Full title: A pupil-linked arousal mechanism for deciding to engage in future physical effort

17 An organism’s behavioral success is determined by its ability to mobilize resources to overcome 18 challenges. This ability involves the noradrenergic system, indicated by the finding that pupil-size 19 increases proportionally with currently exerted effort. However, humans can deliberate in advance 20 whether to engage in effort in the future. It remains unclear how effort is represented in such an 21 anticipatory fashion during decision-making. We investigated this by measuring pupil responses while 22 participants decided whether to accept or reject rewards that required effort execution after the 23 experiment. We found a faster rate of pupillary dilation in decisions to accept high-effort rewards. This 24 was accompanied by stronger fMRI activity in anterior cingulate cortex (ACC) and anterior insula: When 25 accepting high-effort rewards, individuals with faster pupil dilation showed larger activity in these areas. 26 Our results identify a brain process instantiating anticipatory arousal when humans prepare for a 27 physical challenge, potentially reflecting simulated energization. 28


Introduction
Participants made decisions in the scanner about whether to accept or reject a reward offer (1 of 6 155 levels, from 0.50 to 10 CHF) that required exertion of physical effort (1 of 6 levels, from 40% to 90% 156 maximum voluntary contraction--MVC) (Fig. 2). To ensure that participants would not treat the task as any lack of such an effect would unlikely be due to the effort task being too trivial for the subjects. We 168 could thus investigate whether pupil-linked arousal scales with increasing physical effort during mere 169 mental simulation when deciding about future efforts.

171
Behavioral evidence for systematic effort-reward trade-offs 172 Initial analyses confirmed that participants indeed systematically traded off the proposed efforts and 174 rewards when taking choices, as could be expected based on previous work 9,10,36 . Offers were 175 accepted significantly more often when they were coupled with higher rewards (logistic regression of  192 193 Pupillary responses during decision making showed a stereotypical dilation shortly following cue onset, 194 peaking right after response onset, and constricting down to baseline level around cue offset (Fig. 3A). 195 To examine whether anticipated effort indeed engages the arousal system during choice, we compared future. Importantly, a comparable mirrored effect for low-effort trials (reject > accept low effort) was not 205 significant (p=0.90). This shows that the pupil-dilation effect for accepting high-effort trials cannot reflect 206 errors, infrequent occurrences, or surprise 37 , which would be similarly present for accepting high-effort 207 and rejecting low-effort options. 208 To investigate further the specificity of these links between choices to accept high effort-options 209 and ROD, we controlled for all other variables in our design within a logistic regression of choice ( fig. 3C). Thus, subjects with higher effort discounting (i.e., whose overall choice was more strongly affected by increasing effort) indeed showed faster pupil dilation when accepting compared to 238 rejecting high effort. These results fit well with the finding that the cost of effort is represented in a non-239 linearly increasing manner as the effort amount increases, captured by a parabolic discounting shape 240 10,39,40 . This non-linearity is also evident in our observation that the energization effect was only evident 241 for high-effort trials, comprising the most difficult effort levels (80-90% of maximum force). Importantly, 242 across subjects, we find that the energization responses in both pupil and the brain are positively 243 associated with a subject-specific parabolic effort discounting parameter, consistent with the idea that 244 this signal may be relevant for guiding overall choice. Despite the tight relationship between the energization signals evident in the pupil and effort 249 discounting, it is theoretically possible that the energization effect we observe in pupil signals may not 250 just relate to choice outcome but may also be higher for trials that are subjectively difficult, as larger 251 pupil size has been observed for trials that require greater cognitive control 30 . This effect might be 252 confounded with the energization effect, particularly because in some cases, high effort trials may be 253 associated with high rewards, hence making the decision to either select or forego effort more difficult.

254
To investigate this possibility, we directly quantified decision difficulty by calculating trial-wise absolute  Concurrent with behavioral evidence that participants systematically trade off reward with effort, we 276 examined the neural representations of reward, effort, and the interaction. In addition, we also 277 examined brain activity correlating with ROD. We replicated previous findings 7,25 of neural reward 278 representation in the ventral striatum and effort modulation in the frontal pole (FWE p<0.05; Fig S6). In 279 addition, using another GLM, we replicated previous finding that brain activity in the vmPFC is 280 correlated with the computed subjective value based on the amount of reward that is subtracted by the 281 amount of effort (FWE p<0.05; Fig S7). Taken together, our brain results fully replicate previous data 282 identifying cortical and subcortical brain regions that support effort-reward trade-offs.   higher choice-related brain activity in these regions ( Fig. 4C; Table 2). This positive relationship was also confirmed in an analogous ROI analysis, correlating the simulated energization pupil and neural 308 measures extracted from functional ROIs of ACC and bilateral anterior insula that were independently  Fig. 4D). 311 Finally, given that the energization effect in the pupil dilation data was correlated with the 312 individual effort-discounting parameter (Fig 3E), we also inspected whether the brain responses for the 313 decision to accept high effort would be associated with individual differences in effort discounting. To 314 test this, we again took each individual's parabolic effort discounting parameter and used it as a subject-315 specific covariate at the second level for our critical contrast (accept>reject in high effort bin). This  related arousal signals, neural responses in these areas during choices to accept high-effort trials were 320 strongest in people with higher effort discounting.

321
Taken together, our data show that brain activity in ACC and anterior insula shows anticipatory 322 effort signaling in a way that is consistent with simulated energization for high physical challenges.

323
These areas show higher activity during decisions to take on a difficult physical task in the future, and 324 this activation is tightly linked to anticipatory activation of the arousal system and to the weight that 325 participants place on effort when trading off rewards and efforts during choice.

Discussion
We examined how the brain may represent future efforts during choice, motivated by the wealth of data 345 on how it represents effort level during actual physical exertion. Specifically, we directly tested two 346 competing hypotheses against one another: Whether such neurobiological representation of future 347 effort signals simulated cost or energization. Consistent with the latter, our results show stronger activity 348 in the arousal system (as measured in pupil) and ACC-insular brain network for choices that involve 349 anticipating a sizeable amount of effort. This emphasizes that future effort during choice is represented 350 by arousal system in a way that appears to relate to future energization.

351
Our results emphasize that phasic pupil-linked arousal during the decision process is tightly we emphasize that our behavioral data and some aspects of our neural results clearly concur with 418 previous findings that an option is selected based on a trade-off between reward and effort (FigS7).

419
What has been unexplored in previous fMRI work, however, is how the noradrenergic arousal system 420 is sensitive to effort, and in what way this neurobiological representation of effort is functional for choice.

421
Using concurrent pupil-fMRI in an effort discounting task, we were able to scrutinize the precise thus excluded data from one subject whose rate of acceptance was 0.95. The final N was 49. However, 468 in certain analyses in which we had to split the data in accordance with our critical pupil contrast, we 469 had 7 subjects with certain data bins missing. Given the specific emphasis on the effects seen in pupil, 470 we were therefore only able to conduct the neuroimaging analysis with n=42. three 3-s squeezes. Continuous vocal encouragement was given during entire squeeze period (e.g.,

477
Guided by a vertical bar on-screen (Fig. 1A), participants were trained to do set squeezes from 478 force levels of 10%-90% MVC (shown to subjects as level 1-9), alternating between left and right hand.

479
One set consisted of 5 repetitions ('reps') that lasted 3 s interleaved by 3 s rest periods. Participants experienced all levels from 1-8 once, randomly assigned to either left and right, and level 9 twice, once 481 for each hand. The order of force levels was pseudo-randomised. Half of the subjects practiced on 482 levels 1, 3, 5, 7, 9 with left hand and 2, 4, 6, 8, 9 with right hand, and vice versa for the other half of 483 subjects.

484
Following a 5-minute break, they proceeded with a subjective rating task in which they had to 485 squeeze for each hand once at levels 1, 3, 5, and 9 for 5 s without knowing the difficulty levels and Participants made decisions between performing a specific effort level of the force task (between levels 497 4-9) to earn varying reward amounts (0.5, 1, 3, 5, 8, 10 CHF) and performing a counteroffer force task 498 at level 1 to earn either 30% or 40% of the reward of the first offer (Fig. 2C). The force task involves 499 performing one set of 10 'reps' at the selected effort level. Participants were fully aware that they would 500 make successive decisions in the scanner without executing the force task and they were not provided 501 with the dynamometer. 502 We used a factorial design, with six effort and six reward levels (36 cells), and two reward 503 counteroffers per cell (3 exemplars each), totalling in 216 trials. Trials were split in three fMRI runs of 504 72 trials (9 mins); trial order was pseudorandomised per subject per run.

505
During a fixation period of 3-6 s (created using the function gamrnd(0.8,1), mean 3.7s), the text 506 indicating reward and effort levels were masked with a series of letters "X" (Fig. 1B). Following this 507 period, the colour of the + sign at the centre changed and the effort and reward of each of the two 508 options were presented on either side of the fixation point for a fixed duration of 3 s. This prompted the 509 subjects that they were able to press either the left or the right key to indicate their choice. To provide 510 decision feedback, key response was promptly followed by a change in colour for the selected option. as the base hue and blue and purple were randomly assigned trial-by-trial to highlight the selected offer 531 (Fig. 1B).

532
Additionally, in a control experiment, we recorded luminance-driven pupil dilation without any 533 cognitive task. We presented fixation screens with a series of Xs as fixation period and Ys to replace 534 the text that would have indicated the effort and reward levels in the main experiment, each period 535 lasting for 3 s. Participants were instructed to keep their eyes open but were not required to press any 536 key. Just like in the main experiment, green was the base hue during fixation whereas blue and purple 537 were used to highlight the text on one side of the screen. All stimuli were in the same text format as in 538 the main task (Fig. 2B). Order of hue and side assignment were all counterbalanced and 539 pseudorandomised. We found no difference in mean pupil diameter during the presentation of these 540 control stimuli in different hues, confirming that the pupil response in the main task was not driven by 541 differences in text luminance (Fig. S1).  and correction for physiological noise, these are described in supplementary materials. 556 We performed random-effect, event-related statistical analyses. For each subject, we first (temporal derivative). We also added to these GLMs 18 physiological regressors and 6 motion 560 parameters. At the second level, we then tested the significance of subject-specific effects (as tested 561 by t-contrasts at the first level) across the population. For these analyses, we used a grey matter mask 562 as an explicit mask, created by averaging across subjects and smoothing (8mm) all participants' 563 normalized grey matter images (wc1*.nii) from the 'segment' procedure. 564 We built three first level GLMs. In GLM1, to highlight activity correlating with the interaction 565 between choice (accept vs reject) and effort levels (low, mid, high bins), we defined six first-level     analysis shows significant correlation between 'neural energization' (accept > reject high effort trials) 803 and subject covariates 'ROD energization' (C) and z-scored effort discounting parameter (E). D&F) 804 Similarly, ROI analysis shows significantly positive correlations between 'neural energization' contrasts 805 extracted from all three functional ROIs with 'ROD energization' (D) and effort discounting (F). Each 806 data point represents a subject. All scatterplots use the same color-coding scheme for subjects.