Saturday 8 March 2014

If you don't believe that intelligence has declined by about one modern standard deviation in the past 150-200 years - then what DO you believe?

*

1. Do you believe that Victorian simple reaction time (sRT) data are not comparable with modern data? If so, would you be convinced by evidence of rapidly slowing reaction times over recent decades, measured in one laboratory and using only modern RT machines? Because this kind of evidence is in the pipeline.

2. Do you believe that - despite about 140 years consensus that sRT and IQ are significantly correlated, and the general belief that this correlation is because general intelligence is dependent upon processing speed of which sRT is an indirect measure - there is NOT a causal relationship between simple reaction times and intelligence? That, therefore, average sRTs could be getting much, much slower but that this would not necessarily make any difference to average intelligence?

3. Do you believe that the measured decline in average simple reaction time from a Victorian sRT average speed of about 180 milliseconds (in several independent studies) to a modern average speed of 250 milliseconds (or slower), a slowing of 70 milliseconds plus - is not enough to be of interest: that it is too small to not reflect any significant or meaningful reduction in intelligence.

4. Do you believe that because the measured slowing of sRT over the past 150 years seems unexpected, and is larger than you would have supposed possible, strikes you as indeed ludicrous - that therefore we should simply ignore it?

5. Do you believe that - because the data on long term sRTs seems anomalous with your world view, that we should therefore assume that somehow there is something wrong somewhere with the Victorian to Modern comparison; and therefore we should just carry-on just as if we knew nothing about longitudinal changes in sRTs?

6. Do you believe that there has been a significant reduction in average general intelligence over the past 150 years, but that it is much less than one standard deviation - probably more like HALF a standard deviation? And the large size of the sRT slowing is just a Red Herring?

7. Do you believe that average intelligence has NOT changed over the past 150 years - that moderns have the same intelligence as Victorians? And the slowing of average sRT is irrelevant?

8. Do you believe that average intelligence has increased over the past 150 years despite slowing of sRTs, because you believe the pen-and-paper IQ tests are more valid, reliable and/or objective than reaction time data?

9. Or something else, or what?

*

18 comments:

Matthew C. said...

I believe the correlation is not causal, and something about our modern environment has caused our brains to react more slowly (my theory - increased necessity to integrate more knowledge - more connections - slower response).

One reason I am so deeply skeptical of this hypothesis is I test in the bottom third of reaction time and four standard deviations above the average for IQ. That seems highly unlikely if reaction time and intelligence are so closely tied.

And when I look at professional athletes where reaction time is extremely important, I do not see an overrepresentation of Ashkenazi jews, for example.

I also simply don't believe people are so much stupider today. I do agree that certain types of creativity are less prevalent.

MC said...

From the standpoint of a layman, position number 6 doesn't seem unreasonable, although it is possible to believe that the decline was less than 1 SD without believing the RT disparity to be a complete red herring. Just that selection bias or some other error counts for part of the decline while a real reduction in IQ accounts for the rest.

The selection bias argument seems the strongest to me. From what I've read on your various blog posts, the argument for similar study subjects seems to revolve around the fact that the studies were conducted in nearly the same location with similar entrance fees, etc. I just don't know how sure one can be that they represent similar strata of society. Given modern transportation, might many of the modern study subjects be tourists from poorer parts of the country with lower IQs? Or, given economic growth, doesn't disposable income constitute a higher proportion of earnings, so that an entry fee which costs the same after inflation is nonetheless more affordable than in the past?

Maybe you have more data on this that I'm not aware of.

Bruce Charlton said...

@MathC- Simple reaction time is seeing a light and pressing a button.

sRT has nothing to do with athletic ability - except perhaps a few sports where strength matters little compared with pure reaction speed - ping pong and foil fencing perhaps.

And it doesn't measure integration of information - almost the opposite - it tests just about the simplest, quickest neural path.

The correlation coefficient for simple RT and IQ is small. So there is a lot of scatter, a lot of dissociation between the variables. However the correlation is reliably positive - and has been known (and uncontested) for a very long time.

@MC- If something, like class selection, was a significant cause of confounding - in other words when the real cause of the difference was class selection rather than decline - then when controls are introduced for class, then the difference would diminish. But when this was done, the RT difference stayed the same.

So - You do not try, expect or demand perfectly matched samples, because there is no such thing and never has been. What you do is try to evaluate the sensitivity of the outcome measure (sRT) to class controls applied to the data. That is what Michael did.

Also, there is a question of a priori expectations. This is perhaps more important than the above technical point.

Most people (including myself) were very surprised at the results; in the sense the my knowledge of the effects of biases would have suggested that measured reaction times would have got faster.

For example, I assumed that modern RT machines were more responsive, and had less delays - so I expected a speeding up from that cause.

*

The BIG point here is how should we, as scientists, react to clear simple results which at face value refute the standard understanding?

This is Point 4 above.

When some observation stands out very clearly and apparently refuting our basic understanding, do we put tha observation under a microscope to look for micro-methodological flaws to give us an excuse to reject and ignore it?

Or do we take it seriously as a lucky break which may give us a clue to the nature of reality? To me, this very obviously seems the right response.

*

Bruce Charlton said...

(Continued)

*

Every paper ever written has 'flaws' which can give someone who wants (a priori) to reject it, an excuse to reject it. These RT observations are less flawed than most - not least because some of the Victorian data comes from one of the greatest scientists ever (Galton) and this is replicated in other Victorian labs.

If the opposite trend (speeding sRTs) had been reported it would have been accepted without a murmer (as confirming the Flynn Effect or rising IQ test scores, which almost everybody *wants* to believe is true).

To me it is obvious that this observation MUST be taken seriously and followed up. Especially because RT is, for all its imprecision, by far the most objective correlate of IQ which is available over the past 150 years (which was why I looked for this RT data in the first place - http://iqpersonalitygenius.blogspot.co.uk/2013/05/the-e-mail-when-i-had-idea-of-using.html -

And RT is not just a correlate of IQ, but is understood to be measuring (imprecisely, but validly) a major causal component of general intelligence - a constraint on the speed of brain processing.

*

SO, to me this is very simple. ANY real scientist who understood what was at stake would be fascinated by the observation of slowing sRT, and would not ignore it. Insofar as people (who know the field) react by trying to explain-away the observation - by using micro-methodological quibbles which they would never dream of applying to observations they agree with - then they are not behaving as real scientists.

Harsh, I realize, but I think I am being fair! Everybody knows that science may go wrong by believing wrong ideas or data; but science also, and I would say more often - especially nowadays more often - goes wrong by failing to take-seriously and respond to true (or probably true) ideas or data.

This belief of mine was why I was chosen to edit Medical Hypotheses (2004-10) - which was established to try and combat the inbuilt tendency of modern science (but especially medical science) to discourage, ignore and even suppress anomalous findings when they go against consensus.

Imnobody said...

but science also, and I would say more often - especially nowadays more often - goes wrong by failing to take-seriously and respond to true (or probably true) ideas or data.

Well, see my comment in the other thread about disregarding scientific findings that go against PC.

I am not knowledgeable enough to assess this finding but, if it is true, goes against the myth of progress, one of the foundations of PC.

Only with the myth of progress, PC can justify doing away with traditions that have lasted for millennia (for example, the "traditional" family) and replacing them with new PC monstrosities (everything that lives with you is your family).

The ancients were dumb and fanatics, that's all. We are the most smartestest of history so we can destroy everything and put something that looks good to us.

Adam G. said...

I'm also smarter than normal but also have much slower reactions than normal too.

But the evidence in the pipeline sounds pretty convincing. Can't wait to see it.

Bruce Charlton said...

@Adam - Just to clarify - do you mean that you have actually had your sRT measured? Or just that you seem to have 'slow reactions' in the athletics sense of dodging, weaving, and fighting, responding to other people's moves etc?

Adam G. said...

Bruce C.,
I've had it measured. At BYU there is, or used to be, a gizmo in the mechanical engineering department where a light flashed and you were supposed to pull a trigger and it told you how long it took you and provided some data about normal reaction times. It too me much longer than average.

I don't remember the details, but I remember something similar on online tests, where you are supposed to hit a key once you see a flash or something.

Also, in my National Guard time we did a number of combat simulations where you had a laser-equipped rifle and a laser-sensitive big screen that projected a combat scenario and tracked your shots. They had a a lot of different analyses that they applied to the data, one of which was a reaction time analysis. Mine stank.

It's possible I'm just a freak of nature. I do have an outsize skull, though, so I've wondered if my intelligence is less processing speed and more raw processing power.

I believe there is also a positive though not extremely tight correlation between brain size and intelligence. Is there any data that would allow us to determine changes in brain size over time (probably just changes in skull size, assuming the thickness of the bone hasn't changed over time).

Bruce Charlton said...

@Adam - thanks. The measured correlation between intelligence is never very high (it is not a fixed thing, but varies according to the sample, precision of measurements, degree of controls etc) - but doesn't rise much about 0.3 (0 is no correlation and just random scatter, 1 is all points on a straight line) - which means there is a LOT of scatter around the line, and many individual exceptions and dissociations.

As Steve Sailer has said, IQ - and also RT - work very well in group studies, and in studying group differences - where they make very sure predictions of size and direction of differences; but are very imprecise for the prediction of individuals.

(In other words, the opposite way round from how they most often tend to be interpreted and used.)

ajb said...

I think a significant part of the response an individual has to these findings has to do with their assessment of the direction of society (and scientific or mathematical pursuit in particular).

If they think science is stagnating, not producing the results they expect, and so on, they probably will be more receptive to this idea.

If, however, they think science is stagnating but also think that they've already figured out the cause by positing a single or 'economical' solution such as a decrease in 'low-hanging fruit', say, then they will probably be less receptive as well.

So, it seems to me that part of the case to be made is that a) science isn't performing how we should expect, and b) that other explanations like 'low-hanging fruit' aren't adequate, or are less parsimonious and so on.

Slate Slabrock said...

Putting it all down to a decline in IQ seems to be overly simplistic. Part of it probably comes from a decline in IQ, sure.

But there could be other variables that could hypothetically, make say... a 7 point drop in IQ look more like a 15 point drop in IQ.

Toxoplasma gondii makes your reaction time slower.

Low testosterone also seems to make your reaction time slower. There has been a dramatic collapse in testosterone levels over the last however-long-they've-been measuring it.

We may or may not be dolts compared to the Victorians but we're certainly hypo-gonadal nancy boys compared to the men of 70 years ago.

Salivary testosterone and simple reaction time parameters.
http://www.ncbi.nlm.nih.gov/pubmed/7862265

People with high IQs and high T may have a greater tendency to accomplish stuff in real life, while people with high IQs and low T may be more content to accomplish stuff in online roleplaying games.

Luqman said...

I have not paid close attention to the matter beyond mostly cursory reading of the posts here and on the Cochran blog, but my intuition is thus: I dont believe there has been a 1SD decline in IQ since Victorian times, but I dont believe it ultimately matters. Creativity probably correlates with IQ in that you need to be at least somewhat intelligent for that creative impulse to be channelled into something concrete, but it isnt just high intelligence. The premise does not add up for me.

Is something missing though? I agree with you Dr. Charlton that there is something qualitatively different and inferior about science and the arts in the modern era. There is something creatively missing, but the dearth is such that I dont really understand how it can be attributed to a 1SD decline in mean IQ. The argument that the West has just passed below some threshold of IQ that allows for the proliferation of true genius makes little sense to me and seems too convenient.

What seems to be missing is a concrete understanding of creativity and genius as something more than intelligence and a way to quantify it. Something more than philosophical musing. Whatever dysgenic trends have been occuring that you claim may have led to this 1SD decline, well surely they could just as easily affect the related but discrete quality of creativity? Perhaps this is looking in the wrong place for the right result?

Bruce Charlton said...

@SS - That cause doesn't fit the large pattern of data on intelligence.

@L If you think I am over-focused on IQ then maybe you don't know my other blog

http://iqpersonalitygenius.blogspot.co.uk/

Over the years, I have written extensively on the decline of scientific (especially medical) breakthroughs, creativity, education, selectivity in science etc years before I knew anything about reaction time slowing.

http://qjmed.oxfordjournals.org/content/98/1/53.long

It just seems to me that RTs are the best evidence we have, and therefore should be accorded priority.

But the decline of Christianity is even more important; because modern (secular) scientists and other academics are too dishonest (especially with themselves) ever to achieve anything significant - they are Not Even Trying, so how could they succeed?

DrBill said...

Suppose we believe the "processing speed" theory of the correlation between g and reaction time.

There are then three variables of interest: two latent (i.e. unobserved) variables, g and speed, and one observed, reaction time. The idea is that the latent variable speed (S) affects both g and RT, though there are other, unknown, non S, variables affecting each.

If we model these using what I think psychometricians call an SEM and we assume (generously for your theory) that 100% of the increase in reaction time comes from a decrease in speed, then we can show that the number of standard deviations increase in g coming from a decrease in speed sufficient to generate a 1 std deviation increase in RT is:

Corr(RT,g)*Corr(g,speed)/Corr(RT,speed)

Jensen(1993) claims that the right value for Corr(RT,g) is -0.1. The other two I don't know about.

For your intuition that the 1 std dev increase in reaction time should go with about a 1 std dev decrease in g, you would have to believe that Corr(g,speed) is about ten times as big in absolute value as is Corr(RT,speed).

So, I think you can go a long way towards arguing that 6 is the right answer just from the math. And, again, this is being generous to the theory. Jensen's view was that RT mostly represents non-g things like motor skills. If, say, 50% of the increase in RT comes from declining motor skills, then you would have to reduce the amount calculated above by 50%.

It's possible I've got the math wrong, of course. I did not spend a lot of time on it. But I'm pretty sure you can't just go from 1 std dev increase in RT to a 1 std dev decrease in g.

Bruce Charlton said...

DR Bill - My logic was simply that the top 15 percent of sRT would also be the top 15 percent of (true, real, underlying) IQ (etc).

DrBill said...

But that's not right. That would only be true if the correlation between g and RT was -1 (this would lock the other two correlations at 1 also).

The fact that the correlation between RT and g is low means that the top 15% in the one won't be the top 15% in the other, or even similar to it.

I don't recall the analytic formula, so I just quickly ran a simulation. Suppose you have two normals correlated at 0.1, and you look at the top 15% of the first one.

Only 20% of the observations which are top 15% on the first variable will also be top 15% on the second variable (not 100%).

The mean of the second variable over the observations which are in the top 15% of the first variable will only be 0.2 standard deviations above average.

Bruce Charlton said...

@DB - Well, we just don't know what is going on behind the various imprecisions of measurement, imprecisions of concept, assumptions and the rest of it. I don't think its something worth agonizing over! (Not that I am accusing you of doing so!)

In fact, there is a bizarre notion common in intelligence psychometrics (and also in epidemiology - where I first encountered it) that measured quantities are hypotheses - so people might assume that I was stating a 1SD decline as if it was an hypothesis.

But hypothesis are not about effect sizes but about causal processes; and quantitative measures (such as correlation coefficients, or differences between means) vary depending on the properties of samples.

I talk about it here - http://charltonteaching.blogspot.co.uk/2010/10/scope-and-nature-of-epidemiology.html - especially in the section on Modes of Inference. Of course, in trying to get 'people' not to ignore this extremely important observation in sRT slowing, I have been drawn into the kind of focus on effect size which I actually deplore! The main thing about this observation on sRT is arguably that it makes clear the artefactual nature of the Flynn Effect - and reveals it to be 'merely' IQ test score inflation (or, as I say elsewhere 'stagflation').

The amount of intelligence decline this degree of sRT slowing represent is probably not possible to put an IQ number to, since IQ is neither an interval nor a ratio scale - http://iqpersonalitygenius.blogspot.co.uk/2013/02/the-ordinal-scale-of-iq-could-be.html - which means that it cannot really be extrapolated - or not much.

But looked at in milliseconds, it is clearly a *very* significant slowing - something between an extra third to a half the amount of time required to perform the reaction.

DrBill said...

Yes, I agree. It is a huge increase in reaction time and very interesting.