3 ms·
"They are phenomenally accurate at predicting full-scale intelligence scores." I call baloney on this. I want to see the peer-reviewed publication that says so
by tokenadult 17y ago
"They are phenomenally accurate at predicting full-scale intelligence scores."
I call baloney on this. I want to see the peer-reviewed publication that says so. This is the kind of mental test that Galton and James McKeen Cattell did a century ago, and those tests were found to have no valid correlation with intelligence.
http://norvig.com/experiment-design.html http://norvig.com/experiment-design.html
This further paragraph in the submitted article provides better information:
"The team wanted to evaluate several intelligence tests, including their own. So they recruited 77 gifted children through the Parents’ Association for Gifted Children in Switzerland. A previous intelligence test, taken about a year-and-a-half previously, had won them entrance to gifted primary schools. So how many of the kids still classified as gifted just eighteen months later? Only half, no matter what test was used. (And that was using a relaxed cut-off line, to account for standard deviations in testing.)"
This is in accord with many findings by many researchers in many places over decades: preschool IQ tests have remarkably poor reliability for predicting subsequent school-age IQ scores. By the way, the Newsweek blogger's terminology is incorrect. Where Po Bronson wrote, "to account for standard deviations in testing," he should have written, "to account for standard error in estimation," which is a different concept. A "standard error" is something different from a "standard deviation."
The later paragraphs somewhat rescued this casual blog post, but I fear that most news reporting on IQ testing is at the level of excessive credulity found in the earlier paragraphs. It takes time and effort to read and understand the better published literature on IQ testing
http://learninfreedom.org/iqbooks.html http://learninfreedom.org/iqbooks.html
and alas many journalists don't bother to make that effort. Even less do most bloggers bother to check their facts on this contentious issue before posting.
- jerf 17y ago"They are phenomenally accurate at predicting full-scale intelligence scores." You are arguing against the statement without the word "scores" in it. That's not the same statement. The statement as it was made is correct, and there's no need for a study to be cited; this is the study being cited. Note how the authors go on to point out that if this simple test has a 99% correlation with the complicated test, then, logically, they are effectively the same test, and if you can't believe that line sorting is an adequate intelligence indicator, than neither are the conventional tests. Slow down with the shooting-from-the-hip there; this entire article is all about how bad tests are for determining intelligence, if you actually read it. It's not just the "later paragraphs".
- tokenadult 17y agoYou are arguing against the statement without the word "scores" in it. I agree with you that the overall tenor of the blog post is skepticism about early childhood IQ tests, but I am additionally skeptical of the Swiss test-giver's claim that his sensory perception test correlates well with IQ scores of children at the same age. That's just his claim so far. The blog post doesn't include anyone "showing the work" to show that that is replicable result. It would be a very strange result, actually, inconsistent with results that have been replicated many times (as mentioned in a scientific publication I linked to in another reply in this rapidly growing thread). http://www.personalityresearch.org/acton/sense.html http://www.personalityresearch.org/acton/sense.html But, yes, even if the quick-and-dirty test correlates very well with preschool IQ tests, that doesn't mean much, because preschool IQ tests--all of them--correlate poorly with anything of interest, including subsequent IQ scores, that shows up at school age.
- jibiki 17y agoI think this is the paper, unfortunately, it's behind a pay wall. http://psycnet.apa.org/index.cfm?fa=search.displayRecord&uid=2008-12082-007 http://psycnet.apa.org/index.cfm?fa=search.displayRecord&...
- gwern 17y agoI have a copy. The .99 correlation seems to be coming from this: > Paired sample t-tests showed that children diagnosed as gifted in this study (N = 44) achieved equal scores on the HAWIK-IV (M = 126.9, SD = 7.1) and the IDS (M = 128.9, SD = 8.2), t(43) = –1.30, p = .20. Average nongifted children (N = 69) scored equal as well on the HAWIK-IV (M = 101.2, SD = 8.5) and the IDS (M = 99.86, SD = 0.03), t(68) = 1.42, p = .16. (IDS is this new line/weight-based test, the HAWIK-IV the usual IQ battery.)
- tokenadult 17y agoThanks for the reference and for the quotation (to the parent and grandparent of this reply). Yes, the HAWIK would be the usual child IQ battery in a German-speaking country. I'm not following the statistics shown there completely, but I note the sample size. Did the same group of test-givers give both tests? Were they "blind" as to the results of each test when giving the other? I appreciate the references. I'm still doubtful that the general finding would be that the five-minute test would be strongly correlated with full scale child IQ batteries, e.g. the WPPSI. The way to find out would be for other groups of test-givers to attempt to replicate the result.