Group differences in the heritability of items and test scores

J.M. Wicherts, W. Johnson

Research output: Contribution to journalArticleScientificpeer-review

Abstract

It is important to understand potential sources of group differences in the heritability of intelligence test scores. On the basis of a basic item response model we argue that heritabilities which are based on dichotomous item scores normally do not generalize from one sample to the next. If groups differ in mean ability, the functioning of items at different ability levels may result in group differences in the heritability of items, even when these items function equivalently across groups and the heritability of the underlying ability is equal across groups. We illustrate this graphically, by computer simulation, and by focusing on several problems associated with a recent study by Rushton et al. who argued that the heritability estimates of items of Raven's Progressive Matrices test in North-American twin samples generalized to other population groups, and hence that the population group differences on this test of general mental ability (or intelligence) had a substantial genetic component. Our results show that item heritabilities are strongly dependent on the group on which the heritabilities were based. Rushton et al.'s results were artefactual and do not speak to the nature of population group differences in intelligence test performance.
Original languageEnglish
Pages (from-to)2675-2683
Number of pages9
JournalProceedings of the Royal Society B: Biological Sciences
Volume276
Issue number1667
DOIs
Publication statusPublished - 2009
Externally publishedYes

Fingerprint

heritability
Computer simulation
Crows
Computer Simulation
test
computer simulation
matrix

Cite this

@article{eb7141a1f7074bc797f7c414d0c66044,
title = "Group differences in the heritability of items and test scores",
abstract = "It is important to understand potential sources of group differences in the heritability of intelligence test scores. On the basis of a basic item response model we argue that heritabilities which are based on dichotomous item scores normally do not generalize from one sample to the next. If groups differ in mean ability, the functioning of items at different ability levels may result in group differences in the heritability of items, even when these items function equivalently across groups and the heritability of the underlying ability is equal across groups. We illustrate this graphically, by computer simulation, and by focusing on several problems associated with a recent study by Rushton et al. who argued that the heritability estimates of items of Raven's Progressive Matrices test in North-American twin samples generalized to other population groups, and hence that the population group differences on this test of general mental ability (or intelligence) had a substantial genetic component. Our results show that item heritabilities are strongly dependent on the group on which the heritabilities were based. Rushton et al.'s results were artefactual and do not speak to the nature of population group differences in intelligence test performance.",
author = "J.M. Wicherts and W. Johnson",
year = "2009",
doi = "10.1098/rspb.2009.0238",
language = "English",
volume = "276",
pages = "2675--2683",
journal = "Proceedings of the Royal Society B: Biological Sciences",
issn = "0962-8452",
publisher = "Royal Society of London",
number = "1667",

}

Group differences in the heritability of items and test scores. / Wicherts, J.M.; Johnson, W.

In: Proceedings of the Royal Society B: Biological Sciences, Vol. 276, No. 1667, 2009, p. 2675-2683.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - Group differences in the heritability of items and test scores

AU - Wicherts, J.M.

AU - Johnson, W.

PY - 2009

Y1 - 2009

N2 - It is important to understand potential sources of group differences in the heritability of intelligence test scores. On the basis of a basic item response model we argue that heritabilities which are based on dichotomous item scores normally do not generalize from one sample to the next. If groups differ in mean ability, the functioning of items at different ability levels may result in group differences in the heritability of items, even when these items function equivalently across groups and the heritability of the underlying ability is equal across groups. We illustrate this graphically, by computer simulation, and by focusing on several problems associated with a recent study by Rushton et al. who argued that the heritability estimates of items of Raven's Progressive Matrices test in North-American twin samples generalized to other population groups, and hence that the population group differences on this test of general mental ability (or intelligence) had a substantial genetic component. Our results show that item heritabilities are strongly dependent on the group on which the heritabilities were based. Rushton et al.'s results were artefactual and do not speak to the nature of population group differences in intelligence test performance.

AB - It is important to understand potential sources of group differences in the heritability of intelligence test scores. On the basis of a basic item response model we argue that heritabilities which are based on dichotomous item scores normally do not generalize from one sample to the next. If groups differ in mean ability, the functioning of items at different ability levels may result in group differences in the heritability of items, even when these items function equivalently across groups and the heritability of the underlying ability is equal across groups. We illustrate this graphically, by computer simulation, and by focusing on several problems associated with a recent study by Rushton et al. who argued that the heritability estimates of items of Raven's Progressive Matrices test in North-American twin samples generalized to other population groups, and hence that the population group differences on this test of general mental ability (or intelligence) had a substantial genetic component. Our results show that item heritabilities are strongly dependent on the group on which the heritabilities were based. Rushton et al.'s results were artefactual and do not speak to the nature of population group differences in intelligence test performance.

U2 - 10.1098/rspb.2009.0238

DO - 10.1098/rspb.2009.0238

M3 - Article

VL - 276

SP - 2675

EP - 2683

JO - Proceedings of the Royal Society B: Biological Sciences

JF - Proceedings of the Royal Society B: Biological Sciences

SN - 0962-8452

IS - 1667

ER -