Externally manipulating pupil size does not have any
influence on perceptual decision making
A research about the relationship between perceptual decision making and both internally and externally generated fluctuations
in pupil size
Myrte van Kesteren – 11679913 University of Amsterdam
Bachelor Thesis for the Bsc. Psychobiology
Supervisors - dr. Simon van Gaal, Lola Beerendonk Co-assessor – dr. Hannie van Hooff
Contents
Abstract ... 3
Introduction ... 4
Materials & Method ... 6
Participants ... 6
Behavioural task ... 6
Post-questionnaire ... 7
Pupil data acquisition ... 7
Data analysis ... 8 Behavioural data ... 8 Pupil data ... 9 Post-questionnaire data ... 9 Statistical comparisons ... 9 Results ... 10 Exclusion ... 10
Effects of task and colour ... 10
Pre-stimulus pupil size ... 11
Post-stimulus pupil size change ... 12
Post-questionnaire data ... 13
Discussion ... 13
References ... 16
Appendices ... 19
Abstract
Cortical states seem to influence perceptual decision making, since differences in cortical states are associated with variations in perception. Previous research has shown that these cortical states, also referred to as internal fluctuations of the brain, are reflected in pupil size changes. The aim of the current research was to investigate to what extent pupil size
influenced perceptual decision making, thereby looking at internally and externally generated pupil changes. Pupil size was externally manipulated by letting participants perform detection and discrimination tasks either against a white or a black background. White backgrounds made pupils contract, whereas black backgrounds made pupils dilate. Throughout the whole experiment, pupil size, reaction time, volume and behavioural responses were measured. We observed that there was no effect of the external
manipulation on metacognition and volume. Volume was used to indicate performance, since performance was kept constant. Additionally, when looking at internal fluctuations, we found that sensitivity increased for detection and discriminations tasks if the pre-stimulus pupil size was large. Also, post-stimulus pupil size appeared to be larger during correct instead of incorrect decisions, and during unconfident instead of confident decisions. So, externally manipulating pupil size seemed to have no influence on perceptual decision making performance and metacognition, and therefore it was concluded that the relationship between pupil size and cortical states was nonreciprocal. The internal
fluctuations reflected by both pre- and post-stimulus pupil size were, like previous research suggested, associated with perceptual decision making. All in all, these results have
contributed in understanding the relationship between cortical states and decision making behaviour.
Keywords: cortical states, pupillometry, perceptual decision making, detection,
Introduction
We are constantly making choices during the day, in example, think of the variety of chips in a supermarket or the number of clothes in your closet. Even though these are just simple examples, the process of deciding can be crucial in both biological and social situations. This process takes place in our brain, although it is just partly understood. From now on, there will be a focus on perceptual decision making that is the process of making distinctions between multiple options aiming to finally choose one, while mainly sensory information is used (Heekeren, Marrett & Ungerleider, 2008). Even though the available information is exactly the same, people still tend to make different choices. This phenomenon is even visible within individuals, and can be due to fluctuations in perception. Cortical states offer a plausible account for these variations.
Cortical states are momentary, global states of the brain that are associated with different levels of wakefulness (Beerendonk, 2017). Electrophysiological studies demonstrate that these states can have both low and high frequencies. Low frequencies are associated with synchronized states that are characteristic for sleep, while high frequencies are associated with desynchronized states that are characteristic for alertness (Harris & Thiele, 2011). Perceptual decision making performance seems to be better during locomotive behaviour (Bennett, Aroyo & Hestrin, 2013; Bullock, Elliott, Serences & Giesbrecht, 2017). An increase in performance is associated with improved, sensory representations (Hesselmann, Kell, Eger & Kleinschmidt, 2008). These representations are enhanced by increased signal-to-noise ratios that are particularly found during locomotion (Polack et al., 2013). In other words, signal-to-noise ratios and performance are both improved during alertness. Decreased spontaneous firing is found in desynchronized states, which is correlated with locomotion too (Reimer, Froudarakis, Cadwell, Yatsenko, Denfield & Tolias, 2014). Therefore, it seems plausible that less spontaneous firing results in higher signal-to-noise ratios. So, fluctuations in cortical states are associated with differences in the processing of sensory input, as seen in the signal-to-noise ratios, which influence perception.
Pre- and post-stimulus alpha dynamics are widely studied when looking at cortical states during alertness. The alpha rhythm is largely seen during wakeful states and is considered as a marker of attentional state (Hong, Walz, Sajda, 2014; Iemi et al., 2017; Kirstein, 2007). Post-stimulus alpha oscillations are decreased more if a stimulus is perceived than if it is not perceived. So, they seem to be correlated with conscious visual perception (Babiloni,
Vecchio, Bultrini, Romani and Rossini, 2006; Harris, Dux and Mattingley, 2018). However, it is important to keep in mind that post-stimulus waves rely on multiple top-down processes and not just on the stimulus alone (Summerfield & de Lange, 2014). Hence, we will focus on pre-stimulus dynamics from now on. A decreased pre-stimulus alpha rhythm is correlated with an increased number of conscious perceived stimuli during an auditory detection task. This decreased rhythm is also associated with an increased local excitability in the auditory cortex (Leske, Ruhnau, Frey, Lithari, Müller, Hartmann & Weisz, 2015). Based on these findings, it is suggested by Iemi, Chaumon, Crouzet and Busch (2017) that pre-stimulus alpha oscillations are correlated with changes in perception by modulations in baseline neural excitability. However, sometimes there is no effect of pre-stimulus alpha waves found on perceptual decision making performance (Samaha, Iemi and Postle, 2017). These different conclusions can be due to the role of the alpha rhythm that seems more prominent in detection compared to discrimination tasks (Iemi et al., 2017). Moreover, there is found a
correlation between the pre-stimulus alpha rhythm and the level of confidence, since the rhythm is decreased when participants are more confident about their answers (Wöstmann, Waschke & Obleser, 2019). In short, there is evidence for an association of the pre-stimulus alpha rhythm with both perceptual decision making performance and confidence.
Pupillometry is a way to measure pupil size reflecting changes in the cortical states. Hong et al. (2014) show for instance that there is a negative relationship between pre-stimulus alpha oscillations and post-stimulus pupil dilation. Dilated pupils are correlated with increased levels of acetylcholine and noradrenaline (Joshi, Li, Kalwani & Gold, 2017; Reimer, McGinley, Liu, Rodenkirch, Wang, McCormick & Tolias, 2016). This can be explained by the fact that the locus coeruleus plays an important role, through a subcortical pathway, in both the
synthesization of noradrenaline and in the dilation of pupils (Mathôt, 2018). Subsequently, noradrenergic neurons from the locus coeruleus send axons to the basal forebrain where acetylcholine is produced (Larsen & Waters, 2018). Noradrenaline and acetylcholine seem to play an important role in cortical state changes as well (Beerendonk, 2017). Thus, alterations in both cortical state and pupil size are regulated by the same neurotransmitters.
The current research aims to investigate the role of changes in pupil size on perceptual decision making. We will look at both internally and externally generated changes. Some previous studies have already shown the influence of pupil size on behaviour. For instance, increased pupil dilation is associated with better discrimination performance, which resulted from more correct responses next to shorter reaction times (Correa, de Gee, Weaver,
Donner & van Gaal, n.d.). Large pupils are also associated with worse metacognition that is the ability to report your own performance (Correa et al., n.d). A possible mechanism holds that neuromodulation suppresses top-down processes like metacognition, but not bottom-up processes as performance (De Gee, Knapen & Donner, 2014). According to the proposed mechanism, less noradrenaline results in less suppression of metacognition processes, which causes better metacognitive performance. This in agreement with Hauser, Allen, Purg, Moutoussis, Rees and Dolan (2017) who found improved metacognition after administering noradrenaline blockers to participants. The mechanism seems to hold on for the findings of studies that investigated the interplay between noradrenaline and performance.
Noradrenaline concentrations correlate with signal-to-noise ratios in a positive linear way. Since it appears that increased signal-to-noise ratios sharpen neural representations, there is some evidence that high concentrations of noradrenaline enhance even discrimination performance (Ginani, Tufik, Bueno, Pradella-Hallinan, Rusted & Pompéia, 2011; Pinto et al., 2013). However, it seems that acetylcholine influences neural representations via an
inverted U-shaped curve, which challenges the suggested mechanism (Aston-Jones & Cohen, 2005). All in all, increased pupil dilation is associated differently with both meta-cognition and performance, thereby suggesting a role for neuromodulation.
Up to now, most work has focused on internally generated fluctuations in pupil size. This is why Mathôt and Ivanov (2019) manipulated pupil size externally by letting participants perform visual detection and discrimination tasks against bright, grey and dark backgrounds. Pupils dilated through the dark-coloured backgrounds, whereas they contracted through the light-coloured backgrounds. Then, Mathôt and Ivanov (2019) concluded that small pupils led to improved discrimination performance, while large pupils led to improved detection performance.
Whilst the results of Mathôt and Ivanov (2019) are interesting, it cannot be concluded whether they represent a direct effect of pupil size on decision-making, or whether they are merely the result of changing the amount of light on the retina. In our research we
investigate if these results hold for more than only a visual task, in which pupil size
influences perceptual input. Therefore, we let participants perform auditory detection and discrimination tasks including confidence ratings, while their pupil size is externally
manipulated. Since we use auditory instead of visual stimuli, they will not confound the pupil size results. It was already clear that cortical states seem to influence pupil size (Hong et al., 2004; Joshi, Li, Kalwani & Gold, 2017; Mathôt, 2018). However, if it appears that
participants’ performance changes as a consequence of the amount of light on their retina, pupil size seems to influence cortical states too. So, if the results of Mathôt and Ivanov (2019) will be replicated, they give rise to the suggestion that the relationship between pupil size and cortical state is reciprocal because of this two-way manipulation.
Materials & Method
Participants
The study was approved by the ethical committee of the Psychology Department of the University of Amsterdam.
Forty students from the University of Amsterdam (9 male and 31 female) participated in the experiment for course credits. All participants signed the informed consent after reading the information brochure. They were aged between 18 and 34 years and have normal or
corrected-to-normal vision. A participant was excluded if there was no data available from at least one condition, which will be explained later on in more detail.
Behavioural task
Participants came to the lab to perform auditory tone-in-TORC (temporally orthogonal ripple combination) detection and discrimination tasks. The tasks were made in Python software. In the discrimination task, each trial started with a baseline period of 0.6 s. Thereafter, subjects were constantly presented with a noise stimulus, either overlaid with a high or low tone. Both tones were presented in a fixed frequency of 50%. This stimulus interval took 0.5 s. Then, participants had to decide whether the tone was high or low and whether they were confident about their choice or not. Subjects were encouraged to be confident during half of the trials. They had 1 s to give their response by pressing one of the four keys on the
computer keyboard, corresponding to the four possible combinations. Response mapping was consistent throughout the entire experiment but counterbalanced across participants. After every trial, there was a random inter-trial interval ranging from 0 to 0.4 s. During the whole trial, a fixation dot was shown in the centre of the screen with a spatial resolution of 1920 x 1080 pixels. The fixation dot was based on the ABC target mentioned by Thaler, Schütz, Goodale and Gegenfurtner (2013), with an outer circle of 40 pixels, an inner circle of 4 pixels and two lines forming a cross of 6 pixels.
In the detection task, every trial started again with a baseline period. This was followed by a stimulus interval in which there was a noise stimulus all the time, and in 50% of the trials a low tone too. Then, the task of the participants was to indicate whether the target tone was present or absent, and again whether they were confident about their decision or not. Each trial ended with an inter-trial interval. The duration of the intervals in the detection task was the same as in the discrimination task.
At the start of the experimental session, the appropriate difficulty level for each participant was determined by means of a Kaernbach staircase procedure (Kaernbach, 1991) for both detection and discrimination. This staircase held that a correct response led to a decrease in signal volume, while an incorrect response led to an increase in volume that was three times as large. Moreover, the step sizes were larger in the beginning than at the end (from 0.05 to 0.005). During these staircase procedures, subjects performed the tasks described above whilst the volume of the target tones was varied. In the end, a volume was estimated at which the accuracy would be approximately 75%.
After the staircase procedure, participants performed the discrimination and detection tasks either against a white or black background. In other words, there were four conditions: detection on a white background, detection on a black background, discrimination on a white background and discrimination on a black background. The different coloured backgrounds were used to manipulate pupil size while nothing else changed visually. Only one condition was presented during each block. Subjects performed eight blocks in a
random order, which meant that every condition was repeated once. One block consisted of 130 trials and took about five minutes time. During all blocks, the volume was staircased to keep performance at about 75% (Kaernbach, 1991). Throughout the whole experiment, reaction time, volume, pupil size and behavioural responses were recorded. The reaction time was defined as the time from stimulus onset until the key press.
Post-questionnaire
Participants had to fill in a questionnaire (appendix A) after the detection and discrimination tasks, which was used to control for the aim of the experiment. They were asked about the purpose of the experiment and whether they had noticed a difference in white and black coloured backgrounds.
Pupil data acquisition
Subjects were placed in a silent room with dimmed light where they had to put their chin on a chin rest that was 60 cm in front of the computer screen. The diameter of the left eye’s pupil was measured at 500 Hz, using an Eyelink 1000 Desktop Mount (SR Research). A calibration procedure was executed before the start of every block. Participants were asked not to move, and to blink as less as possible.
Data analysis
Behavioural data
The mean and standard deviation of performance, reaction time and volume were
computed per condition. Before, the volumes were normalised by calculating the z-score for every trial. The mean and standard deviation used for calculating the z-score, were different for both detection and discrimination because of different staircase procedures.
Furthermore, sensitivity, criterion, meta d’ and meta-criterion were calculated based on the behavioural responses, concerning first and second orders measures of the signal detection theory.
Figure 1 Discrimination and detection task. (A) The sequence of events during one trial for both discrimination and detection is shown on the left. Participants heard a noise stimulus overlaid with either a high or low tone for the discrimination task. For the detection task they just heard the noise stimulus or the noise stimulus combined with a low tone. During the response interval participants had to decide which stimulus they heard by pressing a key. At the same time, they had to indicate how confident they were about their decision. The background colour could be either black or white. (B) Participants had to perform two staircase procedures, followed by two blocks of each condition. There are four conditions: detection on a white background, detection on a black background, discrimination on a white background and discrimination on a black background.
A complete block was removed from the data if the performance of that block exceeded three times the standard deviation, since the mean performance should be 75%. The standard deviation was calculated per block for all participants. Accordingly, the data of a participant was removed if one condition was missing.
Pupil data
The pupil data were pre-processed according to the script of de Gee et al. (2014). Blinks found by using the standard algorithms of Python, were removed by linear interpolation. The interpolation time window ranged from 100 ms before until 100 ms after the blink. Subsequently, the pupil data were band-pass filtered (passband: 0.01-6 Hz, sample: rate 500) and z-scored.
The continuous mean was calculated per block. Blocks of the same condition were averaged per participant to get absolute means.
Epochs were made from 0.5 s before until 2 s after stimulus onset to isolate single trials. Trials without any response were removed from the data. The baseline that was based on the interval 10 ms before stimulus onset, was subtracted from every trial to make it possible to compare pupil responses between participants. The mean post-stimulus pupil response was calculated as the mean change in pupil size between 1 and 2 s after the start of the trial per participant.
Post-questionnaire data
The percentage of participants that understood the goal of the experiment was calculated.
Statistical comparisons
The independent variables were background colour and task type, while the dependent variables were pupil size, volume, reaction time, behavioural responses and signal detection theory measures. The data were analysed in JASP (Version 0.11.1; Jasp Team, 2019). The significance level was hold on 0.05.
Volume, reaction time, absolute pupil size, meta d’ and meta-criterion were tested in a 2x2 repeated measures ANOVA for both the effects of task and colour. Sensitivity and criterion were tested with a Wilcoxon signed-rank test for colour, since the normality assumption was violated. These two were only tested for the effect of colour since the different outcomes between the tasks cannot be compared. The direction of criterion was tested in a one sample t-test with a test value of 0.5.
The continuous pupil data were split in two bins of trials by using a median split per
condition to differentiate between small and large pre-stimulus pupil sizes. The median split was based on the baseline pupil size from 0.5 seconds before stimulus onset until stimulus onset. Thereafter, sensitivity and criterion were tested in a 2x2 repeated measures ANOVA for baseline pupil size and colour. Meta-criterion and meta-cognition were tested for task and colour while considering large and small pupils in a 2x2x2 repeated measures ANOVA.
The mean post-stimulus pupil responses were tested in a 2x2 repeated measures ANOVA for task and colour. The pupil responses were also tested for correct and incorrect, and
confident and unconfident answers in a 2x2 repeated measures ANOVA.
Results
Exclusion
Nine participants were excluded from the analysis since their data lacked at least one condition completely. Two of them did not finish the experiment and therefore their data was not used. The data of the other seven participants were removed according to the exclusion criterium, as their performance percentage was too high.
Effects of task and background colour
The average performance for all conditions was 75.70% after exclusion (missed trials:
1.80%). The staircase appeared to be successful as there was no difference in colour for both detection (black: 1.53±0.26, white: 1.63±0.56 | z = 219, p = 0.581) and discrimination tasks (black: 1.63±0.62, white: 1.52±0.24 | z = 301, p = 0.308). The manipulation seemed to be successful as well since the absolute pupil size was indeed significantly larger for black than for white backgrounds (detection black: 8.78±1.16 mm white: 4.03±0.51 mm, discrimination black: 8.76±1.32 mm, white: 4.03±0.48 mm | F1,30 = 1539.514, p < 0.001, η2 = 0.866).
Concerning pupil size, there were no effects found of task (F1,30 = 0.151, p = 0.700) and the
interaction between colour and task (F1,30 = 0.017, p = 0.899).
Volume and reaction time are used as variables to describe performance during the different conditions, since the performance measure was kept at 75%. A repeated ANOVA was
conducted to test for the effect of task and colour. A significant effect of colour indicates an influence of the external background manipulation. The test showed that volume was significantly higher for detection (black: 0.039±0.008, white: 0.038±0.010) than for
discrimination tasks (black: 0.029±0.009, white: 0.029±0.011 | F1,30 = 23.822, p < 0.001, η2 =
0.196), suggesting that performance was higher for discrimination tasks. There were no significant effects found of colour (F1,30 = 1.565, p = 0.221) and of the interaction between
task and colour (F1,30 = 0.422, p = 0.521). Likewise, the repeated measures ANOVA for
reaction time showed that reaction time was significantly higher for detection (black:
0.55±0.12 s, white: 0.56±0.11 s) than for discrimination (black: 0.52±0.12 s, white: 0.52±0.10 s | F1,30 = 11.543, p = 0.002, η2 = 0.026). There were again no significant effects found for
colour (F1,30 = 0.641, p = 0.430) and the interaction between task and colour (F1,30 = 0.161, p
= 0.691). These results are shown in figure 2. Thus, there seemed to be no effect of colour on performance, which was indicated by volume and reaction time.
The participants’ decisions were checked for potential biases. It appeared that participants decided in 42.8±0.09% of the detection trials that the signal was present (t30 = -4.646, p <
0.001), and in 55.5±0.09% of the discrimination trials that the tone was high (t30 = 3.457, p =
0.002). So, there seemed to be a small bias towards absent and high tones in respectively the detection and discrimination task. The bias, also called criterion, did not differ between colours (detection black: 0.25±0.41, detection white: 0.34±0.45, discrimination black:
-0.26±0.51, discrimination white: -0.22±0.33| detection: z = 289, p = 0.433, discrimination: z = 262, p = 0.794). So, sensitivity seemed to be independent of bias for the different conditions.
Concerning type 2 measures, we looked at possible differences in the level of confidence, also referred to as meta-criterion, between task and colour. The test showed that there were no effects of task, colour and interaction (detection black: 0.57±0.08, detection white: 0.58±0.11, discrimination black: 0.54±0.08, discrimination white: 0.56±0.10 | task: F1,30 =
3.209, p = 0.083, colour: F1,30 = 1.442, p = 0.239, interaction: F1,30 = 0.099, p = 0.755). We also
investigated if meta-cognition would differ between the conditions. This did not seem to be the case as there were again no effects found of task, colour and interaction (detection black: 0.539±0.623, detection white: 0.569±0.437, discrimination black: 0.653±0.598, discrimination white: 0.728±0.683| task: F1,30 = 1.391, p = 0.247, colour: F1,30 = 0.346, p =
0.561, interaction: F1,30 = 0.111, p = 0.742). All in all, there was no influence of the
background manipulation on type 2 measures of the signal detection theory.
Pre-stimulus pupil size
We will now have a look at pre-stimulus pupil size and performance, also called baseline pupil size, since this tells us something about internally generated fluctuations. The continuous pupil data were split based on baseline pupil size, which resulted in one bin of trials where the baseline pupil size was small, while in the other bin the baseline pupil size was large. Only significant results will be reported, since we are particularly interested in the internal fluctuations.
Sensitivity seems to be an accurate way to define performance. Therefore, we tested for differences in sensitivity between black and white backgrounds, and between small and large baseline pupil sizes for both tasks. It appeared that sensitivity was significantly higher when baseline pupil size was large for both detection (black low: 1.570±0.575, black high: 1.658±0.593, white low: 1.553±0.613, white high: 1.719±0.792| F1,30 = .178, p = 0.050, η2 =
0.010) and discrimination (black low: 1.338±0.339, black high: 1.602±0.350, white low: 1.359±0.339, white high: 1.569±0.439| F1,30 = 6.481, p = 0.016, η2 = 0.084). The results that
are shown in figure 3, suggested that sensitivity improved when baseline pupil size increased.
Figure 2 Volume and reaction time are tested for task and colour. Volume and reaction time are variables used to describe the performance between the different conditions (A) The repeated measures ANOVA showed that the reaction time was significantly higher for detection than for discrimination tasks (F1,30 = 23.822, p < 0.001, η2 = 0.196). The test did not show any significant effects for task and interaction. (B) The reaction time for detection seemed to be significantly higher for detection than for discrimination (F1,30 = 11.543, p = 0.002, η2 = 0.026). Again, the repeated measures ANOVA did not show any other significant effects.
To test for a possible bias that influences the results, we tested criterion for background colour and baseline pupil size as well. The 2x2 repeated measures ANOVA showed that criterion was larger if the baseline pupil size was larger, but only in the discrimination task, which you can see in figure 3 (black low: 0.061±0.413, black high: 0.250±0.361, white low: -0.129±0.345, white high: -0.412±0.783 | F1,30 = 9.702, p = 0.004, η2 = 0.052). The bias
seemed to be towards high tones as mentioned before. So, participants were significantly more biased when the baseline pupil size was large in the discrimination task.
Post-stimulus pupil size change
The mean course of the post-stimulus pupil response during a trial was plotted for the four different conditions in figure 4 on the left. The pupil response was also plotted for correct and incorrect, and confident and unconfident decisions in figure 4 on the right.
Next to the pre-stimulus pupil size, we also tested for post-stimulus pupil size. The mean pupil response was calculated between 1 and 2 s after the start of a trial. Again, only significant results will be reported since we are particularly interested in internally
Figure 3 Sensitivity and criterion were tested for baseline pupil size. The pupil data were split based on a median split per condition that was defined as the baseline pupil size from 0.5 s before stimulus onset until stimulus onset. (A, B) The 2x2 repeated measures ANOVA showed that there were significant effects of pre-stimulus pupil size for sensitivity in both detection (F1,30 = 4.178, p = 0.050, η2 = 0.010) and discrimination (F1,30 = 6.481, p = 0.016, η2 = 0.084). (C,D) Criterion appeared to be larger for large pre-stimulus pupil sizes (F1,30 = 9.702, p = 0.004, η2 = 0.052). The bias seemed to be towards high tones.
generated fluctuations. The test showed that the mean pupil size response was significantly higher for white than for black backgrounds (detection black: 0.554±0.558, detection white: 1.758±1.105, discrimination black: 0.462±0.682, discrimination white: 1.859±1.099 | F1,30 =
89.198, p < 0.001, η2 = 0.336). This suggested that pupils changed more in size when they
contracted through white backgrounds, than when they dilated through black backgrounds. We also tested for significant differences in post-stimulus pupil response between decisions. It seemed that the pupil response was significantly higher for correct than for incorrect answers (correct confident: 1.089±0.707, correct unconfident: 1.390±0.813, incorrect confident: 0.947±0.922, incorrect unconfident: 1.208±1.001 | F1,30 = 5.096, p = 0.031, η2 =
0.005). Likewise, the pupil response was significantly higher for unconfident than for confident decisions (F1,30 = 13.051, p = 0.001, η2 = 0.016). So, a large post-stimulus pupil
response could be associated with both a correct and unconfident answer.
Post-questionnaire data
Seven out of 31 participants expected a manipulation of the different backgrounds on their performance, which is 22.6%. These results are not used for any further analysis, since the groups were too small after classifying the participants.
Discussion
The effect of externally manipulating pupil size was investigated on perceptual decision making performance and meta-cognition in both auditory detection and discrimination tasks. The results showed that there was no effect of the colour manipulation on
metacognition and volume. Volume was used to describe performance, since performance was kept constant. Next to the external manipulation, we looked at internally generated fluctuations reflected by pupil size, during perceptual decision making. We split the data into one bin of trials where the baseline pupil size was small, and into one bin of trials where the baseline pupil size was large. Then, we found that sensitivity increased for detection and discriminations tasks when the pre-stimulus pupil size was large. Additionally, post-stimulus pupil size was generally larger during correct instead of incorrect decisions, and during
Figure 4 The mean course of post-stimulus pupil responses are shown during one trial. (A) The pupil response is presented for the four different conditions. There was significant effect of colour (F1,30 = 89.198, p < 0.001, η2 = 0.336), suggesting that pupils changed more in size when they contracted through white backgrounds, than when they dilated through black backgrounds. (B) The pupil response is presented for correct, incorrect, confident and unconfident decisions. The mean pupil response seemed to be significantly higher for correct instead of incorrect answers (F1,30 = 5.096, p = 0.031, η2 = 0.005), and for unconfident instead of confident answers (F1,30 = 13.051, p = 0.001, η2 = 0.016).
unconfident instead of confident decisions. In conclusion, externally manipulating pupil size did not have any influence on perceptual decision making, while on the other hand
perceptual decision making was associated with pre- and post-stimulus pupil size. The manipulation of background colours in the current research has made it possible to speak about causal relations and reciprocity. The different background colours did not cause a direct difference in performance and meta-cognition measures. This means that we were not able to replicate the findings of Mathôt and Ivanov (2019). Therefore, the results of Mathôt and Ivanov (2019) could be due to changing the amount of light on the retina, which is what we hypothesized. So, we assume that the relationship between cortical states and pupil size is nonreciprocal.
In the current research, it appeared that discrimination performance indicated by volume and reaction time, was higher than detection performance. For now, we assume that shorter reaction times correlate with a better performance (Correa et al., n.d.), which not
necessarily has to be the case. This superiority effect of discrimination over detection is found more often, and also in auditory experiments (Allik, Toom & Rauk, 2014; Lindner, 1968; Thomas, 1985).
Concerning the results of pre-stimulus pupil size, we found that sensitivity increased for detection and discrimination tasks when the pupil size was large. Yet, there has not been done a lot of research about the relationship between pre-stimulus pupil size and
performance. Nevertheless, Mathôt and Ivanov (2019) argued that a large pupil could lead to more emphasis on visual sensitivity, since large pupils were associated with higher arousal situations (Mathôt, 2018, Mathôt & Van der Stigchel, 2015). Our findings are in line with this conclusion of Mathôt and Ivanov (2019) and give some evidence for more emphasis on auditory sensitivity as well. Additionally, it seemed that only during discrimination tasks, participants were more biased when pre-stimulus pupil size was large. The two results of sensitivity and criterion seem contradictory as you would expect that a higher sensitivity is followed by a smaller bias. However, this does not change the result of sensitivity since bias seems to play a much more prominent role in detection instead of discrimination.
With regard to metacognition, there seemed to be no difference between a small and large baseline pupil size. The result seems not to be in line with the study of Correa et al. (n.d.), as these researchers concluded that a large pupil size was associated with worse
metacognition. However, they investigated post-stimulus pupil size whereas we looked for pre-stimulus pupil size. The negative correlation between pre- and post-stimulus size found by Hong, Walz and Sajda (2014) could explain the incoherence in these findings.
The results of the post-stimulus pupil responses showed that a change in pupil size was much larger for white than for black backgrounds. This effect could be explained by the fact that the room where the participants performed the experiment, was quite dark. So, a white background seemed to be of bigger contrast compared to the black background, which resulted in a more extreme change in pupil size. The results also showed that the pupil response was larger for correct than for incorrect answers, which is completely in accordance with the research of Correa et al. (n.d.), where they showed that a positive correlation between pupil size and performance. Furthermore, the pupil response seemed
to be larger for confident than for unconfident answers, which is again in line with Correa et al. (n.d.). Although Heaver and Hutton (2011) could not draw the same conclusion because of their research design, they also pointed in the direction of a correlation between large pupils and high-confident answers.
All in all, the effect of an external pupil size manipulation is investigated on auditory perceptual decision making performance and metacognition for the first time. The given results offer a few insights into the relationship between behaviour and pupil size that seems to reflect changes in cortical states. However, to get a better understanding of these results, it might be interesting in the future to measure both pupil size and internal
fluctuations with EEG, while participants perform the same kind of experiment. This research suggestion can build on the existing literature about the alpha rhythm that seems to be correlated with conscious visual perception (Babiloni, Vecchio, Bultrini, Romani and Rossini, 2006; Harris, Dux and Mattingley, 2018). Moreover, the suggestion will clarify to what extent there is a relationship between changes in pupil size and cortical states. Another research suggestion would be to improve the experiment by lengthening the duration of the inter-trial intervals, to make sure that the inter-trials will not influence each other. Namely, in our study the trials followed each other very quickly. If the duration will be extended, it will be
possible to investigate potential differences in cortical representations between for instance low and high tones, or between perceived and unperceived tones. Finally, previous studies in rats have shown that perceptual decision making performance was better during
locomotion (Bennett, Aroyo & Hestrin, 2013; Bullock, Elliott, Serences & Giesbrecht, 2017). We would suggest to investigate if this finding will be true for humans as well, and to what extent locomotion influences the cortical representations. Thus, more research has to be done to get a more complete account of the given results in the current research.
In summary, the current research does have some implications for the field of pupillometry as it appears that the relationship between pupil size and performance is completely nonreciprocal. So, pupil size has no influence on perceptual decision making, while both internal pre- and post-stimulus fluctuations do show an association. Therefore, the results of the current research have contributed to a better understanding of differences in perception during perceptual decision making.
References
Allik, J., Toom, M., & Rauk, M. (2014). Detection and identification of spatial offset: Double- judgment psychophysics revisited. Attention, Perception, & Psychophysics, 76(8), 2575-2583. doi: 10.3758/s13414-014-0717-0
Aston-Jones, G., & Cohen, J. D. (2005). An integrative theory of locus coeruleus- norepinephrine function: adaptive gain and optimal performance. Annu. Rev. Neurosci., 28, 403-450. doi: 10.1146/annurev.neuro.28.061604.135709
Babiloni, C., Vecchio, F., Bultrini, A., Luca Romani, G., & Rossini, P. M. (2005). Pre-and poststimulus alpha rhythms are related to conscious visual perception: a
high-resolution EEG study. Cerebral cortex, 16(12), 1690-1700. doi: 10.1093/cercor/bhj104 Beerendonk, L. (2017). Cortical state, neuromodulation, and conscious perception.
Unpublished master’s thesis. University of Amsterdam, Amsterdam, the Netherlands. Bennett, C., Arroyo, S., & Hestrin, S. (2013). Subthreshold mechanisms underlying state-
dependent modulation of visual responses. Neuron, 80(2), 350-357. doi: 10.1016/j.neuron.2013.08.007
Bullock, T., Elliott, J. C., Serences, J. T., & Giesbrecht, B. (2017). Acute exercise modulates feature-selective responses in human cortex. Journal of cognitive neuroscience, 29(4), 605-618. doi: 10.1162/jocn_a_O1082
Correa, C.M.C., de Gee, J.W., Weaver, M.D., Donner, T.H. & van Gaal, S. (n.d.). Pupil dilation reflects oppositely perceptual sensitivity and metacognitive insight during human perceptual decision-making. Manuscript in preparation.
De Gee, J.W., Knapen, T. & Donner, T.H. (2014). Decision-related pupil dilation reflects upcoming choice and individual bias. Proceedings of the National Academy of Sciences of the United States of America, 111(5), E618-E625. doi:
10.1073/pnas.1317557111
Gilzenrat, M. S., Nieuwenhuis, S., Jepma, M., & Cohen, J. D. (2010). Pupil diameter tracks changes in control state predicted by the adaptive gain theory of locus coeruleus function. Cognitive, Affective, & Behavioral Neuroscience, 10(2), 252-269.
doi:10.3758/CABN.10.2.252
Ginani, G. E., Tufik, S., Bueno, O. F. A., Pradella-Hallinan, M., Rusted, J., & Pompéia, S.
(2011). Acute effects of donepezil in healthy young adults underline the fractionation of executive functioning. Journal of Psychopharmacology, 25(11), 1508–16. doi: 10.1177/0269881110391832
Harris, A. M., Dux, P. E., & Mattingley, J. B. (2018). Awareness is related to reduced post- stimulus alpha power: a no-report inattentional blindness study. European Journal of Neuroscience. doi: 10.1111/ejn.13947
Harris, K. D., & Thiele, A. (2011). Cortical state and attention. Nature reviews neuroscience, 12(9), 509. doi: 10.1038/nrn3084
Hauser, T. U., Allen, M., Purg, N., Moutoussis, M., Rees, G., & Dolan, R. J. (2017).
Noradrenaline blockade specifically enhances metacognitive performance. Elife, 6, e24901. doi: 10.7554/eLife.24901.001
Heaver, B., & Hutton, S. B. (2011). Keeping an eye on the truth? Pupil size changes associated with recognition memory. Memory, 19(4), 398-405. doi:
10.1080/09658211.2011.575788
Heekeren, H.R., Marrett, S. & Ungerleider, L.G. (2008). The neural systems that mediate human perceptual decision making. Nature Reviews Neuroscience, 9(6), 467. doi: 10.1038/nrn2374
Hesselmann, G., Kell, C. A., Eger, E., & Kleinschmidt, A. (2008). Spontaneous local variations in ongoing neural activity bias perceptual decisions. Proceedings of the National Academy of Sciences, 105(31), 10984-10989. doi: 10.1073/pnas.0712043105 Hong, L., Walz, J. M., & Sajda, P. (2014). Your eyes give you away: prestimulus changes in
pupil diameter correlate with poststimulus task-related EEG dynamics. PLoS One, 9(3), e91321. doi: 10.1371/journal.pone.0091321
Iemi, L., Chaumon, M., Crouzet, S. M., & Busch, N. A. (2017). Spontaneous neural oscillations bias perception by modulating baseline excitability. Journal of Neuroscience, 37(4), 807-819. doi: 10.1523/JNEUROSCI.1432-16.2016
Joshi, S., Li, Y., Kalwani, R. M., & Gold, J. I. (2016). Relationships between pupil diameter and neuronal activity in the locus coeruleus, colliculi, and cingulate cortex. Neuron, 89(1), 221-234. doi: 10.1016/j.neuron.2015.11.028
Kaernbach, C. (1991). Simple adaptive testing with the weighted up-down method. Attention, Perception, & Psychophysics, 49(3), 227-229. doi: 10.3758/bf03214307
Kirstein, C. (2007). Sleeping and Dreaming. In R. Berry and M. Wagner (Ed.), Sleep Medicine Pearls (Third edition). Amsterdam, Netherlands: Elsevier Health Sciences.
Larsen, R. S., & Waters, J. (2018). Neuromodulatory correlates of pupil dilation. Frontiers in neural circuits, 12, 21. doi: 10.3389/fncir.2018.00021
Leske, S., Ruhnau, P., Frey, J., Lithari, C., Müller, N., Hartmann, T., & Weisz, N. (2015). Prestimulus network integration of auditory cortex predisposes near-threshold perception independently of local excitability. Cerebral cortex, 25(12), 4898-4907. doi: 10.1093/cercor/bhv212
Lindner, W. A. (1968). Recognition performance as a function of detection criterion in a simultaneous detection-recognition task. The Journal of the Acoustical Society of America, 44(1), 204-211. doi: 10.1121/1.1911056
Mathôt, S. (2018). Pupillometry: Psychology, physiology, and function. Journal of Cognition, 1(1). doi: 10.5334/joc.18
Mathôt, S. & Ivanov, Y. (2019). The effect of pupil size and peripheral brightness on
detection and discrimination performance. bioRxiv. Advance online publication. doi: 10.1101/624783
Mathôt, S., & Van der Stigchel, S. (2015). New light on the mind’s eye: The pupillary light response as active vision. Current directions in psychological science, 24(5), 374-378. doi: 10.1177/0963721415593725
Pinto, L., Goard, M. J., Estandian, D., Xu, M., Kwan, A. C., Lee, S. H., ... & Dan, Y. (2013). Fast modulation of visual perception by basal forebrain cholinergic neurons. Nature neuroscience, 16(12), 1857. doi: 10.1038/nn.3552
Polack, P. O., Friedman, J., & Golshani, P. (2013). Cellular mechanisms of brain state– dependent gain modulation in visual cortex. Nature neuroscience, 16(9), 1331. doi: 10.1038/nn.3464
Reimer, J., McGinley, M. J., Liu, Y., Rodenkirch, C., Wang, Q., McCormick, D. A., & Tolias, A. S. (2016). Pupil fluctuations track rapid changes in adrenergic and cholinergic activity in cortex. Nature communications, 7, 13289. doi: 10.1038/ncomms13289
Reimer, J., Froudarakis, E., Cadwell, C. R., Yatsenko, D., Denfield, G. H., & Tolias, A. S. (2014). Pupil fluctuations track fast switching of cortical states during quiet
wakefulness. Neuron, 84(2), 355-362. doi: 10.1016/j.neuron.2014.09.033 Samaha, J., Iemi, L., & Postle, B. R. (2017). Prestimulus alpha-band power biases visual
discrimination confidence, but not accuracy. Consciousness and cognition, 54, 47-55. doi: 10.1016/j.concog.2017.02.005
Summerfield, C., & De Lange, F. P. (2014). Expectation in perceptual decision making: neural and computational mechanisms. Nature Reviews Neuroscience, 15(11), 745-756. doi: 10.1038/nrn3838
Thaler, L., Schütz, A. C., Goodale, M. A., & Gegenfurtner, K. R. (2013). What is the best fixation target? The effect of target shape on stability of fixational eye
movements. Vision Research, 76, 31-42. doi: 10.1016/j.visres.2012.10.012
Thomas, J. P. (1985). Detection and identification: how are they related?. JOSA A, 2(9), 1457- 1467. doi: 10.1364/JOSAA.2.001457
perception of errors and its relation to the anterior insula. Brain Structure and Function, 214(5-6), 629-643. doi:10.1007/s00429-010-0261-1
Woestmann, M., Waschke, L., & Obleser, J. (2019). Prestimulus neural alpha power predicts confidence in discriminating identical auditory stimuli. European Journal of
Neuroscience, 49(1), 94-105. doi: 10.1111/ejn.14226
Appendices
Appendix A
Are you left or right handed?
Do you wear glasses or contacts? If so, what is your prescription?
What do you think we have investigated?
Do you think that the black and white screens have had any influence? If so, on what? And what kind of influence do you think they’ve had?
