Despite being interested in intelligence, I am also on guard against judging the mind from the face (there’s no art to find the mind’s construction in the face) while probably doing just that all the time. I assume that I judge mental ability by conversations which go beyond pleasantries. Indeed, perhaps measuring how quickly people turn from pleasantries to content is an ability measure in itself. However, I have never knowingly made a judgement about a person’s intelligence by estimating pupil size. Perhaps I should.
Pupil dilations of the eye are known to correspond to central cognitive processes. However, the relationship between pupil size and individual differences in cognitive ability is not as well studied. A peculiar finding that has cropped up in this research is that those high on cognitive ability have a larger pupil size, even during a passive baseline condition. Yet these findings were incidental and lacked a clear explanation. Therefore, in the present series of studies we systematically investigated whether pupil size during a passive baseline is associated with individual differences in working memory capacity and fluid intelligence.
Across three studies we consistently found that baseline pupil size is, in fact, related to cognitive ability. We showed that this relationship could not be explained by differences in mental effort, and that the effect of working memory capacity and fluid intelligence on pupil size persisted even after 23 sessions and taking into account the effect of novelty or familiarity with the environment. We also accounted for potential confounding variables such as; age, ethnicity, and drug substances. Lastly, we found that it is fluid intelligence, more so than working memory capacity, which is related to baseline pupil size. In order to provide an explanation and suggestions for future research, we also consider our findings in the context of the underlying neural mechanisms involved.
Starting in the 1960s it became apparent to psychologists that the size of the pupil is related to more than just the amount of light entering the eyes. Pupil size also reflects internal mental processes. For instance, in a simple memory span task, pupil size precisely tracks changes in memory load, dilating with each new item held in memory and constricting as each item is subsequently recalled (Hess & Polt, 1964; Kahneman & Beatty, 1966). This research established the use of pupil dilations asan indicator of momentary changes in arousal, mental effort, and attention (Beatty & Lucero-Wagoner, 2000; Hess & Polt, 1960).
Because pupil dilations occur for a wide variety of tasks involving mental effort, psychologists had inferred that the task-evoked pupillary response was reflective of central brain processes (Beatty, 1982). For some, this was seen as providing an opportune way to study the dynamics of cognitive brain function (Beatty & Lucero-Wagoner, 2000). Until more recently, though, the method of measuring pupil size to study brain function did not gain much traction in the field. It was suspected that the reason for this was, ‘‘pupillometry is not widely employed in cognitive psychophysiology because the pupil lacks face validity as a measure of brain function” (Beatty & Lucero-Wagoner, 2000).
In a simple memory span task, pupil size precisely tracks changes in memory load, dilating with each new item held in memory and constricting as each item is subsequently recalled. High memory span subjects had larger pupils than low span subjects even during a ‘‘passive” baseline (in the absence of performing any specific cognitive task). Baseline pupil size was measured during a ‘‘passive” baseline while subjects stared at a fixation on a computer monitor.
The change in pupil diameter over levels of memory load, seen in Fig. 1 reflects the increase in mental effort. The important finding was that pupil diameter increased as a function of memory load by the same amount for high and low working memory subjects.
Being cautious persons, the authors ran a second study to test whether familiarity with the university setting might have accounted for the difference. Their bright subjects were drawn from the university, the less bright ones from the general community, a possible source of considerable bias. Over three sessions they were able to show that familiarity of the environment does not account for the relationship between WMC and pupil size. Both groups got more used to the test setup after a few sessions, at roughly the same rate. Furthermore, the reliability of pupil size over time is high, as indicated by the high correlations ranging from 0.77 to 0.84. In these 102 subjects, average pupil diameter positively correlated with fluid intelligence at r= 0.37 which is a reasonable size for an indirect measure of this sort.
However, the comparison of high and low working memory groups is an extreme group design that can sometimes force a desired result. So, in Study 3 they studied the full range of intelligence on a large experimental sample. This is very welcome, and here are their procedures:
A total of 358 subjects took part in four 2-h sessions in which they were tested on a wide-variety of cognitive tasks. No subject had participated in a study in our lab previously. Subjects were between the ages of 18–35 and had corrected-to-normal vision. Due to technical issues with the eye-tracker, unable to calibrate eye-tracker, or excessive amounts of missing baseline pupil data, the total number of subjects was reduced to 337.
Subjects participated in four sessions that lasted approximately 2 h in which they completed a battery of cognitive tasks. Included in this battery were the measures of working memory capacity (WMC) and fluid intelligence (Gf) described below.
We measured baseline pupil size at the beginning of Session 4 before subjects started any tasks for that day. Immediately following baseline pupil measures subjects performed a simple memory-span task to measure task-evoked pupil dilations.
Measures of working memory consisted of the operation span, rotation span, and symmetry span tasks. Measures of fluid intelligence consisted of the Raven Advanced Progressive Matrices (Raven et al., 1998), Letter Sets (Ekstromet al., 1976), and Number series (Thurstone, 1938).
Given that the size of the pupil is affected by a variety of factors besides locus coeruleus activity, such as age and some drug substances, it is important to account for these. Nine different demographic variables were assessed: Ethnicity, Age (in years), College Student, Nicotine, Medications, Gender, Handedness, Caffeine, Alcohol, Sleep. All demographics were self-reported. At the end of Session 4, the same day as pupil measurements, subjects were asked about: the amount of sleep they got the previous night, their use of nicotine (in the last 10 h), medications (that might affect their attention and memory, in the last 24 h), caffeine (in the last 8 h), and alcohol (more than two drinks in the last 24 h).
Working memory explained 6% of the variance in baseline pupil size and with each 1 SD increase in WMC there was a 0.30 mm increase in baseline pupil diameter, b = 0.30, r = 0.24. Fluid intelligence explained 12% of the variance in baseline pupil size and each 1 SD increase in Gf was associated with a 0.45 mm increase in baseline pupil diameter, b = 0.45, r = 0.35.
Fluid intelligence, however, still predicted baseline pupil size after controlling for WMC, b = 0.45, r partial= 0.27, p< 0.05. These results provide strong evidence that it is fluid intelligence, not working memory, which is uniquely related to baseline pupil size.
What we have shown is that individual differences in fluid intelligence is related to differences in baseline pupil size. However, at this point, our brain story of the intelligence – baseline pupil size relationship is only reasonably informed speculation. Further research is needed to follow up on our findings if we want to draw any definite conclusions about the underlying neural mechanisms.
What are we to make of this? There was a period in which intelligence researchers were drawn to surrogate intelligence measures, which did not run into the storm of criticism which surrounded IQ tests. Reaction time, choice reaction time, tachistoscopic inspection time, analysis of EEGs and the like. Here is a more recent study, using illusory movement to measure a “Motion quotient”.
I see this as a convergence of lines of evidence showing, as if it were necessary to spell this out, that intelligence is tested whenever people are presented with problems, and among all the things going on in the brain, pupil size varies by the difficulty of the task, and varies by the ability level of the subject solving the task even when at rest and not solving problems, dilated pupils being an indicator of higher ability.