Skip to content

Don’t trust anyone over 62

Good morning! Let's start off the day with the latest research results in cognitive decline with age. They come from Annalise A. LaPlume, a postdoc at McGill University, who has collected scores from thousands of people who performed a battery of online tests designed to measure various types of cognitive ability. This was then sifted, sorted, and finally fed through a series of statistical meat grinders that take complex data as input and gradually make it simpler and simpler. Here's the final result:

Here's the deal. A month ago I turned 63. This means that you can expect the sharp political analysis on this blog to rapidly turn to mush over the next few years. Sorry about that. Just don't say you weren't warned.

63 thoughts on “Don’t trust anyone over 62

  1. Tbomber

    Just remember Kevin, it's not that you can't find your car keys but that you are holding them and don't know what they're for.

  2. typhoon

    I’m also 63….all I can say is it’s great that we have a 79-year old President who wants to stay in office until he is 86.

    1. KawSunflower

      We don't need a poll, survey or new Kevin Drum chart to answer that, do we?

      And as for how WORKING for Fox affects cognition, it seems that Lara Logan provides the most recent & compelling proof of that.

  3. MontyTheClipArtMongoose

    So, cognitive decline is why Antiwokeness Crusader Kevin Drum has not piled on the Cancellation of the Left for their SANTIFA arson of the FOXnews Christmas Tree? KD just couldn't remember to hit send on that post?

  4. J. Frank Parnell

    Not to worry Kevin. For many of us our sharp political criticism will also be turning to mush, so we won’t really notice.

  5. Steve_OH

    Did they measure cognitive ability, or was it actually cognitive give a shit?

    At 64, my cognitive ability is fine, but I've lost all give a shit.

    Looking at their raw data, the breakpoint is completely artificial. They used a segmented regression model just because, even though a more realistic nonlinear regression is more plausible. The dispersion is so large across the sampled population that almost any model is going to fit with roughly the same correlation coefficient.

    1. ScentOfViolets

      Why do you think 'nonlinear regression' is more plausible? More to the point, what do you object to in the raw data? IOW, are you saying, "Nuh-uh, the knee is at 63, not 62", or are you saying there is no increase in the slopes at all? If the latter, I'd certainly be entertained to hear your arguments.

      1. Steve_OH

        I'm saying that there is no knee, and that the slope increases smoothly and monotonically.

        Nonlinear regression is more reasonable because the data do show a nonlinear (but not piecewise linear) relationship. The existence of a knee is not supported by the raw data, and a knee would imply some sort of near-universal degeneration process based purely on calendar age, which is not supported in general.

        1. ScentOfViolets

          So you're saying that if the slope of a curve is, say, 1/2 over ten units, but then smoothly goes from 1/2 to 6 in, say, one unit there's no knee? I've never heard that one before. and the wiki seems to disagree with you as well.

          Your other point about a 'near-universal' degeneration process seems to be caveat-loaded and non-standard as well. The giveaway is that 'near-universal'. Finally, aging in and of itself is most generally -- dare I say it? -- a near-universal degeneration process based almost entirely on calendar age. That's just standard cant from the field of gerontology, and more generally, the aging process itself[1].

          I was, quite frankly, expecting a more sophisticated arguement.

          [1] Puberty, for example, is one such phenomenon. Very, very few (if any) people hit puberty at the age of five, or enter it at the age of thiry.

          1. Steve_OH

            Consider the function y = x^2 for x > 0. Its slope increases linearly. Where is the knee in that function? The concept of "knee" is meaningless here.

            Look at the sixth panel in Figure 2 in the paper (Letter-Number Alternation completion time). Where is the knee in that function? Note that the next panel (Letter-Number Alternation accuracy) does exhibit a knee. Using segmented linear regression is appropriate for that one measure, but is inappropriate for most of the others. The use of segmented regression is especially egregious in the case of the third and fourth (Stroop) graphs. Compare the curves in Figure 2 with the corresponding "fits" in Figure 3. The breakpoint in the "Total" graphs is almost entirely the result of the inappropriate application of segmented linear regression, not because of any actual breakpoint in the data.

            re puberty: The normal ranges for the various puberty milestones have a standard deviation of about one year, so the 95% confidence interval is around double that. That's a pretty wide range when you're 10 or 12 years old. But even if we were to proportionally broaden the range with age, we still wouldn't match the cognitive decline data. Take a look at Figure 3 in https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3637655/, for example. There are no knees. Cognitive decline occurs steadily and continuously over time.

            1. ScentOfViolets

              Um, you said this:

              I'm saying that there is no knee, and that the slope increases smoothly and monotonically

              I pointed you directly at a source that says otherwise. Do you concede that you were wrong? There's no point in progressing otherwise, because you are. I'll take up your modified claim as soon as you admit that your characterization of what what is (not) a knee was mistaken. Otherwise, well, I'm not going to engage someone who is unwilling to concede error about even the simplest and most obvious statements.

              1. Steve_OH

                Did you actually read the Wikipedia article? Nowhere does it say that a curve must have a knee, and in fact references a criticism of the concept of a knee (that particular criticism really hits home with exponential curves, where the "obvious" visual location of the knee depends entirely on the axis scales, something I learned the hard way in grad school).

                Let's break down the description in the article:

                In heuristic use, the term may be used informally, and a knee point identified visually, ...

                Where is the visual breakpoint in the curves I pointed out?

                ...but in more formal use an explicit objective function is used, and depends on the particular optimization problem....

                What was the objective function used by the authors of the paper?

                ...A knee may also be defined purely geometrically, in terms of the curvature or the second derivative.

                Well, the second derivative of the quadratic example I gave is everywhere zero, so that's no help. And the curvature is directly related to the second derivative, so that' no help, either.

                There is no knee.

                1. ScentOfViolets

                  One more time: you claimed that the curve couldn't have a knee because it was 'smooth'. My link addresses that point and that point alone.

                  A curve can't have a knee because it's 'smooth' (which I take you to mean continuous and differenctiable)? No. Not true.

                  1. Steve_OH

                    I didn't say "smooth." I said "varies smoothly." It's not the same thing at all. Once again going back to the Wikipedia article you linked (quoting Webster's definition of "knee"): "an abrupt change in direction in a curve..." [emphasis added]

                    A curve that varies smoothly does not have an abrupt change in direction; thus it has no knee. Yes, it means that it is continuous and differentiable; those are necessary but not sufficient. It also means that its derivatives are also smoothly varying. It's smooth all the way down.

                    Look at the curvature of the sigmoid.* There are two clear peaks at approximately x
                    = +/-1.4. Those are the knees. Now do the same with the data in the paper. Where are the peaks? Where are the knees?

                    I did misstate the second derivative of the quadratic. I meant to say that it was constant, with zero slope, but that got muddled into just "zero."

                    *You can plug this into Wolfram Alpha: -(e^x (e^x - 1))/(e^x + 1)^3 / (1 + (e^(-x)/(e^(-x) + 1)^2)^2)^1.5

                    1. ScentOfViolets

                      Sigh. You claimed that there could be no knees because it 'varied smoothly', i.e. (presumably) that it's second derivative was smooth. You. Are. Wrong.

                      Dude. Just admit it and move on. OTOH, why am I wasting my time arguing with someone who can't admit to an obvious mistake? I'm outta here.

                2. ScentOfViolets

                  What? _Still_ no admission of error? Why am I not surprised. Will you at least admit that that the second derivative of f(x)=x^2 is _not_ zero? That maybe, just maybe it's actually two?

                  I'd also mention that the sigmoid 1/(1+e^(-x)) has very sharp knees indeed, but hey, what do the rest of us who aren't named steve_oh know?

              2. KenSchulz

                1. You pointed to a very general wiki article which notes early on that defining a knee point requires an objective function. What would that be in this case?
                2. The cited article includes this in the abstract: “Variability between people (interindividual variability or diversity) and variability within a person across tasks (intraindividual variability or dispersion) also increased gradually until the 60s, and rapidly after.” That does not describe ‘a near-universal degeneration process based almost entirely on calendar age’ (your words).
                3. Aging and cognition is something that should be studied longitudinally; one ought to be very cautious in interpreting a cross-sectional study such as the one cited.
                4. Having a model of one’s data, and having a model of the phenomenon under study, are very different things. The University of Washington IHME gave us a fine example recently, with a predictive Covid-19 model that simply extrapolated from reported data, that was consistently wrong.

                1. ScentOfViolets

                  You do realize that

                  That does not describe ‘a near-universal degeneration process based almost entirely on calendar age’ (your words).

                  is a paraphrase of Steve's words, right? You know, where he said

                  knee would imply some sort of near-universal degeneration process based purely on calendar age

                  I'll give you the same opportunity to show your cognitive decline that I gave Steve: Willing to admit those were his words, paraphrased, not mine? Or are you going to insist, like Steve, that even when you're wrong you're stll right? You're move.

                  1. KenSchulz

                    The grammatical construction Steve used is a subjunctive contrary-to-fact; the rest of his sentence is “which is not supported in general.”
                    You quote thus: “Finally, aging in and of itself is most generally -- dare I say it? -- a near-universal degeneration process based almost entirely on calendar age.” I read that as an assertion of the statement Steve was rejecting, but I will grant that your sentence is not easy to parse.
                    Meanwhile, you’re not addressing any of the substance of my critique.

                    1. ScentOfViolets

                      You don't get it, do you? We're checking to see if it's even worth my time arguing with you. Arguing with someone who can't admit to even the most simple, most obvious errors doesn't strike me as good use of my time. And since you're going forward doggedly insisting you weren't wrong, well, I'm outta here. Happy now?

              3. Pittsburgh Mike

                Your example, of a curve with a slope of 1/2 up to a point and then a slope of 5.5 in the next unit, isn't what I'd call smooth, which is what Steve_OH called for in a counter-example. Specifically, the first derivative is discontinuous.

                Looking at the raw data, it looks to me like the decline for all of the curves is steepest over the entire decade of the 60s, with no obvious knee in that region. That is, the slope is steepest in that area, and then decreases again for a lot of the data points. Indeed, the rate of decline looks to be about the same in the 80s and 50s, which the "refined" figure definitely doesn't show.

                1. ScentOfViolets

                  Well, that isn't what I said, but let's go with that (yes, it's quite easy to patch so the the first derivative is smooth. 1st semester quantum mechanics, the hamonic oscillator, but I'm guessing you want to see if I can cop to being wrong, amirite? That answer is yes.). So instead I'll go with my other example, a sigmoid of the form f(x)=1/(1+e^(-10x)).

    2. Larry Jones

      I think you're on to something here, Steve_OH. By the age of 62, a person has seen it all, three times. The first time, in your twenties, it's shocking. The second time, in your forties, it's mystifying (this couldn't be happening again!. The third time, as you approach 62, it's maddening but you realize it's never going to stop, and none of your ranting will have any effect on it.

      You cruise the rest of the way through the repetitions of history with equanimity, amused at all the shocked twenty-somethings who are seeing it for the first time, and not giving much of a shit.

    3. KenSchulz

      Oh yeah, I definitely saw a decline in give a shit. My career was in engineering psychology; along the way I used a number of programming languages, but finally stopped learning new ones when I came to realize that, for my purposes, the ones I already knew were good enough. I figured I would learn a new language when there was one that was significantly less error-prone, and significantly more readable*, than any I was using. Retirement (at age 69) ended the wait.
      *I was once on the mailing list of a magazine for coders that printed a puzzle on the envelope in which they mailed solicitations for subscriptions - a piece of C code and the question, “What does this code do?” ‘It convinces me not to learn such an opaque language.’

    1. Maynard Handley

      That assumes that
      - the set of concepts captured by "cognitive ability"
      - is more important than concepts like experience
      - for the task performed by these people.

      It's unclear that this is true. We want the people creating the fine details of law and policy to have strengths in abstraction, pattern matching and suchlike; but for the broad outlines, for knowing how people will try to game the system, how things have gone wrong in the past, the limits of rational analysis, and so on, experience is probably more useful.

      Which is not to claim that all people in government are especially good at these aspects of the job.
      - I think the Supreme Court does pretty well, better than they're given credit for.
      - Congress does OK, better than the electorate deserves, and fully cognizant, though it will never be admitted, of the disasters of some previous ambiguous legislation.
      - The presidency however, hmm. That's the one that has been far too enthusiastic about theoretical models of the world, from Domino Theory to New World Order to Axis of Evil to They Hate Us For Our Liberties. If anything the presidency needs a lot less abstract thinking about the world, and a lot more concern with history, experience, and the mismatch between theory vs reality

      1. KenSchulz

        Maynard, happy to agree with you! I couldn’t care less if Joe Biden has slowed down on a Stroop task; his experience in Congress, his personal relationships with national and world leaders, his ability to learn from mistakes, his genuine interest in the wants and needs of others, his vision for a better future are what counts.

  6. iamr4man

    I’m 69. I’ve been rewatching the 1970’s documentary “The World At War”. In the episode that covers Germany’s invasion of France they keep talking about the ancient World War I generals the French relied on and how this was one of the causes of their easy defeat. They were just unable to grasp modern warfare. Then they mentioned these ancient generals ages. Late 60’s and early 70’s.
    Damn.

    1. wvmcl2

      Yes, William Shirer's book "The Collapse of the Third Republic" covers this quite well. The WWI generals were fixated on infantry, and even cavalry, and saw those new-fangled tanks as a passing fad that would never be good for much.

    2. GenXer

      George Marshall was in charge of picking combat commanders for the US in WW2. One of his main rules was no "old" generals. He himself was nearly 61 at the time of Pearl Harbor and he considered himself far too old mentally to command troops in the field.

  7. OverclockedApe

    Tbf the stunning lack of kitten pics has led us to that conclusion. Come on man, you know the whole internet runs on cat blogging.

    1. KawSunflower

      DAILY cat blogging! I mean, what are we paying for?

      Seriously, I'm especially grateful for this spot on Fridays & would be willing to spot the host & family the occasional supply of catfood, especially now that we may get more videos!

  8. Joseph Harbin

    The surest sign of decline is having your teenage son, who used to be content to get a few answers right, run entire categories on Jeopardy! and all you can say is "I never even heard of that."

    Pop culture? Gimme a break.
    https://j-archive.com/showgame.php?game_id=7208

    (I do think the post-Alex regime is targeting a younger audience, which means more niche categories that long-time viewers -- e.g., me -- couldn't care less about.)

  9. Displaced Canuck

    I'm turning 64 this Saturday so I'm experiencing this decline already. What I notice is a decline in attention span.It's harder to solve complex problems if you lose interest after 10 or 15 minutes. Not giving a damm is also a symptom.

    1. ScentOfViolets

      THIS. Many years ago, I read the entirety of Remembrance of Things Past solely to shut up a rather insufferable member of Our Gang. Nowadays? I have to work really hard to hold that level of concentrated attention for even a few hours, let alone six hours a day, day after day for the best part of a month.

    2. Pittsburgh Mike

      This could also be the effect of reading 2 paragraph comments on the Internet (TM) instead of actual books for the last decade or so 🙂

  10. azumbrunn

    Just because it is in a graph does not mean it is true (not only for this graph!).

    We have the saying that if something is too good to be true it is probably not true. but the same token though the opposite is also valid: If something is too bad to be true it isn't true either.

    The author is a post doc; that alone is reason to be skeptical. What is more there is also the "series of statistical meat grinders" aka a black box.

  11. geordie

    I don't think the 62 point is the real pivot point however it seems likely that in the mid-60s the mental activities one engages in regularly would change and therefore a decrease in some abilities would occur.

  12. D_Ohrk_E1

    As with any internet test, the biggest and most glaring shortcoming of their research is participation bias:

    Participants were self-selected and motivated to complete the assessment. The assessment was specifically marketed for older adults with concerns about cognition, thus participants older than 60 may have joined because of cognitive concerns, while younger participants may have joined out of interest in their own cognition. The older participants may be lower-functioning than a random sample (from having more memory concerns than average), or may be higher-functioning (from being more informed about memory, or having high computer literacy).

    1. KenSchulz

      Or neither. Or all over the (cognitive) map. Sheesh. I suspected there would be some self-selection effect, but I suffer from diminished givashit capacity to read that far.

  13. cld

    This is why I grind up caffeine pills and mix it into my coffee grounds in the morning so I have extra perkiness.

    Everybody wins! Particularly the entire Earth.

    1. Vog46

      Top 5 OLDEST senators from oldest on down
      Diane Feinstein(D)
      Chuck Grassley(R)
      Richard Shelby(R)
      Jim Inhofe(R)
      Patrick Leahy(D)

      I am old. I believe there should be an age limit to serving in a political office (and supreme court) not because I don't believe that older Americans cannot do the job but the job requires a more robust memory, intellect, and physical stamina than ever before!

      1. MontyTheClipArtMongoose

        You can count Shelby as a Democrat -- his original affiliation -- to make the Neoliberal Democrat Reactionary Gerontocracy that the #OurRevolution Rising Sun Movement continues to assail more obvious.

  14. Pingback: TheMoneyIllusion » Conservative follies

  15. Pingback: Conservative follies - FX Journo

  16. TriassicSands

    "This means that you can expect the sharp political analysis on this blog to rapidly turn to mush over the next few years."

    It's already way too late. Whatever "sharp political analysis" there was, is long gone. Increasingly, it's one incredibly stupid -- often privileged white guy -- post after another. I've been reading Drum for many years. I just can't take any more.

    Goodbye to Hilbert and Charlie, I'll miss them. (Mostly, Hilbert -- tuxedos are great.)

    Good luck to all of you -- except the the trolls, like the complete idiot "Spades."

    Kevin, good luck with your multiple myeloma. You should retire from blogging and enjoy things that don't involve serious thinking. You have had a good, long run.

Comments are closed.