Brain networks for working memory and factors of intelligence assessed in males and females with fMRI and DTI
Introduction
Neuro-imaging studies of the underlying structural and functional anatomy of intelligence implicate areas throughout the brain, irrespective of the intelligence tests used. Jung and Haier (2007) characterized these findings as mostly, but not exclusively, in frontal and parietal areas. They proposed a model of how these areas may form overlapping networks underlying individual differences in intelligence: the Parieto-Frontal Integration Theory—P-FIT. The P-FIT areas represent stages of information processing from posterior sensory perception to abstraction in parietal areas to anterior hypothesis testing and decision-making. Integration of information among the areas is key. Based on functional imaging studies that found inverse correlations between regional brain activation and performance on intelligence tests (Haier et al., 1988, Neubauer and Fink, 2009) the P-FIT includes the hypothesis that efficient flow of information around these networks is related to intelligence. Similar networks have been identified for performance on fundamental cognitive tasks including aspects of attention and memory (Cabeza and Nyberg, 2000, Naghavi and Nyberg, 2007, Wager et al., 2004, Wager and Smith, 2003), although inverse correlations with such tasks are not reported, possibly because fundamental cognitive tasks used in imaging studies usually are chosen to minimize individual differences in performance.
Since intelligence tests tap more than one cognitive domain, it remains to be seen if the P-FIT or other models of brain networks represent a general factor of intelligence (g), common among all tests, or more specific group factors like memory or spatial ability. So far, only two structural imaging studies have extracted a g-factor score from a battery of tests and then correlated these and other more specific group factor scores (with g removed) to gray matter (Colom et al., 2009, Haier et al., 2009). Both studies showed similar results for a spatial factor, but not for a g-factor, suggesting that there may not be a single neural basis for g (Haier et al., 2009). Johnson et al. (2008) also, for example, extracted other cognitive factors with g removed in a small sample and showed some gray matter correlates different than those associated with g.
Functional imaging studies of networks related to intelligence have not yet used g-factor scores as dependent variables, instead relying on single tests like the Raven Progressive Matrices Test (Gray et al., 2003, Haier et al., 1988, Lee et al., 2006, Prabhakaran et al., 1997). The interpretation of functional imaging data, moreover, is constrained by the task performed during the imaging, unlike structural imaging (Toga & Thompson, 2005), so task selection is a key element of research design. Two functional imaging studies with fMRI (Gray et al., 2003, Waiter et al., 2009) have used the working memory n-back task because working memory is highly related to intelligence (Colom et al., 2008, Colom et al., 2005, Colom et al., 2004, Engle, 2002, Grabner et al., 2004, Kane et al., 2005, Oberauer et al., 2005). These two fMRI studies of intelligence using the n-back test report that activation in frontal and parietal areas is correlated to single intelligence test scores. Inverse correlations are not reported. All the subjects were males.
In addition to issues about the use of single test scores rather than factor scores, another unresolved issue concerns sex differences. A number of imaging studies show male/female differences related to intelligence (Haier et al., 2005, Luders et al., 2008, Schmithorst and Holland, 2007, Sowell et al., 2007) and other cognitive abilities (Haier and Benbow, 1995, Jausovec and Jausovec, 2008). Findings regarding brain efficiency also show strong sex differences (Neubauer, Fink, & Schrausser, 2002) that may be related to any number of brain differences between males and females (e.g. (Luders et al., 2004, Rabinowicz et al., 1999). Due to cost, most imaging studies focus on one sex (usually males) or partial out sex when male and female samples are combined to increase statistical power at the cost of obscuring any actual sex differences. In our view, separate analyses for males and females are required to explore any differences that may be unique to the study sample or to a more general finding.
Here we extend the two previous n-back studies to determine fMRI correlates of intelligence using factors derived from a battery of tests rather than a single test score; these factors are independent of the g-factor. Based on the structural studies of (Colom et al., 2009) and (Haier et al., 2009), we hypothesize that individual factors will have functional correlates different from the g-factor. Further, we present analyses separately for males and females to test whether the P-FIT areas differ in activation during the non-verbal n-back memory task and whether inverse correlations consistent with brain efficiency may be stronger in males, as suggested by Neubauer and Fink (2009). The P-FIT also noted the potential importance of individual differences in white matter connectivity, especially the arcuate fasciculus; and there is some suggestion that white matter may be more important for intelligence in women than in men (Haier et al., 2005). Therefore, we added a second imaging method, Diffusion Tensor Imaging (DTI), to characterize white matter tracts among any areas identified with functional imaging as related to n-back performance. DTI provides information on the integrity of the axonal connections in the brain (Basser, 1997). By combining DTI with fMRI it is possible to provide a more comprehensive picture of structural and functional integration during cognitive performance (Fjell et al., 2008). For example Schmithorst and Holland (2007) found differences in activated brain areas between boys and girls during a verbal task as well as differences in white matter pathways (Schmithorst, Holland, & Dardzinski, 2008). Older girls showed greater inter-hemispheric connectivity. Yu et al (2008) computed correlations between the integrity of several tracts (corpus callosum, cingulum, uncinate fasciculus, optic radiation, and corticospinal tract) and intelligence. The 79 participants (men and women; mean age 23.8) were divided in two groups: average and high intelligence. White matter integrity was assessed by fractional anisotropy (FA). The results, controlling for age and sex, showed that high intelligence participants display more white matter integrity than average intelligence participants only in the right uncinate fasciculus. The authors concluded that the right uncinate fasciculus is an important neural basis of intelligence differences. There were no separate analyses by sex, but a sample of 15 participants with mental retardation was also studied. These participants were compared with the 79 healthy controls and they showed extensive damage in the integrity of the brain white matter tracts: corpus callosum, uncinate fasciculus, optic radiation, and corticospinal tract.
A recent paper (Chiang et al., 2009) reported the first study combining a genetically informative design and a DTI approach for analyzing the relationships between the white matter integrity and human intelligence. Intelligence was assessed by the Multidimensional Aptitude Battery, which provides measures of general intelligence, verbal (information, vocabulary, and arithmetic), and non-verbal intelligence (spatial and object assembly). The sample included 23 pairs of identical twins and 23 pairs of fraternal twins (males and females but all pairs were same sex; mean age 25 years). White matter integrity, quantified using fractional anisotropy (FA), was used to fit structural equation models (SEM) at each point in the brain. They then generated three-dimensional maps of heritability. White matter integrity was found to be under strong genetic control in bilateral frontal, bilateral parietal, and left occipital lobes. FA measures were correlated with the estimate of general intelligence and with non-verbal intelligence in the cingulum, optic radiations, superior fronto-occipital fasciculus, internal capsule, callosal isthmus, and the corona radiate. Further, common genetic factors mediated the correlation between intelligence and white matter integrity. This latter finding suggested a common physiological mechanism and common genetic determination.
DTI studies of intelligence are relatively new and there are not yet data in adult samples, so our analyses are exploratory. Since integrity of white matter could relate to efficient flow of information, we generally expect positive correlations with intelligence factors.
Section snippets
Participants
The sample was the same as reported in a previous study focused on structural MRI assessments of gray matter only (Haier et al., 2009). During 2002–2003, 6889 individuals sought consultation from the Johnson O'Connor Research Foundation (JOCRF), a non-profit organization dedicated to using psychometric assessments for vocational guidance. Each completed the same battery of eight cognitive tests listed below in one of 11 testing centers in major cities throughout the United States. The mean age
fMRI
Prior to analyses, the anatomical scans were read by staff radiologists to screen for any incidental clinical findings; none were found. BOLD data were processed using SPM5 (Wellcome Department of Cognitive Neurology, Institute of Neurology, University College London, London, UK). Functional data was slice-time corrected by interpolation to the middle slice prior to motion correction. Functional images were co-registered to each subject's anatomical scan. Co-registered anatomical images were
fMRI
As shown in Fig. 1, clusters of activation during the n-back task were identified (1,2,3 back vs. 0 back; p < .001, uncorrected) in: Anterior Cingulate (ACC), bilateral Prefrontal Cortex (PFC), bilateral Parietal cortex (PC), bilateral Insular Cortex (IC) and bilateral visual cortex (VC). As noted, we used these areas to correlate percent activation with intelligence factor scores; two regions, although activated during the n-back task, did not produce any significant correlations: namely,
Discussion
This study examined intelligence factors with g removed to determine if the parieto-frontal integration theory of intelligence (P-FIT) characterizes specific cognitive abilities beyond the pervasive influence of the g-factor representing general intelligence. This was examined separately in males and females and uniquely combined fMRI and DTI imaging to study the neuroanatomy of intelligence. Since the sample sizes are relatively small and none of the significant findings survived correction
Acknowledgements
Funding for imaging and for R. Haier was provided by the Johnson O'Connor Research Support Corporation. R. Colom was funded by the grant SEJ-2006-07890 from the “Ministerio de Educación y Cultura” (MEC) [Ministry of Education and Culture, Spain].
References (63)
- et al.
Working memory and intelligence are highly related constructs, but why?
Intelligence
(2008) - et al.
Memory span and general intelligence: A latent variable approach
Intelligence
(2005) - et al.
Gray matter correlates of fluid, crystallized, and spatial intelligence: Testing the P-FIT model
Intelligence
(2009) - et al.
Working memory is (almost) perfectly predicted by g
Intelligence
(2004) - et al.
Characterization of sexual dimorphism in the human corpus callosum
Neuroimage
(2003) - et al.
The relationship between diffusion tensor imaging and volumetry as measures of white matter properties
Neuroimage
(2008) - et al.
Intelligence and working memory systems: Evidence of neural efficiency in alpha band ERD
Brain Research. Cognitive Brain Research
(2004) Neuro-intelligence, neuro-metrics and the next phase of brain imaging studies
Intelligence
(2009)- et al.
Gray matter and intelligence factors: Is there a neuro-g?
Intelligence
(2009) - et al.
The neuroanatomy of general intelligence: Sex matters
Neuroimage
(2005)