Tuesday, 6 May 2014

10 fast questions about processing speed: Ian Deary

We are in a transitional stage as regards the communication of research results. The most common method, the published paper in a high impact, peer-reviewed journal is the current gold standard. As you will see, the next presentation has many of those. Such publications are quality checked, and come with a seal of approval. However, apart from high cost it imposes roughly a one year delay on the dissemination of results, so as a consequence much of the prompt communication of research findings is done at conferences. That is great for those who attend, but bewildering for those left with nothing but the un-explained Powerpoint version. Youtube doesn’t quite bridge the gap, because lectures are less engaging if you are watching them afterwards, and cannot catch the atmosphere of the meeting and interrupt to ask a question. We need holograms.

My task here is to provide some notes, from memory, of the opening talk so that you can get as much as possible from the slides that follow. I have cut out contextual references to the task we were set as invited guests. All you need to know is that the speakers will be asked to contribute to a publication, probably in Intelligence which will look at processing speed and its relation to other cognitive skills. However, the other thing you need to know is that the speaker is the lead singer of rock band Dancing Mice, next gig 15 May. As you see, I keep cool company.  http://www.dancingmice.co.uk/

Here is the presentation:


Slide 2. Donders used reaction time studies to infer differences in more general cognitive processing, found that simple reaction times were faster than choice reaction times, presumably because the latter required more decision-making.

Slide 4. The examples of processing speed tasks include the Wechsler Digit Symbol Coding task; the Deary-Leawald simple and choice reaction time apparatus; the inspection time task (which refers to the length of the brief exposure, not the response itself, which is untimed); the bottom right figure on reaction times shows that they can be de-composed into encoding time, decision time and response time. 

Slide 10 shows the how each age group fares in terms of number of trial correct on IT (Inspection Time) duration. This clearly shows that as children mature they become capable of getting better at detecting briefly exposed stimuli, though it is always the case that the longer durations are easier to see: a real measure of human processing limitations.

Slide 11 looks at the relationship between processing speed and prior intelligence.  Age attenuates a lot of the correlations.

Slide 13 is important: it shows the relationship between choice reaction time and general intelligence measured on the AH4 (Alice Heim 4). Brighter subjects make the correct choices more quickly.

Slide 14 shows what used to be the Verbal IQ and Performance IQ subtests of the Wechsler intelligence test, with Inspection Time loading on the second factor.

Slide 15 shows how a general speed factor and a general intelligence factor map on to a variety of subtests. Unfortunately the subtest names have been lost in a Figure 1 which is not shown in this presentation. The tests in the perceptual speed and reaction time factors are: PS = Symbol Search, Digit Symbol, Inspection Time, Choice Reaction Time standard deviation, Choice Reaction Time Mean. RT = Simple Reaction Time Mean, Simple Reaction Time standard deviation.

Slide 18 is pretty clear, though a mite depressing. Choice reaction time definitely slows with age. This has nothing to do with having a vast store of a lifetime’s accumulated knowledge through which one must serially search (the recently touted “overstuffed mind” hypothesis, as Pat Rabbitt dismissively calls it, showing its errors  http://pmar76.wordpress.com/2014/02/01/hello-world/ ) but a more general process in which decision making, even at a very simple level, takes more time. Slide 19 shows the smoothed data.

Slide 20 succinctly gives Tim Salthouse’s 1996 Processing Speed theory that when processing speed slows respondents run out of time and as a consequence also run out of short term memory. This is a classic paper in the literature (and is also a good example of how to write an abstract). The cognoscenti in this field describe historical progress as Salthouse I, Salthouse II etc

Slide 21 shows how Ian Deary depicts later data from Salthouse. Note that age is positively related to Vocabulary but negatively to processing speed and even more negatively to g as a whole.

Slide 24 shows two different models of relationship between age and processing speed, namely whether the latter could be underlying IT (inspection time) RT (reaction time) and DS (digit symbol) alone, or also BD (block design) PA (picture arrangement) and PC (picture completion).

Incidentally, the best way to measure digit symbol processing speed is to show the symbols and get subjects to fill in the relevant digits. If you do it the other way, subject have to learn how to write the unfamiliar symbols, and that effect confuses the pure measure of processing. Most people have much more practice at writing simple digits, so that is the better way round to do it. Deary says that a two minute symbol-digit task is an excellent, highly time-effective task. Should this be the task we ask all psychologists to give to all of their subjects, before going on to do whatever else is being measured? Symbol-Digit 120?

Slide 29 is the icing on the cake of epidemiology, or the stab in the back of one’s health prospects, depending whether one is interested in pure knowledge or personal survival. Lower intelligence predicts higher all-cause mortality. Cue a frenzied discussion on how bright people read the medicine labels carefully, avoid all health hazards and use their brains to lead healthy lives. Then look at the next line. Slower simple reaction times also predict higher mortality. (This is not because illness slowed them up just before death: the last 5 years of reaction time data were ignored in these results, in case there were any “fading of the light” illness effects, and still the result held up strongly). There was little attenuation after adjustment for sex, smoking, social class & education. You read that correctly. Smoking. Social Class. Education. All these things we hear so much about. However, the simple reaction time data accounts for two thirds of the effect of intelligence. You read that correctly. Simple, not choice, reaction time.

By the way, this slide was news to at least one participant. Not altogether surprising, since it is ignored by many epidemiologists who ought to know better, or who are paid to know better. All this “eat your vegetables” stuff is going to have to be re-evaluated by ensuring that participants in health surveys complete some simple IQ tests.

This one slide is the cornerstone of cognitive epidemiology. Something about intelligence makes you live longer, but that something is also associated with faster reaction time. Deary calls this conjunction “system integrity”. He made that up, off the cuff, to explain a mysterious physical capacity which gives lucky recipients quick reactions, powerful problem solving ability, and long, healthy lives. Print it on a T shirt so that even epidemiologists get to see it: System integrity = quick reactions, bright minds, healthy lives.

Slides 32 to 35 are best covered in a later presentation.

Slide 37 shows that none of the variance in intelligence or inspection time seems to be explained by “shared environment” the very core of sociological theory: families and schools and general culture. The non-genetic is due to the “non-shared environment” for which the plain English would be “environments you create yourself”.


Finally, Slide 49 which show that in terms of health hazards, smoking is in first place and next is choice reaction time.

In summary, Ian Deary believes that there still is life in the hypothesis that processing speed has wider significance, particularly as regards health and lifespan. Processing Speed rides again.


  1. So when are you guys going to have your conferences closer to Maine? Portland and Bangor each have pretty nice spots not far from the local airport... ;)

  2. And the weather is beautiful here in July and August...