The Flynn Effect

What it is

The observation is simple and hard to argue with. Standardised intelligence tests are periodically renormed so that the average score stays fixed, and Flynn's work from 1984 onwards established that this housekeeping was concealing a trend. Score a later cohort against an earlier standardisation and it comes out above average, at a rate of roughly three points per decade. The pattern turns up in nearly every country with long enough test records to check. For a claim about populations it is about as robust as psychology gets, and it did not come out of one laboratory or one dataset.

Its rhetorical force is what makes it worth a page. Every generation produces a confident argument that some new medium is degrading the minds using it, and the Flynn effect is the fact that argument has to get past. A population whose measured performance rose steadily through the century of radio, cinema and television is a difficult population to describe as being made stupid. Most participants in the screen-time argument step around this. Carr does not, which is the most intellectually honest passage in The Shallows: he concedes the rise is real, accepts the figures, and says the gains should hearten us.

His resolution is borrowed from Flynn himself, which is the correct place to borrow it from. Flynn's account, set out in What Is Intelligence? (2007), is that the twentieth century did not make people cleverer in some general sense. It trained them in the classificatory, scientific mode of thought that IQ tests were built to measure. People "began to wear the same scientific spectacles" as the test designers. Asked how two things are related, a respondent can answer with the practical connection between them or with the abstract category that contains them both. It is the second kind of answer that an intelligence test is built to reward, and a century of schooling, urban work and abstract administration taught more and more people to reach for it first. On Flynn's reading the rise records a change in the kind of thinking people do, not in the amount of it they are capable of.

That reading is what lets Carr keep his argument without denying the data. The gains, he notes, are concentrated in abstract, nonverbal, pattern-and-classification tasks and in visual reasoning, while vocabulary, general knowledge, arithmetic and reading scores are flat or falling. His conclusion is that we are smart in different ways rather than smarter, and that the specific capacity he cares about, sustained contemplative deep reading, is not one of the things the rise measures. Whether that conclusion is right is a separate question. What matters methodologically is that he reached it by engaging with the counter-evidence rather than by ignoring it.

There are other explanations for the same rise, and they are not all compatible with Carr's use of it. Patricia Greenfield, writing in Science, attributes the gains to urbanisation and rising societal complexity, which locates the cause in the environment's demands rather than in a change to the style of thinking that tests happen to reward. The effect is settled. What produced it is not.

The other thing to know is that it has stopped, and in places reversed. Since the 1990s the rise has slowed or turned negative in several countries, with Norwegian and Danish data flattening, and a UK study finding teen scores fell two points between 1980 and 2008. That reversal is used enthusiastically by both camps, which should be a warning about how much weight it can carry on its own.

In effect

The finding, and why it counts as strong evidence

The Flynn effect is worth holding up next to the studies that dominate this subject, because it is a different kind of thing. It is not a single striking laboratory demonstration with a surprising result and a press release. It is a trend visible across decades and many countries, in data collected by people who were not testing this hypothesis. The renorming of tests is administrative housekeeping rather than research, which is part of why the pattern is credible: nobody was incentivised to find it. Compare the fate of the Google effect, a single vivid experiment that failed twice under replication. Effects assembled from boring, incidental, large-scale records tend to survive. Effects built from one memorable manipulation often do not.

The argument you have already had

The everyday version of this concept is a conversation, not an experience. Someone says that people cannot concentrate any more, that attention spans have collapsed, that a previous generation could sit with a book and this one cannot. It may well be true about reading specifically. What the Flynn effect supplies is the reason it cannot be stated as a general claim about intelligence: measured performance went up, for a century, everywhere anyone was counting. The honest version has to specify what capacity is supposed to be falling and accept that others were rising at the same time, which is a less satisfying sentence and a more defensible one.

How to use it if you write about technology

For anyone producing arguments about media and minds, the Flynn effect functions as a gate. If your thesis implies that a population is becoming cognitively worse, it has to survive contact with a century of rising scores, and the way to survive it is Carr's way: concede the data, then narrow the claim to a specific capacity the data does not cover. Any argument that reaches the conclusion without going through the gate should be discounted. It is also a fair test of good faith, and Carr's treatment of Flynn is the strongest evidence in The Shallows that its author is not a polemicist.

What it does not say

It does not say people are getting cleverer in any general sense, and Flynn is the one insisting on that most firmly. Both his own account and Greenfield's locate the rise in something other than raw capacity: a learned mode of thinking, or a more demanding environment. A rise in scores on an instrument is a fact about performance on that instrument. Reading it as a rise in intelligence itself imports an assumption about what the test measures that the effect's own discoverer declined to make.

It does not settle the screen-time argument in either direction, and both sides overreach with it. Used against Carr, it is supposed to show that media cannot be harming cognition, which does not follow: the rise predates the internet by most of a century, and a trend that ran through radio and television is not evidence about what happened after 2007. Used by Carr, the flat vocabulary and general-knowledge scores are supposed to show that reading-related capacities are declining while abstract ones rise. Flat is not falling, and a subtest pattern within a rising aggregate is thin material for a civilisational thesis. The effect is real evidence and it is not evidence for very much in this particular dispute.

It does not mean the reversal proves the alarmists right. Slowing or negative trends since the 1990s in Norway, Denmark and the UK are real, but nobody has established a cause, and a reversal that began in the 1990s is awkward evidence for anything about smartphones. A finding used to support opposite conclusions by both camps is usually a finding neither camp has earned.

And it does not tell you anything about individuals. This is a population-level trend in aggregate scores across decades, carrying no implication about any particular person, child or household, and the distance between a cohort trend and a person is the most common place this concept is misused.

---