You will be well aware that ever since brain imaging became cheaper neuro-bollocks has become more prevalent, since sticking a few persons in a scanner reliably leads to a publication. The best approach would be to get scanner-researcher-publishers into a dark basement and not let them out till they had agreed upon a) standard ways of conducting a scan b) standard ways of analysing a scan c) a few standard cognitive tasks to be done in the scanner d) a few cognitive assessments to be done outside the scanner and d) increasing the sample sizes and representativeness. Until that happy day, the best that can happen is that scholarly souls wade through the heterogenous hodgepodge to elucidate some general features.
Ulrike Basten, Kirsten Hilger, Christian J. Fiebach. Where smart brains are different: A quantitative meta-analysis of functional and structural brain imaging studies on intelligence. Intelligence 51 (2015) 10–27
So, let us begin with the caveats: The original studies available for the meta-analyses show heterogeneity with respect to (a) the assessment of intelligence, (b) the cognitive challenges used during the measurement of brain activation, and (c) the consideration of potential moderator variables like sex or age. [We included only] studies that used measures from established tests of intelligence.
The tasks target different cognitive functions such as working memory, reasoning, mental rotation, and set shifting. We argue that despite this heterogeneity in task paradigms, summarising the corresponding studies in meta-analyses is well justified by the fact that many of these diverse cognitive demands are known to trigger very similar patterns of activation in the brain (Duncan & Owen, 2000). At the very least, this approach provides us with a conservative estimate of where in the brain intelligence makes a difference with respect to the strength of activation that is required for successful cognitive performance.
Let me add a bigger caveat: all these analyses show areas of activation which are then averaged. As Rich Haier has pointed out, individual brains show patterns of activation in a dynamic flux, the enchanted loom in action. Finding a way to categorise this choreography of thought is another way into the system, and might give more clues as to what is happening, about which we have no useful functional theories.
Here are the studies used for the structural analyses:
I know most of these researchers, and they are at the very serious end of the business, paying great attention to the reliability of measures, and with above average sample sizes, so they give the best chance of finding signals among the noise.
So, what meta-conclusions can be drawn from all these publications?
We found substantial convergence across studies as well as overlap with theoretical models of a fronto-parietal basis for intelligence. Our meta-analyses of structural grey matter correlates of intelligence identified widespread clusters of convergence across the brain. Notably, there was no overlap between brain regions identified as relevant for intelligence in the functional and structural meta-analyses, respectively. We propose an updated neurocognitive model for the brain bases of intelligence that includes insular cortex, posterior cingulate cortex and subcortical structures in addition to the previously considered frontal, parietal, temporal, and occipital brain lobes, and that explicitly distinguishes between structural and functional brain correlates of intelligence.
In sum, the P-FIT (2007) model is supported, but has needed to be extended. To my mild irritation, the results do not clearly support the finding that brighter brains show less activation because they have higher neural efficiency. Tired of hearing how people were being trained “to use more of their brain” I kept pointing out that the cool thing was to be able to solve problems by using less of your brain. The picture which emerged from these studies was mixed. On a brighter note, they make a good analytic point about the measurement of efficiency:
smart brains do not generally show weaker activation. Importantly, an interpretation of individual differences in brain activation in terms of differences in neural efficiency must take into account the associated behavioural performance. Only if behavioural performance (= effect) is equal across subjects, can efficiency be inferred directly from brain activation (= neural effort).
The authors say that 4 improvements are required:
(i) Continue the trend of studying larger samples. This is particularly important for individual differences analyses of brain imaging data to result in reliable results at the level of the original studies.
(ii) Test for an association between intelligence and brain characteristics across the whole brain — as opposed to using a priori defined regions of interest and thus excluding parts of the brain from the analyses.
(iii) Choose intelligence tests that are comparable across studies and that cover a broad range of cognitive demands that allow deriving measures of ability corresponding to different factors at the different levels of the hierarchy in intelligence structure.
(iv) Systematically test for moderator effects of sex and age on the association between intelligence and brain characteristics.
Here is an offer: I am willing to provide a cellar in central London, disguised as a small lecture theatre, with suitable supplies of refreshments, to let the international scanner gangs assemble and sort out their differences, merge their different territories, agree their enforcement techniques, and then distribute the nectar of neuro-intelligence to a public thirsting for knowledge. Get your people to talk to my people.