Thread by @Scientific_Bird

Are people with bigger brains smarter (on average)?

Here I will summarize some of the biggest studies which test whether there is a correlation between intelligence and brain size (within modern humans).

Thread.

2) In modern intelligence tests, you give individuals a battery of cognitive tests, testing a wide range of cognitive abilities at once. Then a general factor of intelligence can be estimated, in short "g".

Brain size can be measured with modern brain-imaging technology.

3) To address this question, McDaniel (2005) gathered 37 samples with a total of 1530 people who had been intelligence tested and brain-scanned.

He found that, on average, the correlation between brain size and general intelligence was 0.33.

doi.org/10.1016/j.inte…

4) In a later and larger meta-analysis of 148 samples and more than 8000 individuals, Pietschnig and colleagues (2015) estimate the correlation to be slightly smaller, but still highly significant, r = 0.24.

doi.org/10.1016/j.neub…

5) Quality of tests vary (e.g. Brief tests vs longer and more comprehensive tests).

Gignac & Bates (2017) reanalyzed the studies from the previous meta-analysis and found that the correlations were higher when higher-quality intelligence tests were used.

doi.org/10.1016/j.inte…

6) In fact, this is just a statistical phenomenon. When there is a correlation between two variables, then the better measured the two variables are, the higher the correlation will be.

Poor measurements contain lots of noise which weakens the relationship between the variables.

7) Instead of meta-analyses that pool results from many samples, we can also look at individual high-quality studies.

Ritchie and colleagues (2015) look at many brain variables in a sample of N = 672, but total brain volume and "g" correlated r = 0.31.

https://t.co/y0ENghSE73doi.org/10.1016/j.inte…

8) Dubois and colleagues, in a large sample (N = 884), test if resting-state brain activity correlates with intelligence. It does, but they also report other brain correlates. They find that brain volume correlates r = 0.33 with general intelligence.

https://t.co/sFto5YtXa9royalsocietypublishing.org/doi/10.1098/rs…

9) Recently a study with a massive sample (N = 29,004 individuals) was published.

The correlation between brain volume and general intelligence found was 0.276.

They only used 4 subtests to estimate "g", hence this might even be a slight underestimate.

https://t.co/lmJD6GvP7Udoi.org/10.1016/j.inte…

10) To minimize possible effects of different environments people have lived in, you can look within families. Lee and colleagues find that, of siblings raised in the same family, the sibling with the larger brain tends to be more intelligent.

doi.org/10.1016/j.inte…

11) Another source of evidence comes from "genome-wide association studies." Here you take DNA samples from many individuals and see which gene variants tend to be correlated with some chosen variable of interest.

You can then calculate the genetic overlap two traits have.

12) Genome-wide association studies have now been done for many traits, including intelligence and head circumference (a noisy proxy for brain size).

One such study finds a genetic correlation of 0.31 between intelligence and head circumference.

https://t.co/1OcIjompkHnature.com/articles/s4138…

13) To summarize:

In studies with high-quality measurements, the correlation between brain size and intelligence is roughly r ≈ 0.3.

This is a real, but not a strong relationship. If you want to estimate intelligence, test it directly rather than measuring just brain size.

14) But brain size is just a single variable. Other brain variables also correlate with intelligence. In principle, prediction of intelligence from brain imaging could likely become much stronger if we took multiple brain variables into account at once.

pnas.org/content/early/…

15) Brain size is probably the simplest brain variable to measure, and likely that's why it has been included in many studies. What other brain variables might contribute to differences in intelligence? That's for other research to establish.

Thanks for reading.

@rexjung And yes, it must be frustrating working with expensive research with highly noisy measurements. So I can understand why people would prefer working with structural imaging.

@DumoulinBaptis2 @JamesPsychol For example, take the study in tweet 9). This is likely the most representative. They find estimate of 0.276. If we divide by 0.8 for correction for attentuation (which seems reasonable, perhaps even too generous), we get 0.276/0.8 = 0.345.

So I don't think it's higher than 0.35

@DumoulinBaptis2 @JamesPsychol Find me an individual study (not a meta-analysis) that has a large sample (say, N > 500) and a good intelligence measure that finds correlation above 0.4.

These studies should be plenty if you think the *average* correlation is around 0.4.

@DumoulinBaptis2 @JamesPsychol And this challenge is still open. I'm genuinely interested. Because I've read this literature fairly closely, and from my reading the value is closer to 0.3 than to 0.4.

https://twitter.com/Scientific_Bird/status/1155410345663619072

@joanne_mourched You're right that this is different from establishing causality, as the relationship could be confounded.

A study that attempts to address this as well as they can is the following:

sciencedirect.com/science/articl…

New genome-wide association study of brain volume. The genetic correlation between brain volume and intelligence they find is 0.24, quite close to other estimates given in the above thread.

Source:
https://t.co/g1KQs0vatHnature.com/articles/s4146…