Sunday, 18 September 2016

Sex, lies, and videotaped lectures


In terms of information theory, communication is the reduction of uncertainty. Transmitter, channel and receiver are part of a system communicating Shannon bits: in-guessable knowledge. Equi-probable coin tosses are the most informative, because they are hard to guess. The more predictable dross is more easily guessable and therefore less informative. So communications to you should be in-guessable and informative, and all you have to decide is in what format you want your helping of Shannon bits per second.

Printed text is my preference. It is eminently skippable. Fast forward is my favourite button. I usually read the first few sentences of a paper to detect the obfuscation quotient, then scan the references to get a deeper estimation of quality. Then if all is well, I search for the figure or table that gives me what I want: the key finding, the author’s main story. After that it is plod, plod, plod to see if I believe that sparkling jewel of a result offered for my delectation. It means working through Supplementary Appendix 3, and all that. I read while battling boredom and confusion, to check that all is well, or as well as I can bear to find out about, before I lose the will to live. No wonder that scientific papers are more quoted than read. What a perverse art form! Published papers are a confection, everything tidy and shipshape because the Inspector General is calling. No vacillation, no confessions, and sufficient ritual humility about shortcomings to confuse the innocent. I never settled into the writing of papers. Writing a blog is a liberation: like talking to a friend, compared with preparing a tax return. Will the day dawn when a blog counts on an academic CV?

Why bother with lectures? The projector is often a problem. The text is usually too small, voices not always audible, and the pace variable. And yet, and yet, there is a delight to hearing the story unfold, the detective tale of curious events in the night, the skittles of accepted findings being set up so as to be knocked down, the swerves in the path to discovery, the researcher’s art of creation. There in front of us in the lecture room is the very person, showing every aspect of their intellectual endeavours, using their real everyday words, responding to the faces of the audience, making disparaging asides, cheerfully admitting problems and short cuts, cracking jokes, taking the occasional brief interruption and then answering questions at the end so that they become a fellow traveller on the way to find the treasure at the end of the rainbow.

Lectures can have way more impact than papers. They are the royal road to understanding the researcher’s subject: the task is make the story easy to understand, and to establish the speaker as a trusted guide. A good lecture motivates you to read further, on the basis of acquaintance, and even nascent friendship. So I think it is time to sit back and listen to a few lectures.

Let’s have a look at one speaker taking on the received opinion that there are no sex differences in intelligence. Prof Richard Lynn rises to the challenge, suggesting that the majority are wrong and dares to propose a radically different point of view.

His own review of published papers with Wechsler results show (visible at 9:50 on the tape) :

WISC (6-16 years) 32 studies, male advantage 2.85 IQ points

WAIS (adults)        32 studies, male advantage 3.60 IQ points

Intrigued by these differences, he managed to get some disclosures from the test publishers of the Wechsler test by the simple expedient of ringing them up to ask if they had any findings on sex differences in their standardisation samples. Their answer will raise a few eyebrows. Incidentally, Richard Lynn asked how many other callers had asked to see those results, and was told that he was the first to do so. Here are the results (visible at 10.24 on the tape. Ignore the red line, which is just the tape progress indicator):

WAIS standardisation sex differences

It very much looks like the publishers have been sitting on the sex differences which emerge from their standardisation samples. Indeed, though this was admitted in a telephone call and exchange of emails, permission to quote these findings in an upcoming publication was denied. This is disturbing, because the Wechsler is seen as the gold standard of intelligence testing, and if there are unreported sex differences even in the carefully constructed standardisation sample, that is a cause for considerable concern.

Having criticised the stilted format of scientific papers, and the poor quality of some lecturers, I would like to announce a new art form: the integrated slides-in-vision lecture. This sparkling product, produced by Mingrui Wang, gives you a front row seat, with the lecturer in full view, and the slides perfectly visible.









  1. I still wonder if this supposed sex difference is an artifact of sampling bias.

    1. I thought so, but the Scottish samples are the entire population, and at the very highest perfect score levels boys predominate.

    2. Also, the Wechsler standardisation samples should be good enough to get round bias.

  2. In a lecture once, I praised a colleague. The students were astonished - they'd clearly never heard the like.

  3. "the very highest perfect score levels": an exam (of whatever sort) where anybody gets a perfect score is too easy.

    Anyway, wouldn't your point be consistent with a difference in SD rather than any need for a difference in mean?

    1. Yes, but unfortunately consistent with bigger SD and also 2 to 4 point mean difference. Too close to call this one, I think.

  4. Heard Professor Diane Halpern speak on this topic in Melbourne (Australia) last week. Differences at the right-tail of the distribution in Mathematics ability were explored, but not in general ability.

  5. Sorry for the silly question- but is it possible that authors were bright enough insinuatingly arranging the subtests for equal average results for man and women in past and recent IQ tests?

    1. Not a silly question. It is possible that good items in intelligence tests have been dropped because they revealed sex and race differences. The presumption now is that those differences must be due to bias. They might be, but it should not be assumed by default. As far as I know there is no public record of the items which have been rejected, which would at least let us use the on a new standardisation sample to see what their characteristics are.

    2. Thank you for the quick answer. So the next logical question: is it possible to develop a non-PC or „absolute” IQ-test?
      It would be very interesting checking this „absolute” IQ test on different races- I mean the results of Asian (Japanese, Chinese etc.) versus European group, as the former population is more feminine and might has advantage because of the gender bias.

    3. Yes, so long as you accept that there are some concepts which are universal to all cultures, which I do. So, reaction times, digit spans, simple sequence puzzles, inspection time, processing speeds on simple tasks are good candidates.
      Here is a post on one aspect of the culture debate:

    4. Not a silly question. It is possible that good items in intelligence tests have been dropped because they revealed sex and race differences.

      I have read many non-silly people claiming this as a sure fact. And it has been done since the early 20th century. Items that left women behind were removed.

      <a href=">This</a> is the clearest table on the subject.

      Cognitive, and mental differences between men and women go far beyond what IQ can express, however.

  6. "Sex differences": how quaintly scientific. Wot abaht all the BLTGPOs? It should be "gender differences" nowadays, amongst the fifty or so groups. Or should that be a couple of hundred?

    On a different line: lecturing is an awfully slow way to transmit info compared with writing. The beauty of a lecture would lie in seeing the lecturer answering questions. We didn't get that.

    Still and all; there seems to be the problem that men and women are (on average) superior on different types of test. That means that different arbitrary weightings of those tests can give you whichever answer you'd like. Perhaps a more fruitful attack would be to ask which tests reveal more about subsequent progress in life, or length of life, or of healthy life, or whatever.

    Mind you, his early point that males do better, compared to females, only after age 16 or so, seems plausible to me, even though he later recanted somewhat.

    Remember my advice to the university admissions tutor faced with choosing between two candidates who seem equally matched in terms of ability and attainment. If they are both male, admit the younger; if both female, the prettier; if one of each, spin a coin.