Friday, 29 January 2016

Vita brevis, intelligentia longa

Yesterday, at a lakeside party, Bob told me about the evening he drove up to an isolated beachside hotel in Guatemala, some four decades ago, and made his way through the deserted lobby to talk to the sole member of staff, a waiter with a skew-whiff bow tie. “Have I time to swim in the sea before dinner?” Bob asked. The waiter looked at him calmly and said “Tenemos mas tiempo que vida” (we have more Time than Life).

It is curious that there should be any correlation between intelligence and longevity. The contemporary fashionable view of intelligence is that it is a creature of priviledge, a confection of schooling and private tutors, granting acess to good jobs for those who can manipulate a narrow range of logical symbols, no better grounded in real ability than the recitation of classic verse or the oratorical flourishes of rhetoric. Such hothouse flowers bloom in sheltered spaces, and have nothing else to commend them. If they live longer, it is because they dine well and sleep in feathered beds. If this is remotely true, then “correcting” for social class should cancel out any effects of upbringing and living circumstances on lifespan.

To the contrary, much research (Der, Batty, & Deary, 2009; Deary, Whalley, & Starr, 2009; Gottfredson, 2004) finds that measures of intelligence taken in early childhood are good predictors of lifespan, even when social class is taken into account.

Intelligence and Early Life Mortality: Findings From a Longitudinal Sample of Youth

Kevin M. Beaver, Joseph A. Schwartz, Eric J. Connolly,Mohammed Said Al-Ghamdi, Ahmed Nezar Kobeisy, J. C. Barnes & Brian B. Boutwell

DOI: 10.1080/07481187.2015.1137994


The current study examined whether adolescent IQ predicted risk for mortality by the age of 32. Analyses of data from the Add Health revealed that IQ was related to mortality risk, such that respondents with relatively lower IQs were significantly more likely to experience early life mortality when compared with respondents with comparatively higher IQs. This association remained statistically significant even after controlling for a host of covariates such as race, gender, involvement in violent behaviors, levels of self-control, and poverty. The average IQ of deceased respondents was approximately 95 while the average IQ of living respondents was about 100.

Persons with comparatively lower IQ scores have been found, for instance, to be more likely to engage in risky behaviors that have been shown to compromise short- and long-term health (Gottfredson & Deary, 2004). Additionally, research findings have revealed significant direct associations between IQ and a number of health outcomes, including asthma, depression, high cholesterol, and tumor growth to name just a few (Der et al., 2009). Beyond these associations with health outcomes, IQ also appears to be related to the way in which individuals respond to medical advice and directions. To illustrate, once diagnosed with a health-debilitating disorder, individuals with lower levels of intelligence are less likely to take prescription medications as instructed and are less likely to schedule follow-up appointments compared to those with higher levels of intelligence (Gottfredson & Deary, 2004). Taken together, these findings from the cognitive epidemiological literature point to IQ as one of the most important factors that is connected with overall health (Deary, 2009).

Data for this study were drawn from the National Longitudinal Study of Adolescent to Adult Health (Add Health; Udry, 2003) a four-wave prospective study that was initially based on a nationally representative sample of American youth. The sampling frame consisted of all high schools with an 11th grade class and that had an enrollment of at least thirty students. A systematic random sample of these schools was then selected with the end result being the inclusion of 80 schools. The schools were stratified based on region, school type, percentage white, and urbanicity. The largest feeder school for each of these 80 schools was then selected to be included in the study. With this sampling procedure in place, there were a total of 132 schools that were retained. The first wave of data was collected in 1994–1995 when in-school surveys were administered to students who were in attendance at these middle or high schools on a specified day. In addition, an in-home component to wave 1 data was also included when 20,745 youth were selected to be re-interviewed in their homes along with their primary caregiver. The second wave of data was collected approximately 1.5 years later when 14,738 of the original respondents completed the survey instrument. Nearly seven years after the wave 1 data were originally collected, the third round of surveys were administered to 15,197 participants when the majority of the respondents were between the ages of 18 and 26 years old. The fourth and final wave of data was collected in 2008–2009when most of the 15,701 respondents were between the ages of 24 and 32 years of age and 50.5% of the sample was female. Overall, the Add Health data span approximately 14 years of human development (Harris, 2009; Harris et al., 2003).

At wave 3, participants were administered the Picture Vocabulary Test (PVT), which is a shortened version of the Peabody Picture Vocabulary Test-Revised (PPVT-R). The PVT is designed to measure individual variation in verbal skills and receptive vocabulary.




So, even at the early age of 32, a 5 IQ point handicap is making  difference between life and death.

First, and in line with existing research, there was a negative and statistically significant association between IQ scores and the odds of mortality. This significant association was detected in a bivariate rare-events binary logistic regression model as well as in the more fully specified models that accounted for the confounding effects of age, gender, race, involvement in violent behavior, levels of self-control, and poverty. Taken together, these findings suggest that lower IQ scores in adolescence are related to an increased risk of mortality in late adolescence and early adulthood.

The second main finding to emerge from the analyses was that, prior to including controls for violence and self-control, African Americans were about 2.61 times more likely to have experienced death by early adulthood relative to other races. This association was expected as previous research has revealed African Americans—as a group—are at increased risk for unhealthy outcomes and premature death and that their mean life expectancy is lower than that of other races (Crimmins & Saito, 2001). What was particularly surprising about the analyses, however, was that that the influence of race was no longer statistically significant in the fully specified model that accounted not only for IQ, but also for involvement in violent behaviors and for levels of self-control. These findings converge with those focusing on other phenotypes, wherein the effects of race can be fully accounted for when including a complete list of covariates (Beaver et al., 2013; Wright et al., 2014), some of which may serve as mediators.

Nonetheless, the authors are very cautious about intelligence being causal, particularly through a shared genetic pathway, though that is likely from other research. The short Peabody test was given when respondents were 18-26, which is old enough for other things to have influenced their intelligence (though this is probably not a big factor, but cannot be discounted). Moral: test intelligence early, at 4 years of age before kids go to school. That is already a very predictive score. The authors are happy with the Peabody test, saying it correlates well with other kid’s tests. I don’t question that, but I have the feeling it is insensitive for the higher ranges, so the correlations found here may be under-estimates of the real effect. Of course, happily for the subjects, few of them have died, but their good fortune makes it hard for the researchers to be sure of their findings. I am already pretty sure, because it fits in with other findings, but they, quite properly, cannot be.

Read it all here:

In my view this is another finding to strengthen Deary’s “system integrity” hypothesis. In genetic terms, whatever makes us bright makes us healthier and longer-lived.


  1. Regarding blood cholesterol level.

    Higher IQ, higher cholesterol

    Lower IQ, lower cholesterol


    1. " Additionally, research findings have revealed significant direct associations between IQ and a number of health outcomes, including asthma, depression, high cholesterol, and tumor growth to name just a few (Der et al., 2009)."

      You present a contradictory statement here. So who has the truth?


    2. High cholesterol isn't a health outcome, it's merely a measurement.


      "Correlation of IQ with different blood biochemical parameters shows a positive relationship with total cholesterol, high density lipoprotein cholesterol, low density lipoprotein cholesterol, very low density lipoprotein cholesterol and triglycerides"

  2. “Tenemos mas tiempo que vida” (we have more Time than Life).

    Great quote--I'm going to use this....

  3. About correlations, doc. (i) There is a small positive correlation between IQ and height. (ii) There is, you tell us, a positive correlation between IQ and lifespan. So (iii) you might expect long-lived men to be taller than the average for their generation. But my casual observation is that it's the shortarses who live forever, not the men of stature.

    Is my casual observation wrong, or is it right and mildly puzzling?

    1. I think your casual observation is right and mildly puzzling and would also be interested in Dr. Thompson's take on this.

      Here is the correlation pattern (these correlation coefficients are examples, not definitive) for height/IQ/lifespan:
      IQ/lifespan - r = .12 (95% confidence interval .06 to .18) from
      height/lifespan - r = –0.18 to –0.33 quoted in
      height/IQ - phenotypic r = 0.16 and genotypic r = 0.28 from
      Razib discusses height/IQ at

      I have seen this pattern of correlations between three variables (one opposite signed) before and have always been intrigued. Are there any methodologies commonly used in this case? I would think a scatterplot matrix or 3D scatterplot of the variables might be illuminating.

      Can anyone think of other variables (either phenotypic or genetic) which might influence these relationships and be suitable for controlling for statistically? Perhaps measuring genetic load as described in ? Nutrition is clearly important, but I'm not sure how to evaluate in this context. One of my pet hypotheses is that macronutrition (calories/protein, perhaps also calcium and phosphorus) is more important for height with micronutrition (e.g. omega 3 fatty acids, phospholipids) more important for IQ.

      Fun quote on the height/longevity/IQ topic from Slate at
      "Unlike intelligence, which has a merely coincidental relationship with height, there are plausible biological explanations for why short people live longer." ;-)

      This quote from Samaras above was also interesting: "In addition, mortality studies can suffer from
      a crossover problem. For example, a large study that tracked mortality beyond 85 years
      of age finds that below 70 years of age, tall (> 183 cm) men have the lowest mortality,
      but between 70 and 85 years of age, shorter men (from 170 to 183 cm) have the lowest

    2. My preliminary impression is that this requires an entire post. On a minor point, predictive patterns for longevity change at about 75 because adventitious biological events increase, and one is dealing with a select population of "normally long-lived" persons. The predictive power of IQ is less strong at those advanced ages. This from my fallible memory, based on recollections of the Edinburgh work on ageing.

    3. When the correlation factors in the same direction, you divide them. When the correlation in opposite direction, you multiply them. (basic statistic concept).

      IQ vs height and IQ vs lifespan are in the same direction. r should be divided. IQ vs lifespan and height vs lifespan are in opposite direction, their r should multiply. I did calculation long time ago. I remember IQ vs lifespan factor is stronger over height vs lifespan factor. In other words, IQ correlation with lifespan is stronger over IQ correlation with height. Thus when they are in conflict, stronger one win.

      So all these r have different strength.


    4. Thanks for the reply Dr. Thompson. Your observation about the change at 75 sounds similar to the Samaras quote I gave. The observational data seems fairly well explained by early life mortality being related to IQ (which mediates giving the height correlation, another possibility being other height correlates like income) with height being directly relevant to post-75 mortality. I wonder if the literature (and data) agrees with that reasoning.

      I was unable to find the relevant (specific) Edinburgh research in a quick look. I'll try again later. If anyone else wants to investigate, this looks like a helpful link:

      P.S. If you do choose to do a post on this in the future I look forward to reading it.

  4. In my generation those of lesser intelligence were drafted into the infantry and sent to Vietnam.

  5. What ho, doc. Here's your homework for after you've got your tax return in.

    1. Saw it, but decided to go to the beach instead. Got the same lecture from Prof George Brown (social factors in psychological relapse) in the 80s, and conceded a fair bit of what he said, given that I was doing many multiple regression calculations then, and knew of their peculiarities.

    2. Thanks dearieme

      Really enjoy the article. Always interpret association very carefully with strong dose of skepticism. True data or evidence should be able to stand scrutiny from many different angles like correct math answer. Math reason is only rule in the universe that can be tested without any doubt.



    In modelling, it is assumed that except for the variables of interests other things being equal. It is the human who break that rules, as he said,

    "A doctor or an article has told him to take Vitamin E, so he does that, but he’s also the guy who’s watching his weight and his cholesterol, gets plenty of exercise, drinks alcohol in moderation, doesn’t smoke, has a high level of education, and a high income."

    Other things are not being equal. The researcher missed all that important factors.

  7. "In my view this is another finding to strengthen Deary’s “system integrity” hypothesis. In genetic terms, whatever makes us bright makes us healthier and longer-lived."

    Not always, though.
    Or that's what I gathered from Natural History of Ashkenazi Intelligence (Cochran, Hardy, Harpending).

    "As with any regime of strong directional selection on a quantitative trait, genetic variants that were otherwise fitness reducing (ma che aumentavano l'intelligenza) rose in frequency. In particular we propose that the well-known clusters of Ashkenazi genetic diseases, the sphingolipid cluster and the DNA repair cluster in particular, increase intelligence in heterozygotes. Other Ashkenazi disorders are known to increase intelligence."

  8. Yes, some benefits come with costs. Ashkenazi intelligence at the cost of neurological diseases, surviving malaria at the cost of sickle cell anemia. Worth it in both cases, considering the alternative.