The Flynn Effect is a funny mixture: part IQ inflation, part civilizational advantage. It seems that ever since intelligence testing began, people have been getting smarter. That is not too surprising. The first world war partly destroyed Europe, but the slaughter caused by the Gatling gun was matched by the ability to bring food across the world, preserved in tin cans and refrigerated ships. Nutrition improved. The second world war brought destruction world wide, but also spurred innovation. Since 1945 the world has got richer. More of the world was exposed to better nutrition, health and education. Since the 1990s the world has got healthier and better educated and even richer. Complain as we might, most people on the planet are living better lives than ever before. All this should boost the human condition, and as we become more civilised, both mind and body improve. Test scores record this improvement. This is the rosy view of rising intelligence.
The cautious view is that modern life sets us more IQ type questions, and problems which seemed out of the ordinary in the early part of the 20th Century soon became commonplace, because of better and more widespread education. IQ tests have become less puzzling, less of a test, because they look like familiar items in regular school tests. Therefore, the apparent gains are hollow, and there is no real life proof that we are getting as much brighter as the tests suggest. We are suffering IQ inflation, not a real increase in intellectual wealth.
The contrary view is that if you use proper tests: reaction times, digit span, colour perception, then there is slight but unmistakeable evidence of mental decay, and falling real intelligence.
And now to Germany.
A reversal of the Flynn effect for spatial perception in German-speaking countries: Evidence from a cross-temporal IRT-based meta-analysis (1977–2014) Jakob Pietschnig and Georg Gittler. Intelligence 53 (2015) 145–153.
In their abstract Pietschnig and Gittler say:
Generational IQ changes (the Flynn effect) have been shown to be predominantly positive but differentiated according to IQ domains and countries. However, evidence from recent studies points towards a decrease of the Flynn effect globally or even a reversal in some countries. In the present meta-analysis, we show an inverse u-shaped trajectory of IQ test performance changes in a large number of samples (k = 96; N = 13,172) on a well known test for spatial perception (the three-dimensional cubes test, 3DC) in German-speaking countries over 38 years (1977–2014). Assessment of both item response theory-based measures as well as more standard measures of classical test theory showed initial increases and a subsequent decrease of performance when controlling for age, sample type (general population vs. mixed samples vs. university students) and sex. Our results suggest saturation and diminishing returns of IQ increasing factors (e.g., life history speed) whilst negative associations of IQ changes with psychometric g may have led to the observed IQ score decrease in more recent years.
Here are the u-shaped historical trends, shown either as ability parameters or percentage of solved items, both giving the same overall picture.
So, has Germany abolished itself?
The authors give some background:
Worldwide IQ gains have been shown to have started to slow down in the past decades across all domains. This has been interpreted as a potential predecessor for a stagnation or even reversal of gains in the future (Pietschnig & Voracek, 2015). In fact, stagnation of gains has been reported in data from Norway and Sweden (Emanuelsson, Reuterberg, & Svensson, 1993; Sundet, Barlaug, & Torjussen, 2004). More recent findings even indicate a reversal of IQ gains in Denmark, Finland, and France in past decades (Dutton & Lynn, 2013, 2015; Teasdale & Owen, 2005). Interestingly, declines in reaction times over more than one century indicate that contrary to the frequently observed IQ gains, generational changes in psychometric g may have been in fact negative in Western countries (Woodley, te Nijenhuis, & Murphy, 2013, 2014; Woodley of Menie, te Nijenhuis, & Murphy, 2015). Therefore, it seems likely that IQ changes in further countries may eventually show a similar trajectory of an initial stagnation and subsequent decrease of IQ test performance
They go on to make a crucial point:
IRT-based estimates [are] more accurate than conventional raw scores within the framework of classical test theory, where each correct item solution is equivalent to one raw score point, regardless of the item’s difficulty.
We investigate changes in spatial IQ performance of the general population on the three-dimensional cubes test (3DC) over a time-span of 38 years in German speaking countries. In our main analysis, we focus on changes of IRT based mean person parameters of samples and support our findings by supplementary analyses of estimates derived from classical test theory. Moreover, we investigate influences of sample age, sample type (general population vs. mixed samples vs. university students), and sex on spatial task performance.
In contrast to many cognitive ability measures, item difficulty of the 3DC is not progressively increasing throughout the test, but rather easier and more difficult items alternate
In all, 96 independent samples (N = 13,172) from 76 published and unpublished studies were included in the present meta-analysis. Participants in primary studies were predominantly from healthy convenience samples in schools, Universities, and from the general population.
Here is the key result:
Results from our linear regression analyses suggest decreases of about 4.8 IQ points per decade when controlling for age, sample type, and sex, thus indicating a substantial negative Flynn effect that is even stronger compared to previously observed positive trends (e.g., Flynn, 1984, 1987; Pietschnig & Voracek, 2015). This trend was observed in linear regression analyses, but our results showed that the present changes over time may be even better described as a curvilinear function, thus indicating initial increases, followed by stagnation (with performance peaking around the mid-1990s), and subsequent decreases of task performance. This curvilinear relationship emerged for both IRT- as well as non-IRT-based measures and remained robust when controlling for age, sample type, and sex, thus corroborating stability of our results and further corroborating validity of the Rasch model in our samples.
The fall in intelligence is quite severe. The authors soberly go through a list of possible explanations for IQ having been boosted by cultural changes but then subject to diminishing returns, finally tentatively landing on:
One possible reason may be that the above discussed IQ-boosting factors have masked the g-based ability decrease until a point of saturation was reached. If this is so, the zenith of our curvilinear regression would approximately give the point where beneficial factors were eventually outperformed by the negative trend in g.
If life is dandy, particularly in wealthy Germany, workshop of the world, what could be causing a drop in intelligence? The authors comment:
Other potential reasons that have been cited for decreasing IQ test scores such as changing population ability levels due to non-Western immigration (e.g., Rindermann & Thompson, 2014) or dysgenic fertility patterns (e.g., Lynn, 2011) do not seem suitable to contribute substantially to our present findings. On the one hand, effects of immigration on national IQ levels were observed not to be long-lasting with performance gaps diminishing over time (te Nijenhuis, de Jong, Evers, & van der Flier, 2004). On the other hand, both immigration and dysgenic fertility effects have been shown to be too small to provide substantial contributions for our present findings (Meisenberg & Kaul, 2010; Rindermann & Thompson, 2014).
Comment: Rindermann & Thompson (2014) calculate that, on average, the mean natives’ and immigrants’ competence gap is equivalent to 4.71 IQ points, which by coincidence is precisely the drop experienced in Germany. How much of an effect do immigrants contribute?
(By way of a brief aside, as of 2012 Germany was 80% German. Another 3.9% are European, but that includes a substantial 1.9% Polish (IQ92) 1% Italian (IQ97) 0.5% Romanian (IQ91), 0.5% Greek (IQ92) and 2.9% others Spanish(97), Croatian (99), Dutch, Portuguese and Austrians. Many but not all are significantly below the German mean of IQ99. 3.7% of the population are Turkish (IQ88) and this should have a bigger effect.
Here is the entry for Germany in our 2014 paper:
Natives constitute 81% of the population, their competence is (IQ) 100.99
Migrants constitute 19% of the population, their competence is (IQ) 92.75
Competence difference –8.26 IQ points
This certainly looks like it should push Germany back a bit, but at only one fifth of the population the migrant effect is still relatively slight and takes the national mean down only to 99.4. However, if we plot migration numbers by year it might match some of the effect, so it could be a partial contributor, though far from the whole story. Dysgenic trends might be a further contributor, so we can follow that argument by seeing what Woodley makes of a similar drop in ability in France, which I hope to post about shortly.
The authors’ final words are: This may most likely be due to saturation and diminishing returns of IQ boosting factors (e.g., life history speed) and a manifestation of declining psychometric g.
Comment: Saturation and diminishing returns, as the authors make clear, would result in improvement being sustained at a plateau. In fact, in these results abilities have fallen as fast and as far as they had previously risen, which is an alarming finding. All that could explain that (if immigration is taken out of the picture) is declining psychometric g, which presents Germany with a substantial German problem. This is particularly sobering because the sample is younger than the general population (84% are below 30 years of age) so the young generation is duller than the older, not the usual pattern. (Incidentally, it makes me wonder whether immigrants, with greater family size are contributing more to this younger population).
Here is a testable prediction: if Germany takes in the projected numbers of Middle East immigrants, then that will boost the immigrant origins population by at least 1 million. Once their families join them the number could be 3 or 4 million. In all probability this will have a negative effect on German national intelligence levels. (It makes no difference whether they are from Syria or Pakistan, since roughly IQ 83 is expected from either). However, we do not need to wait very long to test all this. If a German researcher is quick about it, they could pop round the reception centres and give the three dimensional cube test to as many new immigrants as possible, and we could have the answer in a few months. Indeed, giving a few more intelligence tests would be predictive for training and occupational placement purposes. A thesis in it for someone?