THP modulates saccadic measures.

a) Trial structure for a high-incentive trial with a salient distractor. After fixation on the pink starting circle, the incentive cue plays, and after a short fixation wait the preparation cue is given which is the fixation cross turning white. 1500ms later, one circle dims, which is the target, and on 50% of trials the other circle brightens (salient distractor). Feedback is given when participants saccade to the target, based on their speed. Timings are given below each screen. b) Eye position as a function of time for a selection of saccades. Saccade reaction time (RT) is the time at which the saccade begins, peak velocity is the maximal speed during the movement (steepest slope here), and amplitude is the distance to the saccade endpoint. c) Plotting peak velocity against amplitude (sample data) shows the main sequence effect (dashed lines) where larger saccades have higher velocity. We regressed velocity on amplitude separately within each drug session (to control for potential drug-effects on amplitude, velocity, or the main sequence), giving residual peak velocity as our measure of vigour (solid vertical lines). d) Mean peak residual velocity for each condition. Incentives increased velocity, distractors decreased it, and THP reduced the invigoration by incentives. This interaction was significant only for the no-distractor trials. Crosses show individual participant means for each condition. E) Mean saccadic RT for each condition (log RT was analysed, raw values plotted here). High incentives decreased RT, distractors slowed RT, and THP reduced the effect of incentive on RT – which was driven by trials with distractors present. Crosses show individual participant means for each condition. The following figure supplements are available:

Figure 1 – Figure supplement 1. Linear mixed-effects single-trial regression outputs for behavioural variables.

Figure 1 – Figure supplement 2. Incentive effects are uncorrelated between saccade RT and velocity, but the THP*incentive interactions are correlated.

Muscarinic blockade increases the pull from salient distractors.

A) Sample saccades showing fixation at the bottom left circle, the target on the right, and the distractor on the top. Distractor pull is the angle of the eye when it leaves the fixation circle, relative to a straight line from the fixation to target circle (positive values reflect angles towards the distractor, zero is flat, negative reflects repulsion). B) Mean distractor pulls for low and high incentives when the salient distractor is and is not present (crosses show individual participant means per condition). Distractor pull was negative (i.e., below the horizontal line in panel a) reflecting repulsion from the distractor when it did not light up. However, when the distractor did light up, distractor pull was positive, reflecting a bias towards it, and this bias was greater on THP than placebo (statistics are given in table in Figure 1 – Figure supplement 1). C) Mean kernel-smoothed density of distractor pulls for all trials with a distractor (averaged across all other conditions) with shading showing the within-subject standard errors. There is a smaller peak centred on the distractor’s orientation (grey dashed line and circle). Negative distractor pulls show the repulsive bias away from the distractor location. D) Mean kernel-smoothed densities showing the effects of incentive (i.e., 50p – 0p) and THP (i.e., THP – incentive) for all ‘with distractor’ trials. Cluster-based permutation testing showed that THP reduced the number of trials biased around -30°, indicating reduced repulsive bias when muscarinic receptors are antagonised. The following figure supplements are available:

Figure 2 – Figure supplement 1. Distractor pull as a function of reaction time.

Mean ERPs to the preparation cue.

a) Grand-average ERPs in electrode Cz split for the four conditions (low & high incentive, placebo & THP). The three time-windows are highlighted in different colours, and correspond to the columns of panels below, with topographies of the mean amplitude within each window superimposed. The ‘incentive’ window is a non-contiguous window of 900-1100ms after the incentive cue, just before the preparation-cue appears, which contains the late negative potential after the incentive cue. b-d) The mean voltages within each time-window for the different incentive and drug conditions (individual participants’ data are shown in Figure Supplement 1). b) Late ERP to the incentive cue (900:1100ms at Cz) was more negative when on THP than placebo, but it was not affected by incentive (p > .05). c) Mean P3a (200:280ms after the preparation cue) is decreased by high incentives but unaffected by THP (note the different y-axis scale to b&d). d) The CNV (1200:1500ms after the preparation cue) is strengthened (more negative) by incentives and THP, with a weak interaction (p = .0213) as THP slightly reduces the incentive effect (flatter slope for the orange line; and THP lines are closer than Placebo lines in panel a). e-g) The β coefficients from regressing each component against each behavioural variable, with stars representing significant (p < .0056; Bonferroni-corrected for 9 comparisons) associations (error bars are 95% CI). h) Difference waves showing the effects of Incentive (50p minus 0p, averaged over other factors) and THP (THP minus placebo, averaged over other factors). Incentive starts decreasing (i.e., strengthening) the CNV early (during the P3a window), while the THP effect starts around 900ms after the preparation cue. i) Difference waves showing Incentive effects within each drug condition separately. Incentives strengthen the CNV for both conditions, with the effect growing more slowly for the THP condition, reflecting the THP*Incentive interaction reported in the main text. The following figure supplements are available:

Figure 3 – Figure supplement 1. Mean individual participants’ ERP component voltages for each time-window

Figure 3 – Figure supplement 2. Linear mixed-effects single-trial regression outputs for P3a and CNV.

Figure 3 – Figure supplement 3. Linear mixed-effects single-trial regression outputs for the effect of pre-preparation cue activity, P3a, and CNV on each behavioural measure

Figure 3 – Figure supplement 4. Testing for effects of incentives, THP or THP*incentives across all electrodes and timepoints.

Regression coefficients from regressing each electrode and time-point against the different behavioural variables.

The time series show the regression coefficients for three chosen electrodes, with the solid bars at the bottom showing significant clusters for those electrodes (FWER=.05). Topographies are shown below the graph at the times written on the x-axis, with the colours showing the regression coefficient and yellow electrodes showing significant clusters. A) Residual velocity is predicted by voltage in the frontal electrodes briefly from 150:190ms after preparation cue, and then again from 280ms onwards, and gradually spreads backwards. B) RT is predicted by almost all electrodes from 250ms onwards, and is strongest centrally. C) Distractor pull is predicted by frontal electrodes from 120ms and this grows stronger over time, the central and posterior electrodes have a brief cluster around 180ms but only become consistently associated from 650ms. The following figure supplements are available:

Figure 4 – Figure supplement 1. Covariate-controlled regression coefficients from regressing each electrode and time-point against the different behavioural variables.

Figure 4 – Figure supplement 2. Regression analyses are relatively unchanged when controlling for eye-movements.

Mediation analyses of late CNV amplitude on residual velocity and RT.

Black lines show significant associations, with regression coefficients and significance indicated by the text, dashed red lines show significant mediations (permutation-testing), and dashed orange lines show significant mediated moderations (permutation testing). A) There was no mediation of the incentive effect on residual velocity by the mean CNV (1200:1500ms). B) However, CNV does partially mediate the effect of incentive on saccadic RT (dashed red line shows the indirect effect, which is the partial mediation). C) Mediated moderation analysis: CNV mediated the moderation of THP on the incentive effect.

Linear mixed-effects single-trial regression outputs for behavioural variables.

Each model also included a random effect of participant. RT was log-transformed for this analysis. Significant effects are shown in red.

Incentive effects are uncorrelated between saccade RT and velocity, but the THP*incentive interactions are correlated.

The correlation is negative as RT is inverse to response speed, meaning in both variables the drug is reducing the incentive-speeding effect. Each dot shows one participant’s mean, the line is the linear correlation, the shading shows the 95% confidence interval of the slope, and the text shows the Spearman’s correlation and p-value. A) Incentive effect is measured as the mean difference in RT & velocity between 50p and 0p conditions, averaged across drug and distractor factors. B) THP*Incentive effect is measured as the mean difference between incentive effects for the drug and placebo conditions, averaged across the distractor levels.

Distractor pull as a function of reaction time.

For all trials with a distractor present, after binning RT into 80 percentile windows for each subject and condition, and plotting mean distractor pull angle for each bin. Distractor pull was greatest for quickest saccades, and incentives sped responses (solid lines shifted leftwards), while THP increased distraction (orange lines shifted upwards) and slowed RT (small rightwards shift). This means that for a given speed, distraction was lessened by incentives and increased by THP.

Mean individual participants’ ERP component voltages for each time-window,

with the means for each individual person superimposed on top.

Linear mixed-effects single-trial regression outputs for P3a and CNV.

Significant effects are shown in red. Each model also included a random effect of participant.

Linear mixed-effects single-trial regression outputs for the effect of pre-preparation cue activity, P3a, and CNV on each behavioural measure

(while controlling for all factors, interactions, and a random effect of participant). Pre-preparation cue activity and P3a did not predict any behavioural measure, while CNV predicted RT and distractor pull. (Bonferroni-corrected threshold: α = .0056).

Testing for effects of incentives, THP or THP*incentives across all electrodes and timepoints.

We used difference waves, and difference-of-difference-waves for the interaction, with cluster-based permutation testing (2500 iterations, family-wise error rate = .05; DMGroppe Mass Univariate toolbox; Groppe, Urbach, & Kutas, 2011). a) T-statistics for the incentive difference wave (i.e., 50p – 0p, averaged across other factors) for three selected channels, with the solid bar at the bottom showing significant clusters (FWER = .05). The topographies below show the t-statistics for all channels at the times written on the x-axis, with the yellow dots representing electrodes in significant clusters. Higher incentives lead to more negative voltages centro-posteriorly from about 400ms after the preparation cue began, and this increases over the epoch. B) t-statistics for the drug difference wave (THP – placebo) shows no significant clusters at any channels or time-points, suggesting that THP did not change the voltage overall. C) Difference of difference waves showing the THP*incentive interaction ( (drug 50p – 0p) – (placebo 50p – 0p) ), also shows no cluster of significant difference.

Covariate-controlled regression coefficients from regressing each electrode and time-point against the different behavioural variables.

The association between the EEG signals and each behavioural measure remained after accounting for the other two measures, with a very similar pattern to the main results (Figure 4). The time series show the regression coefficients for three chosen electrodes, with the solid bars at the bottom showing significant clusters for those electrodes (FWER=.05). Topographies are shown below the graph at the times written on the x-axis, with the colours showing the regression coefficient and yellow electrodes showing significant clusters. a) Residual velocity is predicted by voltage in the frontal electrodes briefly from 150:190ms after preparation cue, and then again from 280ms onwards, and gradually spreads backwards. b) RT is predicted by almost all electrodes from 250ms onwards, and is strongest centrally. c) Distractor pull is predicted by frontal electrodes from 120ms and this grows stronger over time, the central and posterior electrodes have a brief cluster around 180ms but only become consistently associated from 650ms.

Regression analyses are relatively unchanged when controlling for eye-movements.

We included either the number of microsaccades (left), or the mean ocular drift speed (right) during the entire preparation period, as covariates in the regression model. (We did not do the permutation testing for this control analysis, due to time-constraints.) The beta-coefficients are almost unchanged from the analyses presented in Figure 4, either in the time-course of the selected channels, or in the topographies.