Friday, July 28, 2006

Reinventing the Wheel   posted by Alex B. @ 7/28/2006 05:33:00 AM

The NYT has an article up about how supposedly malleable IQs are by a professor trying to get more money for early childhood education programs. Aside from the arguments about funding a program that is not that effective , the merits of the NYT article are questionable.

Kirp's argument is based on two studies by the same researcher: Michel Duyme. Before anything is stated about the studies, I'll point out that while the participants in these studies may be new, the theory and designs are not. Herman Spitz's masterful work goes over many of these types of studies, and the overall result is not too impressive for those thinking IQ is easily changed. If your educational psychology professor didn't make that book a required reading, join Questia and read it, or snag it from the local library.

Now on to the NYT article.......

About the first study, Kirp writes:
Regardless of whether the adopting families were rich or poor, Capron and Duyme learned, children whose biological parents were well-off had I.Q. scores averaging 16 points higher than those from working-class parents. Yet what is really remarkable is how big a difference the adopting families' backgrounds made all the same. The average I.Q. of children from well-to-do parents who were placed with families from the same social stratum was 119.6. But when such infants were adopted by poor families, their average I.Q. was 107.5 - 12 points lower. The same holds true for children born into impoverished families: youngsters adopted by parents of similarly modest means had average I.Q.'s of 92.4, while the I.Q.'s of those placed with well-off parents averaged 103.6. These studies confirm that environment matters - the only, and crucial, difference between these children is the lives they have led.

Darth Quixote has already deeply delved into this study, and there is no need to repeat him here. I will add:

  • the design was non-experimental, post-test only. While these designs have merit, to infer causality is tenuous as the adoptees' IQ could easily have fit the same pattern before adoption. Moreover, since there was no randomness in the adoption process, we have no idea as to why the children were placed where they were.

  • the researcher's did not test to see if the instrument they were using was functioning the same across comparison groups. A (good) way to check this Multigroup confirmatory factor analysis (e.g.,), but with the n so small, it is difficult to do.1 As a proxy, I checked for equality of covariances: B+ vs B-: not significantly different; A+ vs A-: not significantly different; B+A+ vs B-A-: significantly different! (chi-square: 154.79, df= 55, p < .0001). This could mean multiple things. Two that come to mind are: a) the B-A- group's pattern of cognitive functioning has not matured at the same rate as the B+A+ group, so give them some time and it may converge. b) the B- kids were placed into the A- home for a specific reason (i.e., they had some type of deficit that made A+ parents pass on them).

  • So where does leave us? Capron and Duyme sum it up best:
    Although these findings clearly indicate that the biological parents' background contributes to observed differences in IQ between extreme groups, as does that of adoptive parents, more detailed interpretation is difficult (p. 553)

    which is a far cry from Kirp's interpretation:
    the only, and crucial, difference between these children is the lives they have led.

    About the second study, Kirp writes:
    A later study of French youngsters adopted between the ages of 4 and 6 shows the continuing interplay of nature and nurture. Those children had little going for them. Their I.Q.'s averaged 77, putting them near retardation. Most were abused or neglected as infants, then shunted from one foster home or institution to the next.

    Nine years later, they retook the I.Q. tests, and contrary to the conventional belief that I.Q. is essentially stable, all of them did better. The amount they improved was directly related to the adopting family's status. Children adopted by farmers and laborers had average I.Q. scores of 85.5; those placed with middle-class families had average scores of 92. The average I.Q. scores of youngsters placed in well-to-do homes climbed more than 20 points, to 98 - a jump from borderline retardation to a whisker below average. That is a huge difference - a person with an I.Q. of 77 couldn't explain the rules of baseball, while an individual with a 98 I.Q. could actually manage a baseball team - and it can only be explained by pointing to variations in family circumstances.

    One has to do some digging elsewhere, but the study he is referring to is from PNAS. Since the study is free to read, I'll only go over the major points:

  • The research team picked children/teenagers who met five criteria: They had all been (1) neglected and/or abused during infancy, having been definitively removed from their biological family by court order after judicial procedures; (2) placed in many foster families and/or institutions before adoptive placement; (3) had an IQ <86 and >60 in the year preceding adoptive placement; (iv) aged 4-6 at the time of the adoptive placement and (v) aged 11-18 and being raised by the same two adoptive parents at the time of the second IQ test (i.e., 5-14 years with same adoptive parents)

  • The original IQ test was either a) the first or second edition of the Stanford Binet (their wording and reference date makes it hard to discern), b) the original Bayley Scale of Infant Development, or c) "other French tests of Intelligence". The post-adoption IQ was either the WAIS or the WISC-R.

  • The Children/teenagers were classified into one of three groups based on adoptive father's occupation: High, Middle, or Low

  • There was a universal gain in IQ scores for all participants, averaging 13.88 points.

  • There was a (range restriction corrected) correlation of 0.67 between IQs before and after adoption, which "indicates a degree of stability close to the stability found in longitudinal studies of biological children who have not undergone an environmental change. . . Thus, on the basis of IQ at the end of the preschool period, the results show that there is a moderate stability for rank. This is a near-universal finding.

  • As to a critique, first the selection criteria is loaded. No one with any sense about them would think that being abused/neglected from birth so bad that the legal system has to intervene would not decrease cognitive functioning. This is not at all what is meant when discussing "Average Expectable Environment." So right off the bat we know that these kids are not at all the same as a regular Joe (or Jane) off the street who has Borderline IQ. Thus, right off the bat we would expect that after any semblance of stability, the kids' cognitive functioning would improve (i.e., return to its normal state).

    Second, while there is some difference between the IQs across groups, it isn't major (max: Low SES PIQ vs High SES PIQ, d = .9 [r2=.17]; min: Low SES VIQ vs. High SES VIQ, d=.5 [r2=.06]), indicating SES explains somewhere between 6 and 17 percent of the variance in post-adoption IQ scores. While there is definitely an effect, given the extreme nature of the study's categories, it does not appear to be much of one.

    Third, it is very difficult to say with much certainty that IQ scores from scale, developed with one set of theoretical guidelines and one set of 1950s norms are directly comparable other IQ scores from a different scale, developed with a totally different theoretical orientation that uses a set of 1970s/1980s norms. This is even more so when the Bayley scale is used. For those of you that have given the Bayley (all 2 of you!), you know it is an extreme pain to administer and that it is about as unlike the Wechsler instruments as you can find a test that still measures cognitive ability.

    Fourth, while there was an overall mean increase in IQ, in addition, there was a universal gain in variance (stnd. dev. increased approx 9 IQ points). Thus, assuming that the before and after IQs were directly comparable, the majority of the scores fell into these ranges:

    95% Range of IQ (i.e., -/+ 2 SD)
    Low SES64.2391.4350.54118.54
    High SES64.9790.1768.81272

    A few things to notice. 1. There were at least some kids in both categories who, pre-adoption, were in the average range. Given the stark conditions of their pre-adoption upbringing, that is amazing - and likely has little to due with a nurturing environment. 2. There were at least some kids in both categories who, post-adoption, were in the Mild to Moderately Deficient range. Meaning - they still were classified as MR. If the environment is so powerful, why didn't their IQ rise also? 3. It is a fair assumption that even the Low SES households were much better off for the kids than their abusive/neglectful original home. If the environment is so powerful, then why are some children doing worse, IQ-wise, post-adoption?

    Fifth, buried in the next to last paragraph of the study, we find this statement:
    This study shows that stability for rank can be found following a marked environmental change after 4 years of age regardless of the SES of adoptive families. The factors explaining this stability are undoubtedly different from those explaining the gains in mean IQs.

    Interpretation: the worst performers before the adoption tended to be the worst performers after the adoption. The only thing that happened was a linear transformation of scores (i.e., post-score ≈ a*pre-score + b) across all SES groups, which could be due to many things, including: changing IQ scales and/or many years of living in an "average expectable environment."

    So, what we end up with is a picture much more complex and intricate than Kirp allows. What we can definitely conclude is that abusing or neglecting your children so bad that the government has to take them away tends to produce lower IQs (although some kids will still be in the average range), but was this really in question? What we can also conclude is that for some children in the MR range, adoption into high SES environments will not significnalty improve thier IQ. If the environment is so powerful to change cognitive abilities, then how did this happen? Last, I think we can conclude that Kirp's attempt to glorify the ability of the environment to change IQ is not much more than the wheel, reinvented.

    [1.] I did it using the bootstrap feature in AMOS. I can go into more detail by email or in the comment box.