Saturday, December 03, 2005

ISIR blog # 18 - CHC Flynn effect study

Installment #18 of blogging "live" from the Internaltional Society for Intelligence Research (ISIR) 2005 conference in Albuquerque, NM.

Wicherts. Flynn effect in the Woodcock-Johnson Cognitive Ability and Achievement Tests, 1976-1999.

Program Abstract

  • There has been an extensive debate concerning the nature, causes, and implications of the secular increase of scores on cognitive ability tests (i.e., the Flynn Effect). The aim of this study is to compare the scores on unaltered subtests across the US standardization samples of the Woodcock-Johnson (WJ; Woodcock & Johnson, 1977), the Woodcock-Johnson-R (WJ-R; Woodcock & Johnson, 1989) and the Woodcock-Johnson-III (WJ-III; Woodcock, McGrew, & Mather, 2001). Structural equation modeling with mean structure is employed to shed light on the precise nature of the trend in scores on both the cognitive ability and achievement tests. In addition, with tests of measurement invariance (i.e., the absence of measurement bias) with respect to cohorts and periods, it is investigated whether score gains could be attributed to latent increases in ability or to measurement artifacts such as heightened test sophistication. Moreover, we consider trends over time with age held constant (e.g., 30-40 year-olds in 1976 vs. 2000), as well as differences between cohorts (e.g., those born in the 1950s vs. those born in the 1960s). Disentangling cohort effects from time-of-measurement effects may contribute to our understanding of the Flynn Effect, because the causes of these two types of effects are different. For instance, hypotheses concerning the effects of nutrition, family size, and heterosis are related only to cohorts, whereas most hypotheses concerning test artifacts are related to time-of-measurement.

Live Presentation Comments

  • Recent Dutch studies have demonstrated the importance of evaluating the comparability of samples compared - possible sample selection bias may confound Flynn effect findings.
  • Recent research by presentor has suggested that Flynn effect is more subtest (individual test) specific...and may be reflecting changes in specific/narrow abilites (test specificity) over time.

WJR/WJ III findings/comments

  • Advantage over prior research is use of contemporary CHC theory
  • Preliminary findings.....only on subset of cognitive tests. More will be analyzed later.
  • Used common items across editions. Rasch equated scales provided common metric across editions.
  • Author suggests 4 possible level sources of gain/change that need to be investigated (and which he is investigating): g, broad abilities, test specific abilities (Spearman's "s"), DIF/item drift. Presentor criticizes Flynn for dumping all these sources together....ignoring them. Presentor is using multiple group confirmatory factor analysis (MGCFA) in current research.
  • Presentor is using demographic variables from both editions to re-weight norm samples to control for sampling differences across time.

General findings.

  • MGCFA to investigate the 4 level sources above. Concluded -- way to much to summarize. I need to listen...sorry. Wicherts research results will be interesting and need to be monitored by serious intelligence scholars.
  • Flynn effect not present in children and adolescents (Flynn effect over?) but present in adult samples.
  • Mean difference changes (over time) are related to g and some due to changes in subtest specific/narrow and broad abilities.

Technorati tags:

No comments: