Based on an exhaustive review of neuroimaging studies, including volumetric MRI (especially voxel-based morphometry [VBM]), diffusion tensor MRI (DTI), magnetic resonance spectroscopy (MRS), functional MRI (fMRI), and functional PET (fPET), as well as lesion studies, Jung & Haier (J&H) propose a parieto-frontal integration theory (P-FIT) model which underpins individual differences in intelligence in humans. The P-FIT model is consonant with a unitary theory of intelligence, involving general intelligence (g), given that “one of the main insights of cognitive neuroscience is that the ‘functional units’ of higher cognition are networks of brain areas, not single areas” (Gray & Thompson Reference Gray and Thompson2004). J&H's important review can serve as a starting point for future research into the correlation of brain structure with intelligence. Neuroimaging technology and analysis procedures have now progressed to the point where a detailed analysis of the correlation of networks of brain activation with cognitive function is feasible. Such an analysis may, as the authors acknowledge, necessitate further, possibly significant, modifications of the P-FIT. Nevertheless, the P-FIT model provides an important basis for future investigations.

The technology and analyses available for neuroimaging research on intelligence have progressed substantially from simple correlations of whole-brain or whole-head volumes. Only a few of the functional studies cited by the authors have investigated the role of functional or effective connectivity, however, which would seem to be critical for evaluating the role of specific functional networks in intelligence. The P-FIT model would seem particularly amenable to the use of techniques such as linear structural equation modeling (SEM) (Buchel & Friston Reference Buchel and Friston1997; Karunanayaka et al. Reference Karunanayaka, Holland, Schmithorst, Solodkin, Chen, Szaflarski and Plante2007; McIntosh et al. Reference McIntosh, Grady, Ungerleider, Haxby, Rapoport and Horwitz1994), or dynamic causal modeling (DCM) (Friston et al. Reference Friston, Harrison and Penny2003; Penny et al. Reference Penny, Stephan, Mechelli and Friston2004), which allow the investigation of effective, as opposed to functional, connectivity. Whereas functional connectivity refers only to a relationship, or association, between brain regions, effective connectivity assesses the degree of influence of one brain region over another (Patel et al. Reference Patel, Bowman and Rilling2006). Other techniques for connectivity analysis have also been developed, including a Bayesian analysis of functional connectivity (Patel et al. Reference Patel, Bowman and Rilling2006; Schmithorst & Holland Reference Schmithorst and Holland2007) and multivariate autoregressive modeling (MAM; Harrison et al. Reference Harrison, Penny and Friston2003). Together, these techniques will allow a detailed analysis of how cognitive function related to intelligence relies on an integrated network of several regions of the brain.

The research will further benefit from the incorporation of magnetoencephalography (MEG) or electroencephalography (EEG) studies carried out in conjunction with fMRI, as the much greater temporal resolution afforded by MEG and EEG will allow finer distinction of how the activation of functional networks differs in more-intelligent individuals. Each of the imaging studies cited by the authors employed only a single methodology, whether fMRI, VBM, or DTI. An integrated approach combining these methodologies, possibly also in conjunction with MRS, into a single study would constitute a powerful tool to further unravel present questions about the biological bases of intelligence. An example is the combination of fMRI and DTI in analysis of the properties of the auditory cortex (Upadhyay et al. Reference Upadhyay, Ducros, Knaus, Lindgren, Silver, Tager-Flusberg and Kim2006). For instance, is greater neuronal density in men (Haier et al. Reference Haier, Jung, Yeo, Head and Alkire2005) correlating with intelligence, also associated with greater “neural efficiency” (Neubauer et al. Reference Neubauer, Fink and Schrausser2002) and with greater localization of cognitive function (Schmithorst & Holland Reference Schmithorst and Holland2006)? And is this correlation also possibly associated with differences in white matter anisotropy (Schmithorst et al. Reference Schmithorst, Wilke, Dardzinski and Holland2005) and metabolite concentrations, as measured by MRS (Jung et al. Reference Jung, Haier, Yeo, Rowland, Petropoulos, Levine, Sibbitt and Brooks2005)? A combined methodological approach, using the P-FIT model as an a priori hypothesis, would be a powerful tool to answer these types of questions.

These advanced imaging approaches provide the tools to investigate the P-FIT model in a great deal of detail. There obviously remain several unresolved questions to be answered by future studies, which will likely necessitate significant modification of the P-FIT model. Most of the papers reviewed used adult subjects, hence, it is yet to be determined precisely how the P-FIT develops through childhood and adolescence. The authors propose the ASPM and microcephalin genes as candidates for a genetic basis behind general intelligence; however, the relationship among these genes, evolution, and cognition has recently been challenged (Balter Reference Balter2006; Currat et al. Reference Currat, Excoffier, Maddison, Otto, Ray, Whitlock and Yeaman2006) and a significant gene–environment interaction is also likely, further complicating the issue of genetic mechanisms and hereditability. Sex differences may also interact with the P-FIT. Whereas J&H focus on fronto-parietal (presumably intrahemispheric) connections, a recent fMRI study (Schmithorst & Holland Reference Schmithorst and Holland2007) has shown an increasing dependence on interhemispheric functional connectivity with age in girls, and a greater dependence on functional connectivity with Broca's area in boys. Moreover, in boys, a negative correlation of parieto-frontal connectivity with intelligence has been seen to develop with age in the left hemisphere (Schmithorst & Holland Reference Schmithorst and Holland2006), although in that study the superior medial frontal gyrus as well as the left prefrontal cortex were included in the network. It is therefore likely that a characterization of the relevant network merely as “parieto-frontal” lacks sufficient specificity, and a detailed investigation of the contribution of each area in the parietal and frontal lobes will be necessary. Functional segregation has been seen, for instance, in regions including Broca's area (Cannestra et al. Reference Cannestra, Bookheimer, Pouratian, O'Farrell, Sicotte, Martin, Becker, Rubino and Toga2000), and the P-FIT model will likely need to encompass these differences. How the P-FIT model is consistent with “neural efficiency,” where more-intelligent individuals use a more limited set of brain circuits and neurons and fewer neural resources to perform at a given cognitive level (Gray & Thompson Reference Gray and Thompson2004), is another issue to resolve in the future, especially since neural efficiency is modulated by task and sex (Neubauer et al. Reference Neubauer, Fink and Schrausser2002).

In conclusion, the authors provide a convincing motivation for the use of neuroimaging technology in studies of intelligence and a useful model (P-FIT) for further investigation. With current advances in the field of neuroimaging technology and analyses, researchers should now be poised to make substantial advances in our understanding of the biological bases of intelligence.