Driving along a busy street requires you to constantly monitor the behavior of other road users. You need to be able to spot and avoid the car that suddenly changes lane, or the pedestrian who steps out in front of you. How fast you can react to such events depends in part on your brain's level of alertness, or 'arousal'. This in turn depends on chemicals within the brain called neuromodulators.

Neuromodulators are a type of neurotransmitter. But whereas other neurotransmitters enable brain cells to signal to each other, neuromodulators turn the volume of these signals up or down. The activity of brain regions that produce neuromodulators varies over time, leading to changes in brain arousal. These changes take place over different time scales. Sudden unexpected events, such as those on the busy street above, trigger sub-second changes in arousal. But arousal levels also show spontaneous fluctuations over minutes to hours. We can follow these changes in real-time by looking into a participant’s eyes. This is because the brain regions that produce neuromodulators also control pupil size.

Van Kempen et al. have now combined measurements of pupil size with recordings of electrical brain activity. Healthy volunteers learned to press a button as soon as a target appeared on a screen. The larger a volunteer’s pupils were before the target appeared, the more slowly the volunteer responded on that trial. Large baseline pupil size is thought to indicate a high baseline level of brain arousal. By contrast, the larger the increase in pupil size in response to the target, the faster the volunteer responded on that trial. This increase in pupil size is thought to reflect an increase in brain arousal.

The recordings of brain activity provided clues to the underlying mechanisms. In trials with large baseline pupil size – and therefore high baseline arousal – the volunteers’ brains showed more variable responses to the target. But in trials with a large increase in pupil size – and a large increase in arousal – the volunteers’ brains showed less variable responses, as well as stronger signals related to attention.

Neuromodulators thus act on different timescales to influence different aspects of cognitive performance, including attention and target detection. Fluctuating levels of neuromodulator activity may help explain the variability in our behavior. Monitoring pupil size is one way to gain insights into the mechanisms that bring about these changes in neuromodulator activity.

The speed and accuracy with which humans, as well as non-human animals, respond to a stimulus depends not only on the characteristics of the stimulus, but also on the cognitive state of the subject. When drowsy, a subject will respond more slowly to the same stimulus compared to when she is attentive and alert. Central arousal also fluctuates across a smaller range during quiet wakefulness, when the subject is neither drowsy or inattentive, nor overly excited or distractible. Although these trial-to-trial fluctuations can impact on behavioural performance during decision-making tasks (Aston-Jones and Cohen, 2005), it is largely unknown how arousal modulates the underlying processes that support decision formation. Perceptual decision-making depends on multiple neural processing stages that represent and select sensory information, those that process and accumulate sensory evidence, and those that prepare and execute motor commands. Variability in central arousal could affect any one or potentially all of these processing stages, which in turn could influence behavioural performance.

The neuromodulatory systems that control central arousal state, such as the noradrenergic (NA) locus coeruleus (LC) and the cholinergic basal forebrain (BF), have also been suggested to drive fluctuations in endogenous activity linked to changes in cortical (de)synchronization, that is cortical state (Harris and Thiele, 2011; Lee and Dan, 2012), and are linked to cognitive functions such as attention (Thiele and Bellgrove, 2018), both known to affect information processing and behavioural performance. These modulatory systems have both tonic and phasic firing patterns that are recruited on different timescales and support different functional roles (Aston-Jones and Cohen, 2005; Dayan and Yu, 2006; Parikh et al., 2007; Parikh and Sarter, 2008; Sarter et al., 2016). Tonic changes in neuromodulator activity occur over longer timescales that can span multiple trials, whereas fast (task-evoked) recruitment through phasic activation occurs on short enough timescales to influence neural activity and behavioural decisions within the same trial (Aston-Jones and Cohen, 2005; Bouret and Sara, 2005; Dayan and Yu, 2006; Parikh et al., 2007).

Pupil diameter correlates strongly with a variety of measurements of cortical state and behavioural arousal (Eldar et al., 2013; Reimer et al., 2014; McGinley et al., 2015b; McGinley et al., 2015a; Vinck et al., 2015; Engel et al., 2016), and can thus be considered a reliable proxy of central arousal state. Indeed, there is a strong correlation between pupil size and activity in various neuromodulatory centres that control arousal (Aston-Jones and Cohen, 2005; Gilzenrat et al., 2010; Murphy et al., 2014a; Varazzani et al., 2015; Joshi et al., 2016; Reimer et al., 2016; de Gee et al., 2017). Both baseline pupil diameter, reflecting tonic activity levels in neuromodulatory centres (tonic arousal), and task-evoked pupil diameter changes (phasic arousal), have been related to specific neural processing stages of perceptual decision making. Baseline pupil diameter correlates with sensory sensitivity (McGinley et al., 2015a; McGinley et al., 2015b) and is predictive of behavioural performance during elementary detection tasks (Murphy et al., 2011; McGinley et al., 2015a). Pupil diameter also changes phasically in the course of a single decision (Beatty, 1982a; de Gee et al., 2014; de Gee et al., 2017; Lempert et al., 2015; Murphy et al., 2016; Urai et al., 2017), and has been related to specific elements of the decision making process, such as decision bias (de Gee et al., 2014; de Gee et al., 2017), uncertainty (Urai et al., 2017), and urgency (Murphy et al., 2016). This suggests that these neuromodulatory systems do not only dictate network states (through tonic activity changes), but that they are recruited throughout the decision making process (Cheadle et al., 2014; de Gee et al., 2014; de Gee et al., 2017). Although both baseline pupil diameter and the phasic pupil response have been associated with specific aspects of decision-making, the relationship between pupil-linked arousal and the electrophysiological correlates of decision-making are largely unknown.

Recently developed behavioural paradigms have made it possible to non-invasively study the individual electroencephalographic (EEG) signatures of perceptual decision-making described above (O'Connell et al., 2012; Kelly and O'Connell, 2013; Loughnane et al., 2016; Loughnane et al., 2018; Newman et al., 2017). In these paradigms, participants are required to continuously monitor (multiple) stimuli for subtle changes in a feature. Because stimuli are presented continuously, target onset times (and locations) are unpredictable, and sudden stimulus onsets are absent, eliminating sensory evoked deflections in the EEG traces. These characteristics allow for the investigation of the gradual development of build-to-threshold decision variables as well as signals that code for the selection of relevant information from multiple competing stimuli, a critical feature of visuospatial attentional orienting that impact evidence accumulation processes (Loughnane et al., 2016).

Here, we asked how arousal influences EEG signals that relate to each of the separate processing stages described above. Specifically, we tested the effects of pupil-linked arousal on pre-target preparatory parieto-occipital α-band activity, associated with fluctuations in the allocation of attentional resources (Kelly and O'Connell, 2013); early target selection signals measured over contra- and ipsilateral occipital cortex, the N2c and N2i (Loughnane et al., 2016); perceptual evidence accumulation signals measured as the centroparietal positivity (CPP), which is a build-to-threshold decision variable demonstrated to scale with the strength of sensory evidence and predictive of reaction time (RT) (O'Connell et al., 2012; Kelly and O'Connell, 2013); and motor-preparation signals measured via contralateral β-band activity (Donner et al., 2009; O'Connell et al., 2012). Of these signals, we extracted specific characteristics such as the latency, build-up rate and amplitude, and tested whether these were affected by pupil-linked arousal. Additionally, because the variance and response reliability of the membrane potential of sensory neurons varies with pupil diameter (Reimer et al., 2014; McGinley et al., 2015a), we also investigated whether arousal affected the inter-trial phase coherence (ITPC), a measure of across trial consistency in the EEG signal, of the N2 and the CPP.

We found that both baseline pupil diameter as well as the pupil response were predictive of behavioural performance, and that this relationship was best described by non-monotonic, but not U-shaped, second-order polynomial model fits. Furthermore, we found that both tonic and phasic arousal bore a predictive relationship with the neural signals coding for baseline attentional engagement, early target selection, decision processing as well as the preparatory motor response. Although neural activity representing all these stages varied with changes in arousal, unique variability in task performance due to tonic arousal (baseline pupil diameter) could only be explained by the amplitude of target selection signals and the consistency of the CPP, reflecting decision processing. In contrast, variability due to phasic arousal (pupil response) was explained by pre-target α-band activity as well as the consistency of the CPP.

80 subjects performed a continuous version of the random dot motion task in which they were asked to report temporally and spatially unpredictable periods of coherent motion within either of two streams of random motion (Figure 1A). We investigated whether the trial-to-trial fluctuations in behavioural performance and EEG signatures of perceptual decision making could, in part, be explained by trial-to-trial differences in the size of the baseline pupil diameter (reflecting tonic arousal) and the post-target pupil response (reflecting phasic arousal). We quantified this relationship by allocating data into five bins based on the size of either the baseline pupil diameter or the phasic pupil diameter response (Figure 1B & C). Baseline pupil diameter was computed as the average pupil diameter over the 100 ms preceding target onset. The phasic pupillary response was estimated using a single trial general linear model (GLM) approach (Materials and methods). We first assessed the neural input to the peripheral pupil system by applying multiple models with onset and response components as well as various different shapes for the sustained component (Murphy et al., 2016) across all trials for each subject (Figure 1—figure supplement 1). Next we applied the grand average best-fitting model (linear up-ramp) on individual trials (Bach et al., 2018). This provided us with a trial-by-trial estimate of the amplitude of each temporal component. Comparison against several other measures of the pupil diameter response (Figure 1—figure supplement 2), controlling for variance inflation factors (Figure 1—figure supplement 3) and applying the same model across bins of trials, or orthogonalizing the predictors (Figure 1—figure supplement 4) provided support for the reliability of the estimated amplitude of the pupillary response. Here we present the relationship of the amplitude of the target onset component to the behavioural and EEG signatures of perceptual decision-making. We then used sequential multilevel model analyses and maximum likelihood ratio tests to test for fixed effects of pupil bin. We determined whether a linear fit was better than a constant fit and subsequently whether the fit of a second-order polynomial, indicating a non-monotonic relationship between pupil diameter and behaviour/EEG, was superior to a linear fit. We furthermore used a variant of the ‘two-lines’ approach (Simonsohn, 2017) to test whether any non-monotonic relationship was best described by an (inverted) U-shape.

Figure 1 with 6 supplements see all

Download asset Open asset

Figure 1—source data 1

Csv table containing data for Figure 1 panel B.

https://doi.org/10.7554/eLife.42541.010

Download elife-42541-fig1-data1-v2.csv

Figure 1—source data 2

Csv table containing data for Figure 1 panel C.

https://doi.org/10.7554/eLife.42541.011

Download elife-42541-fig1-data2-v2.csv

Phasic arousal has an approximately linear relationship with decision processing

During decision making, perceptual evidence has to be accumulated over time. This accumulation process has long been related to build-to-threshold activity in single neurons in parietal cortex (Gold and Shadlen, 2007); but see Latimer et al., 2015, Latimer et al., 2016; Shadlen et al., 2016). The centro-parietal positivity (CPP) measured from scalp EEG exhibits many of these same properties, including a representation of the accumulation of sensory evidence towards a decision bound (O'Connell et al., 2012; O'Connell et al., 2018; Kelly and O'Connell, 2013). Because in this study we used relatively strong sensory evidence (50% coherence), it is possible that subjects may not have relied upon any temporal integration of this motion signal to reach a decision. Rather, variability in RT could be brought about by variation in the onset transient of target selection due to the temporal and spatial uncertainty of the target stimulus. On single trials, decision formation could be a step-like signal that averaged across trials looks like an accumulate-to-bound signal (Latimer et al., 2015). Although we cannot discount this possibility, aligning the visual early target selection signals (N2c) to response reveals a much lower signal amplitude compared to aligning it to target onset (Figure 2—figure supplement 1). This indicates that there is no fixed delay between target selection and the response, and that there is variability in the duration of the sustained period of this task. This variation could indicate different trial-to-trial strategies (e.g. comparing motion in one stimulus against the stimulus on the other side of the screen), or in addition some variability in accumulation rate. Because of this uncertainty, we refer to the functional significance of the CPP as decision processing.

Here we tested the relationship between the pupil diameter response and the onset, build-up rate, amplitude and consistency (ITPC) of the CPP (Figure 2). We found that the onset latency of the CPP, defined as the first time point that showed a significant difference from zero for 15 consecutive time points, displayed an inverse monotonic relationship with the size of the pupil response (χ²₍₁₎=5.60, p=0.018), such that the fastest onsets were found for the largest pupil response bins (Figure 2A). Likewise, the build-up rate (χ²₍₁₎=4.45, p=0.035), but not the amplitude (p=0.15), of the CPP varied with the pupil response, displaying the steepest slope on trials with the largest pupil dilations. Because the membrane potential of sensory neurons shows the least variance and highest response reliability at intermediate baseline pupil diameter (McGinley et al., 2015a), we additionally investigated the ITPC, a measure of across trial consistency, of the CPP. We computed ITPC with a single-taper spectral analysis in a 512 ms sliding window computed at 50 ms intervals, with a frequency resolution of 1.95 Hz (Materials and methods). Based on the stimulus-locked grand average time-frequency spectrum, we selected a time (300–550 ms) and frequency window (<4 Hz) for further statistical analyses (Figure 2C). We found an approximately linear relationship between pupil diameter response and the consistency of the CPP signal (χ²₍₁₎=41.79, p<0.001), indicating that the CPP signal is less variable for larger pupil response bins (Figure 2D). This relationship was also present when we aligned the CPP to the response (Figure 2—figure supplement 2), indicating that this effect is unlikely to solely reflect variability in the onset of the CPP. Thus, we found that the size of the pupillary response was predictive of both the onset latency, build-up rate as well as the ITPC of the CPP. Moreover, the relationship with the neural parameters of the CPP resembled the relationship between the pupil response and behavioural performance (Figure 1C). Large pupil dilations were predictive of faster responses, earlier CPP onset latencies, as well as steeper build-up rates and more consistent CPP. Next, we asked whether other stages of information processing underpinning perceptual decision making also varied with the pupil response.

Figure 2 with 2 supplements see all

Download asset Open asset

Figure 2—source data 1

Csv table containing data for Figure 2.

https://doi.org/10.7554/eLife.42541.015

Download elife-42541-fig2-data1-v2.csv

The phasic pupil response relates monotonically to spectral measures of baseline attentional engagement, but not motor output

We next investigated pre-target preparatory α-band power (8–13 Hz), a sensitive index of attentional deployment that has been shown to vary with behavioural performance. Specifically, previous studies have found higher pre-target α-band power preceding trials with longer RT, and suggested that fluctuations in α-power may reflect an attentional influence on variability in task performance (Ergenoglu et al., 2004; van Dijk et al., 2008; O'Connell et al., 2009; Kelly and O'Connell, 2013). We first verified the relationship between α-band power and behavioural performance by binning the data into five bins according to α-band power and performing the same sequential regression analysis as described above (Figure 3A). We replicated previous findings (Kelly and O'Connell, 2013) and found an approximately linear relationship between α-band power and RT (χ²₍₁₎=25.27, p<0.001) but not RTcv (p=0.48). In line with previous research (Hong et al., 2014), pupil diameter response was inversely related to α-band power (Figure 3B), displaying an approximately linear relationship (χ²₍₁₎=28.24, p<0.001), suggesting that pre-target attentional engagement is related to phasic arousal.

Figure 3 with 1 supplement see all

Download asset Open asset

Figure 3—source data 1

Csv table containing data for Figure 3 panel A.

https://doi.org/10.7554/eLife.42541.017

Download elife-42541-fig3-data1-v2.csv

Figure 3—source data 2

Csv table containing data for Figure 3 panel B-F.

https://doi.org/10.7554/eLife.42541.018

Download elife-42541-fig3-data2-v2.csv

We next focused on response-related motor activity in the form of left hemispheric β-power (LHB). LHB decreases before a button press and has been shown to reflect the motor-output stage of perceptual decision making, but also to trace decision formation, reflecting the build-up of choice selective activity (Donner et al., 2009). Here we investigated the LHB amplitude and build-up rate preceding response (Figure 3C). We found that neither LHB amplitude (p=0.63) nor LHB slope (p=0.20) varied with phasic arousal, suggesting that phasic arousal does not influence the build-up rate of choice-related activity over motor cortex.

Baseline pupil diameter is inversely related to the consistency of decision processing

Figure 4 illustrates the relationship between baseline pupil diameter and the CPP. Unlike the pupil response, baseline pupil diameter was not predictive of the onset (p=0.20) or build-up rate (p=0.12), but it displayed an inverse relationship with both the amplitude (χ²₍₁₎=7.09, p=0.01) and the consistency of the CPP (χ²₍₁₎=9.34, p=0.002). In line with previous research that revealed increased variability in the rate of evidence accumulation during periods with larger baseline pupil diameter (Murphy et al., 2014b), we found an inverse, approximately linear, relationship in which higher baseline pupil diameter displayed lower EEG signal consistency (Figure 4D). Thus, states of higher arousal are characterized by less consistency, that is more variability, in decision processing. Additionally, states of higher tonic arousal also display lower task performance (Figure 1C), indicating that higher variability in decision processing (due to higher tonic arousal) can affect behavioural performance.

Figure 4

Download asset Open asset

Figure 4—source data 1

Csv table containing data for Figure 4.

https://doi.org/10.7554/eLife.42541.022

Download elife-42541-fig4-data1-v2.csv

Baseline pupil diameter relates to spectral measures of baseline attentional engagement and motor output as well as early target selection signals

We found a relationship between baseline pupil diameter and specific characteristics of multiple neural processing stages of perceptual decision-making. Specifically, as observed before (Hong et al., 2014), pre-target alpha power (Figure 5A) varied with baseline pupil diameter in a non-monotonic, but not inverted-U shaped, manner (χ²₍₁₎=4.49, p=0.034). This suggests that with higher tonic arousal, alpha activity is higher (or less desynchronised). Next, we tested whether baseline pupil diameter was predictive of EEG characteristics representing motor output (Figure 5B). We found an inverse relationship with LHB build-up rate (χ²₍₁₎=10.99, p<0.001), decreasing with larger baseline pupil diameter, but we did not find a relationship with LHB amplitude (p=0.34).

Figure 5

Download asset Open asset

Figure 5—source data 1

Csv table containing data for Figure 5.

https://doi.org/10.7554/eLife.42541.024

Download elife-42541-fig5-data1-v2.csv

Lastly, we investigated whether baseline pupil diameter affected early target selection signals, the N2 (Figure 5C–D). Previous studies have revealed that baseline pupil diameter affected the size and variability of neural responses to visual and auditory stimuli (Reimer et al., 2014; McGinley et al., 2015a). Here we found that baseline pupil diameter was not predictive of the peak latency of the N2c (p=0.75), but that it did display a monotonic relationship with the N2c amplitude (χ²₍₁₎=13.72, p<0.001). Trials with larger baseline pupil diameter displayed smaller N2c amplitudes, suggesting that higher arousal has a negative impact on sensory encoding. N2c ITPC did not vary with baseline pupil diameter (p=0.25), and nor did N2i ITPC (p=0.33), N2i latency (p=0.78) or amplitude (p=0.06). We thus found that, similar to the phasic pupil diameter response, baseline pupil diameter is predictive of specific characteristics of each of the processing stages of perceptual decision-making. Next, we investigated which of these components explained unique variance in task performance across pupil size bins.

Here we investigated whether behavioural and neural correlates of decision-making varied as a function of baseline or task-evoked pupil diameter, indexing tonic and phasic arousal, respectively. The perceptual decision-making paradigm employed (Figure 1A) allowed us to monitor the relationship between pupil diameter and independent measures of attentional engagement, early target selection, decision formation and motor output. We found that the trial-by-trial variability in both tonic and phasic arousal were predictive of behavioural performance (Figure 1B & C). For tonic arousal, this relationship was best described by a non-monotonic polynomial fit with slower RT for higher baseline pupil diameter. Higher phasic arousal, on the other hand, was predictive of better task performance.

We furthermore established that both tonic and phasic arousal were predictive of a subset of EEG signatures, together reflecting discrete aspects of information processing underpinning perceptual decision-making. A hierarchical regression analysis allowed us to determine which of these processing stages exerted an independent influence on behavioural performance associated with central arousal. We found that pre-target α power, indexing baseline attentional engagement, and the consistency of the CPP, reflecting the decision formation, each explained unique variability in task performance that was due to variability in phasic arousal. Variability in task performance due to fluctuations in tonic arousal was explained by the amplitude of the target selection signal N2c and the consistency of the CPP. We thus revealed a direct relationship between both tonic and phasic measures of arousal, and a distinct but overlapping set of EEG signatures of perceptual decision-making, and in particular the CPP.

All data have been deposited at https://figshare.com/s/8d6f461834c47180a444, in association with Newman et al (2017). All analysis scripts are publicly available at https://github.com/jochemvankempen/2019_pupil_decisionMaking (copy archived at https://github.com/elifesciences-publications/2019_pupil_decisionMaking).

1. 1. Aston-Jones G
   2. Rajkowski J
   3. Kubiak P
   4. Alexinsky T

   (1994) Locus coeruleus neurons in monkey are selectively activated by attended cues in a vigilance task

   The Journal of Neuroscience 14:4467–4480.

   https://doi.org/10.1523/JNEUROSCI.14-07-04467.1994

   • PubMed
   • Google Scholar
2. 1. Aston-Jones G
   2. Rajkowski J
   3. Kubiak P

   (1997) Conditioned responses of monkey locus coeruleus neurons anticipate acquisition of discriminative behavior in a vigilance task

   Neuroscience 80:697–715.

   https://doi.org/10.1016/S0306-4522(97)00060-2

   • PubMed
   • Google Scholar
3. 1. Aston-Jones G
   2. Cohen JD

   (2005) An integrative theory of locus coeruleus-norepinephrine function: adaptive gain and optimal performance

   Annual Review of Neuroscience 28:403–450.

   https://doi.org/10.1146/annurev.neuro.28.061604.135709

   • PubMed
   • Google Scholar
4. 1. Bach DR
   2. Tzovara A
   3. Vunder J

   (2018) Blocking human fear memory with the matrix metalloproteinase inhibitor doxycycline

   Molecular Psychiatry 23:1584–1589.

   https://doi.org/10.1038/mp.2017.65

   • PubMed
   • Google Scholar
5. 1. Bates D
   2. Mächler M
   3. Bolker B
   4. Walker S

   (2015) Fitting linear Mixed-Effects models using lme4

   Journal of Statistical Software 67:.

   https://doi.org/10.18637/jss.v067.i01

   • Google Scholar
6. 1. Beatty J

   (1982a) Task-evoked pupillary responses, processing load, and the structure of processing resources

   Psychological Bulletin 91:276–292.

   https://doi.org/10.1037/0033-2909.91.2.276

   • PubMed
   • Google Scholar
7. 1. Beatty J

   (1982b) Phasic not tonic pupillary responses vary with auditory vigilance performance

   Psychophysiology 19:167–172.

   https://doi.org/10.1111/j.1469-8986.1982.tb02540.x

   • PubMed
   • Google Scholar
8. 1. Bellgrove MA
   2. Hester R
   3. Garavan H

   (2004) The functional neuroanatomical correlates of response variability: evidence from a response inhibition task

   Neuropsychologia 42:1910–1916.

   https://doi.org/10.1016/j.neuropsychologia.2004.05.007

   • PubMed
   • Google Scholar
9. 1. Berens P

   (2009)

   CircStat : a MATLAB toolbox for circular statistics

   Journal of Statistical Software 31:.

   • Google Scholar
10. 1. Bokil H
    2. Andrews P
    3. Kulkarni JE
    4. Mehta S
    5. Mitra PP

    (2010) Chronux: a platform for analyzing neural signals

    Journal of Neuroscience Methods 192:146–151.

    https://doi.org/10.1016/j.jneumeth.2010.06.020

    • PubMed
    • Google Scholar
11. 1. Bouret S
    2. Sara SJ

    (2005) Network reset: a simplified overarching theory of locus coeruleus noradrenaline function

    Trends in Neurosciences 28:574–582.

    https://doi.org/10.1016/j.tins.2005.09.002

    • PubMed
    • Google Scholar
12. 1. Britten KH
    2. Shadlen MN
    3. Newsome WT
    4. Movshon JA

    (1992) The analysis of visual motion: a comparison of neuronal and psychophysical performance

    The Journal of Neuroscience 12:4745–4765.

    https://doi.org/10.1523/JNEUROSCI.12-12-04745.1992

    • PubMed
    • Google Scholar
13. 1. Cheadle S
    2. Wyart V
    3. Tsetsos K
    4. Myers N
    5. de Gardelle V
    6. Herce Castañón S
    7. Summerfield C

    (2014) Adaptive gain control during human perceptual choice

    Neuron 81:1429–1441.

    https://doi.org/10.1016/j.neuron.2014.01.020

    • PubMed
    • Google Scholar
14. 1. Costagli M
    2. Ueno K
    3. Sun P
    4. Gardner JL
    5. Wan X
    6. Ricciardi E
    7. Pietrini P
    8. Tanaka K
    9. Cheng K

    (2014) Functional signalers of changes in visual stimuli: cortical responses to increments and decrements in motion coherence

    Cerebral Cortex 24:110–118.

    https://doi.org/10.1093/cercor/bhs294

    • PubMed
    • Google Scholar
15. 1. Dayan P
    2. Yu AJ

    (2006) Phasic norepinephrine: a neural interrupt signal for unexpected events

    Network: Computation in Neural Systems 17:335–350.

    https://doi.org/10.1080/09548980601004024

    • Google Scholar
16. 1. de Gee JW
    2. Knapen T
    3. Donner TH

    (2014) Decision-related pupil dilation reflects upcoming choice and individual bias

    PNAS 111:E618–E625.

    https://doi.org/10.1073/pnas.1317557111

    • PubMed
    • Google Scholar
17. 1. de Gee JW
    2. Colizoli O
    3. Kloosterman NA
    4. Knapen T
    5. Nieuwenhuis S
    6. Donner TH

    (2017) Dynamic modulation of decision biases by brainstem arousal systems

    eLife 6:e23232.

    https://doi.org/10.7554/eLife.23232

    • PubMed
    • Google Scholar
18. 1. Delorme A
    2. Makeig S

    (2004) EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis

    Journal of Neuroscience Methods 134:9–21.

    https://doi.org/10.1016/j.jneumeth.2003.10.009

    • PubMed
    • Google Scholar
19. 1. Donner TH
    2. Siegel M
    3. Fries P
    4. Engel AK

    (2009) Buildup of choice-predictive activity in human motor cortex during perceptual decision making

    Current Biology 19:1581–1585.

    https://doi.org/10.1016/j.cub.2009.07.066

    • PubMed
    • Google Scholar
20. 1. Ebitz RB
    2. Pearson JM
    3. Platt ML

    (2014) Pupil size and social vigilance in rhesus macaques

    Frontiers in Neuroscience 8:.

    https://doi.org/10.3389/fnins.2014.00100

    • PubMed
    • Google Scholar
21. 1. Eldar E
    2. Cohen JD
    3. Niv Y

    (2013) The effects of neural gain on attention and learning

    Nature Neuroscience 16:1146–1153.

    https://doi.org/10.1038/nn.3428

    • PubMed
    • Google Scholar
22. 1. Engel TA
    2. Steinmetz NA
    3. Gieselmann MA
    4. Thiele A
    5. Moore T
    6. Boahen K

    (2016) Selective modulation of cortical state during spatial attention

    Science 354:1140–1144.

    https://doi.org/10.1126/science.aag1420

    • PubMed
    • Google Scholar
23. 1. Ergenoglu T
    2. Demiralp T
    3. Bayraktaroglu Z
    4. Ergen M
    5. Beydagi H
    6. Uresin Y

    (2004) Alpha rhythm of the EEG modulates visual detection performance in humans

    Cognitive Brain Research 20:376–383.

    https://doi.org/10.1016/j.cogbrainres.2004.03.009

    • PubMed
    • Google Scholar
24. Book

    1. Fisher NI

    (1993)

    Statistical Analysis of Circular Data

    cambridge university press.

    • Google Scholar
25. 1. Gilzenrat MS
    2. Nieuwenhuis S
    3. Jepma M
    4. Cohen JD

    (2010) Pupil diameter tracks changes in control state predicted by the adaptive gain theory of locus coeruleus function

    Cognitive, Affective, & Behavioral Neuroscience 10:252–269.

    https://doi.org/10.3758/CABN.10.2.252

    • PubMed
    • Google Scholar
26. 1. Gold JI
    2. Shadlen MN

    (2007) The neural basis of decision making

    Annual Review of Neuroscience 30:535–574.

    https://doi.org/10.1146/annurev.neuro.29.051605.113038

    • PubMed
    • Google Scholar
27. 1. Gritton HJ
    2. Howe WM
    3. Mallory CS
    4. Hetrick VL
    5. Berke JD
    6. Sarter M

    (2016) Cortical cholinergic signaling controls the detection of cues

    PNAS 113:E1089–E1097.

    https://doi.org/10.1073/pnas.1516134113

    • PubMed
    • Google Scholar
28. 1. Harris KD
    2. Thiele A

    (2011) Cortical state and attention

    Nature Reviews Neuroscience 12:509–523.

    https://doi.org/10.1038/nrn3084

    • PubMed
    • Google Scholar
29. 1. Hoeks B
    2. Levelt WJM

    (1993) Pupillary dilation as a measure of attention: a quantitative system analysis

    Behavior Research Methods, Instruments, & Computers 25:16–26.

    https://doi.org/10.3758/BF03204445

    • Google Scholar
30. 1. Hong L
    2. Walz JM
    3. Sajda P

    (2014) Your eyes give you away: prestimulus changes in pupil diameter correlate with poststimulus task-related EEG dynamics

    PLoS ONE 9:e91321.

    https://doi.org/10.1371/journal.pone.0091321

    • PubMed
    • Google Scholar
31. Book

    1. James G
    2. Witten D
    3. Hastie T
    4. Tibshirani R

    (2017)

    An Introduction to Statistical Learning (Eighth ed)

    New York: Springer Science.

    • Google Scholar
32. 1. Joshi S
    2. Li Y
    3. Kalwani RM
    4. Gold JI

    (2016) Relationships between pupil diameter and neuronal activity in the locus coeruleus, Colliculi, and cingulate cortex

    Neuron 89:221–234.

    https://doi.org/10.1016/j.neuron.2015.11.028

    • PubMed
    • Google Scholar
33. 1. Kayser J
    2. Tenke CE

    (2006) Principal components analysis of laplacian waveforms as a generic method for identifying ERP generator patterns: I. evaluation with auditory oddball tasks

    Clinical Neurophysiology 117:348–368.

    https://doi.org/10.1016/j.clinph.2005.08.034

    • PubMed
    • Google Scholar
34. 1. Kelly SP
    2. O'Connell RG

    (2013) Internal and external influences on the rate of sensory evidence accumulation in the human brain

    Journal of Neuroscience 33:19434–19441.

    https://doi.org/10.1523/JNEUROSCI.3355-13.2013

    • PubMed
    • Google Scholar
35. 1. Kloosterman NA
    2. Meindertsma T
    3. van Loon AM
    4. Lamme VA
    5. Bonneh YS
    6. Donner TH

    (2015) Pupil size tracks perceptual content and surprise

    European Journal of Neuroscience 41:1068–1078.

    https://doi.org/10.1111/ejn.12859

    • PubMed
    • Google Scholar
36. 1. Kristjansson SD
    2. Stern JA
    3. Brown TB
    4. Rohrbaugh JW

    (2009) Detecting phasic lapses in alertness using pupillometric measures

    Applied Ergonomics 40:978–986.

    https://doi.org/10.1016/j.apergo.2009.04.007

    • PubMed
    • Google Scholar
37. 1. Latimer KW
    2. Yates JL
    3. Meister ML
    4. Huk AC
    5. Pillow JW

    (2015) NEURONAL MODELING. Single-trial spike trains in parietal cortex reveal discrete steps during decision-making

    Science 349:184–187.

    https://doi.org/10.1126/science.aaa4056

    • PubMed
    • Google Scholar
38. 1. Latimer KW
    2. Yates JL
    3. Meister ML
    4. Huk AC
    5. Pillow JW

    (2016) Response to comment on "Single-trial spike trains in parietal cortex reveal discrete steps during decision-making"

    Science 351:1406.

    https://doi.org/10.1126/science.aad3596

    • PubMed
    • Google Scholar
39. 1. Lee SH
    2. Dan Y

    (2012) Neuromodulation of brain states

    Neuron 76:209–222.

    https://doi.org/10.1016/j.neuron.2012.09.012

    • PubMed
    • Google Scholar
40. 1. Lempert KM
    2. Chen YL
    3. Fleming SM

    (2015) Relating pupil dilation and metacognitive confidence during auditory Decision-Making

    Plos One 10:e0126588.

    https://doi.org/10.1371/journal.pone.0126588

    • PubMed
    • Google Scholar
41. 1. Loughnane GM
    2. Newman DP
    3. Bellgrove MA
    4. Lalor EC
    5. Kelly SP
    6. O'Connell RG

    (2016) Target selection signals influence perceptual decisions by modulating the onset and rate of evidence accumulation

    Current Biology 26:496–502.

    https://doi.org/10.1016/j.cub.2015.12.049

    • PubMed
    • Google Scholar
42. 1. Loughnane GM
    2. Newman DP
    3. Tamang S
    4. Kelly SP
    5. O'Connell RG

    (2018) Antagonistic interactions between microsaccades and evidence accumulation processes during decision formation

    The Journal of Neuroscience 38:2163–2176.

    https://doi.org/10.1523/JNEUROSCI.2340-17.2018

    • PubMed
    • Google Scholar
43. 1. McGinley MJ
    2. David SV
    3. McCormick DA

    (2015a) Cortical membrane potential signature of optimal states for sensory signal detection

    Neuron 87:179–192.

    https://doi.org/10.1016/j.neuron.2015.05.038

    • PubMed
    • Google Scholar
44. 1. McGinley MJ
    2. Vinck M
    3. Reimer J
    4. Batista-Brito R
    5. Zagha E
    6. Cadwell CR
    7. Tolias AS
    8. Cardin JA
    9. McCormick DA

    (2015b) Waking state: rapid variations modulate neural and behavioral responses

    Neuron 87:1143–1161.

    https://doi.org/10.1016/j.neuron.2015.09.012

    • PubMed
    • Google Scholar
45. 1. Murphy PR
    2. Robertson IH
    3. Balsters JH
    4. O'connell RG

    (2011) Pupillometry and P3 index the locus coeruleus-noradrenergic arousal function in humans

    Psychophysiology 48:1532–1543.

    https://doi.org/10.1111/j.1469-8986.2011.01226.x

    • PubMed
    • Google Scholar
46. 1. Murphy PR
    2. O'Connell RG
    3. O'Sullivan M
    4. Robertson IH
    5. Balsters JH

    (2014a) Pupil diameter covaries with BOLD activity in human locus coeruleus

    Human Brain Mapping 35:4140–4154.

    https://doi.org/10.1002/hbm.22466

    • PubMed
    • Google Scholar
47. 1. Murphy PR
    2. Vandekerckhove J
    3. Nieuwenhuis S

    (2014b) Pupil-linked arousal determines variability in perceptual decision making

    PLoS Computational Biology 10:e1003854.

    https://doi.org/10.1371/journal.pcbi.1003854

    • PubMed
    • Google Scholar
48. 1. Murphy PR
    2. Boonstra E
    3. Nieuwenhuis S

    (2016) Global gain modulation generates time-dependent urgency during perceptual choice in humans

    Nature Communications 7:13526.

    https://doi.org/10.1038/ncomms13526

    • PubMed
    • Google Scholar
49. 1. Newman DP
    2. Lockley SW
    3. Loughnane GM
    4. Martins AC
    5. Abe R
    6. Zoratti MT
    7. Kelly SP
    8. O'Neill MH
    9. Rajaratnam SM
    10. O'Connell RG
    11. Bellgrove MA

    (2016) Ocular exposure to blue-enriched light has an asymmetric influence on neural activity and spatial attention

    Scientific Reports 6:27754.

    https://doi.org/10.1038/srep27754

    • PubMed
    • Google Scholar
50. 1. Newman DP
    2. Loughnane GM
    3. Kelly SP
    4. O'Connell RG
    5. Bellgrove MA

    (2017) Visuospatial asymmetries arise from differences in the onset time of perceptual evidence accumulation

    The Journal of Neuroscience 37:3378–3385.

    https://doi.org/10.1523/JNEUROSCI.3512-16.2017

    • PubMed
    • Google Scholar
51. 1. Newsome WT
    2. Britten KH
    3. Movshon JA

    (1989) Neuronal correlates of a perceptual decision

    Nature 341:52–54.

    https://doi.org/10.1038/341052a0

    • PubMed
    • Google Scholar
52. 1. O'Connell RG
    2. Dockree PM
    3. Robertson IH
    4. Bellgrove MA
    5. Foxe JJ
    6. Kelly SP

    (2009) Uncovering the neural signature of lapsing attention: electrophysiological signals predict errors up to 20 s before they occur

    Journal of Neuroscience 29:8604–8611.

    https://doi.org/10.1523/JNEUROSCI.5967-08.2009

    • PubMed
    • Google Scholar
53. 1. O'Connell RG
    2. Dockree PM
    3. Kelly SP

    (2012) A supramodal accumulation-to-bound signal that determines perceptual decisions in humans

    Nature Neuroscience 15:1729–1735.

    https://doi.org/10.1038/nn.3248

    • PubMed
    • Google Scholar
54. 1. O'Connell RG
    2. Shadlen MN
    3. Wong-Lin K
    4. Kelly SP

    (2018) Bridging neural and computational viewpoints on perceptual Decision-Making

    Trends in Neurosciences 41:838–852.

    https://doi.org/10.1016/j.tins.2018.06.005

    • PubMed
    • Google Scholar
55. 1. O’brien RM

    (2007) A caution regarding rules of thumb for variance inflation factors

    Quality & Quantity 41:673–690.

    https://doi.org/10.1007/s11135-006-9018-6

    • Google Scholar
56. 1. Parikh V
    2. Kozak R
    3. Martinez V
    4. Sarter M

    (2007) Prefrontal acetylcholine release controls cue detection on multiple timescales

    Neuron 56:141–154.

    https://doi.org/10.1016/j.neuron.2007.08.025

    • PubMed
    • Google Scholar
57. 1. Parikh V
    2. Sarter M

    (2008) Cholinergic mediation of attention

    Annals of the New York Academy of Sciences 1129:225–235.

    https://doi.org/10.1196/annals.1417.021

    • Google Scholar
58. 1. Rajkowski J
    2. Kubiak P
    3. Aston-Jones G

    (1994) Locus coeruleus activity in monkey: phasic and tonic changes are associated with altered vigilance

    Brain Research Bulletin 35:607–616.

    https://doi.org/10.1016/0361-9230(94)90175-9

    • PubMed
    • Google Scholar
59. 1. Rajkowski J
    2. Majczynski H
    3. Clayton E
    4. Aston-Jones G

    (2004) Activation of monkey locus coeruleus neurons varies with difficulty and performance in a target detection task

    Journal of Neurophysiology 92:361–371.

    https://doi.org/10.1152/jn.00673.2003

    • PubMed
    • Google Scholar
60. 1. Reimer J
    2. Froudarakis E
    3. Cadwell CR
    4. Yatsenko D
    5. Denfield GH
    6. Tolias AS

    (2014) Pupil fluctuations track fast switching of cortical states during quiet wakefulness

    Neuron 84:355–362.

    https://doi.org/10.1016/j.neuron.2014.09.033

    • PubMed
    • Google Scholar
61. 1. Reimer J
    2. McGinley MJ
    3. Liu Y
    4. Rodenkirch C
    5. Wang Q
    6. McCormick DA
    7. Tolias AS

    (2016) Pupil fluctuations track rapid changes in adrenergic and cholinergic activity in cortex

    Nature Communications 7:13289.

    https://doi.org/10.1038/ncomms13289

    • PubMed
    • Google Scholar
62. 1. Sarter M
    2. Lustig C
    3. Berry AS
    4. Gritton H
    5. Howe WM
    6. Parikh V

    (2016) What do phasic cholinergic signals do?

    Neurobiology of Learning and Memory 130:135–141.

    https://doi.org/10.1016/j.nlm.2016.02.008

    • PubMed
    • Google Scholar
63. 1. Shadlen MN
    2. Kiani R
    3. Newsome WT
    4. Gold JI
    5. Wolpert DM
    6. Zylberberg A
    7. Ditterich J
    8. de Lafuente V
    9. Yang T
    10. Roitman J

    (2016) Comment on "Single-trial spike trains in parietal cortex reveal discrete steps during decision-making"

    Science 351:1406.

    https://doi.org/10.1126/science.aad3242

    • PubMed
    • Google Scholar
64. Book

    1. Sheather SJ

    (2009)

    A Modern Approach to Regression with R (Firstst ed)

    New York: Springer.

    • Google Scholar
65. Preprint

    1. Simonsohn U

    (2017) Two-Lines: a valid alternative to the invalid testing of U-Shaped relationships with quadratic regressions

    Ssrn.

    https://doi.org/10.2139/ssrn.3021690

    • Google Scholar
66. 1. Thiele A
    2. Bellgrove MA

    (2018) Neuromodulation of attention

    Neuron 97:769–785.

    https://doi.org/10.1016/j.neuron.2018.01.008

    • PubMed
    • Google Scholar
67. 1. Thut G
    2. Nietzel A
    3. Brandt SA
    4. Pascual-Leone A

    (2006) Alpha-band electroencephalographic activity over occipital cortex indexes visuospatial attention bias and predicts visual target detection

    Journal of Neuroscience 26:9494–9502.

    https://doi.org/10.1523/JNEUROSCI.0875-06.2006

    • PubMed
    • Google Scholar
68. 1. Urai AE
    2. Braun A
    3. Donner TH

    (2017) Pupil-linked arousal is driven by decision uncertainty and alters serial choice bias

    Nature Communications 8:14637.

    https://doi.org/10.1038/ncomms14637

    • PubMed
    • Google Scholar
69. 1. van Dijk H
    2. Schoffelen JM
    3. Oostenveld R
    4. Jensen O

    (2008) Prestimulus oscillatory activity in the alpha band predicts visual discrimination ability

    Journal of Neuroscience 28:1816–1823.

    https://doi.org/10.1523/JNEUROSCI.1853-07.2008

    • PubMed
    • Google Scholar
70. Software

    1. van Kempen J

    (2019) 2019_pupil_decisionMaking, version 212b94f 

    GitHub.

    https://github.com/jochemvankempen/2019_pupil_decisionMaking
71. 1. Varazzani C
    2. San-Galli A
    3. Gilardeau S
    4. Bouret S

    (2015) Noradrenaline and dopamine neurons in the reward/effort trade-off: a direct electrophysiological comparison in behaving monkeys

    Journal of Neuroscience 35:7866–7877.

    https://doi.org/10.1523/JNEUROSCI.0454-15.2015

    • PubMed
    • Google Scholar
72. Book

    1. Venables WN
    2. Ripley BD

    (2002)

    Modern Applied Statistics with S (Fourth ed)

    New York: Springer.

    • Google Scholar
73. 1. Vinck M
    2. Batista-Brito R
    3. Knoblich U
    4. Cardin JA

    (2015) Arousal and locomotion make distinct contributions to cortical activity patterns and visual encoding

    Neuron 86:740–754.

    https://doi.org/10.1016/j.neuron.2015.03.028

    • PubMed
    • Google Scholar

Thank you for submitting your article "Behavioural and neural signatures of perceptual evidence accumulation are modulated by pupil-linked arousal" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by Eran Eldar as the Reviewing Editor, and Richard Ivry as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Tobias H Donner (Reviewer #2); Marieke Jepma (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This paper addresses an interesting and timely topic: the effects of arousal (indexed by pupil size) on different stages of information processing during perceptual decision making. Recent studies have already linked variation in pupil size to various components of decision making, but the systematic assessment of the link between pupil size, performance, and a set of EEG signals related to specific information-processing stages in this study is novel.

The authors re-analyze a large amount of data – mainly from one previous study (N=80; Newman et al., 2017), and, for some specific questions, also from other previously published studies. They quantify a variety of EEG markers as a function of pupil-linked arousal: markers of attention, decision processing, and motor preparation, whose properties have previously been characterized extensively.

The relationship between trial-to-trial variations in pupil measures and these EEG markers reveals a non-monotonic relationship of baseline pupil diameter with both reaction time and pre-trial alpha power, which accords with a body of work on baseline arousal and neuromodulatory tone. Interestingly, the authors also find EEG markers (CPP ITCP, N2c, LHB) that scale linearly with baseline pupil diameter, which illuminate the difference between slow performance associated with low and high baseline pupil diameter.

The analyses also yield non-monotonic relationships with task-evoked changes in pupil diameter. This contradicts the monotonic relationships between phasic neuromodulatory or pupil responses and cortical signals or parameters of choice behavior observed in a substantial body of previous work.

We find the topic of this paper important, and we sympathize with the general approach pursued in this study. The data look extremely clean, and the analyses are generally sophisticated. That said, we have some fundamental concerns about the validity of the study's conclusions, in particular those conclusions drawn based on the phasic pupillary responses. In this respect, we think that the conclusions are not sufficiently supported by the data, and the conclusions drawn from these results may, in fact, mislead the community. We thus feel it would be irresponsible to recommend it for publication at eLife in its current form. Below, we explain our concerns in detail and make specific suggestions for how to address them. Most importantly, we propose that the authors focus the paper on the baseline pupil diameter findings, which in our estimation are more well founded.

Essential revisions:

1) The increase of pupil response amplitude with RT can be explained by the characteristics of the peripheral system controlling pupil diameter rather than central arousal state – any central arousal signal shaping cortical state (i.e. a response of the LC, basal forebrain, or any other center contributing to central arousal and affecting pupil diameter) passes through a sluggish peripheral system (peripheral nerves/ganglia, smooth pupil muscles) to produce a measured pupil response. Previous work has quantified this transformation as a simple linear convolution with a pupil impulse response function (IRF, e.g., Hoeks and Levelt, 1993; Korn and Bach, JoV, 2016).) This linear systems approach may, or may not, be accurate for the intermediate range of pupil diameters measured in human laboratory settings. Regardless of the specifics of the transformation, the point is that, just as for fMRI, the measured pupil response is only a low-pass filtered version of the central input signal. Obviously this study is about the inferred amplitude of the central arousal signal. If that central signal is sustained, the amplitude of the peripheral pupil response increases with central signal duration – even if the amplitude of the central signal does not change.

Previous work shows that pupil dilation during perceptual decisions is driven by a sustained central input – either a plateau or a slow build-up – throughout the interval of decision formation (Cheadle et al., 2014; de Gee et al., 2014, 2017; Murphy et al., 2016). Whenever there is such a sustained component in the input, it can trivially predict the right (steeper) flank of the plot in Figure 1E – hence, it can also explain part of the subsequent effects of pupil response on EEG markers. This trivial effect may superimpose with an interesting, genuine effect of phasic arousal (perhaps negative slope), to produce the non-monotonic pattern shown in Figure 1E and the subsequent ones. The authors themselves refer to the above conclusion that some component of the central arousal signals may be sustained. (It might not be the LC-component, but DA neurons, some of which exhibit sustained activity during decision formation.) But then the above 'RT-confound' is inescapable. The possibility of such a trivial confound renders the results difficult to interpret.

One possible way to resolve this confound could be to fit different GLMs convolving each input regressor with a "canonical" IRF (Hoeks and Levelt). Within this framework the authors could model various different kinds of input to the pupil system, including both linear and non-linear parametric modulation by RT, and use model comparison to actually test what form the input to the pupil system is actually likely to take place in their data. However, even if this confound is resolved, the present interpretation of the pupil response findings would still be problematic due to its claim to causality (see comment #2).

2) The claim of causality. The paper, as it stands, interprets the findings associated with lower or higher pupillary responses as being caused by differences in phasic arousal. This would be one way to explain the findings. However, there is another way which does not assume causality and seems at least as plausible. Specifically, some long RTs may reflect the subject doing less and other long RTs may reflect the subject doing more. 'Doing more' here could be for instance a more sluggish processing of the stimulus, which leads to lower phasic arousal/effort, and 'doing less' here could be taking extra care to verify detection before responding, and comparing against the stimulus on the other side of the screen (as suggested by Figure 4A), which leads to greater or more sustained phasic arousal/effort. A combination of these two types of trials can parsimoniously explain a U-shape relationship between pupil response and RT while assuming that arousal is simply a result of certain characteristics of the decision process.

It is true that the issue of causality is common to a lot of papers in the field. However, we think this caveat needs to be weighed based on: (1) does the paper strongly argue for a causal interpretation? (2) is there converging evidence supporting the causal perspective (e.g., from both RT and accuracy)? and (3) is there a non-causal interpretation that seems more plausible than the causal one? Our feeling is that all three factors work against the pupil response findings in this paper. Therefore, due to the issues described in the first two comments, we propose that the authors either remove the "phasic arousal" part or relegate it to the supplementary material (the latter option requires that the authors perform the additional analysis proposed in comment 1), and instead focus on the "baseline arousal" part. The findings with respect to baseline pupil diameter do not suffer from the same interpretational issues, at least not to the same degree, and they provide a natural extension to the impact of baseline arousal on previous work on the responses of sensory cortical neurons (McGinley et al., 2015). This is a more specific issue than what the authors are alluding to in the title and Abstract, but it is what it is. In sum, with the above, rather fundamental change, the study might provide insight into the role of (baseline) arousal state in some form of visuo-motor decision processing and selective attention.

3) The authors note that participants performed at ceiling and therefore do not analyze accuracy. Does this mean there were no missed responses/errors at all? Please specify the accuracy data.

4) Throughout the manuscript, the authors infer non-monotonic relationships from the superior fit a quadratic model compared to a linear one. This is not sufficient. Consider the two vectors: x = [1 2 3 4 5] and y = [10 5 1 1 1]. A quadratic function would fit the relationship between x and y better than a linear function even though the relationship is monotonic. This issue plagues many of the analyses, and visual inspection of some of the relevant figures does not convince that there is a significant non-monotonic relationship (e.g., Figure 4E).

5) The details and appropriateness of the hierarchical mediation analysis reported in the subsection “The impact of phasic arousal on task performance is mainly mediated by the consistency in evidence accumulation” do not seem entirely clear. Our main concern here is that the approach may be biased by certain assumptions about the "temporal order of the decision-making process" and the associated "successive stages of neural processing". The point is that motor beta power suppression and CPP seem to behave analogously in all analyses. Yet, it is entered in the hierarchical analysis only after CPP. So, the CPP is identified as having a unique role in mediating arousal effects. Does that conclusion change if the order of entering motor beta and CPP is reversed? We assume that the authors' underlying assumption is as follows: (i) The CPP is generated at a processing stage (e.g. LIP) "prior to" motor beta and (ii) it encodes the accumulating decision variable, which is only subsequently translated into a motor preparatory signal evident in motor beta-band activity. We are convinced that the CPP is a very interesting signal worth studying, but we do not think these assumptions are supported by the available evidence. For example, one of the authors' recent studies (Twomey et al., 2016) has shown that motor beta behaves more similar to the LIP signal than the CPP in a delayed version of the classic Shadlen and Newsome task: encoding upcoming choice throughout the delay. Plus, the authors are aware of all the recent provocative animal work from several different labs (inactivation studies and sophisticated analyses, in rodents and monkeys) that cast doubt even on the notion that LIP/PPC is "the evidence accumulator". We are not saying that this notion is "wrong", but we feel it cannot be treated as an established fact, on which to base the logic of an analysis like the above.

6) Implications for accounts of evidence accumulation are unclear – we do not think that much can be concluded from these results about "evidence accumulation", a term that occurs throughout the title, Abstract, conclusions, etc. Neither do authors present behavioral evidence for such accumulation, nor is it clear why that would be necessary to solve the task at hand: coherence change detection task with strong sensory evidence (50% coherence) that is constant across trials. Detecting such a strong motion signal, and discriminating its coarse direction, is trivial; the only performance-limiting factors are spatial and temporal uncertainty about the occurrence of the change. Indeed, accuracy seems to be "at ceiling" (in Newman et al., it ranged between 92% and 100%). Especially if the change is as large as here, such a task can be solved without any meaningful temporal integration (i.e., over timescales beyond 100 ms, as used for near-threshold motion discrimination). Transient detection, for which there are mechanisms in V4 (Gardner et al., Neuron, 2005; Costagli et al., 2014), would suffice. Please note that the uncertainty about signal onset also renders standard drift diffusion unsuitable for this task (Ossmy et al., 2013; Glaze et al., 2015). Thus, any conclusions about evidence accumulation should be reframed in terms of "decision processing", "motion detection", etc. At the same time, the lack of choice variability across trials severely limits the mechanistic insight that can be gained from this study, because no other behavioral readout than RT is available and fitting models as in the Ossmy et al., 2013 or Glaze et al., 2015 work will not work (simulations form such models, or fits from simpler models, might be possible and informative though).

7) In general, we feel that the current presentation rests on several strong and doubtful assumptions about the internal process underlying behavior, and about the functional nature of the different EEG markers. Some of those assumptions are made explicit, others remain implicit. Here are some paraphrased examples: "CPP equals the decision variable"; "if we detect a CPP in our task, this implies evidence accumulation, because the CCP was linked to evidence accumulation in previous work using other tasks". But here, CPP/motor beta could just be an abrupt step-like signal upon transient motion detection, but variability in the motor component of non-decision time would smooth the associated trial-averaged traces to make it look as if they were gradually building. These assumptions should be made explicit. In fact, we would encourage the authors to try re-frame the paper without these assumptions, most of are not necessary for a presentation of the findings. Thus, Though we appreciate the careful previous work of these labs to characterize these EEG markers, we think it is sufficient to refer to this previous work to motivate the focus on these markers, and then simply report how these markers behave in the current work.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Behavioural and neural signatures of perceptual decision-making are modulated by pupil-linked arousal" for further consideration at eLife. Your revised article has been favorably evaluated by Richard Ivry (Senior Editor) and Eran Elder (Reviewing Editor).

The reviewers’ comments were comprehensively addressed, and the manuscript has been substantially improved. There are only minor issues left, brought about by the updated results, that need to be addressed before acceptance.

First, there is remarkable similarity between the baseline diameter and pupil response results. These two measures are correlated, and both are associated with high baseline alpha power and with a similar behavioral effect that is in both cases mediated by the consistency of the CPP. This is all the more notable since the authors made sure to remove variance in the pupil responses that is attributable to differences in baseline diameter. This suggests that both might be in large part two noisy measures of baseline arousal, from which neither measure can be fully dissociated by removing variance attributable to the other measure. Importantly, there were also differences between the measures, but this issue – both the overlap and dissociation – deserves to be pointed out and discussed, especially considering other work indicating that pupil responses are inversely related to baseline LC function.

Second, in the last sentence of the Abstract, confusion can be prevented by pointing out that the terms "attentional engagement, sensory processing and decision processing" relate to EEG correlates of these processes.

Essential revisions:

1) The increase of pupil response amplitude with RT can be explained by the characteristics of the peripheral system controlling pupil diameter rather than central arousal state – Any central arousal signal shaping cortical state (i.e. a response of the LC, basal forebrain, or any other center contributing to central arousal and affecting pupil diameter) passes through a sluggish peripheral system (peripheral nerves/ganglia, smooth pupil muscles) to produce a measured pupil response. Previous work has quantified this transformation as a simple linear convolution with a pupil impulse response function (IRF, e.g., Hoeks and Levelt, 1993; Korn and Bach, JoV, 2016).) This linear systems approach may, or may not, be accurate for the intermediate range of pupil diameters measured in human laboratory settings. Regardless of the specifics of the transformation, the point is that, just as for fMRI, the measured pupil response is only a low-pass filtered version of the central input signal. Obviously this study is about the inferred amplitude of the central arousal signal. If that central signal is sustained, the amplitude of the peripheral pupil response increases with central signal duration – even if the amplitude of the central signal does not change.

Previous work shows that pupil dilation during perceptual decisions is driven by a sustained central input – either a plateau or a slow build-up – throughout the interval of decision formation (Cheadle et al., 2014; de Gee et al., 2014, 2017; Murphy et al., 2016). Whenever there is such a sustained component in the input, it can trivially predict the right (steeper) flank of the plot in Figure 1E – hence, it can also explain part of the subsequent effects of pupil response on EEG markers. This trivial effect may superimpose with an interesting, genuine effect of phasic arousal (perhaps negative slope), to produce the non-monotonic pattern shown in Figure 1E and the subsequent ones. The authors themselves refer to the above conclusion that some component of the central arousal signals may be sustained. (It might not be the LC-component, but DA neurons, some of which exhibit sustained activity during decision formation.) But then the above 'RT-confound' is inescapable. The possibility of such a trivial confound renders the results difficult to interpret.

One possible way to resolve this confound could be to fit different GLMs convolving each input regressor with a "canonical" IRF (Hoeks and Levelt). Within this framework the authors could model various different kinds of input to the pupil system, including both linear and non-linear parametric modulation by RT, and use model comparison to actually test what form the input to the pupil system is actually likely to take place in their data. However, even if this confound is resolved, the present interpretation of the pupil response findings would still be problematic due to its claim to causality (see comment #2).

We acknowledge that our scalar measure of pupil dilation may indeed have been confounded by the duration of the trial. We took up the reviewers’ advice and tested the fit of various GLMs in order to investigate the form of the input to the pupil system. With this approach, we focused on the different forms previously tested by Murphy et al., 2016. Next, we extended this analysis by applying the best fitting GLM on each individual trial by considering each trial as a separate condition (similar to the approach by Bach et al., 2018. Molecular Psychiatry). This allowed us to estimate the presence of each of the three components (target onset, a sustained period and the response) for each trial. We then used the estimates of the target onset response as scalar values by which we sorted the data in order to investigate their relationship to behavioural performance as well as the EEG measures associated with perceptual decision-making. This analysis has allowed us to reduce the confounding effect of RT on the scalar measure of the pupil response, as well as distinguishing between the target onset and sustained component, whilst still having the single-trial resolution to investigate its relationship to the behavioural and EEG signatures under investigation. With this approach, we found that the relationship between the pupil response to target onset and task performance (and EEG components) are no longer best described by a non-monotonic (U-shaped) relationship. Rather, they are best described by an inverse monotonic relationship, where the largest pupil responses are predictive of the best task performance. The changes to the analysis can now be found in the revised manuscript (Materials and methods), with Figure 1—figure supplements 1-4 relating to this new analysis. Altogether, we believe that this change to the approach of estimating the pupil response deals with the concerns mentioned above, as well as (some of) the concerns raised by the reviewers in comment 2 below.

2) The claim of causality. The paper, as it stands, interprets the findings associated with lower or higher pupillary responses as being caused by differences in phasic arousal. This would be one way to explain the findings. However, there is another way which does not assume causality and seems at least as plausible. Specifically, some long RTs may reflect the subject doing less and other long RTs may reflect the subject doing more. 'Doing more' here could be for instance a more sluggish processing of the stimulus, which leads to lower phasic arousal/effort, and 'doing less' here could be taking extra care to verify detection before responding, and comparing against the stimulus on the other side of the screen (as suggested by Figure 4A), which leads to greater or more sustained phasic arousal/effort. A combination of these two types of trials can parsimoniously explain a U-shape relationship between pupil response and RT while assuming that arousal is simply a result of certain characteristics of the decision process.

It is true that the issue of causality is common to a lot of papers in the field. However, we think this caveat needs to be weighed based on: (1) does the paper strongly argue for a causal interpretation? (2) is there converging evidence supporting the causal perspective (e.g., from both RT and accuracy)? and (3) is there a non-causal interpretation that seems more plausible than the causal one? Our feeling is that all three factors work against the pupil response findings in this paper. Therefore, due to the issues described in the first two comments, we propose that the authors either remove the "phasic arousal" part or relegate it to the supplementary material (the latter option requires that the authors perform the additional analysis proposed in comment 1), and instead focus on the "baseline arousal" part. The findings with respect to baseline pupil diameter do not suffer from the same interpretational issues, at least not to the same degree, and they provide a natural extension to the impact of baseline arousal on previous work on the responses of sensory cortical neurons (McGinley et al., 2015). This is a more specific issue than what the authors are alluding to in the title and Abstract, but it is what it is. In sum, with the above, rather fundamental change, the study might provide insight into the role of (baseline) arousal state in some form of visuo-motor decision processing and selective attention.

Indeed, given this dataset, we cannot control for the behavioural approach taken by the subject on individual trials. Differences in approach, however, would likely produce differences between target onset and response, during the sustained period of the task. The target onset component, arguably reflecting a bottom up stimulus processing response (potentially a transient locus coeruleus response), does not suffer from the same interpretational issues. Moreover, these results are in line with previous studies revealing an increase in performance on trials with a larger phasic arousal response, and reveal new insights into how this phasic response influences different neural processing stages of perceptual decision-making. We believe that by focusing on the pupillary response to target onset (separately from the sustained period and response related component), we have addressed the first two of the reviewers’ concerns, and we therefore included these results in the main text, illustrated in Figure 1-3. We, however, welcome any further comments, and if the reviewers still request that we move these results, we are willing to do this.

3) The authors note that participants performed at ceiling and therefore do not analyze accuracy. Does this mean there were no missed responses/errors at all? Please specify the accuracy data.

We specified the accuracy data in the Results, under the header “Both tonic and phasic arousal are predictive of task performance”.

4) Throughout the manuscript, the authors infer non-monotonic relationships from the superior fit a quadratic model compared to a linear one. This is not sufficient. Consider the two vectors: x = [1 2 3 4 5] and y = [10 5 1 1 1]. A quadratic function would fit the relationship between x and y better than a linear function even though the relationship is monotonic. This issue plagues many of the analyses, and visual inspection of some of the relevant figures does not convince that there is a significant non-monotonic relationship (e.g., Figure 4E).

To test for the statistical presence of (inverted) U-shaped relationships in the case of a non-monotonic relationship, we implemented the ‘two-lines’ approach (Simonsohn, 2017). Using this method, we tested whether there is a significant sign change from low to high pupil diameter values. This is done by estimating two linear regression lines separately to low and high pupil diameter values. To classify a relationship as U-shaped, the regression coefficients need to be independently significant and be of opposite sign. The details for this analysis are described in the Materials and methods, subsection “Statistical analyses”.

5) The details and appropriateness of the hierarchical mediation analysis reported in the subsection “The impact of phasic arousal on task performance is mainly mediated by the consistency in evidence accumulation” do not seem entirely clear. Our main concern here is that the approach may be biased by certain assumptions about the "temporal order of the decision-making process" and the associated "successive stages of neural processing". The point is that motor beta power suppression and CPP seem to behave analogously in all analyses. Yet, it is entered in the hierarchical analysis only after CPP. So, the CPP is identified as having a unique role in mediating arousal effects. Does that conclusion change if the order of entering motor beta and CPP is reversed? We assume that the authors' underlying assumption is as follows: (i) The CPP is generated at a processing stage (e.g. LIP) "prior to" motor beta and (ii) it encodes the accumulating decision variable, which is only subsequently translated into a motor preparatory signal evident in motor beta-band activity. We are convinced that the CPP is a very interesting signal worth studying, but we do not think these assumptions are supported by the available evidence. For example, one of the authors' recent studies (Twomey et al., 2016) has shown that motor beta behaves more similar to the LIP signal than the CPP in a delayed version of the classic Shadlen and Newsome task: encoding upcoming choice throughout the delay. Plus, the authors are aware of all the recent provocative animal work from several different labs (inactivation studies and sophisticated analyses, in rodents and monkeys) that cast doubt even on the notion that LIP/PPC is "the evidence accumulator". We are not saying that this notion is "wrong", but we feel it cannot be treated as an established fact, on which to base the logic of an analysis like the above.

The hierarchical mediation analysis allows for the sequential testing of likelihood ratio tests, which could indeed, as the reviewers suggested, be based upon false assumptions about the temporal order of the EEG components in the decision making process. To test whether any components that were excluded from the final model based on this hierarchical approach exerted any significant influence on variability in task performance, we additionally investigated the coefficients of the final regression model (the last step in the hierarchical model analysis) that included all EEG components. This revealed no additional components to those included based on the hierarchical model analysis. These results are summarised in Supplementary files 3 and 5.

6) Implications for accounts of evidence accumulation are unclear – we do not think that much can be concluded from these results about "evidence accumulation", a term that occurs throughout the title, Abstract, conclusions, etc. Neither do authors present behavioral evidence for such accumulation, nor is it clear why that would be necessary to solve the task at hand: coherence change detection task with strong sensory evidence (50% coherence) that is constant across trials. Detecting such a strong motion signal, and discriminating its coarse direction, is trivial; the only performance-limiting factors are spatial and temporal uncertainty about the occurrence of the change. Indeed, accuracy seems to be "at ceiling" (in Newman et al., it ranged between 92% and 100%). Especially if the change is as large as here, such a task can be solved without any meaningful temporal integration (i.e., over timescales beyond 100 ms, as used for near-threshold motion discrimination). Transient detection, for which there are mechanisms in V4 (Gardner et al., Neuron, 2005; Costagli et al., 2014), would suffice. Please note that the uncertainty about signal onset also renders standard drift diffusion unsuitable for this task (Ossmy et al., 2013; Glaze et al., 2015). Thus, any conclusions about evidence accumulation should be reframed in terms of "decision processing", "motion detection", etc. At the same time, the lack of choice variability across trials severely limits the mechanistic insight that can be gained from this study, because no other behavioral readout than RT is available and fitting models as in the Ossmy et al. (2013) or Glaze et al. (2015) work will not work (simulations form such models, or fits from simpler models, might be possible and informative though).

We addressed this concern in this revision by investigating the early target selection signals (N2c) aligned to both target onset as well as the response. If variability in RT was indeed mainly caused by variability in the target onset transient, and there would be a fixed delay between the target onset transient and the response, then the amplitude of the N2c would be much larger when aligned to response compared to target onset. We, however, found that this is not the case. The amplitude of the response-aligned N2c was significantly lower than when aligned to target onset. Insofar as the N2c reflects the target onset transient, this indicates that there is variation in the duration between early target selection and the response. Whether this variation does indeed reflect variability in the duration of evidence accumulation can, however, still be questioned because of the points raised above. Subjects could employ a different strategy on different trials. On some trials they could verify coherent motion in one of the stimuli by comparing it against the other stimulus, whereas on other trials this approach may not have been taken. We address this limitation both in the Results and in the Discussion of the revised manuscript. Additionally, we have substantially reduced our usage of the term evidence accumulation throughout the manuscript.

7) In general, we feel that the current presentation rests on several strong and doubtful assumptions about the internal process underlying behavior, and about the functional nature of the different EEG markers. Some of those assumptions are made explicit, others remain implicit. Here are some paraphrased examples: "CPP equals the decision variable"; "if we detect a CPP in our task, this implies evidence accumulation, because the CCP was linked to evidence accumulation in previous work using other tasks". But here, CPP/motor beta could just be an abrupt step-like signal upon transient motion detection, but variability in the motor component of non-decision time would smooth the associated trial-averaged traces to make it look as if they were gradually building. These assumptions should be made explicit. In fact, we would encourage the authors to try re-frame the paper without these assumptions, most of are not necessary for a presentation of the findings. Thus, Though we appreciate the careful previous work of these labs to characterize these EEG markers, we think it is sufficient to refer to this previous work to motivate the focus on these markers, and then simply report how these markers behave in the current work.

As indicated above (comment 6), we have addressed the limitations of our measure of the decision variable both in the Results and the Discussion. As part of this, we have provided more background, such as making explicit the assumption that the CPP could be a build-to-threshold decision variable, but might equally be a step-wise function that appears like a build-to-threshold decision variable when averaged across trials.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

First, there is remarkable similarity between the baseline diameter and pupil response results. These two measures are correlated, and both are associated with high baseline alpha power and with a similar behavioral effect that is in both cases mediated by the consistency of the CPP. This is all the more notable since the authors made sure to remove variance in the pupil responses that is attributable to differences in baseline diameter. This suggests that both might be in large part two noisy measures of baseline arousal, from which neither measure can be fully dissociated by removing variance attributable to the other measure. Importantly, there were also differences between the measures, but this issue – both the overlap and dissociation – deserves to be pointed out and discussed, especially considering other work indicating that pupil responses are inversely related to baseline LC function.

We addressed the correlation between baseline pupil diameter and the pupil response as well as the overlap in their effect on the investigated EEG signatures of decision-making in the Discussion (section “The overlap and dissociation between baseline pupil diameter and the pupil response”).

Second, in the last sentence of the Abstract, confusion can be prevented by pointing out that the terms "attentional engagement, sensory processing and decision processing" relate to EEG correlates of these processes.

We addressed this issue by explicitly mentioning the term “EEG markers of” attentional engagement, sensory processing and decision processing in the last sentence of the Abstract.

Author details

1. Jochem van Kempen

   1. Institute of Neuroscience, Newcastle University, Newcastle upon Tyne, United Kingdom
   2. Monash Institute for Cognitive and Clinical Neurosciences, School of Psychological Sciences, Monash University, Melbourne, Australia

   Contribution

   Conceptualization, Formal analysis, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing

   For correspondence

   jochem.van-kempen@ncl.ac.uk

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-0211-9545

2. Gerard M Loughnane

   1. School of Engineering, Trinity College Dublin, Dublin, Ireland
   2. Trinity College Institute of Neuroscience, Trinity College Dublin, Dublin, Ireland

   Contribution

   Conceptualization, Data curation, Methodology, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-1961-5294

3. Daniel P Newman

   Monash Institute for Cognitive and Clinical Neurosciences, School of Psychological Sciences, Monash University, Melbourne, Australia

   Contribution

   Conceptualization, Data curation, Methodology

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-8240-1876

4. Simon P Kelly

   School of Electrical and Electronic Engineering, University College Dublin, Dublin, Ireland

   Contribution

   Supervision, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-9983-3595

5. Alexander Thiele

   Institute of Neuroscience, Newcastle University, Newcastle upon Tyne, United Kingdom

   Contribution

   Supervision, Funding acquisition, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-4894-0213

6. Redmond G O'Connell

   1. Monash Institute for Cognitive and Clinical Neurosciences, School of Psychological Sciences, Monash University, Melbourne, Australia
   2. Trinity College Institute of Neuroscience, Trinity College Dublin, Dublin, Ireland
   3. School of Psychology, Trinity College Dublin, Dublin, Ireland

   Contribution

   Conceptualization, Resources, Supervision, Funding acquisition, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-6949-2793

7. Mark A Bellgrove

   1. Monash Institute for Cognitive and Clinical Neurosciences, School of Psychological Sciences, Monash University, Melbourne, Australia
   2. School of Psychology, Trinity College Dublin, Dublin, Ireland

   Contribution

   Conceptualization, Resources, Supervision, Funding acquisition, Writing—original draft, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-0186-8349

• 3,061

  Page views

• 490

  Downloads

• 29

  Citations

Article citation count generated by polling the highest count across the following sources: Scopus, Crossref, PubMed Central.