“Is there a biology of intelligence which is characteristic of the normal human nervous system?” Here we review 37 modern neuroimaging studies in an attempt to address this question posed by Halstead (1947) as he and other icons of the last century endeavored to understand how brain and behavior are linked through the expression of intelligence and reason. Reviewing studies from functional (i.e., functional magnetic resonance imaging, positron emission tomography) and structural (i.e., magnetic resonance spectroscopy, diffusion tensor imaging, voxel-based morphometry) neuroimaging paradigms, we report a striking consensus suggesting that variations in a distributed network predict individual differences found on intelligence and reasoning tasks. We describe this network as the Parieto-Frontal Integration Theory (P-FIT). The P-FIT model includes, by Brodmann areas (BAs): the dorsolateral prefrontal cortex (BAs 6, 9, 10, 45, 46, 47), the inferior (BAs 39, 40) and superior (BA 7) parietal lobule, the anterior cingulate (BA 32), and regions within the temporal (BAs 21, 37) and occipital (BAs 18, 19) lobes. White matter regions (i.e., arcuate fasciculus) are also implicated. The P-FIT is examined in light of findings from human lesion studies, including missile wounds, frontal lobotomy/leukotomy, temporal lobectomy, and lesions resulting in damage to the language network (e.g., aphasia), as well as findings from imaging research identifying brain regions under significant genetic control. Overall, we conclude that modern neuroimaging techniques are beginning to articulate a biology of intelligence. We propose that the P-FIT provides a parsimonious account for many of the empirical observations, to date, which relate individual differences in intelligence test scores to variations in brain structure and function. Moreover, the model provides a framework for testing new hypotheses in future experimental designs.

The target article provides a thoughtful review and synthesis of studies examining the neural basis of cognitive abilities associated with intelligence test performance. In its attempt to present a new or generative theory of the neural basis for intelligence, however, the review faces specific limits to its theoretical model that relate to processes of development and the role of automaticity in cognition.

In this commentary we make two rejoinders to Jung & Haier (J&H). First, we highlight the response selection component in tasks as a confounding variable that may explain the parieto-frontal involvement in studies of human intelligence. Second, we suggest that efficient response selection may be an integral part of the definition of intelligence.

Neuroimaging evidence, both within and between research strategies, is largely heterogeneous. This results from the way the construct of interest (i.e., intelligence) is measured. Every single available measure comprises several cognitive abilities, although the so-called g factor is always present. Here I suggest that studies must always control for this empirical fact to arrive at solid conclusions.

This commentary compares the P-FIT model with psychometric and developmental models of intelligence and shows that there are isomorphisms and divergences between them. All three models involve some common dimensions, but the P-FIT model lacks many of the dimensions of the other models. Then we point to research that can lead to the integration of brain models with cognitive-developmental models.

The P-FIT model is a major step forward in understanding biological causes of intelligence. It is consistent with evidence on the influence of working memory and speediness upon intelligence, and with models that emphasize the role of interaction between modules to produce intelligence. The contribution to understanding genetic contributions is problematical, due to the difficulty of isolating the genes involved.

A large number of experimental results clearly indicate that sleep has an important role for human intelligence. Sleep-wake stages and their specific patterns of brain activation and neuromodulation subserve human memory, states of consciousness, and modes of information processing that strongly relate to intelligence. Therefore, human intelligence should be explained in a broader framework than is implicated by neuroimaging data alone.

General intelligence is largely based on two distinguishable mental abilities: crystallized intelligence (gC) and fluid reasoning ability (gF). The target article authors' P-FIT model emphasizes a network of regions throughout the brain as the neural basis for fluid reasoning and/or working memory. However, it provides little significant insight into the neural basis of gC, or how or why gC is more stable than gF across the life span.

A large body of evidence supports the idea that a common fronto-parietal network is activated across a range of diverse cognitive functions. Jung & Haier's (J&H's) review demonstrates a very similar pattern of activity, which correlates with individual differences in intelligence. We propose that these converging lines of evidence are best interpreted as a general role of the fronto-parietal network in integration and control.

The difficulty of making reliable interpretation from a dense cloud of unreliable correlations means that the grounds for making a testable “biological,” or brain-based, theory of intelligence remain very shaky. We briefly discuss the conceptual and methodological problems that arise and suggest one possible alternative interpretation of the data.

Lest the conjunction “intelligence and reasoning” seduce readers into supposing that the two are of a piece, we point out that analyses made at the superset level concerning intelligence do not readily align with or outperform the scientific advances made via investigations of reasoning, which at best can be viewed as a subset of intelligent behaviour.

Recent brain imaging points to differences in brain structure that relate to intelligence, but how do we model their causal relationship within a coherent framework that circumvents classic dualist traps? A bottom-level nonlinear, dynamic, multifactor, multiplicative, multidimensional, molecular (ND4M) trait-covariance time-space model may accomplish this better than traditional approaches.

A well-recognized framework for modeling human intelligence centers around Spearman's g, a common central factor accounting for individual differences in cognitive performance across a variety of complex tasks (Spearman 1904). The neural basis of g may be better characterized by posterior-frontal integration, rather than parietal, which may be just one of many posterior regions that are controlled by the prefrontal cortex (PFC).

Intelligence studies are confounded by an inability to image the mind, as well as by heterogeneity in intelligence constructs, gender, and age. The ghost (of future, not past) sitting at the table is a molecular one. Biochemistry and molecular biology factors can contribute to or take away from intelligence to a great and not yet fully explored extent.

Performance on abstract unfamiliar tasks used to measure intelligence has not been found to correlate with individual differences in highly skilled and expert performance. Given that cognitive and neural structures and regions mediating performance change as skill increases, the structures highlighted by parieto-frontal integration theory are unlikely to account for individual differences in skilled cognitive achievement in everyday life.

A parieto-frontal integration theory (P-FIT) model of human intelligence has been proposed based on a review of neuroimaging literature and lesion studies. The P-FIT model provides an important basis for future research. Future studies involving connectivity analyses and an integrative approach of imaging modalities using the P-FIT model should provide vastly increased understanding of the biological bases of intelligence.

Jung & Haier (J&H) have done an admirable job of solving the wrong problem. Their article does not show “where in the brain is intelligence,” because intelligence resides not in the brain but, rather, in the interaction of brain and environment. I describe four reasons why the article does not adequately localize intelligence in the brain.

High-order constructs such as intelligence result from the interaction of numerous processing systems, one of which is language. However, in determining the role of language in intelligence, attention must be paid to evidence from lesion studies and, in particular, evidence of dissociation of functions where high-order cognition can be demonstrated in face of profound aphasia.

“Where in the brain is intelligence?” is an intriguing question, and Jung & Haier (J&H) resist the temptation to provide a simple answer. Their concept tries to integrate and transcend previous findings, and, while omitting the contributions from complementary methods, results seem to converge. Whether similarities or differences in such networks are more important, though, is still open for discussion.

Overall mean performance on intelligence tests by brain-damaged patients with focal lesions can be misleading in regard to localization of intelligence. The widely used WAIS has many subtests that together recruit spatially distant neural “centers,” but individually the subtests reveal localized functions. Moreover, there are kinds of intelligence that defy the localizationist approach inferred from brain damage.