In the race for eye-catching headlines, this mainstream media coverage arguably missed the more fascinating story of the hunt for cause and effect. Here lead author Dr Stuart Ritchie explains the science:
"Causality, it turns out, is really difficult to prove. Correlational studies, while interesting, don’t give us information about causation one way or another. The randomised controlled trial is the ‘gold standard’ method of telling whether a manipulation has an effect on an outcome. But what if a randomised experiment isn’t possible, for practical or ethical reasons? Thankfully, there is an entire toolkit of study designs that go beyond correlation, and can be used to take steps up the ladder closer to causation.
Say you wanted to find interventions that cause intelligence to increase. Since childhood intelligence test scores are so powerfully predictive of later educational success, as well as health and wealth, it’s of great importance to find out how they might be improved. All sorts of nutritional supplements and training programmes have been tried, but all have failed (so far) to reliably show benefits for IQ. However, one factor that has been convincingly shown to cause improvements in intelligence test scores is education. It wouldn’t exactly be ethical to remove some children from school at random and see how they do in comparison to their educated peers. But in a step up the aforementioned causal ladder, researchers in 2012 used a ‘natural experiment’ in the Norwegian education system (where compulsory years of education were increased in some areas but not others) to show that each year’s worth of extra education added 3.6 IQ points.
What is it about education that’s driving these effects? Could it be that a very basic process like learning to read is causing the improvements in IQ? Keith Stanovich and colleagues showed, in a number of studies in the 1990s, that earlier levels of reading interest (though not ability) were predictive of later levels of verbal intelligence, even after controlling for children’s initial verbal intelligence. In a 1998 review, they concluded that “reading will make [children] smarter”.
On the ladder of causation, a control for pre-existing ability in a non-experimental design is important, but problems remain. For instance, since we know that common genes contribute to reading and intelligence, any study that fails to measure or control for genetic influences can’t rule out that the possibility that the early reading advantage and the later intelligence benefit are due simply to a shared genetic basis that is, say, expressed at different times in different areas of the brain. If only there were a way of cloning children – comparing one “baseline” version of each child against a second version with improved reading ability, and then seeing if the better reading translated to higher intelligence later in development...
This sounds like a far-fetched fantasy experiment. But in a recent study, my colleagues and I did just that, though we left it to nature to do the cloning. Tim Bates, Robert Plomin, and I analysed data from 1,890 pairs of identical twins who were part of the Twins Early Development Study (TEDS). The twins had their reading ability and intelligence tested on multiple measures (averaged into a composite) at ages 7, 9, 10, 12, and 16. For each twin pair at each age, we calculated the difference between one twin and the other on both variables. Since each pair was near-100 per cent identical genetically, and was brought up in the same family, these differences must have been caused purely by the ‘non-shared environment’ (that is, environmental influences experienced by one twin but not the other).
We found that twins who had an advantage over their co-twin on reading at earlier points in their development had higher intelligence test scores later on. Because this analysis controls for initial IQ differences, as well as genetics and socioeconomic circumstances, it is considerably more compelling than previous results that used less well-controlled designs. It’s important to note that we found associations between earlier reading ability and later nonverbal intelligence, as well as later verbal intelligence. So, beyond the not-particularly-surprising finding that being better at reading might help with a child’s vocabulary, we made the pretty-surprising finding that it might also help with a child’s problem solving and reasoning ability. Why?
We now enter the realm of speculation. It might be that reading allows children to practise the skills of assimilating information and abstract thought that are useful when completing IQ tests. The process of training in reading may also help teach children to concentrate on tasks—like IQ tests—that they’re asked to complete. Our research doesn’t shed light on these mechanisms, but we hope future studies will.
One should not give our study a criticism-free ride just because it tells a cheery, ‘good news’ story. A step up toward causation is not causation. Could there have been alternative explanations for our findings? Certainly. It is possible that, for instance, teachers spot a child with a reading advantage and give them additional attention, raising their intelligence 'without', as we say in the paper, 'reading doing the causal "work"'. It may also have been that our controls were inadequate – as I said above, identical twins are nearly genetically identical, but a small number of unique genetic mutations might occur within each pair. The largest lacuna in our study, though, was the cause of the initial within-pair reading differences. Whether these were caused by teaching, peers, pure luck, or some other process, we couldn’t tell, and it’s of great interest to find out.
We hope that our study encourages researchers in three ways. First, in the eternal quest for intelligence-boosters, instead of looking to flashy new brain-training games or the like, they might wish to examine, and maximise, the potentially IQ-improving effects of ‘everyday’ education. Second, they could attempt to answer the questions raised by our study. Why do identical twins differ in reading, and are the reasons under a teacher’s control? What are the specific mechanisms that might lead from literacy to intelligence? Third, and more generally, we hope it will inspire them to consider new methods, including the twin-differences design, that edge further up the causal ladder, away from the basic correlational study. The data are, of course, far harder to collect, but the stronger inferences found there are well worth the climb."
Ritchie, S., Bates, T., & Plomin, R. (2014). Does Learning to Read Improve Intelligence? A Longitudinal Multivariate Analysis in Identical Twins From Age 7 to 16 Child Development DOI: 10.1111/cdev.12272
Post written by Stuart J. Ritchie, a Research Fellow in the Centre for Cognitive Ageing and Cognitive Epidemiology at the University of Edinburgh. Follow him on Twitter: @StuartJRitchie