PRETEND INTELLIGENCE 93assimilating equally complex conceptual meanings, yet as an item in an
IQ test, some would be deemed correct and others wrong. Again the true
source of diff erences is really one of specifi c learning rather than general
cognitive strength.
Of course, IQ test items can look quite plausible: they have, aft er all,
been concocted in the minds of experts precisely because they look plau-
sible, not out of a theoretical model of cognitive functions. Particularly
popu lar, and widely respected, are analogical reasoning items like
“DIPLOMAT is to TACT as VIRTUOSO is to... ?” (and the word
“SKILL” has to be selected from four or fi ve alternatives presented).
Does failure on such items really signify little analogical reasoning
ability in general (or low g)? In a number of reports since the early 1990s,
Usha Goswami argued that failure on these items could simply arise from
lack of experience with the specifi c relations. Th e solution is, she suggests,
to design items based on relations equally familiar to all group(s) being
tested. Th en we will know that they are truly being tested for complexity
of cognition and not culturally related experience. When this is done,
indeed, it transpires that even very young children are capable of “com-
plex” analogical reasoning.^24
Th e diff erence is that those items in IQ tests have been selected because
they help produce the expected pattern of scores. A mere assertion of
complexity about IQ test items is not good enough.
CULTURE- FREE (LEARNING- FREE) TESTS?One attempted solution to these prob lems with verbal items has been to
devise nonverbal test items that are assumed to be free of prior learning.
To that end, the much- used Wechsler test includes fi ve per for mance
subtests. Th e Stanford- Binet does likewise. But the best example of such a
test is deemed to be the Raven’s Matrices (or just the “Raven”).
Th e IQ defender Arthur Jensen described the Raven as a test of “pure
g,” dependent only on powers of induction from the information pre-
sented, and having nothing to do with prior learning. Th e example in
fi gure 3.1 illustrates how a correlation or rule has to be induced across the
rows but is conditioned by other rules down the rows. Th e combination
This content downloaded from 139.184.14.159 on Tue, 17 Oct 2017 13:52:12 UTC