Wednesday, 1 May 2013

Flynn effect raises all boats, but some are leaky

Woodley, te Ninjehuis and Murphy (2013) have put together an interesting paper. In looking for the Flynn effect they have eschewed intelligence test data, and instead gone for simple, old-fashioned (1884) simple reaction times. This is about the most basic thing you can measure in the nervous system: a signal is presented and the organism responds. It is a crude measure of whether the organism can sense and respond to its environment, and how long it takes to do so.

Much has been made of more complex measures. For example, the complexity of the signal has an impact on response times. It is easier to respond to one light on its own than to watch out for one of two lights, even if they are relatively close to each other, because having to monitor two potential sources of signals imposes a slight load on the nervous system. If you make the task even more complex by requiring responses to patterns of five lights, then reaction times slow even further, following Hick’s Law (that the slope of reaction times will follow a binary choice log2 function of the number of choices).  

Other sophisticated variations are to distinguish between movement time (taken up by moving your arm) and decision time (time spent looking at the shining light and working out that it is now time to respond).

Simple reaction times are back in fashion. Among other things, they are good at predicting lifespans. See “Can I have a reaction” in order to test your own reaction times, and guess at your lifespan, though we need to go to another data set from Ian Deary for that.

By way of background, both intelligence and reaction speed are heritable at 0.58 and have a correlation with each other of about 0.31. If one corrects for restriction in age range and for lack of reliability the measure probably account for half of the variance in intelligence.  This paper suggests that something seems to be slowing us up. If the Flynn effect is absolutely real, then that should not be the case. We should be super-fast at simple reaction times.

Woodley, te Nijenhuis and Murphy  have conducted a meta-analysis of 14 studies involving Western subjects from 1884 to 2004, which provides them with 16 data points with which to draw the linear regression function showing increased reaction times over 120 years. Interpreting the cause is a different matter. It is possible, but unlikely, that we have been poisoned by some neuro-toxin which is making us sluggish: lead, perhaps. However, lifespans are going up in most of the world, and certainly steadily up in the rich West. Also, lead does not have its main effects on measures of general intelligence. Another possibility is that many people who in earlier eras would have died have now survived their illnesses because of modern medicine, but survive with sluggish reactions. The last option considered by the authors is increasing mutation load, perhaps reducing myelination of neurones, with a consequent reduction in the efficiency of signal transmission and information processing in the brain.

It is possible that at least two contradictory factors are at play here: a welcome general increase in living standards, with much better access to food, housing, health and education boosting intelligence on the one hand; and on the other hand a smaller but persistent increase in mutational load with deleterious intellectual effects, thus cancelling some of the gains. (The rising tide of the Flynn effect raises all boats, but many of those boats are leaky).  If the latter is true, then it ought to be possible to find further evidence to support the hypothesis.  Disentangling the Flynn effect from the Woodley effect will be difficult, but also very illuminating, because we will learn much about the future of human intelligence.

Please cite this article as: Woodley,M.A., et al.,Were the Victorians cleverer than us? The decline in general intelligence estimated from a meta-analysis of the slowing of simple reaction time, Intelligence (2013),



  1. So, escaping the Malthusian Trap appears to be inherently dysgenic - intelligence becomes maladaptive, inasmuch as the less intelligent have more surviving children, and population natural inherent intelligence decreases over time even if nurtured apparent intelligence rises. Interesting that this effect clearly dates from at least the late 19th century, well before the modern Welfare State. I guess those 19th century eugenicists were onto something.
    I would suspect that something similar happened in the later Roman Empire, and that the former Western Roman Empire eventually returned to eugenic fertility in the Dark Ages, although there was also significant population replacement by Germanic and later Arab invaders.
    Long term, I would suspect that most likely our inherent population intelligence will continue to decline through dysgenics and population replacement until there is a return to the Malthusian Trap. If selection for intelligence then becomes positive, the cycle will repeat. There will not be an 'Idiocracy' unless humanity is maintained in its unintelligence by automated self-repairing systems.

  2. Very interesting points. We cannot yet be sure how the balance of forces is working out: benign environments increase ability, but probably eventually plateau (though that is not a universal finding) whilst other forces may be reducing our fundamental ability to deal with complex new problems. Optimistically, it may be that with a very small Smart Fraction we can keep the whole show on the road, so long as there are others who are willing to implement their insights and suggestions. Possibly.

  3. Sigh.
    Just, no.
    Look at the actual study.
    See it?
    Look closer.
    Look at the data points.
    See it?
    (or better still:

    People aren't growing stupider, bloggers are just getting worse at actually checking out the material they reference.

  4. There has been further work on the data points, and on the assumptions and technology behind the original results. The Journal Intelligence has a new paper" Is there any evidence of historical slowing of reaction time? No, unless we compare apples and oranges" by Dodonova and Dodonova. There will a reply by Woodley et al in due course. Deducing change from few data points is difficult but not impossible, and certainly interesting when one has rare data.