Wednesday, 5 November 2014

Intelligence probably declines by considerably more than a standard deviation from age 16 to 63


IJ Deary, G Der. Reaction Time, Age, and Cognitive Ability: Longitudinal Findings from Age 16 to 63 Years in Representative Population Samples. Aging, Neuropsychology, and Cognition  2005; 12: 187-215.


Deary and Der's paper referenced above seems to be the best available estimate of the effects of ageing on simple reaction times (sRTs).

Simple reaction time correlates with general intelligence (g) and there is, I believe, a lot of reason to believe that group-averaged sRTs are the best, most valid measure of long-term changes in intelligence.

At the individual level, the modest correlation of sRT with IQ makes the sRT a relatively poor predictor of cognitive; but this is dealt with be averages the sRT measure in a large group, and of course the sRT is a real, physiological measure on a ratio scale - while IQ is only a measure of relative performance in tests, and is merely an ordinal scale with has no fixed interval measure or zero.


It is well known that general intelligence declines from late teens/ early twenties and into old age - for example as measured by fluid intelligence. In other words, raw scores of fluid intelligence in IQ tests will decline.

But the real magnitude of this decline cannot be obtained from IQ testing - and only a ratio scale, physiological functional measure such as sRT can measure the real magnitude of decline.

Deary and Der 2005 make possible this measure. They measure Men and Women in three cohorts: aged 16 retested at 24; age 36 retested at 44; aged 56 retested at 63. So, there are six data points for men, and another six for women.


Simple visual Reaction Times in milliseconds (rounded to nearest integer)
- Mean (Standard Deviation)


16- 293 (72)
24- 294 (78)

36- 304 (75)
44- 316 (90)

56- 348 (109)
63- 373 (124)

Total decline 16-63 - 373 minus 293 = 80 milliseconds.

Using age 16 average as a baseline value with its standard deviation of 72 - this 80 ms decline represents an intelligence decline of slightly more than one standard deviation - i.e. slightly more than 15 IQ points.

So, an average man of average IQ would decline from 50th centile age 16 to somewhat below the 16th centile aged 63.



16- 295 (57)
24- 306 (73)

36- 314 (79)
44- 333 (95)

56- 346 (101)
63- 375 (126)

Total decline 16-63 - 375 minus 295 = 80 milliseconds.

So, using age 16 as a baseline value with its standard deviation of 57 - this 80ms decline represents an intelligence decline of significantly more than one standard deviation - i.e. significantly more than 15 IQ points.

So, an average woman of average IQ would decline from 50th centile to significantly below the 16th centile aged 63.


A decline of more than one standard deviation - or 15 IQ points, represents the minimum average decline in general intelligence from 16 to 63 - in women the real value is likely to be even larger, because the three cohorts of 16-24, 36-44 and 56-63 very probably had different starting levels for intelligence - with the oldest age group having had a starting (age 16) sRT of about 36ms faster than the measured value for the 16-24 group.  See:

This may be taken imply that among women the true decline of sRT from 16-63 is more like 113 ms instead of 80 ms - and 113 ms decline would be two standard deviations.

However, 2SDs is likely to be an overestimate, because the distribution of sRT is not really a normal distribution, but positively skewed such that there is a longer tail of higher values - so the standard deviation breaks down as a valid description after about one standard deviation.


Nonetheless the data presented in Dear and Der 2005 seems to be measuring a very significant degree of decline in real, underlying, physiological general intelligence/ fluid intelligence between ages 16 and 63; suggesting a significant decline in those cognitive aptitudes underpinned by g.

In practice, this decline in general intelligence may well be obscured by increased specific or 'crystallised' intelligence, due to accumulated specific knowledge, skills and expertise. But the decline in g would be apparent in reduced cognitive flexibility, e.g. slowing of the learning of new knowledge and skills, reduced capacity at solving novel problems and so on.

This data set also suggests that the effect of declining intelligence with age may also be obscured, in this group of women, by declining average intelligence over time, with older generations having a had a higher starting point of for intelligence.

But however the data is adjusted or corrected, the basic finding seems to be that average intelligence declines by more than one standard deviation from age 16 to 63.


Ref: see also