13, 14-16). Psychometric network analyses were conducted based on the correlation matrices provided for the five age groups in the WISC–V interpretive manual. Psychometric network models revealed 3 clusters (6-7, 10-11, 12-13, 14-16) or 4 clusters of nodes (8-9). The two stable clusters were processing speed and verbal comprehension. The distinction between fluid reasoning and visuospatial ability was never found. Data were not consistent with the five-factors recommended by the publisher of the WISC–V. Psychometric network analysis is an alternative to investigate the underlying structure of intelligence data.