Process Overlap Theory: A Unified Account of the General Factor of Intelligence

ABSTRACT

The most replicated result in the field of intelligence is the positive manifold, which refers to an all-positive pattern of correlations among diverse cognitive tests. The positive manifold is typically described by a general factor, or g. In turn, g is often identified as general intelligence, yet this explanation is contradicted by a number of results. Here we offer a new account of g: process overlap theory. According to the theory, cognitive tests tap domain-general executive processes, identified primarily in research on working memory, as well as more domain-specific processes. Executive processes are tapped in an overlapping manner across cognitive tests such that they are required more often than domain-specific ones. The theory provides an account of a number of findings on human intelligence. As well, it is formalized as a multidimensional item response model and as a structural model, and the neural mechanisms underlying the proposed overlapping processes are discussed.

Keywords:

• Cognitive abilities
• differentiation
• factor analysis
• goal neglect
• individual differences
• intelligence
• prefrontal cortex
• working memory
• worst performance rule

Notes

¹ In Figures 1 to 5 we use the contemporary notation of unique variance (delta) instead of s.

² More precisely: to evaluate the theoretical status of reflective latent variables, see “Process Overlap Theory.”

³ We are aware that there are several important models of intelligence other than the Gf/Gc model (e.g., Johnson & Bouchard, 2005). Yet in practically the entirety of research on working memory and intelligence, as well as on goal neglect and intelligence, Gf-Gc is the model that was applied, and this line of research lays the foundations of our theory. Hence our focus on Gf/Gc is motivated by its proliferation of recent cognitive research on intelligence through providing the comprehensive framework of “fluid reasoning,” which is readily interpretable by cognitive psychologists. See, for instance, Blair (2006); Heitz et al. (2006); and Kovacs, Plaisted, and Mackintosh (2006).

⁴ This is a typical (albeit incorrect; see Mackintosh, 2011b) line of criticism against the importance of the positive manifold.

⁵ Formative and reflective measurement is drastically different, but this issue cannot be dealt with in this article. The interested reader is referred to Bagozzi (2007); Edwards (2011); and Howell, Breivik, and Wilcox (2007).

⁶ Working memory serves only as a comprehensible illustration here: Scores on working memory tasks are nondichotomous, and the actual IRT model, described by the preceding equation, is applicable only to dichotomous test scores.

⁷ With notable exceptions: Horn (1989), for instance, in his categorization of ability tests according to the Gf-Gc model, put “inductive reasoning, measured using letter series, number series and/or figure series” as the first example of indicators of Gf, “matrices reasoning with visual patterns” comes only second (p. 79).

⁸ This assumption by Thomson (1916) was, in fact, more practical than substantive: “Note that I do not for one moment suggest that psychological ‘factors,' if they exist, can be added together like dice: I merely intend to apply Professor Spearman's formulae to dice throwing” (p. 275).

⁹ Even though the article does not refer to Thomson or to the concept of sampling.