[tt] Flynn is bunk

[Anders Sandberg]

Turns out that the main source that Flynn is bunk is... James R. Flynn!

Eugen Leitl wrote:
>
> (ok, it Guardian, but still)
> http://www.guardian.co.uk/education/2008/oct/27/teenagers-less-bright

The paper that led to the Guardian article is likely this,
http://www.ingentaconnect.com/content/bpsoc/bjep/2008/00000078/00000001/art00001
which is a review of intelligence in educational psychology. The relevant
section of the paper is this:

"The Flynn effect

Flynn (1987, 1994), surveying data on the standardization and
re-standardization of various psychometric tests from the late 1940s on,
reported gains of about 9 points per generation for tests of crystallized
intelligence, and 15 points or one standard deviation for tests of fluid
intelligence (Raven’s matrices). This general apparent improvement has
entered the literature as ‘the Flynn effect’.

In a later review, Flynn (1998), examining evidence going back to the 19th
century, had become sceptical of claims that this meant that people’s
intelligence – children or adults – was improving generation by
generation.
He wondered whether the concept of ‘intelligence’ was still viable.
Certainly, people’s test-taking ability has improved; but after
discounting as minor factors such as nutrition, urbanization, and TV and
pointing out that children’s gains on arithmetic reasoning, vocabulary,
creativity and speed of learning were far less, he was tempted to suggest
that school and environmental stimulus in some undefined way were
improving children’s
decontextualized problem-solving skills.

Yet, this may be only a historic effect: Grissmer, Williamson, Kirby, and
Berends (1998, pp. 262–265) argue that there is some evidence suggesting
that gains have been levelling off from the late 1980s. In this country,
the NFER-Nelson CAT test showed no change in norms between the 1984 and
2000 standardizations. Flynn (2007) wonders whether IQ gains will cease in
all highly industrialized nations, citing Sundet, Barlaug, and Torjussen
(2004) as evidence that gains have ceased in Norway; and Emanuelsson,
Reuterberg, and Swenson (1993) and Teasdale and Owen (2000) that gains
have flattened since the late 1980s in Sweden and Denmark, respectively.
Flynn argues that in the Scandinavian countries the whole population has
by now adjusted to 20th century industrialization and its implications.

Yet Flynn (2007) also shows, when the subtests of arithmetic, information,
and vocabulary in the WISC are examined, between 1947 and 2002 children
gained only an average of 3 IQ points when compared with a change on
Raven’s of 27.5 points. If ‘real intelligence’ is that which enables
school learning, children’s intelligence has hardly changed at all over 55
years. An alternative explanation that the schools have got that much
worse in teaching the 3Rs seems hardly likely.

The Flynn effect is a major scandal for the psychometric programme – while
50 years of sophisticated work of measurement was going on in the world of
psychometrics their ruler was made of rubber, and stretched over a
standard deviation (nearly two on tests of fluid intelligence). Moreover,
different measures have been sliding by different proportions.

Contrary evidence in Shayer et al. (2007) on some 10,000 Year 7 pupils on
the criterion-referenced CSMS Piagetian test Volume & Heaviness – a test
of conservations and density – showed that between 1975 and 2003 there had
been a drop in performance of 1.04 SD for boys, and 0.55 SD for girls, and
that an initial boy–girl differential of 0.5 SD had completely
disappeared. If Flynn is right in saying that the change in the general
environment is responsible for the gains in decontextualized
problem-solving skills for psychometric tests, then the general
environment change has been very unfavourable for boys (and not all that
favourable for girls either) on Piagetian measures, at least those
concerning conservations and density. Both boys and girls were down, also,
on the 3 items testing early formal performance on volume/mass relations.

Flynn (2007) also cites British evidence that from 1990 to 2003, British
children lost 0.4 SDs on the WISC Arithmetic subtest – a rate, he argued,
which is identical to the rate of decline shown in Shayer et al. (2007)
for the Piagetian conservations."



-- 
Anders Sandberg,
Future of Humanity Institute
Philosophy Faculty of Oxford University