Pupillary Responses to Words That Convey a Sense of Brightness or Darkness
Pupillary Responses to Light
It might seem as if we don’t have much voluntary control over the size of our own pupils. Their tiny muscles appear to typify an innate response, mediated by the sympathetic nervous system, with their shape or size pretty much unknown to us. What we can literally see with our own eyes is that we have the ability to decipher a vast range in light intensity, however our pupils may not only respond to these physical changes. New research is revealing their dependence upon cognitive factors, aka, our own thoughts!
We are able to see our way across a dark room to open a curtain and subsequently to make out the clouds of a sunlit sky. The light intensities that we can see, in fact range across 9 orders of magnitude, referred to as the eye’s dynamic range, spanning from the scotopic threshold to the glare limit (minimum and maximum light intensities). These values differ from each other by 1,000,000,000, a value greater than the number of people on Earth.
The light reflex of the pupil is a well-known phenomenon, easily demonstrated by changing the lighting in a room while watching your eyes in the mirror. That simple! The diameter of the pupil can vary between around 2mm to 8mm, allowing simultaneous accommodation of both pupils to light conditions, through dilation or constriction. The pathway through which detection and accommodation occur is well understood. For example, beginning with retinal photoreceptors being stimulated, followed by ganglion cells and pretectal neurons, with stimulation then reaching the nucleus of Edinger-Westphal on both sides. (Felten, O’Banion & Maida, 2016). However, could there be more to this that meets the eye? New research is asking whether it is possible that the mere thought of light or dark could cause this physical change? Laeng & Sulutvedt revolutionized ideas about the nature of pupillary responses in 2014, when they found that pupils constrict when imagining a bright light. The implication of this is that high-level cognitive control can affect our so called “involuntary” reflexes.
Role of Language in Vision
The extent to which language is intertwined with sensory and motor systems is a matter under investigation. Strong theories of embodiment of language specify that understanding language relies on mental simulation. For example, comprehension of the word cat involves forming of a mental picture of this animal. Weak embodiment theories and traditional views give less emphasis to the importance of simulation. Embodied theories predict that word meaning alone can trigger brain activity associated with the simulation and therefore also accompanying actions. In alignment with this view, the paper to be explored in this article, Mathôt et al (2017), hypothesized that word meaning alone could trigger this ‘reflex’ response of the pupils, tying together the ideas of embodiment of language and of high-level cognitive control of pupillary response. If comprehending words related to brightness or darkness activates brain areas involved in processing these concepts, then pupillary responses could be triggered in the same way as if these stimuli were imagined or actually attended to.
First, a group participants were presented with a sequence of words and asked to rate brightness and valence of these in separate blocks. Words were therefore rated on two scales of one to five, ranging from negative to positive and bright to dark. Categories of words were therefore produced, for example, those related to brightness or those with negative connotations.
Control Experiment Main Experiment
The experiment aimed to compare responses to the sets of words associated with brightness to those related to darkness. Therefore, accurate matching for characteristics other than semantic meaning was needed for words in the two categories. The frequencies with which each word occurs in books was matched for this purpose, as was the total presented luminance of the words in each category for the visual task, for example through choice of words with similar numbers of letters and alterations to font size etc.
There were two main experiments carried out, in order to test both auditory and visual word comprehension. Words were therefore presented either in black writing, or in a synthetic voice. The level of brightness conveyed by the word (semantic brightness) varied, each falling in one of four categories: conveying brightness, neutral, conveying darkness or animal names. A control experiment was also carried out in which words did not vary in their association to brightness or darkness, but in terms of valence.
Overall, 86 participants took part in the experiments, with roughly 30 in each of the visual, auditory and control tasks. A video-based eye tracker recorded pupil diameter as a proportion of the size at word onset. Initially a central fixation dot would be displayed for three seconds, followed by the presentation of a word for the same length of time, each in a random order. Participants were informed that the task was to press the space bar whenever they heard or saw an animal name.
Results and Evaluation
For visually presented words, dilation occurred around 600 ms after presentation of the word followed by constriction, regardless of semantic meaning (see figure 1), due to the visual stimulation. The time following this expected response holds the key to addressing the experimental hypothesis.
Figure from Mathôt et al (2017). Graph of change in pupil size with time following word presentation. Visual experiment results are displayed in a) and auditory results in b). Vertical dotted lines represent mean response time to animal names. Shaded areas show ± 1 standard error. Horizontal lines indicate times when pupil size differed between bright and dark word categories at each of the significance levels shown.
Overall, results confirmed that participants’ pupils were smaller when presented with a word conveying brightness and larger when processing a word associated with darkness (see figure 1). From a default, Bayesian one-sided, independent-samples t test, the Bayes factor was found to be 6.7 for the visual task and 3.8 for auditory presentation, giving a combined Bayes factor of 25.4, strongly supporting the hypothesis. The greatest changes in pupil diameter were observed between 1-2 seconds after the word first appeared, with time delay being slightly shorter for auditory presentation (see figure 2).
From Mathôt et al. (2017). Graphs of the magnitude of difference (pupil size for words conveying darkness minus size that for words conveying brightness) of individual participants for the a) visual experiment and b) auditory experiment. Graphs of change in pupil size over the 1-2 second window from time of word onset, presented for each word for the c) visual experiment and d) auditory experiment. Bars all show ± 1 standard error.
A follow up analysis examined the trend in pupil response over smaller time intervals. A linear-mixed effects model, was generated, using ‘pupil size’ as the ‘dependent variable’ and ‘semantic brightness’ as the ‘fixed effect’. This was carried out for each 10 ms window following word presentation, finding the effect to be reliable from 1,310 to 2,410 and 2,440 to 2,760 ms for the visual task and 1,030 to 1,360 ms in the auditory task. Further support for the general effect on pupil size across the sample was found through a Bayesian one-sided, one-sample t test.
The correlation found between valence and semantic brightness also needed to be addressed, as words associated with brightness are rated more positively and vice versa. As this correlation was so strong, a control task was necessary to confirm that no pupillary response was correlated with valence. Statistical tests confirmed a strong correlation between emotional intensity and pupil size, however, whether the word was positive or negative was unimportant. Therefore, only the absolute magnitude of the emotional intensity correlated with pupil size, meaning valence could not be responsible for the trend observed. However, it would not be entirely valid to ignore this correlation when interpreting the results as a weak correlation was found between words conveying brightness and higher emotional intensity. Regression analysis was used in order to account for this, with emotional intensity as control predictor, still finding brightness to be significant in its effect on pupil diameter.
A key strength of this study was the use of much larger sample sizes than in related experiments. This was a precaution taken due to the expectation of only small changes in pupillary response being observed. Only words associated with brightness or darkness were included in the final statistical analysis. The validity of results was, however, confirmed through finding similar outcomes when all words were included. The use of a variety of statistical techniques is also an important strength of this study. Through the controls and analyses listed, it is possible to reliably determine that the observed effects were not due to valence and emotional intensity of the words, but that brightness level was significant.
However, there were several limitations to this study, for example there was no consideration of cross-cultural or language differences, as all words were presented in French to French speaking participants. Further to this, several words used in the trials were extremely semantically similar, meaning variation was limited, as a high degree of matching between categories was prioritized by experimenters. The age of participants was also not considered as a separate factor, with an age range of 18-54 included in the visual experiments and 18-31 in the auditory tasks. It is known that the maximum dilation of the pupil tends to decline with age, so this may have been a worthwhile consideration.
Further limitations exist at a wider scale, beyond this single study. For example, the theory of embodiment of language is potentially a subject of selective reporting and publication bias. The scale of studies demonstrating embodied-cognition is also fairly limited and further replication would help increase reliability.
Discussion and Conclusion
A key starting point of this study was the idea that language comprehension is associated with the formation of sensory representations, similar to those arising when the concept or object of the word is itself attended. This would rely on involvement of brain areas unrelated to linguistic processing and instead related to visual information. The ability to trigger responses associated with these words, such as pupillary changes is therefore logical. This is the first time word meaning has been demonstrated to be sufficient to generate these effects.
Our use for these mechanisms is largely unknown and there is a potential that they are simply byproducts of comprehending language. It is, however, interesting to speculate what the functionality might be. The authors of this study recognize that these physiological changes could have aided word comprehension. This is because, occurring 1-2 seconds after presentation, responses would be too slow to be beneficial.
It seems that the responses triggered in this experiment were irrelevant to the task, so it is uncertain what other function they could serve. One possibility is that embodied language and consequent responses can be beneficial preparing the body for following conditions, in this case, light intensity changes.
A further discovery has come to light this year, related to the complex and predictive nature of the pupillary response in a different context. Zavagno et al. (2017) monitored pupil diameters while participants observed static or dynamic images of grey-scale gradients. These results suggested that pupil constrictions occurred in anticipation of brightness increases, protecting from damage and pain. Anticipatory dilation in response to expectation of darkness can also be argued to be an advantageous feature, due to shortening the time taken for adaptation to occur, potentially offering an ultimate evolutionary explanation, rather than proximate mechanism for the pupillary results observed in the Mathôt et al. study.
Overall, previous results have demonstrated the relationship between higher-level cognition and pupillary responses. There is also building evidence to support the idea that language comprehension involves generation of simulations. The combination of these two ideas has been shown through the demonstration that pupillary reflexes are triggered in response to word presentation alone, suggesting a tighter link between bodily responses and language than is commonly understood. Whether this is a functionally useful property of our sensory system is unknown, but there is some research to suggest it serves an advantageous anticipatory role.
Felten, D. L., O’Banion, M. K., Maida, M. E. (2016). Netter’s Atlas of Neuroscience (Third Edition). (pages 353-389). Ney York: Elsevier inc.
Laeng, B., & Sulutvedt, U. (2014). The eye pupil adjusts to imaginary light. Psychological Science, 25, 188–197.
Mathôt, S., Grainger, J., Strijkers, K. (2017). Pupillary Responses to Words That Convey a Sense of Brightness or Darkness. Psychological Science, 28(8) 1116–1124.
Zavagno, D., Tommasi, L., Laeng, B. (2017). The Eye Pupil’s Response to Static and Dynamic Illusions of Luminosity and Darkness. i-Perception, 11;8(4).