November 20, 2013

CBC: The g Factor glass isn't half full, it's half empty!

From the Canadian Broadcasting Company last year:
IQ myth debunked by Canadian researchers 
Human intelligence consists of different components, researchers say 
Dec 19, 2012  
A study from researchers at Western University say that there is little evidence for the concept of general intelligence. Instead, human intelligence is made up of multiple and distinct components. 
An individual's IQ score — long-held as the standard measure of human intelligence — is not a valid way of assessing brainpower, say Canadian researchers. 
A team from Western University is debunking the concept of general intelligence, saying that there is no single component that can account for how a person performs various mental and cognitive tasks.  
Instead, human intelligence is made up of multiple and distinct components, each of which must be looked at independently. 
'We have shown categorically that you cannot sum up the difference between people in terms of one number.' 
The study, published today in the journal Neuron, included the largest online intelligence survey on record, which recruited more than 100,000 participants. 
... The results showed that how people performed at the tests could only be explained with at least three distinct components: short-term memory, reasoning and verbal ability. 
No single measure, such as an intelligence quotient, or IQ score, could account for how well, or how poorly, people did.
IQ test: Which way's table tilted?
The concept of a general intelligence factor dates back to at least 1904, when psychologist Charles Spearman suggested that there was a correlation between seemingly unrelated tasks, such as memorization, reading and performing arithmetic. 
He called this link the 'g' factor, or general factor, and proposed that is accounted for an individual's performance across different mental tasks. Various intelligence tests, using a wide variety of methods, were developed throughout the 20th century as a way to evaluate children, students, military recruits and even potential hires. ...
"We have shown categorically that you cannot sum up the difference between people in terms of one number, and that is really what is important here," said Owen, adding that further tests still need to be done.

37 comments:

  1. "We have shown categorically that you cannot sum up the difference between people in terms of one number, and that is really what is important here," said Owen, adding that further tests still need to be done.

    It must be nice to get paid for destroying straw men.

    ReplyDelete
  2. We get it.
    You can't use one number.
    You can't use one number.
    You can't use one number.
    Apparently you can use three numbers however, rather than, say, Howard Gardner's eight.
    And the three numbers show quite a bit of correlation.

    So, fine then. We appreciate the reminder not to oversimplify. Ron Hoeflin's High-IQ society Mega Test apparently does get in the main points.

    ReplyDelete
  3. 'Western' University used to be the 'University of Western Ontario' where Rushton was a prof.

    ReplyDelete
  4. Okay. Let's get rid of 'general intelligence'. Let's identify the various kinds of intelligences, and then let's test all sorts of people for their aptitudes in these multi-intelligence.

    The result will be what? Whites outscore blacks on most of them.

    It's like this. Let's say there is no general strength but only many different kinds of strengths. Leg strength, arm strength, neck strength, and etc.
    Well, guess what. Blacks are gonna beat Mexicans on most if not all of them.

    So, the end result is the same thing.

    ReplyDelete
  5. The results showed that how people performed at the tests could only be explained with at least three distinct components: short-term memory, reasoning and verbal ability.

    I saw this before.

    1. No one ever doubted that intelligence could be broken down into further components. Speculations about short term memory, for example, are all over the literature.

    2. I remains to be seen which of the three factors has more correlation with real world outcomes.

    ReplyDelete
  6. It must be nice to get paid for destroying straw men.

    Yup.

    ReplyDelete
  7. Let me guess.

    Dribbling a basketball is going to be said to be a form of intelligence.

    ReplyDelete
  8. OK, so now we have three numbers instead of one. If we make 100 the norm for each of these abilities then in the future the average person will have an IQ of 300 instead of 100. Take it from there.

    It would be a good PR move at least, and might not be such a bad idea anyway. It would be like a combined SAT score plus a test of short-term memory.

    ReplyDelete
  9. foreign Expert11/20/13, 6:45 PM

    I have shown categorically that there is no such thing as the "temperature of the earth." There are many uncorrelated temperatures, any one of which may be rising or falling at any given time.

    ReplyDelete
  10. Apparently, factorisation is impossible.

    ReplyDelete
  11. I bet the researchers in question were completely unconcerned about getting thier paper in a high impact-factor journal. Because numbers really aren't all that important, after all.

    ReplyDelete
  12. Auntie Analogue11/20/13, 8:02 PM


    Bet that none of these researchers has an IQ under 120.

    ReplyDelete
  13. The black-white IQ gap is the only reason people care about this stuff.

    Here is what Flynn has to say about it:

    "...This is manifestly false.
    Jensen’s arguments would bite no matter whether blacks suffered from a score deficit on one or 10 or 100 factors. I attribute no intent or motive to Gould, it is just that you cannot rebut arguments if you do not acknowledge and address them."
    (Flynn 1999a: 373)

    So nothing at all changes in terms of the black-white IQ gap, but most lay-men and liberals will see the results as a refutation of the black-white IQ gap.

    ReplyDelete
  14. Unlikely they've found anything new. Studies tracking the association between a general intelligence score and 'life outcome x' has been done enough to make us confident that it is 'general' intelligence (i.e. the average of their 3 components) which matters. At most they'd be quibbling over the weightings on these components within a particular context.

    -Jostein Gronvand

    ReplyDelete
  15. I read the paper. I don't feel qualified to try summarizing it, but I will note that their three factors are still all intercorrelated (figure 3).

    Regarding the news piece's title "IQ myth debunked by Canadian researchers" -- well... no... The established scientific consensus doesn't become a "myth" after one primary research paper.

    ReplyDelete
  16. http://www.nytimes.com/roomfordebate/2013/11/20/racism-in-the-age-of-obama/obama-has-become-a-code-word-for-racists

    ReplyDelete
  17. "Instead, human intelligence is made up of multiple and distinct components, each of which must be looked at independently."

    No, the word "independent" actually means something in scientific/mathematical language. Components that are correlated with each other are not independent, by definition.

    ReplyDelete
  18. Full text of the article:

    http://scottbarrykaufman.com/wp-content/uploads/2012/12/Hampshire-et-al.-2012.pdf

    ReplyDelete
  19. Arthur Jensen is hereby demolished and refuted and buried for the 765th time! Odd how he remains in circulation and is whispered about all over.

    ReplyDelete
  20. "Arthur Jensen is hereby demolished and refuted and buried for the 765th time! Odd how he remains in circulation and is whispered about all over."

    Those who control the media/academia can knock down the boulder all over again, and Sisyphus has to push it up again.. until it's knocked down again.

    Libs control the reset button.

    ReplyDelete
  21. "We have shown categorically that you cannot sum up the difference between people in terms of one number, and that is really what is important here," said Owen

    You did no such thing, dumbo. Their paper is a sorry example of people using methods they don't understand and then getting published in a journal like Neuron whose editors and reviewers also lack psychometric expertise.

    Firstly, they make much of the fact that they had (poor quality) test data from thousands of individuals. However, the sample size in the crucial brain scan part of the study was actually 16. Yes, 16. The idea that you could determine anything about the structure of intelligence based on a brain scan of 16 individuals is laughable.

    Secondly, if you look at the Supplemental Information of their paper, you'll notice that the first component in fact explained a whopping 79.7 percent of the variance in their brain scan data. Similarly, in their online test data, there was a first component that accounted for 27.3 percent of the variance, while the two other significant components together accounted for 17.7 percent. (The fact that the common variance is as low as this suggestst that their test battery has an excessive amount of error variance, probably reflecting inadequate sampling of participants and psychometrically unsound tests.)

    So, how did they get from these data that are consistent with there being a general factor to the conclusion that there is no such thing? Well, they simply rotated the components so that that the variance of the first component was dispersed to several components. In other words, they mathematically transformed the data so that the large first components went away. This is, of course, a perfectly circular way of claiming that there is no general factor. It proves nothing about the structure of intelligence. They argue that the correlations between the component loadings from the tiny brain scan study and the big online testing sample confirm their preferred component structure, but the correlations reported are 0.79 and 0.64, which means that the loadings are definitely not the same (a congruence coefficient of >0.95 is generally the standard for component/factor identity; furthermore, it seems that they mistakenly used Pearson's r to calculate the correlations).

    Using the very same data and the same sort of reasoning, they could have also concluded that the g factor is real and that intelligence cannot be conceived without reference to g. That would also have been an erreneous conclusion, because their methods and data are completely inadequate for the task of determining the structure of intelligence.

    ReplyDelete
  22. Their political goal in doing the research was not to replace IQ with two (or three or five) more accurate tests, of course.

    Rest assured, no lefty prof will ever attempt to define the new multiple intelligences or devise a test for them.

    Their goal was simple. Destroy "IQ" and don't replace it with anything.

    ReplyDelete
  23. i came here to comment & you astute commenters have already said everything i would say & more. seriously well-done! no wonder steve trusted the comments section to take the yeoman's job of tearing that article a new one (or a correlated three:)

    ReplyDelete
  24. From the abstract:

    We propose that intelligence is an emergent property of anatomically
    distinct cognitive systems, each of which has its own capacity.


    This is just like saying that "computer speed" is an emergent property of structurally distinct hardware subsystems: integer operation units, floating point operation units, memory controllers, memory chips, and etc.

    How do you conclude from this that there is no such thing as computer speed? How do you conclude from this that statements like "Computers from the 1990s are slower than computers from the 2000s" are false or meaningless?

    It's not a strawman argument so much as a non-sequitur.

    ReplyDelete
  25. Speaking of threes - about fifty years ago I was an undergraduate in psychology. I took courses in statistics, genetics and testing.

    Everything I learned in genetics was wrong or inadequate. Everything has now changed.

    Everything I learned in statistics is exactly the same as it was then. Computers do the calculations faster these days but the theory and the tests are all the same.

    In psychological testing we had the Stanford-Binet, the WAIS, the SAT, and the GRE. All the same. Also the same then was discussion of 'g' versus multifactorial theories of intelligence and the public's discomfort with racial differences in IQ.

    That's why I try to only read about genetics nowadays. At least there's some new stuff occasionally. These IQ arguments are real nostalgia items like that issue of National Lampoon you ran earlier in the week.

    Multi factorial IQ. Ahh yes! It takes me back.

    Albertosaurus

    ReplyDelete
  26. The Glass is of Liberman[Blue] [Democrat AKA Liberal] leaning Right. Hope it is not Anti-Freeze.

    ReplyDelete
  27. Re: IQ Test
    The glass is clearly accellerating towards the right.

    ReplyDelete
  28. The non-existence of g was obvious to any thinking person for a long time. We certainly didn't need a new study to show this - existing well known facts about IQ demonstrate this well enough. That Jews could score 120 on verbal and 100 or less on spatial, and a host of similar facts, makes that blindingly clear. It was only the strange conspiracy to ignore these blatant facts that kept the g myth alive. What seems to be changing now is that - perhaps - scientists are beginning to actually come up with theories that make sense of current data lol.

    g was always a case of wishful thinking and overstretch, a product of the desire to find one overarching factor that explains everything and eliminates messy complexity, which is strong in science, as well a projection of common sense conclusions - we all sort of believe that a single thing like intelligence exists until we examine it closely - onto recalcitrant facts.

    Now I have only to wait till some g zealot tells me I just "don"T understand" the concept of g. Right, because it's soooo complex. I notice more and more the favored tactic among scientist types who don't want to admit they're wrong is to claim they are not understood. I suppose to an intellectual, the strongest put down is to say you don't understand, and I suppose there is no answer to that because if you didn't understand but thought you did you wouldn't know. But then, maybe they don't understand and don't know. lol.

    ReplyDelete
  29. Seen this one yet, Steve?

    MCDONALD'S restaurant turns to opera to drive out loitering teens...

    Probably more SBPDL than iSteve, but funny, no? Classical and country music, they're like kryptonite to blacks in particular, and NAMs in general.

    ReplyDelete
  30. The results showed that how people performed at the tests could only be explained with at least three distinct components: short-term memory, reasoning and verbal ability.

    Inelegance is the capacity to see and predict movement. It is the eye and the mind’s eye that matter most. Everything in the universe is moving, nothing is static --- the future is everything - it is our capacity to predict that movement, that marks intelligence.

    Whether hitting a baseball or predicting the movement of planets – it is the mind’s eye that does the work. When working a puzzle it is the mind’s eye that matches up its different parts – it sees those parts meshing. Words define different objects that have grown to be what they are. And words define the different types of movement the effect the future of those objects. These differences are all taken in by the eye and stored in the mind’s eye.

    The capacity to see something and predict its future is intelligence. That is what should be measured.

    p.s. Hitting a baseball requires concentration, predicting the movement of planets requires concentration, doing a puzzle requires concentration. Can concentration be taught – can that be learned through effort?

    ReplyDelete
  31. Kern, your thinking that Jewish IQ profiles disprove g indeed shows that you don't understand this stuff.

    ReplyDelete
  32. Neal Armstrong had a cool line: "A pessimist says the glass is half empty. An optimist says the glass is half full. An engineer says the glass is twice as big as it needs to be."

    http://twitpic.com/37yrqq

    ReplyDelete
  33. g is a variable that generally explains most of most people's cognitive performance on a range of tests. It explains a large part of the variance.

    It isn't a variable that explains everybody's performance perfectly. Some people can show very little correlation between tests. It's just that people don't do so often.

    The problem with how people see g is that they misinterpret it as a statement that *everybody* has to have their cognitive scores on one test within a certain, fairly close, fixed range of certain other scores.

    But this isn't true. People generally *do* show high correlation but they don't *have* to and some people *don't*.

    For this to be the case, g has to be a population level abstraction, rather than a real property of the brain.

    ReplyDelete
  34. I don´t feel like reading that stuff, but can it be falsified?

    ReplyDelete
  35. Anonymous 11:31 has just given the best explanation of g that I have ever read on all the HBD blogs. Problem is I have not yet encountered a single person - not one - who understands g as anonymous does. Everyone uses it as synonymous with intelligence and as a property of the brain and pertaining to individuals, not populations.

    But it overstates the case for g. "g" is not even a factor, it is a metaphor, a figure of speech, another way of saying "there is a high correlation between scores". To conclude from this that there is an underlying factor is a leap of faith. Ketchup usually shows up with mayonnaise on a hamburger, but doesn't always, and there is no underlying factor responsible for the make up of ketchup and mayonnaise. Since the two usually show up together to then postulate that they are really the same kind of thing would obviously be a leap of faith, a kind of religious thinking. Nor does it gain in plausibility to say they are partly the same thing - that's simply a leap. Yet since multiple kinds of intelligence often show up together - but not always - we feel entitled to say they share an underlying nature, not just partly, but largely. Yet the "not always" should be enough to show that they do not share an underlying nature and appear together for different reasons!

    The other problem is that the correlation that g represents is vastly, vastly overstated - probably because people want so much to find it. The fact that Jews can score 120 on V and 100 or less on spatial is hardly *high* correlation, and women also have widely disparate scores on subtests, and the Asian difference of 110 on S and between 96-97 on V also tends to support the claim for *high* correlation. And this is populations - examples of individuals with weak correlation are legion!

    But people see what they want to see.

    ReplyDelete
  36. http://drjamesthompson.blogspot.co.uk/2013/01/what-makes-good-iq-story.html

    Seems I had covered most of the points, but had forgotten I had done so. Hmm.

    ReplyDelete
  37. http://drjamesthompson.blogspot.co.uk/2013/01/what-makes-good-iq-story.html

    ReplyDelete

Comments are moderated, at whim.