334
WANG, FISHER, SIEGAL, FALCK, CARLSON
With six known variances and covariances of the observed variables, the six
unknowns in the earlier equations can be solved:
f332srqponmlkjihgfed = Cov(yI,edcbaZYXWVUTS Y3)/COgfedcbaZYXWVU v(y I, Y2)
Vet) = V(Y2) - COV(Y2, Y3)/f332
V(SI) = V(YI) - Vet)
f321 = Cov(yI, Y2)/V(SI)
V(S2) = V(Y2) - [~ICOV(yI' Y2) + Vet)]
V(l;J) = V(Y3) - [f332COV(Y2, Y3) + Vet)]
Based on the definition of reliability, the reliabilities of the observed
variables at Time I, 2, and 3 are estimated as fOllowing:
_ V(SI)
PI - [V(SI) + Vet)]
- [132N(SI) + V(S2)]
P2 - [f3iN(SI) + V(S2) + Vet)]
rp~2rpiN(sl) + V(S2)] + V(S3)]
P3 = [f3~2[132N(sl) + V(S2)] + V(S3) + Vet)]
STRUCTURAL EQUATION MODELING, 2(4), 335-366
Copyright © 1995, Lawrence ErIbaum Associates, Inc.CBA
Modeling Complexity and Difficulty
in M easures of Fluid Intelligence
Lazar Stankov
University of Sydney
Tenko Raykov
University of Melbourne
This article applies structural equation modeling to the simultaneous study of
task complexity and difficulty in measures of fluid intelligence. The interrela-
tion between 3 tests of fluid intelligence, 2 experimental tasks, and age is
examined. Each experimental task is decomposed into 4 subtasks, with the most
intellectually demanding ones being measures of fluid intelligence, and the
remaining 3 levels constructed to be gradually simpler. The findings suggest
that the experimental manipulations produced systematic changes in 3 sets of
parameters. These are: (a) the mean intercept parameters of the structural
equation model-interpreted as "pure test difficulties" reflecting task-specific
demands placed on elementary cognitive processes of the "same kind"; (b)
factor loadings on 2 subtask-specific narrow factors (SWAPS and TRIPLETS)
that reflect demands placed on processes of "similar kinds," that is, processes
common to variations within the same experimental task; and (c) factor loadings
on a fluid intelligence factor (Gf) that reflect demands placed on processes of
"diverse kinds," which are common to measures of a broad range of different
cognitive tasks. The results indicate that as task requirements become more
demanding, cognitive processes of the same kind, which are involved in solving
the easier subtasks, are no longer critical to performance, but the demand on
processes of similar and diverse kinds is increased. It is suggested that for
human subjects, increase in task complexity may be associated with lapses of
attention in the course of carrying out a series of interconnected steps.
Requests for reprints should be sent to Lazar Stankov, Department of Psychology, University
of Sydney, Sydney, New South Wales 2006, Australia.
.,.--.
"\