PRETEND GENES 63
ing opportunistically rather than by careful, methodological design. In
addition, the “fi eld” nature of studies with children and parents has re-
quired improvised testing with a vast variety of shortened mea sures and
ad hoc scales. Th ese methods have sacrifi ced precision for expedience.
Typical of the approach is the large- scale Twins Early Development
Study used by Robert Plomin and colleagues to advise us on genes and
human potential. It is based on the use of shortened tests (shortened from
their standard forms) variously administered by telephone, by post, over
the internet, and by parents in the home. Reassurance about the validity of
such scales is off ered by correlations with more standard tests administered
to small subsamples of the children. But the modest level of these, mostly in
the range 0.5–0.6, in fact indicates considerable unreliability. Th ese values
mean that 75 percent of the time, the test may measure diff e r e n t t h i n g s.
Indeed, the choice of test has oft en been based on con ve nience rather
than princi ples of empirical precision. So disparate are the tests in some
studies that it is common practice to estimate a “general factor (intel-
ligence, or g) statistically from pooled results. Th is practice is known
as meta- analy sis and assumes that, because scores on diff er ent tests
intercorrelate to some extent, the resulting score is necessarily one of
intelligence: another indication of an inexact science proposing exact
conclusions.
However, the statistical common factor may well not mea sure the same
“ thing” and may not even be cognitive in origin (see chapter 3). As Kevin
Murphy says, the assumption that these mea sures, with disparate proper-
ties, distributions, and so on, can be combined as if they describe a single
uniform variable can lead to serious prob lems in the meta- analy sis, in-
cluding “lack of clarity in what population pa ram e ter is being estimated.”
Again, this is not a basis for strong conclusions about anything.^32
TRIUMPHS OF THE TWIN METHOD:
BIZARRE HERITABILITIESTh ese fl aws of twin studies prob ably explain why nearly all human traits
investigated by the twin method have yielded substantial heritability es-
timates. Here are some individual diff erences that, according to twin
studies, are largely or mainly due to ge ne tic diff erences:
This content downloaded from 139.184.14.159 on Tue, 17 Oct 2017 13:51:26 UTC