An example of the ‘Mensa fallacy’—using a pathologically self-selected self-diagnosed sample—would be Karpinski et al 2018,
“High intelligence: A risk factor for psychological and physiological overexcitabilities”, which takes Mensa survey results at face-value while ignoring the fact that Mensa has attracted losers since its founding (a fact that I know was pointed out to the authors well before publication, and which they defend merely by saying that self-report data is common in many other areas while ignoring all contradictory evidence, which they pass over in silence).
The results are a priori unlikely as all population samples show that good things increase (eg.
Brown et al 2021), and dysfunctionality & mental illness rates drop steeply with increasing IQ definitely at least to the top decile (often the top percentile) with no quadratic or other curve observable indicating asymptotes or impending reversals; these results are so large and well known that Karpinski et al cannot doubt them, and so they attempt to ‘save the appearances’ with ad hoc invocations of nonlinear thresholds at high intelligence, while still relying on the relatively low IQ threshold of Mensa membership. Their results are so absurd as to discredit any attempt to claim that a Mensa sample can tell us anything at all about high intelligence, as (with the exception of the modest allergies finding) they are completely inconsistent with
genetic & phenotypic correlations, almost impossible to reconcile with the universal life-expectancy/
SES/education/IQ/wealth/mental-health correlations observed everywhere in psychology/sociology/medicine, and non-self-selected high-IQ samples (whether
Terman,
SMPY, FLS, Munich Longitudinal Study, SET, HCES, Scottish survey, Scandinavian population registry-based,
UKBB etc)—including a self-reported Asperger’s relative risk of
223!
It’s unclear how these are even arithmetically reconcilable with the population estimates, as such large risk increases ought to push the averages way up at the top decile or percentiles, overwhelming the modest lower risks for non-Mensa-level individuals. Further, if any of the relative risks were true, higher intelligence would be one of the strongest risk factors ever discovered for illness, far exceeding the effects of minor things like smoking. If such relative risks were true of Mensans, who are merely ~+2.3SD (being generous & taking their 1% criterion at face-value), then the RRs of groups like SMPY, MIT, Nobelists, or Fields Medalists, who are 3–6SD, would be off the charts, and it would be difficult to so much as run a SMPY summer-camp without dealing with multiple suicide attempts or psychotic breaks, or find a single child who seemed at all socially-well-adjusted, or an eminent scientist who had not been institutionalized, or… Of course, this is not the case. No reported statistics from SMPY or other high-IQ samples
not suffering from self-selection into exaggerated self-diagnoses agree with this, and researchers & journalists who interact with SMPYers and similar high-IQ cohorts fail to mention that the entire cohort is nuttier than a Snickers bar, and often mention that the members defy stereotype by seeming quite healthy, well-adjusted, and happy. (The implications continue to go beyond that—considering just genetics, such pathology would force stabilizing selection, which we do not observe.)
So all Karpinski et al 2018 has to offer is a cautionary warning about GIGO: Mensa members are either remarkably selected for pathologies, or are not responding honestly (perhaps due to the trendiness of self-diagnosing autism as an excuse for failure).
