Wednesday, 15 May 2013

More than one third decline in general intelligence since Victorian times?


Having reflected on the title of an earlier post which described the decline in intelligence since Victorian times in terms of approximately 15 IQ points (or one standard deviation):

And then thinking about the idea of using reaction times as a scale of general intelligence: 

I realized that it would make more sense to most people if the decline in intelligence was expressed in terms of percentages.


Of course this is only approximate, but if the data from Silverman's paper is used the (median) average Victorian men's reaction time was 183ms (milliseconds), while the (mean) average Modern men's reaction time was 250 ms.

So the slowing from Victorian to Modern reaction times is 67ms

which is approx 37 percent of the Victorian reaction time of 183 ms; and 37 percent is more than one third.


So, in a phrase and using Victorian intelligence as the baseline; it would be reasonable to say that average intelligence had declined by more than one third since Victorian times.


And what does this 'one third' refer to?

MA Woodley talks of reaction times as a measure of core efficiency of the central nervous system - and this slowing of reaction times therefore suggests that our average thinking is biologically less efficient than the Victorians such that Victorian brains could perform an extra thirty seven percent more intelligent processing per unit time.



James Purcell said...

This might be simpler for average layman to understand and care about the decline of intelligence(most people don't know that i.q. actually measures the distribution of intelligence in a population through normalization of scores),but we know that it's all about the proportion of smart and dumb that really matters.

I had great fun reading the comments!

Jonathan said...

Bruce, no. If you want to use the phrase "intelligence had declined by" (some fraction), you have to use a measure of intelligence for which the absence of intelligence receives a score of zero, not a score of infinity.

Therefore, the reciprocal of reaction time; call it "reaction frequency."

If a quantity increases by one third, its reciprocal decreases by one quarter. So reaction time increased by one third, but reaction frequency (i.e. intelligence) decreased by one quarter.

Bruce Charlton said...

@Jonathan - I know what you mean, it is not entirely satisfactory, which is why I hedged by using the Victorian average explicitly as a baseline to make it clear what I was saying.

But reaction times in relation to intelligence are an interval scale (like Celcius) not a ratio scale (like Kelvin), so we can legitimately talk about size of differences but not ratios of differences.

An alternative would perhaps be to use a minimum reaction time/ maximum intelligence - say 150 ms to make a ratio scale.

What I was working towards was the last sentence, which is implicitly a reciprocal measure along the lines you describe - using unit time as the reciprocal - as a way of trying to explain and quantify what is at issue; since it is clear that few people outside intelligence research understand the way that IQ scores are generated and what it implies

(e.g. see this piece where the intro talks about a decline of 14 percent of IQ -

Ronald Calitri said...

Thanks for this; but I'm a little confused at using reaction time to proxy for g. As an economist, I like to illustrate intelligence development to my students using the films from Inoue & Matsuzawa Current Biology 2007. Asking whether they could match chimp performance leads to helpful discussions of training. The point here is that their supplemental has a nice figure, and, "In human subjects, there was a trade-off between the percentage of correct trials and the response latency. The accuracy of the three chimpanzees was located within human variation, while all of humans were slower than the chimpanzees."
Now, human average accuracy was at chimp maximum; but still, it makes one wonder whether reaction time is the best overall measure of g? Also, given students' responses, and own feeble efforts, some sort of training effect could be present, slowing overall reaction times to allow for needed processing in the increasingly complex mechanical environment of the past century?

Bruce Charlton said...

@RC - I suppose you will need to read the formal literature on this subject - where people like Arthur Jensen and Ian Deary survey the field and endorse reaction times.

This idea of RT as a measure of g is mainstream in intelligence research, and has been since the era of Galton - what *is* surprising is that people did not previously think about using RT to track long term changes, a job for which it is well-suited.

Of course the correlation between RT and measured IQ is loose, there is a lot of scatter on the graph - and to make comparisons between individuals, a population-normed pen-and-paper test would be much more precise.

But for comparing large populations there are great advantages to objective measures such as RT, or measures of brain size.

For measuring long term changes and national differences, I think RT is the best I know of.

But of course we need to look at convergent information. Is there anything comparably objective to RT which seems to *refute* the RT-implied hypothesis of a significant decline in 'g' in England over the past century plus?

I can't think of anything which does refute the idea of a significant decline in g, quite the opposite - but maybe somebody will come-up with an example? That is what needs to be done next.