# Dopamine promotes instrumental motivation, but reduces reward-related vigour

1. Nuffield Department of Clinical Neurosciences, University of Oxford, United Kingdom
2. Department of Psychology, University of Cambridge, United Kingdom
3. Oxford Parkinson’s Disease Centre, University of Oxford, United Kingdom
4. Department of Experimental Psychology, University of Oxford, United Kingdom
Research Article

## Abstract

We can be motivated when reward depends on performance, or merely by the prospect of a guaranteed reward. Performance-dependent (contingent) reward is instrumental, relying on an internal action-outcome model, whereas motivation by guaranteed reward may minimise opportunity cost in reward-rich environments. Competing theories propose that each type of motivation should be dependent on dopaminergic activity. We contrasted these two types of motivation with a rewarded saccade task, in patients with Parkinson’s disease (PD). When PD patients were ON dopamine, they had greater response vigour (peak saccadic velocity residuals) for contingent rewards, whereas when PD patients were OFF medication, they had greater vigour for guaranteed rewards. These results support the view that reward expectation and contingency drive distinct motivational processes, and can be dissociated by manipulating dopaminergic activity. We posit that dopamine promotes goal-directed motivation, but dampens reward-driven vigour, contradictory to the prediction that increased tonic dopamine amplifies reward expectation.

## Introduction

Organisms expend more effort when their actions can lead to rewards, as the value of the reward offsets the extra effort expended to attain them (Kool and Botvinick, 2018; Manohar et al., 2015; Niv et al., 2006; Shenhav et al., 2017). They will even do so if the extra effort does not increase the reward they receive (Glaser et al., 2016; Milstein and Dorris, 2007; Xu-Wilson et al., 2009), indicating that mere expectation of reward is enough to justify the effort cost. Motivation, which promotes this effort expenditure, has two facets: it allows actions to be directed towards goals, and it energises our actions when rewards are expected (Niv et al., 2006). These two aspects are not always coupled. For example, employees might be salaried, where a fixed reward is guaranteed irrespective of achievements, or they might receive merit-based pay that is contingent on meeting performance targets (Lazear, 2000).

Contingent rewards motivate us because we understand the causal relation between successful actions and reward. This is instrumental, in that we apply knowledge of action-outcome associations. For instance, people must realise that merit-based pay depends on their performance for it to incentivise them. In animals, dopaminergic input to dorsal striatum is necessary for instrumental motivation (Lex and Hauber, 2010b).

In contrast, reward that is independent of what an agent does might motivate us because in a variable environment, we capitalise on rewards while they are available (Niv et al., 2007). One proposed mechanism for this is that tonic dopamine encodes expected reward rate, such that in a rich environment agents are motivated to respond faster to maximise the rewards they receive (Niv et al., 2007). Equally, dopamine can be viewed as signalling an opportunity cost– time is more costly when reward is available, and so organisms act faster (Otto and Daw, 2019; Shadmehr et al., 2010). The dopaminergic drive has not only generalised motivating effects, termed vigour (Beierholm et al., 2013; Guitart-Masip et al., 2011; Niv et al., 2007), but also context-specific effects. For example, a stimulus that predicts rewards drives conditioned responses that are uncoupled with reward (Lovibond, 1981) – similar to how salary increases might improve job performance. This phenomenon, known as Pavlovian-to-Instrumental transfer, requires dopamine projections to nucleus accumbens (Hall et al., 2001; Kelley and Delfs, 1991; Talmi et al., 2008; Wassum et al., 2013; Wyvell and Berridge, 2000). Similarly, animals tend to approach stimuli associated with rewards, even in the absence of action-contingency, a behaviour called autoshaping or sign-tracking, which also relies on nucleus accumbens dopamine (Day et al., 2006; Di Ciano et al., 2001).

The dopaminergic basis of instrumental and Pavlovian motivation could potentially explain the impaired motivation seen in PD patients and the rescue of such deficits by rewards (Ang et al., 2018; Chong et al., 2015; de Wit et al., 2011; Kojovic et al., 2014). However in certain situations, motivation by reward can paradoxically be stronger in patients with low dopamine (Aarts et al., 2012; Timmer et al., 2018), making dopamine’s exact role in motivation unclear.

These two effects of contingent and expected rewards frequently overlap in real life and in previous experiments – higher stakes raise reward expectation, but also mean that actions carry more weight. However, experimental control of expectation and contingency allows them to be dissociated (Manohar et al., 2017), which reveals that both contingency and expectation can separately motivate behaviour, and that these effects are independent rather than correlated or antagonistic, suggesting distinct mechanisms.

We used this incentivised saccade task (Manohar et al., 2017) here to test PD patients ON and OFF their dopaminergic medication, along with healthy age-matched controls. We tested the two predictions that dopamine is involved in motivation by expected rewards, and by contingent rewards.

## Results

### Dopamine promotes contingent motivation and attenuates reward-expectation motivation

Figure 1
Table 1

Dopaminergic medication significantly modulated how contingent and guaranteed motivation affected motor vigour (Figure 2a, three-way interaction on peak velocity residuals, p=0.0023; see Table 2 for statistics). This was because, when ON medication, patients were motivated by contingency but not reward expectation (separate two-way ANOVA in PD ON: contingency*motivation, p=0.0170; see Supplementary file 1A), whereas after overnight withdrawal of medication there was a borderline significant interaction in the opposite direction, as PD OFF were motivated by reward expectation but not contingency (PD OFF ANOVA: p=0.0501; Supplementary file 1A). This indicates that when PD patients were ON medication, motivation was strongest when reward was contingent on performance, but when they were OFF medication, patients were motivated by guaranteed rewards.

Figure 2
Table 2

To confirm that the effects on peak velocity residuals were not driven by changes in other aspects of saccades, the same 3-way ANOVA was run on each of the other saccade measures. There were no significant effects on saccadic amplitude (see Table 2 and Figure 2c). Saccadic RT had an effect of contingency as saccades started faster for Performance and Random conditions than 10 p or 0 p conditions (Figure 2d, p=0.0396). Endpoint variability had a contingency*motivation interaction (Figure 2e, p=0.0482) as variability was higher for 0 p condition. Raw peak velocity had an effect of motivation, as both types of motivation increased speed (Figure 2f, p=0.0110), although this will include effects of changes in amplitude (via the main sequence) which showed a borderline significant effect of motivation (Figure 2c, p=0.0607).

The HC peak velocity residuals were not affected by contingency, motivation or the interaction (p>0.05; see Table 3), suggesting that healthy older adults do not adjust their response vigour for contingent or guaranteed rewards. There were also no significant effects on amplitude, saccadic RT, or raw peak velocity in HC, although endpoint variability did have a significant contingency*motivation interaction (p=0.0048; see Table 3). Post-hoc pairwise comparisons showed this was due to guaranteed rewards significantly reducing variability (p=0.0316), while contingent rewards did not (p=0.1219).

Table 3

We also compared both PD ON and OFF separately against the HC with three-way mixed ANOVA, to see under which conditions patients deviated from healthy behaviour. As expected, HC had overall larger amplitudes, quicker saccadic RTs and lower endpoint variability than both PD ON or OFF (Figure 2, see Supplementary file 1B-C for statistics). The use of peak velocity residuals rather than raw velocity factors out the effects of PD on movement amplitude, allowing comparison of the motivational changes in velocity while controlling for differences in the main sequence (Bahill et al., 1975; Manohar et al., 2017). HC did not significantly differ from PD ON or OFF in peak velocity residuals, although their pattern was numerically closest to PD ON with greater contingent motivation.

We additionally checked whether there were practice effects in the PD patients, in case patients behaved differently on their second session due to different expectations. We found no effects or interactions of session on any measure in PD patients (p>0.05).

### Velocity profiles

The effects above demonstrate peak velocity shows strong effects of reward and dopamine, so next we examined the time-course of how velocity was modulated during a saccade. We computed the velocity across time within the movements, and compared the reward effects for PD ON and OFF using cluster-wise permutation tests. Contingent rewards (Performance – Random) did not significantly affect velocity or acceleration for PD ON or OFF, as permutation tests for each condition and the difference between conditions found no significant clusters (cluster-wise permutation tests: p>0.05; Figure 3a&b). However, guaranteed rewards (10 p – 0 p) lead to greater velocity early in the saccade for PD OFF (p<0.05; Figure 3c), which was significantly different from PD ON (p<0.05). Acceleration traces showed this was due to PD OFF having greater acceleration early in the movement (Figure 3d, p<0.05). HC showed no effects of contingent or guaranteed rewards on velocity or acceleration, perhaps unsurprising as there were no differences in overall velocity as reported above. Permutation testing revealed no differences between HC and PD ON or OFF for velocity or acceleration (p>0.05).

Figure 3 with 1 supplement see all

Faster movements are known to be more error-prone (Harris and Wolpert, 1998; Harris and Wolpert, 2006), but motivation can attenuate this effect, making movements more accurate (Manohar et al., 2019). Autocorrelation of eye position over time within saccades provides an indicator of corrective motor signals during movements: noise accumulates during movements, so that variability early in a movement causes endpoint error. This is manifest in autocorrelation, where across trials the eye position at early time-points predicts late time-points. Negative feedback signals correct movement errors during the saccade, and manifest as reductions in this autocorrelation (Codol et al., 2020; Manohar et al., 2019). This feedback, provided by corrective motor signals, can be increased by incentives (Codol et al., 2019; Manohar et al., 2019). In the current study, guaranteed rewards led to greater autocorrelation early in the saccades for PD OFF than ON (Figure 4e & g). This coincides with the greater acceleration PD OFF patients had at the beginning of saccades to guaranteed rewards (Figure 3d), as faster movements have greater motor noise (Harris and Wolpert, 1998; Harris and Wolpert, 2006). Notably, this reward-related autocorrelation did not persist until the end of the saccade, suggesting that negative feedback corrected it. However, as we did not find decreased autocorrelation around the end of the saccades, this represents only indirect evidence of negative feedback.

Figure 4 with 1 supplement see all

### No correlation of the velocity effects for the distinct motivational processes

Previous work had shown that motivation by contingent and guaranteed reward did not correlate across participants (Manohar et al., 2017), so we asked whether dopamine’s effects upon these two types of motivation was also uncorrelated. We found no correlation between effects of contingent and guaranteed rewards on peak saccade velocity residuals in PD ON, PD OFF or HC separately, nor a correlation between medication states, nor between the drug-induced changes in the effects (p>0.05; see Figure 5 legend for statistics). This suggests that the two effects are separate and independent, and not antagonistic within the same person. In particular, the degree to which dopamine improved performance-contingent motivation did not predict the degree to which it reduced motivation by guaranteed rewards.

Figure 5

Source data are available in Figure 5—source data 1.

### Pupil dilatation

We examined pupil dilatation after the cue onset and before the target appeared (after 1400 ms). Previous research has shown a greater effect of contingent than guaranteed reward on pupil dilatation, maximal around 1200 ms after the cue (Manohar et al., 2017), so we used a window-of-interest analysis on the mean pupil dilatation 1000–1400 ms after the cue. There were no significant effects or interactions (p>0.05; Figure 6, see Supplementary file 2A-C for statistics), suggesting that dopamine and reward did not affect pupil responses in PD patients.

Figure 6 with 2 supplements see all

We also used a hypothesis-free analysis, using cluster-wise permutation testing across the whole time-course to look for significant differences between conditions and groups, which also found no significant effects (p>0.05).

We found no correlations between pupil dilatation and motivation effects in any group, or overall (p>0.05; Figure 6—figure supplement 1). Thus, the vigour effects were not related to pupillary dilatation before the movement.

### PD severity

We looked to see whether the dopaminergic effects on velocity residuals could be tied to PD symptom expression. The UPDRS (Martínez-Martín et al., 2015) is a measure of PD symptom severity and was performed in each session; part III measures motor symptom severity. We found no correlations between UPDRS-III scores and reward effects on peak velocity residuals in PD ON (Guaranteed: ρ?=??0.1256, p=0.5410; Contingent: ρ?=??0.2327, p=0.2527) or OFF (Guaranteed: ρ?=??0.2067, p=0.3110; Contingent: ρ?=?0.1553, p=0.4487). Thus, the reward effects were unrelated to PD symptom severity.

### Depression and apathy

We gave participants questionnaires measuring apathy, the AMI (Ang et al., 2017) and depression, BDI-II and HADS (Beck et al., 1996; Zigmond and Snaith, 1983). We found no significant correlations between these questionnaires and contingent or guaranteed motivational effects on peak velocity residuals in PD ON or OFF (p>0.05, see Supplementary file 4 for statistics).

### Fixation period

We looked at whether motivation was affecting behaviour during the fixation period (1400 ms between condition cue onset and target onset) differently, which could potentially lead to differences during the movements. We excluded trials with saccades, blinks, deviations greater than 1.8° and segments with velocities greater than 30°s?1.

PD OFF had more microsaccades (<1°) during the 1400ms fixation period than PD ON (F (1, 201) = 5.0451, p =?.0258, $ηp2$ =?. 0245), but there were no other effects or interactions (p >.05, Supplementary File 3A for statistics). Conversely, ocular drift speed was higher in PD ON than OFF (F (1, 216) = 5.4327, p =?.0207, $ηp2$ =?.0245), but there were no other significant effects or interactions (p >.05, see Supplementary File 3B). Importantly, the lack of interactions means that while patients may have differed in their fixation activity, this was unaffected by motivation conditions, and thus a different pattern to the main effects shown above.

To quantify ocular tremor, we performed Fourier transforms on the eye position in the early (200–700 ms) and late (700–1200 ms) fixation periods, and compared these between conditions with cluster-wise permutation tests to look for clusters of frequencies where patients differed. We found no significant clusters (p>0.05).

## Discussion

In this study, we tested two competing theories of dopaminergic motivation – that dopamine improves instrumental, contingent motivation, and that dopamine improves guaranteed reward motivation via reward expectation. Patients with PD made more vigorous responses, measured by peak saccade velocity residuals (Figure 2a), when rewards were either contingent on performance or guaranteed, but these two effects were differentially affected by dopaminergic medication. When ON medication, PD patients were motivated by rewards contingent on performance, but not by guaranteed rewards. In contrast, when patients were OFF their dopaminergic medication, the opposite pattern was observed; they were motivated by guaranteed rewards, but not by rewards contingent on performance. In this study, older healthy controls were not significantly invigorated by either guaranteed or contingent rewards, although they showed a numerically similar pattern to PD ON. Guaranteed rewards led to PD OFF having earlier increases in velocity and acceleration (Figure 3c & d), which was not seen in PD ON or when rewards were contingent, and this was accompanied by increased autocorrelation of eye position (Figure 4), suggesting increased motor noise early in the saccade. The two motivational effects were uncorrelated across people and between medication states (Figure 5) indicating that dopamine does not promote one type of motivation over another in a competitive fashion, and were not associated with changes in pupil dilatation (Figure 6). Rather, reward expectation and contingency provide distinct motivational drives (Figure 7), which can be dissociated by dopaminergic medication.

Figure 7

The results suggest that dopamine is necessary for contingent motivation. Contingent motivation requires the use of stimulus-action-outcome associations for goal-directed behaviour (Daw and Dayan, 2014; Dickinson, 1985), while reward expectation can occur via stimulus-outcome associations (Niv et al., 2007) that do not require understanding the causal role of action. Our results align with rodent work demonstrating that dorsomedial striatum dopaminergic lesions impair action-outcome associations, such that animals continue to respond to previously rewarding cues even when action-contingency is removed (Lex and Hauber, 2010b). At a more general level, our result is also consistent with dopamine being necessary for behaviours involving a causal state-action-state model (Sharpe et al., 2017), but not simple value-guided actions (Sharp et al., 2016).

Our finding reveals?that dopaminergic medication attenuates the cue-driven reward expectation effect on vigour can be contrasted with previous work suggesting that tonic dopamine couples vigour to average reward rate (Beierholm et al., 2013; Niv et al., 2007). Our adaptive reward schedule held the average reward rate constant over time, while manipulating the average reward rate within each condition, such that the guaranteed 10 p and 0 p trials had different expected rewards. Dopamine might reduce these expectation effects through a different mechanism; the guaranteed cues elicit Pavlovian signals that track expected rewards across states and cues rather than time. Our result implicates dopamine in this signalling, but the direction of effect contrasts with na?ve predictions. Dopamine is necessary for Pavlovian-to-Instrumental transfer (Hall et al., 2001; Kelley and Delfs, 1991; Wyvell and Berridge, 2000) via the nucleus accumbens. In contrast, we show that reward expectation influences vigour when dopaminergic tone is low, yet does not when dopaminergic tone is high. This aligns with the finding that slow, tonic dopaminergic activity is not related to Pavlovian-to-Instrumental transfer (Wassum et al., 2013). A possible explanation is that being ON dopamine led to a saturation in tonic dopamine leaving little room for phasic cue-related reward expectation signals. But if this were the case, one might expect generally higher velocities when ON, compared to PD OFF, which was not seen. Because our contingent and random conditions were matched for average reward rate, and thus opportunity cost, invigoration by contingent reward indicates a truly instrumental effect.

An alternative explanation for the discrepancy with previous research showing dopamine encodes reward rate, is that the previous studies did not fully decouple contingent and non-contingent motivation. In many studies, expected rewards were only given for successful performance (Beierholm et al., 2013; Niv et al., 2007), meaning the rewards were still contingent on performance. However, when separated, contingent motivation has larger effects on vigour than reward expectation (Manohar et al., 2017), and so it is possible that some previously reported effects of average reward rate on vigour were due to the greater contingency, separate from or in addition to, reward expectation. Indeed, vigour may be reduced by dopamine in PD, though reward sensitivity is increased (Muhammed et al., 2016). An additional challenge to the tonic dopamine theory of reward expectation comes from the finding that fast phasic dopaminergic responses in the nucleus accumbens encode average reward rate, but slow tonic responses do not (Mohebi et al., 2019). That study suggests that reward expectation signals are independent of ventral tegmental area dopaminergic neuron firing, and may instead be due to ‘local’ control over nucleus accumbens core dopamine release. As dopamine is depleted in PD via dopamine-neuron death in the substantia nigra and ventral tegmental area, local dopamine release in other areas may be relatively preserved, and thus still able to influence vigour when PD patients are without dopamine.

The effect of reward-expectation on peak velocity was accompanied by greater velocity, acceleration, and autocorrelation early in the saccade for PD OFF than ON. Greater autocorrelation at this point is expected, as greater velocity increases noise (Harris and Wolpert, 2006; Fitts, 1954). However, this noise increase did not persist until the end of the saccade, as there was no increase in autocorrelation at the end of the saccade (Figure 4) and no greater endpoint variability (Figure 2e) – indeed, guaranteed rewards actually decreased endpoint variability, although this was not affected by dopamine. This offers some indirect evidence that the increased noise in this condition was attenuated via negative feedback (c.f. Manohar et al., 2019).

PD patients had slower saccadic RTs, and slower, smaller and more variable saccades compared to age-matched controls. The pattern of invigoration also differed from controls, who did not show significant effects of either contingent motivation or reward-expectation on speed. Instead, controls had lower motor variability when rewards were guaranteed, but no other significant motivation effects. This leads to a pattern where PD ON show contingent motivation, PD OFF show reward-expectation effects, and HC show neither. As these effects themselves are not statistically different between groups, we are limited in the conclusions that we can draw about them. Numerically, controls show a similar pattern to PD ON (Figure 2a), with faster velocity residuals for contingent rewards, which could suggest that dopaminergic medication is restoring healthy function, but care must be taken with this interpretation. The lack of either type of motivation in the older HC is surprising given that in healthy young adults, both contingent and guaranteed rewards increase saccade velocity (Manohar et al., 2017). This could suggest ageing decreases both contingent-motivation and reward-expectation, although a study directly comparing ages would be needed to conclude this.

The motivational effects reported here were not related to any pupillary responses, unlike our previous findings in young people, which may be due to both ageing and PD decreasing the influence of rewards on pupil size (Manohar and Husain, 2015; Muhammed et al., 2016). Additionally, while the two distinct motivational effects on velocity were uncorrelated within PD patients, it is possible that subgroups of patients showed different effects. For example, whether patients were on D2 agonists (Bryce and Floresco, 2019) or had tremor-dominant disease (Wojtala et al., 2019) might be relevant. However, this study was not powered to detect such differences as only six patients were taking agonists in addition to levodopa.

Considering the neuroanatomical differences between contingent motivation and reward expectation may help to explain our results. The nucleus accumbens and ventral pallidum modulate their activity by reward expectation (Mohebi et al., 2019; Tachibana and Hikosaka, 2012), while the caudate nucleus is active when rewards are contingent on behaviour (Lex and Hauber, 2010a; Tricomi et al., 2004). Both the caudate and accumbens/pallidum project to the output nuclei of the basal ganglia, allowing saccade initiation via the superior colliculus, which controls not only the direction of saccades, but also their instantaneous velocity during the movement (Smalianchuk et al., 2018). We propose contingent motivation and reward expectation both lead to motivational signals affecting the superior colliculus’ activity controlling the velocity and acceleration of saccades, and these are differentially affected by dopamine (Figure 7), although we remain agnostic as to the mechanism for this difference. Possibilities include the two systems receiving input from separate regions of the dopaminergic system which are differentially depleted in PD (e.g. dopamine overdose hypothesis [Cools, 2006]), differences in ‘global’ and ‘local’ dopamine signals (Mohebi et al., 2019), or differences in D1-like and D2-like receptor expression within these systems (Surmeier et al., 2007; Yetnikoff et al., 2014). Further studies should address this question of the underlying mechanism.

We have shown that in PD, dopaminergic medication boosts motivation by contingent rewards, but reduces motivation by expected reward. Nonspecific invigoration by reward may thus be generated by a different neural system than goal-directed motivation. This suggests that dopaminergic medication may be a potential treatment for impairments in contingent motivation, but not for deficits related to reward expectation.

## Materials and methods

### Participants

Thirty PD patients were recruited from volunteer databases in the University of Oxford. They were all taking levodopa medication, and some were also taking monoamine oxidase inhibitors and/or dopamine agonists (Table 3). They were randomly assigned to be tested ON or OFF medication first, and withdrawn from standard release medication for 16+ hours and controlled-release medication for 24+ hours. Two patients did not complete both sessions, and two did not have enough trials that passed all the criteria (see Analysis section) so were excluded, leaving 26 patients. Thirty healthy controls (HC) were recruited from volunteer databases also, and tested once, and one HC was excluded for insufficient trials passing the criteria. We recruited 30 participants in each group as this was the sample size used in previous experiments with this task and yielded robust effects (Manohar et al., 2017). Sensitivity power calculations showed this would detect effect sizes above 0.46 (Faul et al., 2009) (α?=?0.05, power?=?0.8, sample size?=?30), although as we only included 26 PD in the analysis, this effect size rose to 0.5.

All participants gave written informed consent, and ethical approval was granted by the South Central Oxford A REC (18/SC/0448).

### Procedure

Request a detailed protocol

The task was run in Matlab (www.mathworks.com, version 7) using the Psychophysics toolbox (Kleiner et al., 2007), on a Windows XP computer with a CRT monitor (1024 × 768 pixels, 40 × 30 cm, 100 Hz refresh rate) at 70 cm viewing distance. Eye movements and pupil size were recorded with Eyelink1000 at 1000 Hz.

On each trial of the task a fixation dot (0.3° radius) was presented at the centre of the screen, with two empty circles (1.1° radius) shown 9.3° to the left and right of the fixation dot. After 500 ms of fixation, a cue was given by a voice over the speaker, indicating the type of trial the participant was in:

• ‘Performance’ indicated that fast response times would win 10 p, while slow response times would win 0 p

• ‘Random’ indicated a 50% probability of 10 p or 0 p, regardless of response time

• ‘Ten pence’ indicated a guaranteed 10 p, regardless of response time

• ‘Zero pence’ indicated guaranteed 0 p, regardless of response time

A delay of 1400, 1500 or 1600 ms was given (with equal probability), after which one of the two circles turned white (50% probability of left or right) and participants had to saccade to this circle to complete the trial and receive the outcome.

Participants could only affect the outcome in the Performance condition (by moving faster); all others were independent of their speed. In the Performance condition, rewards were based upon response time (i.e. total time between the target appearing and gaze arriving at the target), which is only minimally influenced by saccade velocity. Participants were rewarded when response time was quicker than their recent median response time for the last 20 Performance trials, which thus yielded a 50% reward rate overall. The Random condition acts as a control to these trials, with a random 50% of trials rewarded, and thus equal expected value but with no performance-contingency. Rewards in the guaranteed conditions also had zero contingency on performance, but yielded different expected rewards (10 p vs 0 p), thus comparing them indexes the pure effect of expecting reward.

When rewards are contingent, people get feedback about how they performed. This itself is known to increase motivation, independent of reward – a phenomenon termed intrinsic motivation. To control for this, we ensured participants always received feedback on their speed (fast/slow, using median split over 20 previous trials in that condition – i.e. the same criteria as for contingent rewards), regardless of reward. This should equate the level of intrinsic motivation across conditions, providing that the feedback is as noticeable as the reward. In order to ensure the speed feedback and reward were matched in physical salience, the feedback modalities were counterbalanced. Two blocks gave auditory feedback for speed and visual feedback for reward, and vice versa for the other two blocks, with order randomised across participants. This counterbalancing accords with our previous study (Manohar et al., 2017) which found no modality effects, suggesting participants were attending to audio and visual feedback equally. We also found no effects of modality on any of the measures of interest (p>0.05), so collapsed across the two modalities for all the analyses.

There were 12 types of each trial in a block, in a random order, and participants completed four blocks.

### Analysis

The Performance and 10 p conditions are high motivation conditions. The difference between Performance and Random conditions gives the effect of contingent motivation, while the difference between 10 p and 0 p conditions gives the effect of reward expectation.

As in previous studies (Manohar et al., 2017), our primary measure of interest was saccadic vigour. We measured peak saccade velocity on each trial. We took the first saccade after target onset which was greater than 1° in amplitude, and used a sliding window of 4 ms width to calculate velocity, excluding segments faster than 3000°s?1 or where eye tracking was lost. Saccades with peak velocities outside 80–2500°s?1 were excluded, as were trials where participants reached the target before 180 ms or after 580 ms. Two PD patients and one HC had fewer than 10 trials that passed these criteria for one condition, so were excluded from the analysis.

To remove the main sequence effect of amplitude on velocity (Bahill et al., 1975; Harris and Wolpert, 2006), we regressed velocity against amplitude and took the peak velocity residuals as our measure of interest. This measures the difference between the velocity predicted by the main sequence, and the velocity actually recorded, with positive (negative) values meaning faster (slower) velocity. This was done for each participant’s separate session. This approach has been used before, by us and others (Blundell et al., 2018; Manohar et al., 2017; Muhammed et al., 2020; Muhammed et al., 2016; Van Opstal et al., 1990), and it is similar to simply including amplitude as a covariate when analysing raw peak velocity, but it does not reduce the degrees of freedom and yields simpler to interpret results. Moreover since motivation increases amplitude (Manohar et al., 2019), including amplitude as a covariate would mean that amplitude would compete with motivation to explain variance in velocity, potentially resulting in overestimation of motivation effects. Our findings did not qualitatively change when we used the covariate approach instead.

We also measured amplitude, saccadic reaction time (RT), and endpoint variability of these saccades. Saccadic RT is the time between the target onset and the start of the saccade.

To analyse velocity and acceleration traces, and autocorrelation and covariance of the eye movements we linearly interpolated 50 points along each saccade to move them into the same units. Instantaneous velocity was smoothed across three time-points, while acceleration was smoothed across 5. We also calculated velocity and acceleration traces on the raw (non-interpolated) traces and then interpolated them afterwards, which gave very similar results.

Pupil dilatation was measured in arbitrary units (a.u.) relative to the baseline pupil size at the cue onset. Blinks under 500 ms were linearly interpolated, steps in pupil size above 2.5 a.u./ms were removed, and data were averaged in 20 ms bins for plotting.

We used rmanova from the matlib toolbox (https://github.com/sgmanohar/matlib;?Manohar, 2020) to perform analyses – this uses fitglme to perform the repeated-measures test and anova to perform hypothesis tests on the GLME. We used three-way repeated-measures ANOVA to compare effects of motivation, contingency and dopaminergic medication in PD patients, and followed this up with two-way ANOVA when a three-way interaction was found. These analyses were also performed using a full linear mixed effects model including each trial, which produced qualitatively identical results. To compare each PD condition against HC we used mixed ANOVA. We also used cluster-wise permutation tests for the time-course data (velocity, acceleration, pupil dilatation, autocorrelation and covariance), to control the family-wise error rate at. 05.

### Data and code availability

Request a detailed protocol

Analyses were performed in Matlab using custom scripts, which are available on GitHub (https://doi.org/10.5281/zenodo.4032711). Anonymous data are available on OSF (https://osf.io/2k6x3), as is the experiment file (osf.io/y9xhp)?https://doi.org/10.5281/zenodo.4032711.

## References

1. 1
2. 2
3. 3
4. 4
5. 5
Beck depression Inventory–II (1996)
In the Psychological Corporation.
https://doi.org/10.1037/t00742-000
6. 6
Dopamine modulates reward-related vigor (2013)
Neuropsychopharmacology 38:1495–1503.
https://doi.org/10.1038/npp.2013.48
7. 7
8. 8
9. 9
10. 10
11. 11
12. 12
13. 13
The algorithmic anatomy of model-based evaluation (2014)
Philosophical Transactions of the Royal Society B: Biological Sciences 369:20130478.
https://doi.org/10.1098/rstb.2013.0478
14. 14
15. 15
16. 16
17. 17
Actions and habits : the development of behavioural autonomy (1985)
Philosophical Transactions of the Royal Society B 308:67–78.
https://doi.org/10.1098/rstb.1985.0010
18. 18
19. 19
20. 20
21. 21
22. 22
23. 23
24. 24
25. 25
26. 26
What’s new in Psychtoolbox-3? A free cross-platform toolkit for Psychophysics with Matlab & GNU/Octave
(2007)
Perception 36:1.
27. 27
28. 28
Mental labour (2018)
Nature Human Behaviour 2:899–908.
https://doi.org/10.1038/s41562-018-0401-9
29. 29
Performance pay and productivity (2000)
American Economic Review 90:1346–1361.
https://doi.org/10.1257/aer.90.5.1346
30. 30
31. 31
32. 32
33. 33
34. 34
35. 35
36. 36
Matlib (2020)
GitHub.
37. 37
38. 38
39. 39
40. 40
41. 41
42. 42
43. 43
A normative perspective on motivation (2006)
Trends in Cognitive Sciences 10:375–381.
https://doi.org/10.1016/j.tics.2006.06.010
44. 44
45. 45
46. 46
47. 47
48. 48
49. 49
50. 50
51. 51
52. 52
53. 53
54. 54
55. 55
56. 56
57. 57
58. 58
59. 59
60. 60
61. 61
62. 62

## Decision letter

1. Shelly B Flagel
Reviewing Editor; University of Michigan, United States
2. Michael J Frank
Senior Editor; Brown University, United States
3. Shelly B Flagel
Reviewer; University of Michigan, United States
4. Roshan Cools

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The manuscript describes a unique and elegantly designed experiment to parse the role of dopamine in reward learning using human subjects – namely, patients with Parkinson's Disease, on and off dopaminergic medication. The results are novel, showing that patients with Parkinson's Disease that were ON medication had greater response vigor for contingent rewards; while those OFF medication had greater vigor for guaranteed rewards. These findings support the long-standing notion that reward expectation and contingency learning represent distinct motivational and neurobiological processes.

Decision letter after peer review:

Thank you for submitting your article "Dopamine promotes instrumental motivation, but reduces reward-related vigour" for consideration by eLife. Your article has been reviewed by two peer reviewers, including Shelly B Flagel as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Michael Frank as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Roshan Cools (Reviewer #2).

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

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (http://www.asadoresteatre.com/articles/57162). Specifically, we are asking editors to accept without delay manuscripts, like yours, that they judge can stand as eLife papers without additional data, even if they feel that they would make the manuscript stronger. Thus the revisions requested below only address clarity and presentation.

Summary:

This is a timely paper on a topic of great interest with intriguing findings that incidentally also reconcile some paradoxical findings in extant literature. A clever novel rewarded saccade task was used to disentangle effects of dopaminergic medication in Parkinson's disease on two distinct types of motivation: contingent (instrumental) motivation, where reward depends on performance versus non-contingent (non-instrumental) motivation, where reward is guaranteed. Results reveal a unique double dissociation, with contingent motivation (indexed by peak velocity) being greater ON than OFF meds, but noncontingent motivation being greater OFF than ON meds. Thus PD patients OFF their medication showed a positive effect of non-contingent reward on vigor, but no effect of contingency, whereas the same patients ON their dopaminergic medication showed a positive effect of contingent vs non-contingent reward on vigor, but no effect of non-contingent reward.

This is a robust set of findings and a beautiful cross-over effect, not often seen, obtained using a well-controlled paradigm. The authors provide thorough auxiliary and control analyses to validate the specificity of their effect of interest.

Revisions:

Some concerns raised by both reviewers, regarding data analysis, interpretation and conclusions are outlined below. Of particular note, the authors need to better incorporate the reported findings into the extant literature.

1) Given the involvement of dopamine in "expectation", did the authors investigate any potential impact of order effects? That is, if the subjects were tested ON or OFF medication first, did that impact subsequent responding? It may be that the sample size is too low for this analysis and/or for the correlations to be meaningful; nonetheless, this possibility should be discussed.

2) As peak saccade velocity residuals were the primary outcome measure affected, this metric warrants further description and justification in the primary text, prior to the results.

3) It seems that HC subjects are not included in the primary analyses, but only compared to PD ON and PD OFF, in separate analyses, in the supplemental text. HC subjects should be included in the primary analyses, as they are shown in Figure 2. In relation, the authors should further interpret the fact that HC subjects differ from PD patients on measures of amplitude, saccadic RT and endpoint variability, but seem to resemble PD ON for peak velocity residuals (as shown in Figure 2). While the authors briefly discuss the negative results in HC subjects, it is not clear why there is no effect of motivation in these subjects, as one might expect. The pattern observed in HC subjects might help to explain the data from the PD subjects and therefore should be better incorporated into the manuscript. HC data should be included in Figure 3 and the accompanying analyses, as well.

4) Additional "discussion should be provided regarding the conclusion that these data suggest that dopamine is necessary only for model-based, and not model-free learning; as there are other potential explanations; especially since the HC data don't seem to support model-based learning in this task.

5) Considerable parts of the Introduction and Discussion should be rewritten. Specifically, the data do not undermine the (i) theory that dopamine is key for reward expectation (Abstract: “challenging the theory that tonic dopamine encodes reward expectation”) or (ii) average reward rate (Discussion: “this contrasts with previous research which suggests tonic dopamine encode the average reward rate”) or (iii) model-free behavior (“thus the work fits with previous research in which dopamine is necessary only for model-based, not model-free learning”). These conclusions are not supported by the current findings, because:

i) Detrimental effects of dopamine implicate dopamine as much as do beneficial effects as the authors conclude later on in the discussion.

ii) Average reward rate was not manipulated and its effects not measured. Related to this issue, I do not think that the Niv view is optimally characterized by attributing to it the prediction that dopamine would leave contingent motivation unaltered.

iii) The noncontingent Pavlovian procedure tested here doesn't address model-free learning, as measured in the papers referenced. Moreover, evidence for impaired model-based vs intact model-free control in PD (first evidenced by de Wit et al. in JoCN) is not directly relevant here, given the fairly wide gap between the type of inference that is measured in e.g. the two-step task and similar other learning tasks and the contingent reward motivation manipulation measured here.

Some of these conclusions (e.g. (i)) also counter the hypothesis put forward later in the discussion and illustrated in Figure 6 about the two forms of motivation implicating distinct dopaminergic mechanisms.

6) Two sets of literature that are directly relevant to this research are absent from the manuscript and need to be incorporated.

? The first is (pharmacological and lesion) evidence from work with experimental rodents, but also humans, on dopamine's role in the Pavlovian control of behavior, e.g. on conditioned reinforcement (Parkinson et al; Taylor and Robbins), Pavlovian-to-instrumental transfer (Dickinson et al., 2001; Wyvell and Berridge, 2000; Talmi et al., 2008), autoshaping, goal vs sign tracking etcetera.

? The second is prior evidence for enhanced effects of reward motivation in PD patients OFF but not ON meds in a manner that depends on dopamine cell loss (Aarts et al., 2012, and Timmer et al., 2018). The Introduction should reflect this prior evidence. In fact, the rationale for the current study is strengthened by the apparent discrepancy between this evidence for enhanced motivation in PD OFF meds (somewhat reminiscent of paradoxical kinesia) and the evidence previously obtained by the authors for reduced motivation in PD OFF meds.

7) The current version of the Introduction (as well as some of the Discussion) is structured based a contrast between model-based versus Pavlovian control of behavior that does not do justice our understanding of these concepts from classic learning theory. For example, the statement that “Motivation by reward expectation, independent of what an agent does, is model-free” should be rephrased. Pavlovian control can be model-free or model-based (Dayan and Berridge, 2014, CABN or Bornstein and Daw, 2011, Curr Opin Neurobiol). Related to this is my issue with the conclusion in the Discussion (see above).

8) Some of the statistical choices need to be justified and, data presentation optimized.

? The use of peak velocity residuals, corrected for amplitude, should be justified more clearly.

? It would also be good to clarify the rationale for using these residuals rather than correcting the velocities for amplitude with a covariate of no interest e.g. in a trial-wise mixed linear regression analysis.

? The use of mixed linear modeling would allow both between- as well as within-subject variability to be taken into account. Was this considered?

? I suggest that readers get access to uncorrected, observed velocity data rather than only the residuals.

? While individual datapoints are presented in the supplement, I see no principled reason to present aggregate dataplots in the main text. Why not move the supplementary figures with actual datapoints to the main text? Raincloud plots might increase the readability of these plots?

https://doi.org/10.7554/eLife.58321.sa1

## Author response

Revisions:

Some concerns raised by both reviewers, regarding data analysis, interpretation and conclusions are outlined below. Of particular note, the authors need to better incorporate the reported findings into the extant literature.

We thank the reviewers for their time and their helpful comments. We have addressed all of them, and believe it has made the manuscript more robust and readable. We address each of the points below.

When adding individual data points and HC data to some figures, we had to split some figures into new separate ones. We also noticed that some participants had not been excluded from the autocorrelation and pupillometry data, so have redone those analyses to match the exclusions applied to the other data. This has resulted in a new cluster of significance in PD OFF which matches up with the previous ON vs OFF difference, so has not changed our findings. We also corrected a typo in the demographics Table 3 (PD ACE was wrong), and have moved legends and tables to the end of this file and uploaded figures separately as requested by eLife editorial support.

1) Given the involvement of dopamine in "expectation", did the authors investigate any potential impact of order effects? That is, if the subjects were tested ON or OFF medication first, did that impact subsequent responding? It may be that the sample size is too low for this analysis and/or for the correlations to be meaningful; nonetheless, this possibility should be discussed.

We agree with the reviewer that order effects could be important. Including which session (ON/OFF) was performed first as a factor in the analysis revealed no effects of session order, or any interactions with session order, on any of the saccade measures reported. We have stated this in the Results section:

“We additionally checked whether there were practice effects in the PD patients, in case patients behaved differently on their second session due to different expectations. We found no effects or interactions of session on any measure in PD patients (p >.05).”

2) As peak saccade velocity residuals were the primary outcome measure affected, this metric warrants further description and justification in the primary text, prior to the results.

We have further explained this measure:

“A saccade’s velocity is tightly governed by its amplitude, a relation known as the “main sequence” (Bahill, Clark and Stark, 1975). […] We did this regression for each participant and session separately, but across conditions.”

We have also added a panel to Figure 1 to have an illustrative example of residuals using sample data, and have updated the legend to:

“Example of the main sequence and velocity residuals – the points show a subset of individual trials illustrating the “main sequence” relationship where larger saccades have greater velocity, shown by the regression line. The distance from each point to its line is the velocity residual, which we take as out main measure of response vigour.”

3) It seems that HC subjects are not included in the primary analyses, but only compared to PD ON and PD OFF, in separate analyses, in the supplemental text. HC subjects should be included in the primary analyses, as they are shown in Figure 2. In relation, the authors should further interpret the fact that HC subjects differ from PD patients on measures of amplitude, saccadic RT and endpoint variability, but seem to resemble PD ON for peak velocity residuals (as shown in Figure 2). While the authors briefly discuss the negative results in HC subjects, it is not clear why there is no effect of motivation in these subjects, as one might expect. The pattern observed in HC subjects might help to explain the data from the PD subjects and therefore should be better incorporated into the manuscript. HC data should be included in Figure 3 and the accompanying analyses, as well.

We agree with the reviewers that it would be useful to add interpretation to the HC results. We have therefore moved the paragraphs describing the HC-only findings to earlier in the first Results section, and rewritten it to be clearer. We have explained why residual velocity did not show a main effect of PD vs HC, and added a sentence evaluating the comparison to HC in the Results section:

“The HC peak velocity residuals were not affected by contingency, motivation or the interaction (p >.05; see Table 2), suggesting that healthy older adults do not adjust their response vigour for contingent or guaranteed rewards. […] HC did not significantly differ from PD ON or OFF in peak velocity residuals, although their pattern was numerically closest to PD ON with greater contingent motivation.”

And have expanded the Discussion of these results:

“PD patients had slower saccadic RTs, and slower, smaller and more variable saccades compared to age-matched controls. […]This could suggest ageing decreases both contingent motivation and reward-expectation, although a study directly comparing ages would be needed to conclude this.”

We have added HC data to all data figures, and the accompanying analyses. We have moved the autocorrelation data to a new separate figure in order to include the HC data, and we now present two correlation matrices (for contingent and guaranteed reward effects) for each group (PD ON, OFF, HC) and the ON-OFF difference.

4) Additional "discussion should be provided regarding the conclusion that these data suggest that dopamine is necessary only for model-based, and not model-free learning; as there are other potential explanations; especially since the HC data don't seem to support model-based learning in this task.

We have rewritten this part of the discussion, discussing PIT and instrumental model-free associations, and have clarified that model-based associations are only one possibility here:

“The results suggest that dopamine is necessary for contingent motivation. […] At a more general level, our result is also consistent with dopamine being necessary for behaviours involving a causal state-action-state model (Sharpe et al., 2017), but not simple value-guided actions (Sharp et al., 2016).”

5) Considerable parts of the Introduction and Discussion should be rewritten. Specifically, the data do not undermine the (i) theory that dopamine is key for reward expectation (Abstract: “challenging the theory that tonic dopamine encodes reward expectation”) or (ii) average reward rate (Discussion: “this contrasts with previous research which suggests tonic dopamine encode the average reward rate”) or (iii) model-free behavior (“thus the work fits with previous research in which dopamine is necessary only for model-based, not model-free learning”). These conclusions are not supported by the current findings, because:

i) Detrimental effects of dopamine implicate dopamine as much as do beneficial effects as the authors conclude later on in the discussion.

We have thought carefully about the interpretation of our findings, and we agree with the reviewers that our main result does not fit exactly where we placed it relative to the extensive animal literature.

We have changed this claim:

“We posit that dopamine promotes goal-directed motivation, but dampens reward-driven vigour, contradictory to the prediction that increased tonic dopamine amplifies reward expectation.”

ii) Average reward rate was not manipulated and its effects not measured. Related to this issue, I do not think that the Niv view is optimally characterized by attributing to it the prediction that dopamine would leave contingent motivation unaltered.

The reviewer correctly argues that although we manipulated the average reward rates between conditions, the reward rate over time was not manipulated. It is therefore not possible for us to directly address the Niv view. We have clarified this and added a sentence pointing out this difference may be relevant for the lack of effect. We also mention that the fact dopamine reduces guaranteed motivation is in fact consistent with dopamine’s being involved with this kind of motivation:

“Our finding that dopaminergic medication attenuates the cue-driven reward expectation effect on vigour can be contrasted with previous work suggesting that tonic dopamine couples vigour to average reward rate (Beierholm et al., 2013; Niv et al., 2007). […] Because our contingent and random conditions were matched for average reward rate, and thus opportunity cost, invigoration by contingent reward indicates a truly instrumental effect.”

iii) The noncontingent Pavlovian procedure tested here doesn't address model-free learning, as measured in the papers referenced. Moreover, evidence for impaired model-based vs intact model-free control in PD (first evidenced by de Wit et al. in JoCN) is not directly relevant here, given the fairly wide gap between the type of inference that is measured in e.g. the two-step task and similar other learning tasks and the contingent reward motivation manipulation measured here.

We have shifted our focus away from “model-based vs model-free” and acknowledge that our task does not help understand the traditional distinction between forward planning vs reinforcement learning. We now clarify that the task is probing motivation through action-outcome knowledge, and is related to Pavlovian vs instrumental drives (see response to point 4 for pasted text).

Some of these conclusions (e.g. (i)) also counter the hypothesis put forward later in the discussion and illustrated in Figure 6 about the two forms of motivation implicating distinct dopaminergic mechanisms.

We have adjusted these claims, and believe the new literature added to the discussion fits with the proposal of two separate systems.

6) Two sets of literature that are directly relevant to this research are absent from the manuscript and need to be incorporated.

? The first is (pharmacological and lesion) evidence from work with experimental rodents, but also humans, on dopamine's role in the Pavlovian control of behavior, e.g. on conditioned reinforcement (Parkinson et al; Taylor and Robbins), Pavlovian-to-instrumental transfer (Dickinson et al., 2001; Wyvell and Berridge, 2000; Talmi et al., 2008), autoshaping, goal vs sign tracking etcetera.

? The second is prior evidence for enhanced effects of reward motivation in PD patients OFF but not ON meds in a manner that depends on dopamine cell loss (Aarts et al., 2012 and Timmer et al., 2018). The Introduction should reflect this prior evidence. In fact, the rationale for the current study is strengthened by the apparent discrepancy between this evidence for enhanced motivation in PD OFF meds (somewhat reminiscent of paradoxical kinesia) and the evidence previously obtained by the authors for reduced motivation in PD OFF meds.

We thank the reviewers for pointing out these highly relevant bodies of literature. We have rewritten the Introduction to reflect this, and added them to the Discussion (see above):

“Contingent rewards motivate us because we understand the causal relation between successful actions and reward. […] We tested the two predictions that dopamine is involved in motivation by expected rewards, and by contingent rewards.”

7) The current version of the Introduction (as well as some of the discussion) is structured based a contrast between model-based versus Pavlovian control of behavior that does not do justice our understanding of these concepts from classic learning theory. For example, the statement that “Motivation by reward expectation, independent of what an agent does, is model-free” should be rephrased. Pavlovian control can be model-free or model-based (Dayan and Berridge, 2014, CABN or Bornstein and Daw, 2011, Curr Opin Neurobiol). Related to this is my issue with the conclusion in the Discussion (see above).

We thank the reviewer for advising on these very relevant distinctions, which we now make in our narrative. We have rewritten the Introduction and the discussion to modify these claims. We had previously attempted to suggest that model-based reasoning could underlie the PD ON benefit, although we were not clear enough that this was only one possibility. We have refocused the paper on the more clear distinction between goal-directed action-outcome and Pavlovian stimulus-outcome associations (see responses above).

8) Some of the statistical choices need to be justified and, data presentation optimized.

? The use of peak velocity residuals, corrected for amplitude, should be justified more clearly.

We have justified this (please see response to point 2 above).

? It would also be good to clarify the rationale for using these residuals rather than correcting the velocities for amplitude with a covariate of no interest e.g. in a trial-wise mixed linear regression analysis.

We have added the following to the Materials and methods section, justifying our use of residuals in ANOVA:

“To remove the main sequence effect of amplitude on velocity (Bahill et al., 1975; Harris and Wolpert, 2006), we regressed velocity against amplitude and took the peak velocity residuals as our measure of interest. […] Our findings did not qualitatively change when we used the covariate approach instead.”

? The use of mixed linear modeling would allow both between- as well as within-subject variability to be taken into account. Was this considered?

We planned in advance to use repeated measures ANOVA (i.e. a mixed linear model on the condition means). But as the reviewer suggests, trial-wise mixed models can provide a much more sensitive test and account for the different numbers of valid trials between conditions and subjects. We performed this analysis. For velocity residuals we obtained the following results shown in Author response table 1:

Author response table 1

In fact, for each variable, the pattern of significant effects turned out the same for both methods. Thus we added the following to the Materials and methods/analysis section:

“All statistical analyses were also performed using a full linear mixed effects model including each trial, which produced in qualitatively identical results.”

? I suggest that readers get access to uncorrected, observed velocity data rather than only the residuals.

We have included raw velocity as an extra panel in Figure 2, included the statistics in all main and supplementary results tables, and discussed this measure in the Results:

“Raw peak velocity had an effect of motivation, as both types of motivation increased speed (Figure 2F, p =.0110), although this will include effects of changes in amplitude (via the main sequence) which showed a borderline significant effect of motivation (Figure 2C, p =.0607).”

? While individual datapoints are presented in the supplement, I see no principled reason to present aggregate dataplots in the main text. Why not move the supplementary figures with actual datapoints to the main text? Raincloud plots might increase the readability of these plots?

We have added individual data points to the line graphs in Figure 2 (raincloud plots were very busy due to the 2*2*3 design). A few HC had strong motivation effects, which expanded the y-axes a lot, making the effects harder to see (see Author response image 1 for example),

Author response image 1

So we have put the HC data on a separate panel (Figure 2B) to the PD data (Figure 2A) which we hope will be acceptable.The correlation plots already show individual data, but we were unable to include individual participants’ timecourses of velocity, acceleration, autocorrelation, or pupillometry on the respective figures as they swamped the means. Instead, we have added figure supplements showing these individual data, which we hope will be acceptable.

Figure 3—figure supplement 1 and Figure 6—figure supplement 2 are the new supplementary figures showing individual velocity and acceleration data, and pupil dilatation.

While making the individual pupillometry and autocorrelation plots, we realised we had not excluded the 4 PD and 1 HC, as we had for the main analyses. We have now updated our analyses and figures to have 26 PD and 29 HC to match the main saccade data. This affects Figures 4 and 6 only, and has not changed the findings except that PD OFF now have a significant cluster of higher autocorrelation at the beginning of the saccade for guaranteed motivation (Figure 4).

This increase when off (panel g) now matches the significant effect of dopamine (panel e) found when comparing PD ON vs OFF autocorrelations, so has not changed our interpretation. We have added:

“This coincides with the greater acceleration PD OFF patients had at the beginning of saccades to guaranteed rewards (Figure 3D), as faster movements have greater motor noise (Harris and Wolpert, 1998, 2006).”

The pupil size text has not changed.

https://doi.org/10.7554/eLife.58321.sa2

## Article and author information

### Author details

1. #### John P Grogan

Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom
##### Contribution
Resources, Data curation, Software, Formal analysis, Investigation, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing
##### For correspondence
john.grogan@ndcn.ox.ac.uk
##### Competing interests
No competing interests declared
2. #### Timothy R Sandhu

1. Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom
2. Department of Psychology, University of Cambridge, Cambridge, United Kingdom
##### Contribution
Data curation, Investigation, Methodology, Writing - review and editing
##### Competing interests
No competing interests declared
3. #### Michele T Hu

1. Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom
2. Oxford Parkinson’s Disease Centre, University of Oxford, Oxford, United Kingdom
##### Contribution
Resources, Writing - review and editing
##### Competing interests
MTH is a consultant advisor to the Roche Prodromal Advisory, Biogen Digital Advisory Board, Evidera, and CuraSen Therapeutics, Inc.
4. #### Sanjay G Manohar

1. Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom
2. Department of Experimental Psychology, University of Oxford, Oxford, United Kingdom
##### Contribution
Conceptualization, Resources, Supervision, Funding acquisition, Methodology, Writing - review and editing
##### Competing interests
No competing interests declared

### Funding

#### MRC (MR/P00878X)

• Sanjay G Manohar

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

### Acknowledgements

We would like to thank the patients and participants for their time in taking part in this study, and the funders for their support (MRC Clinician Scientist Fellowship to SGM, MR/P00878X).

### Ethics

Human subjects: Ethical approval was granted by the South Central Oxford A REC (18/SC/0448). All participants gave written informed consent.

### Senior Editor

1. Michael J Frank, Brown University, United States

### Reviewing Editor

1. Shelly B Flagel, University of Michigan, United States

### Reviewers

1. Shelly B Flagel, University of Michigan, United States
2. Roshan Cools, Radboud University Nijmegen, Netherlands

### Publication history

2. Accepted: September 30, 2020
3. Accepted Manuscript published: October 1, 2020 (version 1)
4. Version of Record published: October 30, 2020 (version 2)

? 2020, Grogan et al.

## Metrics

• 1,099
Page views
• 151
• 0
Citations

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

A two-part list of links to download the article, or parts of the article, in various formats.

### Open citations (links to open the citations from this article in various online reference manager services)

1. Chromosomes and Gene Expression
2. Neuroscience

# Conditional protein tagging methods reveal highly specific subcellular distribution of ion channels in motion-sensing neurons

Sandra Fendl et al.
Tools and Resources Updated

Neurotransmitter receptors and ion channels shape the biophysical properties of neurons, from the sign of the response mediated by neurotransmitter receptors to the dynamics shaped by voltage-gated ion channels. Therefore, knowing the localizations and types of receptors and channels present in neurons is fundamental to our understanding of neural computation. Here, we developed two approaches to visualize the subcellular localization of specific proteins in Drosophila: The flippase-dependent expression of GFP-tagged receptor subunits in single neurons and ‘FlpTag’, a versatile new tool for the conditional labelling of endogenous proteins. Using these methods, we investigated the subcellular distribution of the receptors GluClα, Rdl, and Dα7 and the ion channels para and Ih in motion-sensing T4/T5 neurons of the Drosophila visual system. We discovered a strictly segregated subcellular distribution of these proteins and a sequential spatial arrangement of glutamate, acetylcholine, and GABA receptors along the dendrite that matched the previously reported EM-reconstructed synapse distributions.

1. Neuroscience
2. Stem Cells and Regenerative Medicine

# 16p11.2 microdeletion imparts transcriptional alterations in human iPSC-derived models of early neural development

Julien G Roth et al.
Tools and Resources

Microdeletions and microduplications of the 16p11.2 chromosomal locus are associated with syndromic neurodevelopmental disorders and reciprocal physiological conditions such as macro/microcephaly and high/low body mass index. To facilitate cellular and molecular investigations into these phenotypes, 65 clones of human induced pluripotent stem cells (hiPSCs) were generated from 13 individuals with 16p11.2 copy number variations (CNVs). To ensure these cell lines were suitable for downstream mechanistic investigations, a customizable bioinformatic strategy for the detection of random integration and expression of reprogramming vectors was developed and leveraged towards identifying a subset of 'footprint'-free hiPSC clones. Transcriptomic profiling of cortical neural progenitor cells derived from these hiPSCs identified alterations in gene expression patterns which precede morphological abnormalities reported at later neurodevelopmental stages. Interpreting clinical information—available with the cell lines by request from the Simons Foundation Autism Research Initiative—with this transcriptional data revealed disruptions in gene programs related to both nervous system function and cellular metabolism. As demonstrated by these analyses, this publicly available resource has the potential to serve as a powerful medium for probing the etiology of developmental disorders associated with 16p11.2 CNVs.