Need for cognition moderates the relief of avoiding cognitive effort

Davide Gheza; Wouter Kool; Gilles Pourtois

doi:10.1371/journal.pone.0287954

. 2023 Nov 16;18(11):e0287954. doi: 10.1371/journal.pone.0287954

Need for cognition moderates the relief of avoiding cognitive effort

Davide Gheza ^1,^2,^*, Wouter Kool ², Gilles Pourtois ¹

Editor: Poppy Watson³

PMCID: PMC10653461 PMID: 37972115

Abstract

When making decisions, humans aim to maximize rewards while minimizing costs. The exertion of mental or physical effort has been proposed to be one those costs, translating into avoidance of behaviors carrying effort demands. This motivational framework also predicts that people should experience positive affect when anticipating demand that is subsequently avoided (i.e., a “relief effect”), but evidence for this prediction is scarce. Here, we follow up on a previous study [1] that provided some initial evidence that people more positively evaluated outcomes if it meant they could avoid performing an additional demanding task. However, the results from this study did not provide conclusive evidence that this effect was driven by effort avoidance. Here, we report two experiments that are able to do this. Participants performed a gambling task, and if they did not receive reward they would have to perform an orthogonal effort task. Prior to the gamble, a cue indicated whether this effort task would be easy or hard. We probed hedonic responses to the reward-related feedback, as well as after the subsequent effort task feedback. Participants reported lower hedonic responses for no-reward outcomes when high vs. low effort was anticipated (and later exerted). They also reported higher hedonic responses for reward outcomes when high vs. low effort was anticipated (and avoided). Importantly, this relief effect was smaller in participants with high need for cognition. These results suggest that avoidance of high effort tasks is rewarding, but that the size off this effect depends on the individual disposition to engage with and expend cognitive effort. They also raise the important question of whether this disposition alters the cost of effort per se, or rather offset this cost during cost-benefit analyses.

Introduction

Daily life constantly presents us with challenging tasks that we need to perform to achieve our goals. Successful completion of these tasks requires us to invest cognitive effort. Recent theories [2, 3] suggest that such decisions are implemented as simple cost-benefit analyses, where reward is discounted by the anticipated cost of effort. The idea that effort is costly is captured by the ‘law of least effort’ [4]. This law predicts that, all else being equal, people will tend to avoid effortful actions. There is ample empirical evidence for this principle. For example, Kool and colleagues [5] demonstrated that human participants prefer choice options with the least cognitive demands (even when time on task and error likelihood were matched). Such effort avoidance has been demonstrated for a wide range of demands, such as response conflict [6, 7], task switching [5, 8], complex task policies [9], and short-term memory load [5]. In fact, a recent study demonstrated that people are willing to endure physical pain to avoid effort [10]. Across these studies, the task demands invoke the need for cognitive control or controlled information processing [11, 12], which are core features in models of mental effort [13]. In short, mental effort carries a cost, and this cost is used to devaluate anticipated reward [14–16], resulting in an expected value of control’ [2].

If reward is discounted by effort, then it should also be enhanced if effort is anticipated, but eventually avoided. That is, the avoidance of expected effort should be experienced as relief. Interestingly, relief (from anticipated pain) appears to involve the same reward processing systems that underly representations of effort costs [17, 18]. Together with the observation that effort is costly, this predicts that the avoidance of anticipated effort should be experienced as rewarding. To our best knowledge, there is a scarcity of studies that investigate this phenomenon. Here, we report two studies that test this hypothesis.

Earlier work from our group [1] has provided some initial, but ultimately inconclusive, support for this hypothesis. In this experiment, participants performed a well-validated gambling paradigm [19, 20] that was combined with an effort task [21]. On each trial of this task, participants chose between four options that each afforded a chance to win a reward. Critically, on some of the trials, participants were informed that after a loss (a ‘no reward’ outcome), they would be given the choice to exert effort to repeat the gamble. Therefore, on these ‘special’ trials, there was a prospect of effort both prior to and during gambling, while for the rest of trials, there was not. Moreover, rewards on special trials were associated with the avoidance of effort (i.e., it rendered the opportunity to spend effort to repeat the gamble moot). Interestingly, the participants reported more positive affect (pleasantness), as well as increased relief for reward delivered on special trials compared to standard trials. In addition, at the electrophysiological level, we found that reward delivered after gambling on special trials elicited a larger reward positivity event-related potential [22] as well as enhanced power in delta [23] and beta-gamma [24–26] frequency bands. Together, these results suggest that relief from effort registers as rewarding [4, 27, 28]. However, an alternative interpretation exists.

To see this, note that people did not just avoid mental effort on rewarded special trials, but they also avoided the extra time spent on deciding whether to repeat the gamble and on the effort task itself. This creates an opportunity cost [29]: special trials with no reward lasted longer, and thus were associated with reduced time to accrue reward. In other words, participants may have experienced more positive affect for rewards on special trials simply because it reduced their time on the experiment. Thus, though these results are suggestive, it remains to be shown whether relief from effort enhances subjective reward.

To test this, we made several key changes to our experimental design [1], so that only demands for mental effort, and not opportunity costs, differed between conditions. In this new task, each trial with no reward was followed by a cognitive effort task [30] performed to recuperate the missed reward. This cognitive effort task required mental arithmetic. At the start of each trial, a cue indicated the amount of demanded effort for the arithmetic task (i.e., easy vs. hard). Comparing self-report ratings between these trials allowed us to isolate the effect of effort avoidance, whereas in our previous study the analogous comparison conflated effort and time. It is worth drawing attention to the fact that each trial with no reward lead to the effort task. Participants did not decide whether or not to perform the task. Therefore, the expectation of effort was high and constant across all trials. These methodological changes allowed us to compare the subjective experience of reward when either an easy or difficult task was expected, with opportunity cost equal between conditions. In line with our previous study, we assessed the latent construct of the “hedonic response” (i.e., the subjective experience of the gambling trial outcome) using three probes, each addressing a different affective dimension (“pleasure”, “frustration”, and “relief”). We refer to the interaction of trial outcome with effort anticipation on the hedonic response as a “relief effect”.”

The second goal of our study was to assess the source of individual differences in effort-related relief. In addition to objective task demands, valuations of effort are influenced by people’s state and trait dispositions. For example, stress increases effort avoidance [31] and depression/anhedonia decreases reward processing [32]. Here, we tested whether relief was predicted by participants’ ‘need for cognition’ (NFC; [33]). The NFC is defined as the tendency to engage in and enjoy effortful cognitive activities (e.g., those requiring thinking and problem solving; see also [34]) and can be conceived as a motivational drive [35]. We hypothesized that an increased NFC would decrease the relief of effort avoidance, and hence behave in an opposite manner compared to stress, depression or negative affect [36]. In other words, the NFC might mitigate the impact of effort (anticipation) on reward processing, with a relatively reduced effort-related relief for participants with a high NFC.

Methods

Participants

Twenty-three undergrad students from Ghent University (17 females; median age: 21 years, range: 18–30) participated in Experiment 1. They had normal or corrected-to-normal vision and did not report any history of neurological or psychiatric disorders. Sample size was determined to be at least as large as in our previous experiment [1] where a similar experimental manipulation was used, and where a significant effect of cost anticipation on reward was found (increased pleasantness [p = .008, d = 0.57] and relief [p = .033, d = 0.27]).

Seventy-nine young adults participated in Experiment 2. None of them participated in Experiment 1. To explore individual differences in NFC, the sample size was determined to be as large as available time and resources would allow. Three participants were excluded from further analyses due to low precision in their ratings of the reward-related feedback (see exclusion criteria below). Hence, the final sample consisted of 76 participants (62 females; median age: 21 years, range: 18–46).

These two experiments were part of a more general research project investigating effects of motivation on reward that was approved by the local ethics committee at Ghent University. All participants gave written informed consent prior to the start of the experiment, were debriefed at the end, and received a monetary compensation for their participation. Participants’ data was collected between 2018 and 2019 and was anonymized at the time of collection.

Stimuli and task

For Experiment 1, we adapted a widely used gambling task [1, 37] and combined it with a cognitive effort task [30]. At the start of each trial, participants were informed about the cognitive effort level with a text cue (“easy” vs “hard”) located at the center of the screen (1000 ms). Following a fixation dot (1500 ms), four doors appeared on the screen, and participants had to choose one of them by pressing with their left hand the corresponding numeric key (1 to 4) on a keyboard. After another fixation dot (700 ms), this choice was followed by reward-related feedback (1000 ms), indicating either a reward (green “+”) of 6 cents, or a no-reward outcome (red “o”). Participants were instructed to guess and select a door containing a reward in order to maximize their payoff. However, unbeknown to them, the outcome was unrelated to the choice and reward probability was set to exactly 50%. Participants were instructed that in case of a reward, the trial would end and a new trial would follow. Hence, no additional effort would be necessary. However, when there was no reward, a second task would follow, which could be hard or easy (as indicated by the previous cue), so that they could get another chance to win reward. Thus, receiving no reward during the gambling task resulted in the prospect of effort. More specifically, after 1000 ms (fixation), a mental arithmetic task started (i.e., the ‘effort task’). This task required participants to complete two calculations (two additions or an addition and a subtraction, all of which with single-digit numbers) (see Fig 1). In the hard condition, every operation required carrying or borrowing. In the easy condition, none of the two operations required carrying or borrowing. This manipulation results in two levels of difficulty, as shown in previous studies [38–40] and confirmed by subjective ratings (see Results below).

Fig 1 — Participants were first informed about the cognitive effort level with a text cue (word “easy” or “hard”). After they picked one door, they received reward or no-reward feedback (50% reward probability). Only in case of no-reward feedback, the effort task ensued. For both conditions, the initiation of the effort task was probed by an uninformative pound symbol (#), followed by a serially-presented list of digits and operations. Participants were instructed to integrate them, and based on this arithmetic select the correct one out of two possible answers. Based on their answer, “performance” feedback was given, providing reward in case of a correct answer. On a small proportion of trials, the subjective hedonic value of the reward-related feedback was assessed with probes presented 1000 ms after its offset.

The effort task was structured as follows: a pound symbol indicated the start (400ms); digits and arithmetic signs were then presented serially, each lasting 500ms and interleaved with blank screens (200ms in Experiment 1, 100ms in Experiment 2); finally, two possible solutions were presented simultaneously, and the participants had to choose the correct one by pressing the corresponding key with the right hand (i.e. numeric keypad; “1” for the leftmost or “2” for the rightmost solution). They were instructed to select the correct answer as quickly as possible, with a time limit of 4000ms. After this choice, a blank slide was presented (1000 ms), followed by a new feedback screen related to their performance (1000 ms): a reward outcome (green “+”) indicated a correct response and a win of 6 cents, and a no-reward outcome for incorrect responses (red “o”) was presented. If participants missed their response, or if they responded too late, a screen indicated that there was “no response detected”. Serial presentation of digits and operations was chosen in order to equate the time spent observing stimuli between conditions, as well as to avoid calculation strategies [30]. Across trials, different combinations of digits and arithmetic signs were used to avoid learning or habituation. After the effort task, a new trial of the gambling task followed. The intertrial interval was fixed and set to 1000 ms.

The subjective value of the first feedback screen (after gambling) was assessed by specific probes. In a few trials (n = 48), 1000 ms after the offset of the gambling outcome, three questions were presented, probing the perceived pleasantness, frustration, and relief of the outcome. Participants answered them using visual analog scales (VAS). These three ratings were submitted 12 times for each effort level (low vs. high) and outcome (reward vs. no reward) condition. By using three questions, which assessed a range of different affective dimensions, we aimed at measuring the latent construct of a ‘hedonic response’ along a negative-positive valence axis [41]. Multiple self-report items allowed us to validate the consistency of participants’ responses, and to increase statistical power by controlling for nuisance variability between affective dimensions.

Experiment 1 consisted of 208 trials, including an equal amount of easy and hard trials. The gambling task had a pre-set reward probability of 50%, and the effort task had to be completed only in case the choice did not deliver reward. Therefore, the effort task was administered 52 times for each difficulty level.

Seven self-paced breaks were distributed equally throughout the experiment. At the end of each break, participants were asked to rate the difficulty of the effort task, its pleasantness, their motivation to complete it, and their satisfaction with correct performance on the task. Each question was submitted twice, for each of the two difficulty levels, using a VAS. Participants received a fixed €8 compensation for their participation. Depending on their accuracy with the effort task, a maximum payoff of €12.48 could be earned (mean = €11.91).

For Experiment 2, where the NFC questionnaire was also administered [42], the same procedure was used but a few changes were made. The reward feedback with the gambling task indicated that 5 cents were won (instead of 6). The three ratings about the gambling outcome were submitted each 40 times in total (instead of 48), 1000ms after its offset (10 times for each combination of difficulty level and outcome). In Experiment 2 we also probed the subjective value of the performance feedback (after the effort task). Analogous to the reward-related feedback, three ratings probing the perceived pleasantness, frustration, and relief of this feedback. These were presented directly after the performance feedback. Performance feedback ratings were submitted up to 26 times, equally split between high and low effort, but were omitted in case of an incorrect response. At this stage we only probed affective ratings for correct feedback as the effort task was tuned to induce only very few errors. Reward-related feedback (gambling task) and performance feedback (effort task) were never probed in the same trial. Experiment 2 consisted of 200 trials, with an equal amount of easy and hard trials (100 each). For each of these, there were 50 trials where the effort task had to be performed (no reward) and 50 trials where it was avoided (reward). Participants were compensated with €8 for their participation. Depending on their accuracy with the effort task, they could earn up to €10 (mean = €9.63). Last, in Experiment 2 we added 16 “catch trials” to promote and assess participant’s attention to the effort cues presented at the beginning of the trial. More precisely, after door selection participants were asked to report how hard a following effort task would be. After their response, participants received catch-related feedback (correct, incorrect, or too late response) and the catch trial terminated. Eight catch trials, for each of the two effort levels, were randomly interspersed among the 200 task trials.

The experiments’ duration was approximately 60 minutes, including instructions and a short practice. In Experiment 2, the NFC questionnaire was administered after the task. All stimuli were shown against a grey homogenous background on a 21 inch CRT screen and controlled using E-Prime [43].

Data analysis

We evaluated the effectiveness of the cognitive effort manipulation by comparing performance on the effort task (i.e., accuracy and speed) between the easy and hard conditions. Moreover, we also compared their subjective value by analyzing the ratings of difficulty, pleasantness, motivation to perform well, and pleasure in performing well. These ratings were first transformed to percentages, setting anchors to the boundaries of the scales, and were averaged across the seven repetitions.

The subjective ratings of the reward-related feedback obtained for each difficulty level (easy vs. hard), outcome (reward vs. no reward), and affective dimension (pleasantness, frustration, and relief) were also first transformed to percentages, setting anchors to the boundaries of these scales. VAS scores were calculated by first identifying the relative x-axis position of the mouse with respect to the leftmost position of the scale at the time of participant’s click. We then converted this measure from pixels distance to percentage over the scale range (396 pixels). For the “frustration” scale, we reverse-scored the percentages in order to provide comparable ratings for the three affective dimensions. Subjective ratings for the three affective dimensions were used to assess the hedonic response as a whole, along a negative-positive valence axis. In all analyses, we predicted single trial subjective ratings with a mixed modelling approach, in which participant and affective dimension were treated as random effects (i.e., considered nuisance factors).

For Experiment 2, subjective ratings of the performance feedback were analyzed in the same way. Additionally, for statistical analyses the VAS scores and NFC data were centered by subtracting the group means from the individual scores.

Participant’s data were excluded from further analysis if the average mouse-click x-coordinate of the reward-related feedback rating was above 105% or below -5% of the rating range, in any of the 3 affective dimensions, and any of the 4 levels of outcome by difficulty.

Statistical analyses

Subjective ratings data were analyzed using Bayesian model comparison. Inference about their generative processes was based upon Bayes Factors (BFs), computed for alternative explanatory models in ANOVA designs ([44], see also [45]). The analyses’ pipeline was implemented in R v4.0.5 [46] with the package BayesFactor v0.9.12–4.2 [47], and involved: I) defining theoretically sound probability models; II) computing BFs, i.e. the ratio between the likelihood of each model of interest (the probability of the observed data, given the model/hypothesis) and the likelihood of the null model; III) model selection based on the highest BF; IV) characterizing the direction of follow-up contrasts by means of Bayesian t-test between conditions of interest. The models’ likelihood was estimated using Markov-Chain Monte Carlo simulations with 10,000 iterations, and BFs were computed assuming a wide Cauchy prior centered on zero: d ~ Cauchy (0, 0.707).

For the gambling task feedback ratings, used to estimate the underlying hedonic response, the models of interest included the effects of 1) outcome, 2) difficulty level, 3) outcome + difficulty level, and 4) outcome x difficulty level. Additionally, for Experiment 2 we tested a model including 5) the three-way interaction of outcome x difficulty level x NFC score. Moreover, the factors Subject, Affective dimension, and their interaction were treated as nuisance variables (i.e., varying effects to control for). To do this, they were entered in all models, including the null model. Additionally, the null model included a single (i.e., constant) intercept. This procedure allowed us to control for the nuisance effects of 1) individual differences in VAS biases (e.g., preference for a given section of the scale), 2) overall differences in rating across affective dimensions (e.g., overall higher responses for a specific scale), and 3) individual variations in the latter differences.

In Experiment 2, to further explore the effect of NFC on reward processing, we also ran a full Bayesian estimation of parameter values for model 5, using the “brms” R package [48]. To analyze performance feedback ratings (Experiment 2), we used a one-tailed Bayesian t-test to estimate the evidence in favor of increased positive evaluation of a reward outcome after high vs. low effort, as compared to a null model.

Results

Experiment 1

Accuracy on the effort task was higher for the easy (M = 98%, SD = 14) compared to the hard condition (M = 87%, SD = 34; BF+0 = 2.20 x 10³). Mean reaction time was larger for the hard (M = 1091 ms, SD = 476) compared to the easy condition (M = 621 ms, SD = 144; BF+0 = 4.11 x 10³). These results indicated that the difficulty manipulation was successful.

The subjective ratings of the effort task indicated these two difficulty levels were experienced as clearly different. The hard compared to easy condition was perceived as more difficult (M easy = 6.1, SD = 7.3; M hard = 31.2, SD = 18.7; BF-0 = 1.09 x 10⁵) and less pleasant (M easy = 83.4, SD = 12.5; M hard = 62.7, SD = 21.0; BF+0 = 1.92 x 10³), while participants reported similar levels of motivation to perform them correctly (M easy = 85.1, SD = 14.8; M hard = 86.6, SD = 13.6; BF01 = 3.58), as well as pleasure in performing them correctly (M easy = 81.8, SD = 17.1; M hard = 76.4, SD = 18.5; BF01 = 2.00).

The hedonic responses to the gambling outcome (i.e., reward-related feedback ratings; Fig 2) were best explained by a model that contained the outcome x difficulty level interaction. Under this model the observed data were BF10 = 5.28 x 10⁶⁴⁸ times more likely to be produced than under the null model. Moreover, the outcome x difficulty model explained the observed data 284 times better than the second-best model, which only included the main effect of outcome (Table 1). Follow-up Bayesian one-tailed t-tests showed strong evidence for the hypothesis that no-reward outcomes were evaluated as more positive in low-effort compared to high-effort trials (BF+0 = 6.00 x 10⁵). Conversely, for reward outcomes, there was only weak support (BF+0 = 2.22) for more positive evaluations in the high- compared to low-effort trials. In other words, participants rated a no reward outcome as more positive when they anticipated low effort.

Table 1. Bayesian model comparison (Experiment 1).

Model	BF10	% pe
[1] Subject * AffDim + efflev	0.05	±3.39%
[2] Subject * AffDim + outcome + difflev	2.33 x 10^645	±2.58%
[3] Subject * AffDim + outcome	1.85 x 10^646	±2.22%
[4] Subject * AffDim + outcome * difflev	5.28 x 10^648	±4.32%
Against denominator:
percent ~ 1 + Subject * AffDim

Open in a new tab

Note. Bayes factors (BF₁₀) and percentage of proportional errors (% pe) for each model relative to the null. The models are sorted from top to bottom with ascending Bayes Factor.

As can be seen from the S1-S6 Figs in S1 File, participants showed a robust internal consistency in their rating of the reward-related feedback across outcome levels: Their rating was consistently lower for no-reward compared to reward outcome. Moreover, we observed very robust correlations between pairs of affective dimensions (group-level statistics of these correlations are reported in the S1 File).

Experiment 2

Participants paid attention to the effort cues presented at the beginning of each trial, as suggested by high accuracy in reporting them during catch trials (M easy = 96.2, SD = 8.9; M hard = 91.8, SD = 13.8).

Similarly to Experiment 1, the accuracy for the effort task was higher for the easy (M = 99%, SD = 12) compared to the hard condition (M = 86%, SD = 34; BF+0 = 3.23 x 10¹⁶). Mean reaction time was larger for the hard (M = 1267 ms, SD = 467) compared to the easy condition (M = 694 ms, SD = 198; BF+0 = 7.21 x 10¹⁹). This demonstrated that the difficulty manipulation was successful.

The subjective ratings of the effort task again indicated these two difficulty levels were experienced as clearly different. The hard compared to easy condition was perceived as more difficult (M easy = 14.0, SD = 15.7; M hard = 40.2, SD = 23.1; BF-0 = 4.39 x 10¹⁴) and less pleasant (M easy = 76.7, SD = 19.9; M hard = 57.0, SD = 21.7; BF+0 = 6.46 x 10⁷). In line with Experiment 1, participants reported similar levels of motivation to perform the easy and hard tasks correctly (M easy = 82.0, SD = 16.1; M hard = 84.5, SD = 15.7; BF01 = 2.01), and also similar pleasure in performing them correctly (M easy = 76.3, SD = 19.3; M hard = 79.1, SD = 17.8; BF01 = 3.39).

Next, we turned our attention to the analyses of hedonic responses as a function of both gambling outcome and difficulty (Fig 3). Those ratings were again best explained by a model that included the outcome x difficulty level interaction. Under this model, the observed data were BF10 = 7.52 x 10¹²⁰⁹ times more likely to be produced than under the null model. This model also explained the observed data 2.11 x 10²⁰ times better than the second-best model, which only included the main effects of outcome and difficulty level. The follow-up Bayesian one-tailed t-tests showed strong evidence for the hypothesis that no-reward outcomes were evaluated as more positive on low-effort compared to high-effort trials (BF+0 = 1.55 x 10¹⁴). Conversely, for reward outcomes, there was strong evidence that evaluations were more positive in the high compared to low effort trials (BF+0 = 2.08 x 10⁵). In sum, we replicated Experiment 1: participants rated the no reward outcomes as more positive when they anticipated low effort. In extension of Experiment 1 (Fig 2), participants also rated the reward outcomes as more positive when they anticipated high effort (Fig 3). This result suggests that participants experienced relief when high effort was anticipated but avoided.

To test the hypothesis that a predisposition towards cognitive effort (as measured with the NFC questionnaire) would reduce this relief effect, we included participants’ NFC score as a continuous predictor in the best-fitting model described above (the two-way model including the interaction between outcome and difficulty level). Under this new three-way interaction model (Table 2), observed data were more likely compared to the former best model (BF = 2.05 x 10¹²) and the null model (BF = 1.54 x 10¹²²²). In other words, there was strong evidence that the disposition towards cognitive effort moderated the interaction effect between outcome and difficulty level.

Table 2. Bayesian model comparison (Experiment 2).

Model	BF10	% pe
[1] Subject * AffDim + efflev	0.20	±3.92%
[2] Subject * AffDim + outcome	2.22 x 10^1189	±3.89%
[3] Subject * AffDim + outcome + difflev	3.55 x 10^1189	±4.37%
[4] Subject * AffDim + outcome * difflev + NFC	1.07 x 10^1209	±4.36%
[5] Subject * AffDim + outcome * difflev	7.52 x 10^1209	±4.53%
[6] Subject * AffDim + outcome * difflev * NFC	1.54 x 10^1222	±9.71%
Against denominator:
percent ~ 1 + Subject * AffDim

Open in a new tab

Note. Bayes factors (BF₁₀) and percentage of proportional errors (% pe) for each model relative to the null. The models are sorted from top to bottom with ascending Bayes Factor.

Furthermore, when selectively predicting the feedback ratings for the ‘relief’ affective dimension, data were still best explained by a model that included the three-way interaction of outcome x difficulty level x NFC, followed by a model including the two-way outcome x difficulty level interaction (see S1 File).

To further explore the effect of NFC on reward processing, we conducted a full Bayesian estimation of parameter values. We fit model 5) with the “brms” R package [48], and we inspected the posterior probability distributions for every interaction parameter that included NFC. In Table 3, we report all the estimated model parameters. Confirming the previous model comparison, we observed a negative three-way interaction outcome x difficulty level x NFC (estimate = -0.15; 95% credible interval = [-0.27 0.04]; posterior probability = 0.98), indicating that the outcome x difficulty level interaction was moderated by NFC. As can be seen in Fig 4, the differences in subjective ratings between difficulty conditions were attenuated, in both reward conditions, for participants with higher NFC scores. This effect can also be observed in the raw data in Fig 5, where the outcome x difficulty level crossover interaction is shown for two subsamples of participants scoring at the extremes of the NFC scale (participants within the 0–30 percentiles were included in the Low NFC group; participants within the 70–100 percentiles were included in the High NFC group). Both the more positive ratings of reward feedback in the difficult condition, as well as the more positive ratings of no reward feedback in the easy condition were attenuated for people in high in NFC.

Table 3. Posterior parameter estimation summary (Experiment 2).

Parameter	Median	95% CI	pd	Rhat	ESS
(Intercept)	-15.56	[-18.78–12.33]	100%	1.002	1137
outcomereward	34.36	[33.11 35.60]	100%	1	11017
difflevhigh	-5.56	[-6.80–4.31]	100%	1	11471
NFC	0.07	[-0.11 0.27]	77.24%	1	2447
outcomereward : difflevhigh	8.57	[6.80 10.34]	100%	1	10051
outcomereward : NFC	-0.24	[-0.34–0.14]	100%	1	10621
difflevhigh : NFC	0.09	[-0.01 0.19]	95.76%	1	10693
outcomereward : difflevhigh : NFC	-0.15	[-0.29–0.01]	98.47%	1	9637

Open in a new tab

Note. Model 5 parameter estimation using the R package “brms”. Summarized posterior distributions for constant effects, presenting median values, 95% Credible Interval, probability of direction (pd; proportion of posterior samples with the same sign as the median). Rhat and Effective sample size (ESS) are diagnostic metrics (at convergence, Rhat = 1). Factors are dummy coded.

Fig 4 — Interaction effects of predictors outcome, difficulty level, and NFC on the estimated feedback rating. The regression lines represent the mean of posterior probability samples, for each condition and NFC score. The shading represents the 95% credible interval around them.

Fig 5 — Crossover interaction between outcome and difficulty level for two subsamples of participants with extreme NFC scores. Low NFC = participants within 0–30 percentiles. High NFC = participants within 70–100 percentiles. The three affective dimensions (pleasure, reverse-scored frustration, relief) and the ten repetitions were aggregated. Error bars represent the within-subject confidence intervals. The inter-subject variance has been removed by subtracting the subject mean from each rating, and adding the group mean [49].

Additionally, we observed a negative two-way interaction outcome x NFC (estimate = -0.24; 95% credible interval = [-0.32–0.16]; posterior probability = 1), suggesting that the main effect of outcome was also moderated by NFC. As can be seen in Fig 4, participants low on NFC reported more extreme ratings for both reward and no-reward feedback.

To investigate this pattern of results further, we also verified whether NFC predicted cognitive performance in the arithmetic task. We analyzed reaction times and accuracy of the arithmetic task’s responses and compared models including the effects of 1) difficulty level, 2) NFC, 3) difficulty level + NFC, 4) difficulty level x NFC. The null model was a simple intercept model, and all models included the random factors participants as nuisance. Reaction times (for correct responses) were best explained by the difficulty level X NFC interaction model, under which the observed data were BF10 = 1.62 x 10³³⁶ times more likely to be produced than under the null model. This model also explained the observed data BF10 = 8.22 times better than the second-best model, which only included the main effect of difficulty level. Similarly, accuracy was best explained by the difficulty level X NFC interaction model, under which the observed data were BF10 = 3.47 x 10⁷⁴ times more likely to be produced than under the null model. This model explained the observed data BF10 = 8.59 times better than the second-best model, which only included the main effect of difficulty level. In other words, for both reaction times and accuracy, there was moderate evidence in support of the hypothesis that NFC interacted with the difficulty level in predicting performance. As can be seen in Fig 6, while the two subsamples of participants scoring at the extremes of the NFC scale showed similar performance for the easy condition, participants scoring high on NFC were more accurate and faster in the hard condition.

Fig 6 — Reaction times (top) and accuracy (bottom) are reported for two subgroups of participants scoring below the 30^th percentile (low NFC, N = 20, NFC < 33) or above the 70^th percentile (high NFC, N = 23, NFC > 46.5) of the group distribution of NFC scores, separately for the easy and the hard arithmetic conditions. Error bars represent the within-subject confidence intervals. The inter-subject variance has been removed by subtracting the subject mean from each rating, and adding the group mean [49].

Finally, we analyzed the subjective ratings of the performance feedback (Fig 7). Participants rated positive feedback following the high effort task as more positive compared to following the low effort task. The Bayesian one-tailed t-tests showed strong evidence for this hypothesis (M hard = 74.8, SD = 23.2; M easy = 71.4, SD = 24.3; BF+0 = 6.99 x 10⁴).

Discussion

Exerting cognitive effort carries a cost [50]. Over the last decade there has been an abundance of empirical evidence that supports this notion, showing that effort exertion is avoided [5, 7, 8] or traded for reward. Here, by extension, we hypothesized that the avoidance of anticipated cognitive effort carries positive value, because it provides relief from an anticipated cost, which in turn bolsters reward [1, 51–57]. We tested this hypothesis by adapting a paradigm that we previously used to show that rewards were perceived as more positive when they signaled avoidance of a demanding task [1]. However, in this previous experiment, avoidance of the demanding task meant that no task would be performed at all. Therefore, this result could also be driven by opportunity costs [29] rather than effort per se. Here, we adapted this experimental paradigm to overcome this limitation. In our new task, we measured subjective ratings to estimate hedonic responses to outcomes that signaled whether an upcoming arithmetic task [30] could be avoided. On each trial, this arithmetic task carried either low or high effort demands, which allowed us to specifically measure the effect of effort context on ratings. In addition, we ran a replication of this study which also investigated whether NFC [33] would act as a moderator on any relief effects. A number of important new results emerged from this study.

First, in both Experiments 1 and 2, the performance measures and subjective ratings suggested that people found the effortful version of the arithmetic task to be more demanding (replicating Vassena et al. [30]). Participants were slower and made more errors on the hard task, and rated it as more difficult and less pleasant than the easy one. However, at the same time, the motivation to perform both task’s versions was high and similar, and participants also reported the same amount of pleasure when successfully completing it. These results suggest that the hard task yielded more negative ratings because of the costs associated with increased effort exertion [36], yet this negative evaluation did not impinge on the participants’ motivation. We believe that our data shows that participants exerted more cognitive effort in the hard condition. In response to a substantial difference in difficulty, they only showed a relatively minor drop in accuracy. Most importantly, we saw no clear shift in reported motivation. Nevertheless, it is important to note that our findings provide only preliminary support for this conclusion, as both task demands and the precision of the information-processing operations (i.e., performance) differed, to some extent, between difficulty conditions [13].

Second and central to our main hypothesis, we found that rewards were perceived as more positive when they meant avoidance of the harder, more effort-demanding, task. This results extends our prior findings [1], but now more clearly indicating that the avoidance of cognitive effort per se is perceived as rewarding, translating to a form of relief. Indeed, previous neuroscientific research suggests that the relief of pain involves reward processing pathways [17, 18]. Our finding also suggests that reward is evaluated in relation to prospective effort. Importantly, because we ensured that opportunity costs [29] were equal between effort conditions, we can be sure that this relief effect is truly driven by effort anticipation.

In both experiments, we also found that effort anticipation influenced ratings of unrewarded outcomes, which signaled that effort could not be avoided. Specifically, people viewed no reward outcomes as more positive if they led to a low- compared to a high-effort arithmetic trial. This result suggests that the prospect of exerting low amounts of effort is perceived as more positive compared to exerting high amounts of effort. This finding is compatible with earlier empirical studies and theoretical models that have associated effort with avoidance and/or reward devaluation/discount (e.g., [2, 5, 58, 59]).

Interestingly, in Experiment 2 we found that this relief effect was modulated by participants’ NFC [33]. As expected, people with high NFC showed a smaller relief effect. This finding is important because it shows that the cost-benefit analyses carried out during effort allocation do not just consider structural factors (such as reward probability or objective task difficulty), but also specific dispositions or motivational states [58]. Therefore, our results inform theoretical models of metacontrol, and of the role of motivation in decision making [60]. They clearly signal the need for such models to consider intrinsic disposition towards effortful tasks, and show how they may influence the processing of reward (see [61]).

It is tempting to think that the attenuated relief effect for individuals high in NFC suggests that such individuals are not averse to cognitive effort, or perhaps even value its exertion [58]. This would pose a challenge to the, presumably universal, law of least effort [5]. In fact, recent research suggests that explicitly rewarding the selection of high-effort actions biases people to choose more effortful lines of action [62, 63]. However, an increased NFC may simply reflect a heightened subjective value of succeeding at demanding tasks, or of the rewards obtained after demanding tasks. Alternatively, people with high NFC may believe that completing hard tasks demonstrates self-efficacy or competence. Under this view, these factors would offset the intrinsic cost of effort during decision making [64]. In other words, a NFC may reflect heightened stakes for cognitive success, rather than lower cognitive costs. Future research may specifically address these open questions by probing in a more granular fashion which of these factors affect effort-based decision making. Finally, we hasten to mention that people high on NFC showed an attenuated relief effect, but not a reversed one. This finding is not inconsistent with the idea that these people simply carry a smaller, but still positive, cost of mental effort.

Interestingly, the results from Experiment 2 also revealed that participants high on NFC solved the hard arithmetic task faster and better than those low on this dimension. This result appears to be at odds with previous studies available in the literature. A range of studies have not found any systematic relationship between the NFC and behavioral performance (using a variety of tasks that demand cognitive control functions such as conflict processing or response inhibition; [65]). Accordingly, some caution is needed when interpreting the modulatory influence of the NFC on the relief effect. While it is possible that this effect is driven by an increased intrinsic motivation to engage in effortful tasks (as explained above), it is also possible that it was caused by the level of required effort being “objectively” lower for participants high on NFC. In other words, the increased task performance on the hard arithmetic task by this group of participants may reflect that it was not difficult enough for them, leading to an attenuated relief effect. Another possibility is that participants scoring low in NFC, and performing sub-par in the arithmetic task, may have developed lower expected values for high difficulty trials, corresponding to lower estimates of reward probability. Therefore, the relief for avoiding high effort may be compounded by the relief for avoiding risky arithmetic trials. To tentatively rule out these hypotheses, in the Supporting Information we report two control analyses which confirm the relief effect, and the role of NFC in moderating it, even when controlling for individual differences in arithmetic performance.

However, because individual differences in performance may index relative differences in expected effort between conditions, our experimental design is not suited to ultimately adjudicate between these competing, but not mutually exclusive, accounts. Future research may address this issue with different experimental procedures, for example by omitting performance feedback. That being said, we would like to note that the questionnaire was administered after task completion, so it is not possible that it induced biases or expectations about effort exertion that influenced behavior on the task. Rather, it is possible that experiencing the task may have biased participants’ responses in the NFC questionnaire. Perhaps, observing their performance on the task influenced the self-evaluations prompted by the NFC questionnaire, leading their scores to reflect a combination of stable trait features with transient affective reactions. More generally, the correlational nature of this finding prevents us from drawing any causal inferences between the relief effect and NFC. Hence, additional studies are needed to determine the mechanistic nature of the modulation of the relief effect by NFC. For example, one group of participants may be trained to value effort using one of the manipulations introduced by Clay et al. [62] and Lin et al. [63]. If our results were truly driven by altered representations of effort costs, then this group should also show a similar attenuation of the relief effect.

Finally, in Experiment 2 we also found that participants valued positive feedback after the arithmetic task more highly when they had successfully performed the hard compared to easy task. This result suggests that effort expenditure (as opposed to anticipation only) can also influence reward processing (see also [66, 67]), retrospectively increasing the value of rewards when more effort is exerted. This finding bears resemblance to the IKEA effect [58, 68], according to which people value a product more if it is produced by their own efforts compared to the same product acquired by other means. In addition, it calls to mind the idea that self-efficacy [69], typically achieved through effort, carries value, and contributes to a positive sense of competence (c.f. self-determination theory [70]).

Some methodological limitations warrant a final note. The current studies used a relatively small sample of student participants. This is especially noteworthy for Experiment 2, where we aimed to assess associations between individual differences. Our results invite a replication with a larger and more heterogeneous sample. Second, our effects emerged in the context of a simple gambling task, and so it is unknown whether the relief effect would emerge in other contexts. For example, what would happen if reward feedback, unlike the current study, would be useful for decision making (e.g., in the context of reinforcement learning)? Third, we established a robust relief effect based only on subjective ratings of the reward-related feedback, but it is unclear whether this novel paradigm would still elicit its neural correlates [1]. Neuroimaging work using electrophysiology would not only be useful for validating the relief effect at the single trial level, but also for developing formal neurocomputational models of cost-benefit tradeoffs in effort allocation.

In sum, this study provides novel evidence for the notion that cognitive effort carries a cost, and that this cost is tightly linked to both reward processing and cost-benefit analyses of effort allocation. More precisely, our results suggest that if anticipated effort is avoided, it is returned as a reward, and experienced as relief. Intriguingly, this effect is attenuated in people who report being more disposed towards effortful tasks, suggesting they are less sensitive to cognitive costs. This raises the outstanding question of whether differences in NFC mitigate the aversive nature of cognitive effort, or whether they result in heightened stakes of engaging with, and succeeding at, hard cognitive tasks.

Supporting information

S1 File

(PDF)

Click here for additional data file.^{(583.1KB, pdf)}

Data Availability

All data and analysis scripts are available from the OSF (osf.io/gwhzf/).

Funding Statement

The authors received no specific funding for this work.

References

1.Gheza D, De Raedt R, Baeken C, Pourtois G. Integration of reward with cost anticipation during performance monitoring revealed by ERPs and EEG spectral perturbations. NeuroImage. 2018;173:153–64. doi: 10.1016/j.neuroimage.2018.02.049 [DOI] [PubMed] [Google Scholar]
2.Shenhav A, Botvinick M, Cohen J. The expected value of control: An integrative theory of anterior cingulate cortex function. Neuron. 2013;79(2):217–40. doi: 10.1016/j.neuron.2013.07.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Silvestrini N, Musslick S, Berry AS, Vassena E. An integrative effort: Bridging motivational intensity theory and recent neurocomputational and neuronal models of effort and control allocation. Psychol Rev. 2022; doi: 10.1037/rev0000372 [DOI] [PubMed] [Google Scholar]
4.Hull C. Principles of Behavior. New York, USA: Appleton-Century-Crofts; 1943. [Google Scholar]
5.Kool W, McGuire JT, Rosen ZB, Botvinick MM. Decision Making and the Avoidance of Cogntive Demand. J Exp Psychol Gen. 2010;139(4):665–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Bustos B, Colvett JS, Bugg JM, Kool W. Humans do not avoid reactively implementing cognitive control [Internet]. 2023. [cited 2023 Jun 13]. Available from: doi: 10.31234/osf.io/x5vgj [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Schouppe N, Ridderinkhof KR, Verguts T, Notebaert W. Context-specific control and context selection in conflict tasks. Acta Psychol (Amst). 2014. Feb 1;146:63–6. doi: 10.1016/j.actpsy.2013.11.010 [DOI] [PubMed] [Google Scholar]
8.Sayalı C, Badre D. Neural systems of cognitive demand avoidance. Neuropsychologia. 2019;123(October 2017):41–54. doi: 10.1016/j.neuropsychologia.2018.06.016 [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Sayali C, Rubin-McGregor J, Badre D. Policy abstraction as a predictor of cognitive effort [Internet]. PsyArXiv; 2022. Jun [cited 2023 Jun 13]. Available from: https://osf.io/by7gk [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Vogel TA, Savelson ZM, Otto AR, Roy M. Forced choices reveal a trade-off between cognitive effort and physical pain. eLife. 2020. Nov 17;9:e59410. doi: 10.7554/eLife.59410 [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Cohen JD. Cognitive Control. In: The Wiley Handbook of Cognitive Control [Internet]. John Wiley & Sons, Ltd; 2017. [cited 2023 Jun 12]. p. 1–28. Available from: https://onlinelibrary.wiley.com/doi/abs/10.1002/9781118920497.ch1 [Google Scholar]
12.Miller EK, Cohen JD. An integrative theory of prefrontal cortex function. 2001;167–202. [DOI] [PubMed] [Google Scholar]
13.Shenhav A, Musslick S, Lieder F, Kool W, Griffiths TL, Cohen JD, et al. Toward a Rational and Mechanistic Account of Mental Effort. Annu Rev Neurosci. 2017;40(1):annurev-neuro-072116-031526. doi: 10.1146/annurev-neuro-072116-031526 [DOI] [PubMed] [Google Scholar]
14.Botvinick MM, Huffstetler S, McGuire JT. Effort discounting in human nucleus accumbens. Cogn Affect Behav Neurosci. 2009; doi: 10.3758/CABN.9.1.16 [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Massar SAA, Libedinsky C, Weiyan C, Huettel SA, Chee MWL. Separate and overlapping brain areas encode subjective value during delay and effort discounting. NeuroImage. 2015. Oct 15;120:104–13. doi: 10.1016/j.neuroimage.2015.06.080 [DOI] [PubMed] [Google Scholar]
16.Salamone JD, Correa M. The Mysterious Motivational Functions of Mesolimbic Dopamine. Neuron. 2012;76(3):470–85. doi: 10.1016/j.neuron.2012.10.021 [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Leknes S, Lee M, Berna C, Andersson J, Tracey I. Relief as a Reward: Hedonic and Neural Responses to Safety from Pain. PLOS ONE. 2011. Apr 7;6(4):e17870. doi: 10.1371/journal.pone.0017870 [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Porreca F, Navratilova E. Reward, motivation, and emotion of pain and its relief. Pain. 2017. Apr;158 Suppl 1(Suppl 1):S43–9. doi: 10.1097/j.pain.0000000000000798 [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Hajcak G, Holroyd CB, Moser JS, Simons RF. Brain potentials associated with expected and unexpected good and bad outcomes. Psychophysiology. 2005;42(2):161–70. doi: 10.1111/j.1469-8986.2005.00278.x [DOI] [PubMed] [Google Scholar]
20.Gheza D, Paul K, Pourtois G. Dissociable effects of reward and expectancy during evaluative feedback processing revealed by topographic ERP mapping analysis. Int J Psychophysiol. 2018;132:213–25. doi: 10.1016/j.ijpsycho.2017.11.013 [DOI] [PubMed] [Google Scholar]
21.Klein ED, Bhatt RS, Zentall TR. Contrast and the justification of effort. Psychon Bull Rev. 2005;12(2):335–9. doi: 10.3758/bf03196381 [DOI] [PubMed] [Google Scholar]
22.Proudfit GH. The reward positivity: From basic research on reward to a biomarker for depression. Psychophysiology. 2015;52(4):449–59. doi: 10.1111/psyp.12370 [DOI] [PubMed] [Google Scholar]
23.Webb CA, Auerbach RP, Bondy E, Stanton CH, Foti D, Pizzagalli DA. Abnormal neural responses to feedback in depressed adolescents. J Abnorm Psychol. 2017;126(1):19–31. doi: 10.1037/abn0000228 [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Cohen MX, Elger CE, Ranganath C. Reward expectation modulates feedback-related negativity and EEG spectra. NeuroImage. 2007;35(2):968–78. doi: 10.1016/j.neuroimage.2006.11.056 [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Marco-Pallares J, Cucurell D, Cunillera T, García R, Andrés-Pueyo A, Münte TF, et al. Human oscillatory activity associated to reward processing in a gambling task. Neuropsychologia. 2008;46(1):241–8. doi: 10.1016/j.neuropsychologia.2007.07.016 [DOI] [PubMed] [Google Scholar]
26.Mas-herrero E, Ripollés P, Hajihosseini A, Rodríguez-fornells A, Marco-pallarés J. NeuroImage Beta oscillations and reward processing: Coupling oscillatory activity and hemodynamic responses. NeuroImage. 2015;119:13–9. [DOI] [PubMed] [Google Scholar]
27.Botvinick MM, Braver TS. Motivation and Cognitive Control: From Behavior to Neural Mechanism. Annu Rev Psychol. 2015;66(1):83–113. doi: 10.1146/annurev-psych-010814-015044 [DOI] [PubMed] [Google Scholar]
28.Westbrook A, Braver TS. Dopamine Does Double Duty in Motivating Cognitive Effort. Neuron. 2016;89(4):695–710. doi: 10.1016/j.neuron.2015.12.029 [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Kurzban R, Duckworth A, Kable JW, Myers J. An opportunity cost model of subjective effort and task performance. Behav Brain Sci. 2013;36(06):661–79. doi: 10.1017/S0140525X12003196 [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Vassena E, Silvetti M, Boehler CN, Achten E, Fias W, Verguts T. Overlapping neural systems represent cognitive effort and reward anticipation. PLoS ONE. 2014;9(3). doi: 10.1371/journal.pone.0091008 [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Bogdanov M, Nitschke JP, LoParco S, Bartz JA, Otto AR. Acute Psychosocial Stress Increases Cognitive-Effort Avoidance. Psychol Sci. 2021. Sep 1;32(9):1463–75. doi: 10.1177/09567976211005465 [DOI] [PubMed] [Google Scholar]
32.Pizzagalli DA. Depression, stress, and anhedonia: toward a synthesis and integrated model. Annu Rev Clin Psychol. 2014;10(January):393–423. doi: 10.1146/annurev-clinpsy-050212-185606 [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Cacioppo JT, Petty RE. The need for cognition. J Pers Soc Psychol. 1982;42:116–31. [Google Scholar]
34.Zerna J, Strobel A, Strobel A. The role of Need for Cognition in well-being–Review and meta-analyses of associations and potentially underlying mechanisms [Internet]. PsyArXiv; 2021. [cited 2023 Jun 4]. Available from: https://psyarxiv.com/p6gwh/ [Google Scholar]
35.Cacioppo JT, Petty RE, Feinstein JA, Jarvis WBG. Dispositional differences in cognitive motivation: The life and times of individuals varying in need for cognition. Psychol Bull. 1996;119(2):197–253. [Google Scholar]
36.Yee DM, Leng X, Shenhav A, Braver TS. Aversive motivation and cognitive control. Neurosci Biobehav Rev. 2022. Feb;133:104493. doi: 10.1016/j.neubiorev.2021.12.016 [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Hajcak G, Moser JS, Holroyd CB, Simons RF. It’s worse than you thought: The feedback negativity and violations of reward prediction in gambling tasks. Psychophysiology. 2007;44(6):905–12. doi: 10.1111/j.1469-8986.2007.00567.x [DOI] [PubMed] [Google Scholar]
38.Vassena E, Cobbaert S, Andres M, Fias W, Verguts T. Unsigned value prediction-error modulates the motor system in absence of choice. NeuroImage. 2015;122:73–9. doi: 10.1016/j.neuroimage.2015.07.081 [DOI] [PubMed] [Google Scholar]
39.Vassena E, Deraeve J, Alexander WH. Task-specific prioritization of reward and effort information: Novel insights from behavior and computational modeling. Cogn Affect Behav Neurosci. 2019; doi: 10.3758/s13415-018-00685-w [DOI] [PubMed] [Google Scholar]
40.Vassena E, Gerrits R, Demanet J, Verguts T, Siugzdaite R. Anticipation of a mentally effortful task recruits Dorsolateral Prefrontal Cortex: An fNIRS validation study. Neuropsychologia. 2018;123(October 2017):106–15. doi: 10.1016/j.neuropsychologia.2018.04.033 [DOI] [PubMed] [Google Scholar]
41.Russell JA. A circumplex model of affect. J Pers Soc Psychol. 1980;39(6):1161–78. [Google Scholar]
42.Cacioppo J. T., Petty R. E., Kao C. F. Need for Cognition Scale. Measurement Instrument Database for the Social Science. [Internet]. 2013. [cited 2023 May 23]. Available from: Retrieved from www.midss.org [Google Scholar]
43.Psychology Software Tools, Inc. [E-Prime 2.0] [Internet]. Sharpsburg, PA: Psychology Software Tools, Inc.; 2008. Available from: https://support.pstnet.com/.
44.Rouder JN, Morey RD, Verhagen J, Swagman AR, Wagenmakers EJ. Bayesian analysis of factorial designs. Psychol Methods. 2017;22(2):304–21. doi: 10.1037/met0000057 [DOI] [PubMed] [Google Scholar]
45.Schindler S, Schettino A, Pourtois G. Electrophysiological correlates of the interplay between low-level visual features and emotional content during word reading. Sci Rep. 2018; doi: 10.1038/s41598-018-30701-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
46.R Core Team. R. R Core Team. 2017.
47.Morey RD, Rouder JN. Bayesfactor: Computation of Bayes factors for common designs. R package version 0.9.12–2. BayesFactor Comput Bayes Factors Common Des. 2015; [Google Scholar]
48.Bürkner PC. brms: An R Package for Bayesian Multilevel Models Using Stan. J Stat Softw. 2017;80(1):1–28. [Google Scholar]
49.Morey RD. Confidence Intervals from Normalized Data: A correction to Cousineau (2005). Tutor Quant Methods Psychol. 2008. Sep 1;4(2):61–4. [Google Scholar]
50.Kool W, Botvinick M. Mental labour. Nat Hum Behav. 2018;2(12):899–908. doi: 10.1038/s41562-018-0401-9 [DOI] [PubMed] [Google Scholar]
51.Devine S, Vassena E, Otto AR. More than a feeling: physiological measures of affect index the integration of effort costs and rewards during anticipatory effort evaluation. Cogn Affect Behav Neurosci [Internet]. 2023. Apr 14 [cited 2023 Jun 13]; Available from: https://link.springer.com/10.3758/s13415-023-01095-3 [DOI] [PubMed] [Google Scholar]
52.Kool W, Gershman SJ, Cushman FA. Cost-Benefit Arbitration Between Multiple Reinforcement-Learning Systems. Psychol Sci. 2017;28(9):1321–33. doi: 10.1177/0956797617708288 [DOI] [PubMed] [Google Scholar]
53.Kool W, Shenhav A. Cognitive Control as Cost‐Benefit Decision Making. 2017; [Google Scholar]
54.Padmala S, Pessoa L. Reward Reduces Conflict by Enhancing Attentional Control and Biasing Visual Cortical Processing. J Cogn Neurosci. 2011. Nov 1;23(11):3419–32. doi: 10.1162/jocn_a_00011 [DOI] [PMC free article] [PubMed] [Google Scholar]
55.Treadway MT, Buckholtz JW, Schwartzman AN, Lambert WE, Zald DH. Worth the “EEfRT”? The effort expenditure for rewards task as an objective measure of motivation and anhedonia. PLoS ONE. 2009;4(8):1–9. doi: 10.1371/journal.pone.0006598 [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Westbrook A, Kester D, Braver TS. What Is the Subjective Cost of Cognitive Effort? Load, Trait, and Aging Effects Revealed by Economic Preference. PLoS ONE. 2013;8(7):1–8. doi: 10.1371/journal.pone.0068210 [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Westbrook A, Van Den Bosch R, Määttä JI, Hofmans L, Papadopetraki D, Cools R, et al. Dopamine promotes cognitive effort by biasing the benefits versus costs of cognitive work. Science. 2020. Mar 20;367(6484):1362–6. doi: 10.1126/science.aaz5891 [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Inzlicht M, Shenhav A, Olivola CY. The Effort Paradox: Effort Is Both Costly and Valued. Trends Cogn Sci. 2018;22(4):337–49. doi: 10.1016/j.tics.2018.01.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
59.McGuire JT, Botvinick MM. Prefrontal cortex, cognitive control, and the registration of decision costs. Proc Natl Acad Sci. 2010;107(17):7922–6. doi: 10.1073/pnas.0910662107 [DOI] [PMC free article] [PubMed] [Google Scholar]
60.Vassena E, Deraeve J, Alexander WH. Predicting Motivation: Computational Models of PFC Can Explain Neural Coding of Motivation and Effort-based Decision-making in Health and Disease. J Cogn Neurosci. 2017;29(10):1633–45. doi: 10.1162/jocn_a_01160 [DOI] [PubMed] [Google Scholar]
61.Thompson EP, Chaiken S, Hazlewood JD. Need for cognition and desire for control as moderators of extrinsic reward effects: a person x situation approach to the study of intrinsic motivation. J Pers Soc Psychol. 1993. Jun;64(6):987–99. doi: 10.1037//0022-3514.64.6.987 [DOI] [PubMed] [Google Scholar]
62.Clay G, Mlynski C, Korb FM, Goschke T, Job V. Rewarding cognitive effort increases the intrinsic value of mental labor. Proc Natl Acad Sci. 2022. Feb;119(5):e2111785119. doi: 10.1073/pnas.2111785119 [DOI] [PMC free article] [PubMed] [Google Scholar]
63.Lin H, Westbrook A, Fan F, Inzlicht M. An experimental manipulation increases the value of effort [Internet]. PsyArXiv; 2021. Feb [cited 2023 Jun 13]. Available from: https://osf.io/gnk4m [Google Scholar]
64.Kool W, Botvinick M. The intrinsic cost of cognitive control. Behav Brain Sci. 2013. Dec;36(6):697–8. doi: 10.1017/S0140525X1300109X [DOI] [PubMed] [Google Scholar]
65.Gärtner A, Grass J, Wolff M, Goschke T, Strobel A, Strobel A. No relation of Need for Cognition to basic executive functions. J Pers. 2021. Dec;89(6):1113–25. doi: 10.1111/jopy.12639 [DOI] [PubMed] [Google Scholar]
66.Bogdanov M, Renault H, LoParco S, Weinberg A, Otto AR. Cognitive effort exertion enhances electrophysiological responses to rewarding outcomes. Cereb Cortex. 2022. Sep 19;32(19):4255–70. doi: 10.1093/cercor/bhab480 [DOI] [PubMed] [Google Scholar]
67.Ma Q, Meng L, Wang L, Shen Q. I endeavor to make it: Effort increases valuation of subsequent monetary reward. Behav Brain Res. 2013;261:1–7. doi: 10.1016/j.bbr.2013.11.045 [DOI] [PubMed] [Google Scholar]
68.Norton MI, Mochon D, Ariely D. The IKEA effect: When labor leads to love. J Consum Psychol. 2012. Jul 1;22(3):453–60. [Google Scholar]
69.Bandura A. Self-efficacy: toward a unifying theory of behavioral change. Psychol Rev. 1977. Mar;84(2):191–215. doi: 10.1037//0033-295x.84.2.191 [DOI] [PubMed] [Google Scholar]
70.Ryan RM, Deci EL. Intrinsic and Extrinsic Motivations: Classic Definitions and New Directions. 2000;67:54–67. [DOI] [PubMed] [Google Scholar]

PLoS One. doi: 10.1371/journal.pone.0287954.r001

Decision Letter 0

Poppy Watson

31 Jul 2023

PONE-D-23-18706Need for cognition moderates cognitive effort aversionPLOS ONE

Dear Dr. Gheza,

Thank you for submitting your manuscript to PLOS ONE. Two reviewers have read your manuscript and as you can see below they both enjoyed reading your study but had a number of questions regarding analyses and interpretations that will need your attention. Thus, we feel that your manuscript has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. We invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Sep 14 2023 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Poppy Watson

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. In your Data Availability statement, you have not specified where the minimal data set underlying the results described in your manuscript can be found. PLOS defines a study's minimal data set as the underlying data used to reach the conclusions drawn in the manuscript and any additional data required to replicate the reported study findings in their entirety. All PLOS journals require that the minimal data set be made fully available. For more information about our data policy, please see http://journals.plos.org/plosone/s/data-availability.

"Upon re-submitting your revised manuscript, please upload your study’s minimal underlying data set as either Supporting Information files or to a stable, public repository and include the relevant URLs, DOIs, or accession numbers within your revised cover letter. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories. Any potentially identifying patient information must be fully anonymized.

Important: If there are ethical or legal restrictions to sharing your data publicly, please explain these restrictions in detail. Please see our guidelines for more information on what we consider unacceptable restrictions to publicly sharing data: http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions. Note that it is not acceptable for the authors to be the sole named individuals responsible for ensuring data access.

We will update your Data Availability statement to reflect the information you provide in your cover letter.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Overview: This paper investigates whether avoiding anticipated cognitive effort is itself rewarding. The paper shows, via two experiments using a gambling paradigm, that peoples' sense of relief when avoiding effort is dependent on how much effort an individual was anticipating. Relatedly, the authors argue this sense of relief is moderated by a persons’ Need for Cognition.

Overall, the paper is clearly written and I believe the research question is interesting. The results are well presented and generally align with the authors’ hypotheses—people do seem to be more relieved when avoiding high effort tasks as opposed to low effort tasks. The analysis approach is also appropriate (although could be improved) and the figures present the data transparently. I do however have some reservations about the strengths of the conclusions drawn, plus some other questions and suggestions.

1. Accuracy and self-report scores. It is possible that differences in subjective ratings (pleasantness, frustration, relief) arose due to differences in participants’ expectations of reward, rather than as a direct consequence of avoiding effort. Consider this rough calculation using the accuracy data from Exp 1: on a ‘hard’ trial, the average participant has a 50% chance of immediate reward, a 43.5% chance of reward following effort (87% avg accuracy) and a 6.5% chance of no reward; on an ‘easy’ trial a participant has a 50% chance of immediate reward, a 49% chance of reward following effort (98% avg. accuracy), and 1% no reward. The EVs for the two trial types, hard and easy, are therefore 5.61 points and 5.94, respectively. A participant may therefore feel more relieved on hard trials as it was less likely, as a function of how hard the arithmetic task was, they would receive reward. This hypothesis is somewhat supported by your finding that participants are more relieved after getting ‘hard’ trials correct than ‘easy’ trials (Figure 7). One solution would be to control for participants’ accuracy in the experiment. If a model containing ‘difficulty level’ accounted for the data better than a model containing participants’ accuracy (all other model components being equal), it would provide stronger evidence that effort avoidance is driving the ‘relief effect’. Note that Westbrook et al. (2013) use such an approach to assess effort discounting.

2. Analysis methods. I’ve assumed, as it isn’t entirely clear from the paper, that you are currently averaging participants’ responses and then using participant averages to fit your models. Given each participant gives repeated estimates, it may be more appropriate to use a Bayesian mixed modelling approach which eliminates the need for such aggregation. This can be done vis the brms package in R and would allow the assessment of both fixed and random effects.

3. Relationship between NFC and accuracy. I’m glad the authors included this analysis. I do however think the authors could introduce some scepticism as to the causal direction of this relationship. Given participants fill out the NFC scale immediately after the gambling task is completed, it is possible their NFC scores are affected by their accuracy, rather than NFC affecting accuracy. I also think it would be worthwhile including accuracy as a factor in the analysis which assesses the moderating role of NFC on people’s self-report ratings (pleasantness, frustration, relief). At the very least this potential limitation/confound should be discussed.

4. I wondered why you chose to aggregate across the different subjective rating types (i.e., pleasantness, frustration, relief) given your primary interest, as far as the introduction is concerned, is in a sense of relief. This effect is explicitly referred to as the ‘relief effect’ but comprises frustration and pleasantness scores in addition to relief. Naturally, you have more statistical power by combining these ratings, but it seems like an odd choice to deliberately include 3 different rating types and then aggregate across them in your analysis. I think the authors could include a clearer rationale for their current approach.

Minor points

• In lines 431-432 the authors argue the results show people exerted more effort in the hard condition. This doesn’t seem obvious to me. Perhaps participants exerted the same amount of effort (which itself is difficult to define) and that is why they performed worse at the task.

• What was the scale/range of the VAS ratings? They have been converted to percentages for analysis, but what was the original scale? This is not reported.

• I don’t think Figure 4A adds much and the authors could just report Figure 4B. I’m also not sure if it is necessary to include Figure 5 as it doesn’t add much in addition to Figure 4, but, if you are going to keep it you should explain how the participants in the two extremes are categorised (i.e., bottom and top 30%) in the main text as opposed to only in the Figure caption.

• I don’t think the current title does a great job of representing the work. The fact that NFC is related to effort aversion is well established. I think you could focus on the relief aspect, which is framed as your primary interest in the abstract.

• At a few points the authors alternate between using a period (.) and a comma (,) as a decimal separator (e.g., line 195). I imagine the norm for this journal would be a period.

• Consistency with the number of decimal places would also be good. For example, the accuracy scores are reported to the nearest integer, but subjective ratings are to 1 decimal place. Reporting to two decimal places throughout makes sense to me.

• Line 459-60 refers to the “law of least mental effort”, given Hull was not specifically referring to cognitive effort in his work it may be better to drop the “mental” from this phrase.

Jake Embrey

I sign all of my reviews

Reviewer #2: The current study aimed to test previously published findings from the same group about a relief effect that occurs when participants avoided having to spend mental effort in a gambling task. Across two experiments, they found that participants reported more pleasure, less frustration, and more relief for outcomes that signaled successful avoidance of effort, especially if the anticipated effort was high compared to low. This effect was reversed for outcomes signaling that effort had to be exerted. In the second experiment, the authors show that NFC scores moderate the strength of these results.

Overall, I think the manuscript is well-written and that this study tackles a relevant topic that will be of interest to the broader readership of the journal. The rationale is well-explained, and the design is well-suited to answer the research question. Although I think that the study deserves to be published, I have a few comments that should be addressed before I can ultimately recommend the manuscript to be accepted. These comments are mostly related to the statistical analyses and the authors’ interpretation of the data.

Major comments:

- It was a little unclear to me what exactly served as the dependent variable of interest here. Plots depict participants’ ratings for pleasure, frustration, and relief separately, but it seems as if the authors actually used an average score across these dimensions in their ANOVAs. Is this correct? If not, please ignore the following points. If it is, it should be explained more specifically. The resulting effect of these ANOVAs is sometimes referred to as a relief effect, but if all outcome-related ratings are averaged, it is not really (only) about relief but more generally about the affective response (and sometimes it is actually referred to as such). The authors also seem to interpret this as a measure of effort aversion (hence the title of the manuscript). So, my point here is, please explain how you arrived at your DV and use a consistent term to refer to this effect throughout the manuscript. I suggest it should not be the term relief, though, as it includes more than the specific feeling of relief.

- Related to the previous point, what was the reason to combine these ratings instead of looking at them separately?

- A strength of the paper is that the authors use Bayes Factor comparisons to find the most appropriate models for their analyses. The corresponding numbers are in the text, but it might help the reader to put them in a table (similar to the guidelines published here: https://osf.io/wae57). This applies to both experiments. In experiment 2, a table for the output of the brms model might be good, too.

- I was also unclear about the random effects specification of the models. On page 11, the authors state that they included participant, affective dimension, and their interaction in their models. How exactly did this look like? Would it have made sense to include affective dimension as a fixed effect and nest everything within participants instead, so that potential differences between these dimensions could have been analyzed? Here, again, I am mostly arguing for more clarity and justification (instead of doubting/criticizing the authors’ decisions).

Minor comments:

- Title: I was wondering if the title of the manuscript (that really emphasizes the moderating effect of NFC) truly captures the main findings of the study well, given that NFC is only measured in the second experiment and that the main purpose across both experiments was more to validate positive affective responses associated with not having to exert effort. To be clear, I wouldn’t object the current title if the authors choose to stick with it, I just think there might be better alternatives.

- On page 20, lines 485 – 487, the authors mention that NFC was measured after the task and that the questionnaire was thus unlikely to influence performance in the task itself. This is reasonable. However, I was wondering whether having experienced the task and the subjective responses to potential or actual effort exertion may have affected how participants responded to these questions (i.e., if they did not enjoy completing the math trials, they might be inclined to rate their overall motivation to engage in mentally demanding behavior lower in the questionnaire). I know NFC is supposed to be a trait, but the task might still have some transient effect on participants’ ratings that might explain the relationship between ratings and behavior. This is in no way a major problem for the study, but it might be worth a sentence or two in the discussion.

- This is more a question out of curiosity than a comment: Figures 2 and 3 show that in both experiments, there were people who rated relief low in reward feedback trials and high in no-reward feedback trials. This is a little surprising, although it is unclear from the plots whether these participants were at least internally consistent (i.e., a person who rated relief as let’s say 25 for hard trials with reward feedback would rate it even lower for hard trials with no-reward feedback). Did participants show this consistency or were there some who displayed ratings that ran completely against expectations (i.e., more relief, pleasure etc. for no-reward trials than for reward trials)?

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Jake R. Embrey

Reviewer #2: No

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

PLoS One. 2023 Nov 16;18(11):e0287954. doi: 10.1371/journal.pone.0287954.r002

Author response to Decision Letter 0

30 Sep 2023

Please refer to the Response to Reviewer pdf file for a better formatting and readability.

PONE-D-23-18706

Need for cognition moderates cognitive effort aversion PLOS ONE

Reviewer 1

Overview: This paper investigates whether avoiding anticipated cognitive effort is itself rewarding. The paper shows, via two experiments using a gambling paradigm, that peoples' sense of relief when avoiding effort is dependent on how

much effort an individual was anticipating. Relatedly, the authors argue this sense

of relief is moderated by a persons’ Need for Cognition.

other questions and suggestions.

We thank the reviewer for the kind words, and for the thoughtful points. We believe that they have allowed us to improve the paper.

We thank the reviewer for bring this concern to our attention. They correctly point out that the difference in accuracy between effort conditions results in a difference in reward probability (albeit a relatively small difference, as well quantified by the reviewer). Therefore, we have run control analyses that control for experienced difficulty when

predicting the feedback ratings. More precisely, we computed the average accuracy scores for each difficulty level and participant, and we use these as a continuous predictor (instead of the factor difficulty level used before), as a proxy of expected values for each condition. Following the reviewer’s logic, we reasoned that if the model including ‘difficulty level’ outperformed the model including ‘accuracy’, it would be unlikely that reward expectations (rather than expected effort) drove our effect. One caveat to this reasoning is that individual difference in performance may as well act as a more sensitive proxy of expected effort for the two difficulty conditions, because it models variance in the relative difference in difficulty between conditions across participants. Therefore, this control analysis should not be interpreted as decisive in ultimately disentangling these two alternative interpretations.

Nevertheless, we compared two models including either the outcome x difficulty or outcome x accuracy for both experiments. In Experiment 1, we found that data where more likely under a model including outcome x accuracy than under a model including outcome x difficulty. In Experiment 2, we found the opposite result. The data were more likely under a model including outcome x difficulty than under a model including outcome x accuracy. We report these results in the Supporting Information.

In short, it is possible that the effect of difficulty level affected the feedback ratings through changes in expected value in Experiment 1. On the other hand, we find evidence against this hypothesis for Experiment 2. Importantly, the most important claims of our paper are based on this second experiment. We conclude that our experimental design is not suited to fully address this potential confound. Future research may address this issue with different experimental procedures, for example by omitting performance feedback. We discuss these points in a new paragraph in the General Discussion, where we also reference the analyses in the Supplemental Materials:

"Another possibility is that participants scoring low in NFC, and performing sub-par in the arithmetic task, may have developed lower expected values for high difficulty trials, corresponding to lower estimates of reward probability. Therefore, the relief for avoiding high effort may be compounded by the relief for avoiding risky arithmetic trials. In the Supporting Information we report two control analyses which confirm the relief effect, and the role of NFC in moderating it, even when controlling for individual differences in arithmetic performance.

Please note that this text refers to the results of the control analysis reported above (comparing the use of difficulty as factor vs. accuracy as a continuous predictor), but also to another control analysis suggested by the reviewer, which we discuss in detail in point 3 below (where we use the difference in average accuracy between difficulty levels as additional or alternative moderator to NFC).

We thank the reviewer for bringing this concern to our attention. It has allowed us to clarify our paper. In fact, our analytic approach already fully relies on (hierarchical) mixed-effect models. In all main analyses, we model single-trial data using hierarchical regression with subjects and affective dimensions as random effects. Because we strived to keep these main analyses simple and accessible, we opted for the Bayes Factor approach. However, when we turned our attention to the more complex 3-way interaction (including outcome x difficulty level x NFC; cf. model 5), we corroborated the main analyses with full Bayesian hierarchical regressions using the BRMS package (like the reviewer suggested). We have added text to the Statistical Analyses section to highlight this. On page 10, we now say:

“Subjective ratings for the three affective dimensions were used to assess the hedonic response as a whole, along a negative-positive valence axis (41). In all analyses, we predicted single trial subjective ratings with a mixed modelling approach, in which participant and affective dimension were treated as random effects (i.e., considered nuisance factors).”

On page 12, we add:

“In Experiment 2, to further explore the effect of NFC on reward processing, we also ran a full Bayesian estimation of parameter values for model 5, using the “brms” R package.”

We agree with the reviewer, and have added the following text to the General Discussion (page 21):

“Rather, it is possible that experiencing the task may have biased participants’ responses in the NFC questionnaire. Perhaps, observing their performance on the task influenced the self-evaluations prompted by the NFC questionnaire, leading their scores to reflect a combination of stable trait features with transient affective reactions”.

We added this sentence to a paragraph in which we already placed some caution about interpreting changes in the relief effect as purely driven by difference in the motivation to engage with effort (since we found a relationship between NFC and arithmetic performance).

We also ran the new analyses suggested by the reviewer, and we report them in the Supporting Information. Specifically, we ran an additional control analysis including accuracy when assessing the moderating role of NFC on people’s self-report ratings. We defined a new continuous predictor, “delta_performance”, corresponding to the difference in mean accuracy in the hard vs. easy conditions. Next, we added this predictor in a new model “outcome*effort*delta_accuracy”. We reasoned that the larger the delta_performance, the larger the difference in expected reward between conditions. We expected this interaction to compete for some of the variance explained by the outcome*effort*NFC, but we predicted that a model including the interactions with delta_accuracy (outcome*effort*delta_accuracy) and the interactions with NFC (outcome*effort*NFC) would explain the data better than one including the interactions with delta_accuracy alone. Results confirmed this prediction. Moreover, even a simple three-way interaction model including NFC performed much better than the three-way interaction model including delta_accuracy.

As mentioned above, we briefly describe this analysis in the General Discussion and refer to the Supporting Information there.

We thank the reviewer for pointing out this ambiguity, and for giving us the opportunity to better clarify our rationale. Our goal was to assess a broad hedonic response towards the anticipation or likely avoidance of effort. Therefore, we used three affective dimensions, with different semantic connotations, so that we could collect converging evidence along a single axis (positive to negative). In addition, this approach allowed us to validate the consistency of participants’ reports. We now justify this choice more explicitly in the Stimuli and task section (page 8):

“By using three questions, which assessed a range of different affective dimensions, we aimed at measuring the latent construct of a ‘hedonic response’ along a negative-positive valence axis (41). Multiple self-report items allowed us to validate the consistency of participants’ responses, and to increase statistical power by controlling for nuisance variability between affective dimensions.”

We feel this choice is especially warranted because the measures show extremely robust correlations. The average correlation for frustration and pleasure was 0.81 (Exp. 1) and 0.67 (Exp. 2). The average correlation for frustration and relief was 0.71 (Exp. 1) and 0.57 (Exp. 2). The average correlation for pleasure and relief was 0.85 (Exp. 1) and 0.80 (Exp. 2). We report these correlations in the Supporting Information, and refer to them in the text on page 14:

“Moreover, we observed very robust correlations between pairs of affective dimensions (group-level statistics of these correlations are reported in Supporting Information).”

We also clarify the distinction between the latent construct that we wanted to measure (hedonic responses) and the interpretation of its modulation by the factors Outcome and Difficulty level (the ‘relief effect’). In the Introduction (page 5), we now say:

“In line with our previous study, we assessed the latent construct of the “hedonic response” (i.e., the subjective experience of the gambling trial outcome) using three probes, each addressing a different affective dimension (“pleasure”, “frustration”, and “relief”). We refer to the interaction of trial outcome with effort anticipation on the hedonic response as a “relief effect”.”

We also added a clarification on the Statistical analysis section (pages 11-12):

“Moreover, the factors Subject, Affective dimension, and their interaction were treated as nuisance variables (i.e., varying effects to control for). To do this, they were entered in all models, including the null model. Additionally, the null model included a single (i.e., constant) intercept. This procedure allowed us to control for the nuisance effects of 1) individual differences in VAS biases (e.g., preference for a given section of the scale), 2) overall differences in rating across affective dimensions (e.g., overall higher responses for a specific scale), and 3) individual variations in the latter differences.”

Finally, we report the results of analyses when only modeling the relief component in the Supporting Information. These results were qualitatively and quantitively similar to the results reported in the main paper. We have included a footnote in the main text that refers the reader to this analysis:

“When selectively predicting the feedback ratings for the ‘relief’ affective dimension, data were still best explained by a model that included the three-way interaction of outcome x difficulty level x NFC, followed by a model including the two-way outcome x difficulty level interaction (see Supporting Information).”

Minor points

- In lines 431-432 the authors argue the results show people exerted more effort in the hard condition. This doesn’t seem obvious to me. Perhaps participants exerted the same amount of effort (which itself is difficult to define) and that is why they performed worse at the task.

This is a valid point worth of further discussion. To clarify, we believe that our results are only suggestive of such an interpretation (i.e., that “people allocated more effort in the hard condition”). We base this claim on the relatively minor difference in accuracy, in the face of substantial difference in difficulty, and (importantly) a lack of clear change in reported motivation. In other words, we think that the overall pattern of results lends some support for the conclusion that participants exerted more effort in the hard condition, but this conclusion is not fully granted. We now clarify the caveat of this interpretation, by adapting the concluding sentence of that paragraph:

“We believe that our data shows that participants exerted more cognitive effort in the hard condition. In response to a substantial difference in difficulty, they only showed a relatively minor drop in accuracy. Most importantly, we saw no clear shift in reported motivation. Nevertheless, it is important to note that our findings provide only preliminary support for this conclusion, as both task demands and the precision of the information-processing operations (i.e., performance) differed, to some extent, between difficulty conditions (13).”

- What was the scale/range of the VAS ratings? They have been converted to percentages for analysis, but what was the original scale? This is not reported.

The visual analog scale (VAS) was presented as a continuous line segment at the center of the screen, bounded with vertical shorter segments. It ranged 396 pixels of the screen. To compute the VAS percentage, we converted the mouse coordinates on the screen along the VAS line. We first calculated the relative pixel position of the mouse at time of click, with respect to the leftmost point of the scale. We then converted this point in percentage, considering the 396 pixels range. We have added a sentence in the Data analysis section (page 10) to explain this in more detail:

“VAS scores were calculated by first identifying the relative x-axis position of the mouse with respect to the leftmost position of the scale at the time of participant’s click. We then converted this measure from pixels distance to percentage over the scale range (396 pixels).”

- I don’t think Figure 4A adds much and the authors could just report Figure 4B. I’m also not sure if it is necessary to include Figure 5 as it doesn’t add much in addition to Figure 4, but, if you are going to keep it you should explain how the participants in the two extremes are categorised (i.e., bottom and top 30%) in the main text as opposed to only in the Figure caption.

We agreed with the reviewer that Figure 4A was redundant, and have removed it. However, we have decided to keep Figure 5, insofar it provides different information (i.e., raw data) compared to Figure 4 (i.e., estimates based on posterior distributions). Of course, if the reviewer feels strongly about this, we’d be happy to consider removing it.

We now also explain in text how the extremes are categorized (page 17):

“(participants within the 0-30 percentiles were included in the Low NFC group; participants within the 70-100 percentiles were included in the High NFC group)”

- I don’t think the current title does a great job of representing the work. The fact that NFC is related to effort aversion is well established. I think you could focus on the relief aspect, which is framed as your primary interest in the abstract.

We agree with the reviewer and have changed the title to “Need for cognition moderates the relief of avoiding cognitive effort”.

- At a few points the authors alternate between using a period (.) and a comma (,) as a decimal separator (e.g., line 195). I imagine the norm for this journal would be a period.

We thank the reviewer for spotting these typos. We now consistently use periods as the decimal separator.

- Consistency with the number of decimal places would also be good. For example, the accuracy scores are reported to the nearest integer, but subjective ratings are to 1 decimal place. Reporting to two decimal places throughout makes sense to me.

We have made these changes.

- Line 459-60 refers to the “law of least mental effort”, given Hull was not specifically referring to cognitive effort in his work it may be better to drop the “mental” from this phrase.

Agreed.

Reviewer 2

The current study aimed to test previously published findings from the same group about a relief effect that occurs when participants avoided having to spend mental effort in a gambling task. Across two experiments, they found that participants reported more pleasure, less frustration, and more relief for outcomes that signaled successful avoidance of effort, especially if the anticipated effort was high compared to low. This effect was reversed for outcomes signaling that effort had to be exerted. In the second experiment, the authors show that NFC scores moderate the strength of these results.

accepted. These comments are mostly related to the statistical analyses and the authors’ interpretation of the data.

We thank the reviewer for their positive remarks, and also for their more critical comments. We believe these comments strengthened the manuscript.

1. It was a little unclear to me what exactly served as the dependent variable of interest here. Plots depict participants’ ratings for pleasure, frustration, and relief separately, but it seems as if the authors actually used an average score across these dimensions in their ANOVAs. Is this correct? If not, please ignore the following points. If it is, it should be explained more specifically. The resulting effect of these ANOVAs is sometimes referred to as a relief effect, but if all outcome-related ratings are averaged, it is not really (only) about relief but more generally about the affective response (and sometimes it is actually referred to as such). The authors also seem to interpret this as a measure of effort aversion (hence the title of the manuscript). So, my point here is, please explain how you arrived at your DV and use a consistent term to refer to this effect throughout the manuscript. I suggest it should not be the term relief, though, as it includes more than the specific feeling of relief.

We thank the reviewer for pointing out this ambiguity, and for giving us the opportunity to clarify. We should note that Reviewer 1 raised a very similar point (see above). Therefore, we address them together.

First, we should clarify that we do not average scores across the three affective dimensions, but rather we predict each individual rating, separately for each affective dimension. As explained below, we control for differences across dimensions by adding the factor Affective dimension to all our models, including the null.

This said, it is true that we aim at assessing the latent construct of “hedonic response” as a whole. More specifically, we used three affective dimensions, with different semantic connotations, so that we could collect converging evidence along a single axis (positive to negative). In addition, this approach allowed us to validate the consistency of participants’ reports. We now justify this choice more explicitly in the Stimuli and task section (page 8):

2). We report these correlations in the Supporting Information, and refer to them in the text on page 14:

“Moreover, we observed very robust correlations between pairs of affective dimensions (group-level statistics of these correlations are reported in the Supporting Information).”

We also added a clarification on the Statistical analysis section (pages 11-12):

In some of the Figure captions (Fig. 2,3,7) we now clarify that we combine the three affective dimensions in the statistical analysis:

“Reponses to the three affective dimensions were used to estimate the underlying hedonic response along a single valence axis.”

With these changes, we hope to have better clarified the distinction between our dependent variable and the latent construct we aimed to measure. To emphasize this, we now use the term “hedonic response” more consistently across the manuscript.

2. Related to the previous point, what was the reason to combine these ratings instead of looking at them separately?

We believe we have already addressed this question in our previous answer, and we elaborate on it below (point 4). In short, we leveraged the statistical strength of mixed models to control for consistency across responses.

3. A strength of the paper is that the authors use Bayes Factor comparisons to find the most appropriate models for their analyses. The corresponding numbers are in the text, but it might help the reader to put them in a table (similar to the guidelines published here: https://osf.io/wae57). This applies to both experiments. In experiment 2, a table for the output of the brms model might be good, too.

We have added two tables reporting the model comparison results for each experiment, and a table for the parameter estimates from “brms” for Experiment 2.

4. I was also unclear about the random effects specification of the models. On page 11, the authors state that they included participant, affective dimension, and their interaction in their models. How exactly did this look like? Would it have made sense to include affective dimension as a fixed effect and nest everything within participants instead, so that potential differences between these dimensions could have been analyzed? Here, again, I am mostly arguing for more clarity and justification (instead of doubting/criticizing the authors’ decisions).

We thank the reviewer for giving us the opportunity to better explain how we specified the nuisance effects (i.e., the varying effects, also referred to as ‘random’). We have made substantial changes to Statistical Analysis paragraph referenced by the reviewer. On pages 11-12, we say:

It was our deliberate intention to treat that Affective dimension factor as nuisance, since we did not have sufficient theoretical ground to make predictions about potential differences between its levels. Rather, as explained above, we used three different Affective dimension levels with the aim of assessing the broad hedonic response of the participants along a ‘valence’ axis (from negative to positive). First, probing the same latent construct (hedonic response) with three different questions allowed us to increase the statistical power of our analysis. Second, the three dimensions allowed us to ensure the consistency of participant’s reports. Specifically, in case of inconsistent responses across dimensions, their variance would have been explained by the nuisance effects, and the participant’s contribution to the constant effects of interest would have been hindered.

We also noticed a mistake in the statistical analyses of Experiment 2 (R script for Exp 2 only). In particular, the random term “Subject” in some of the models was treated as a numeric variable, rather than a factor. We have corrected this error and the text reflects the new results (pp. 15-16). Importantly, this correction does not alter the pattern of results described before, but rather provides even stronger evidence in the same direction.

Minor comments

We agree with the reviewer and have changed the title to “Need for cognition moderates the relief of avoiding cognitive effort”.

We thank the reviewer for suggesting this subtle but relevant element. We agree that experiencing the task, which includes mental arithmetic, may have induced a transient affective reaction. Responses to the NFC questionnaire, albeit putatively assessing a stable trait, may have been influenced by these affective aftereffects of the task. We included this consideration in the discussion, on page 23.

it even lower for hard trials with no-reward feedback). Did participants show this consistency or were there some who displayed ratings that ran completely against expectations (i.e., more relief, pleasure etc. for no-reward trials than for reward trials)?

To address this point, we have added Supplementary Figures 1-6 in the Supporting Information:

100 75 50 25 0

Com

150 positera

100 ting(%

)

reward−noreward conditions

) %(g nitaR

Easy Easy reward no−reward

Hard Hard reward no−reward

Relief

e.g., Supplementary Figure 3 (Experiment 1)

They recapitulate the results in the main Figs 2-3, but each Figure refers to a single affective dimension, and compares reward to no-reward conditions for each difficulty level. Moreover, we connected participant’s datapoints with colored segments to highlight the pattern of ratings across conditions. Last, the secondary plot on the right shows a composite rating, obtained by summing ratings from reward conditions and subtracting the ones from no-reward conditions. As can be seen from the figures, there is overall internal consistency across reward levels (i.e., lower affective value for no-reward vs reward). This is especially the case for the affective dimensions “relief” and “pleasure”, while a few participants showed a modest reversal of this pattern for the dimension “frustration”, albeit in Experiment 2 only (negative values in the composite rating). Some participants showed less variance than others in their responses across different conditions, demonstrating a preference for a narrow section of the VAS scale (i.e., showing high precision in their ratings across conditions). This translates into a flatter pattern across conditions. Nevertheless, the data shows a robust internal consistency. We added a reference to these supplementary results in the Result section (page 14):

“As can be seen from the Supplementary Figures 1-6, participants showed a robust internal consistency in their rating of the reward-related feedback across outcome levels: Their rating was consistently lower for no-reward compared to reward outcome.”

Attachment

Submitted filename: Response to Reviewers.pdf

Click here for additional data file.^{(166.4KB, pdf)}

PLoS One. doi: 10.1371/journal.pone.0287954.r003

Decision Letter 1

Poppy Watson

12 Oct 2023

PONE-D-23-18706R1Need for cognition moderates the relief of avoiding cognitive effortPLOS ONE

Dear Dr. Gheza,

Thank you for resubmitting your manuscript to PLOS ONE. As you can see, both reviewers feel that the manuscript is vastly improved and is nearly ready for publication. There is one final point which reviewer 1 has raised and it would be great if you could consider this additional analysis for inclusion. I do not intend to send the manuscript back out for re-review, provided that you can clearly address the reviewer's comment in your response letter/manuscript.

Please submit your revised manuscript by Nov 26 2023 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

We look forward to receiving your revised manuscript.

Kind regards,

Poppy Watson

Academic Editor

PLOS ONE

Journal Requirements:

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: (No Response)

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Reviewer #1: 1. I appreciate that the authors saw my concerns here and agreed this is worthy of addressing. The analysis which was run however isn’t exactly what I had in mind, which may be due to a lack of clarity on my part. The more appropriate analysis would compare iterative models —rating ~ outcome vs. rating ~ outcome + accuracy vs. rating ~ outcome + accuracy + difficulty. The crucial step here is comparing the last two models as it allows you to assess whether adding difficulty improves the model fit above accuracy alone. This is roughly equivalent to what Westbrook et al. (2013) + others have done.

While I don’t think not doing this precludes publication, I think the authors should consider this approach and include it in the main manuscript (or at least supporting information) if they agree. The current approach in the Supporting Information is OK, but it doesn’t allow you to assess whether difficulty adds anything over and above accuracy. The authors, however, are of course correct in saying the current experiment doesn’t allow this issue to be resolved entirely and it’s a good suggestion to note this in the General Discussion, as they have.

I believe all the other amendments to both mine and Reviewer 2’s suggestions are appropriate and have hopefully improved the paper!

Good work again on an interesting pair of experiments. I’ve enjoyed reading this work.

Reviewer #2: I thank the authors for thoroughly addressing all of my comments. I have no further concerns and I am happy to recommend this manuscript for publication.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Jake Embrey

Reviewer #2: Yes: Mario Bogdanov

**********

PLoS One. 2023 Nov 16;18(11):e0287954. doi: 10.1371/journal.pone.0287954.r004

Author response to Decision Letter 1

25 Oct 2023

Please refer to Response to Reviewers.pdf for a better formatting.

PONE-D-23-18706

Need for cognition moderates the relief of avoiding cognitive effort PLOS ONE

Reviewer 1

I appreciate that the authors saw my concerns here and agreed this is worthy of addressing. The analysis which was run however isn’t exactly what I had in mind, which may be due to a lack of clarity on my part. The more appropriate analysis

would compare iterative models —rating ~ outcome vs. rating ~ outcome + accuracy vs. rating ~ outcome + accuracy + difficulty. The crucial step here is comparing the last two models as it allows you to assess whether adding difficulty improves the model fit above accuracy alone. This is roughly equivalent to what

Westbrook et al. (2013) + others have done.

suggestion to note this in the General Discussion, as they have.

I believe all the other amendments to both mine and Reviewer 2’s suggestions are

appropriate and have hopefully improved the paper!

Good work again on an interesting pair of experiments. I’ve enjoyed reading this

work.

Response:

We thank the reviewer for his positive evaluation of our revised manuscript, and for wanting to further clarify how to best approach this analysis. Indeed, comparing iterative models including additional predictors, keeping every other predictor equal, should allow to verify the existence of additional explanatory power of the added predictors.

We implemented the suggestion of the reviewer, but rather than adding the simple main effects of outcome, accuracy, and difficulty iteratively, we decided to compare a model including the main effects and interaction of outcome and accuracy, to a model including those same predictors and the interaction of outcome and difficulty level. We believe that this approach does capture the iterative logic suggested by the reviewer (and implemented in Westbrook et al., 2013), but is better suited to answer our research question. In fact, the main effects of difficulty level and accuracy alone do not explain much variance in the data, but they do show strong interactions with outcome.

Our results confirm the pattern of results previously reported in the Supplementary Information. For Experiment 1, a simple model including the interaction of accuracy with outcome alone performed better than the model including both interaction terms,

suggesting the interaction with difficulty level does not have explanatory power above the interaction of outcome with accuracy. Rather, the more complex model with both interaction terms performed better than one including the simple interaction with difficulty level. For Experiment 2, the opposite pattern was found, in line with the analysis reported in the previous round of revision. More specifically, the model including two interaction terms performed much better than the one including the interaction of outcome with accuracy alone. In other words, the interaction of outcome with difficulty level survived after controlling for the interaction of outcome with accuracy. We have included these results in the Supplementary Information.

We are grateful to the reviewer for pointing out this additional approach which further clarifies that, in the second experiment, there is no evidence that individual differences in arithmetic performance (and therefore, in reward expectations) confounded the relief effect of avoiding high vs. low effort.

Reviewer 2

I thank the authors for thoroughly addressing all of my comments. I have no further concerns and I am happy to recommend this manuscript for publication.

Response:

We are very grateful to the reviewer for the insightful suggestions offered in the previous round of revision, that strongly helped us improving the paper.

Attachment

Submitted filename: Response to Reviewers.pdf

Click here for additional data file.^{(80.8KB, pdf)}

PLoS One. doi: 10.1371/journal.pone.0287954.r005

Decision Letter 2

Poppy Watson

2 Nov 2023

Need for cognition moderates the relief of avoiding cognitive effort

PONE-D-23-18706R2

Dear Dr. Gheza,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Poppy Watson

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

PLoS One. doi: 10.1371/journal.pone.0287954.r006

Acceptance letter

Poppy Watson

8 Nov 2023

PONE-D-23-18706R2

Need for cognition moderates the relief of avoiding cognitive effort

Dear Dr. Gheza:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Poppy Watson

Academic Editor

PLOS ONE

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

S1 File

(PDF)

Click here for additional data file.^{(583.1KB, pdf)}

Attachment

Submitted filename: Response to Reviewers.pdf

Click here for additional data file.^{(166.4KB, pdf)}

Attachment

Submitted filename: Response to Reviewers.pdf

Click here for additional data file.^{(80.8KB, pdf)}

Data Availability Statement

All data and analysis scripts are available from the OSF (osf.io/gwhzf/).

[pone.0287954.ref001] 1.Gheza D, De Raedt R, Baeken C, Pourtois G. Integration of reward with cost anticipation during performance monitoring revealed by ERPs and EEG spectral perturbations. NeuroImage. 2018;173:153–64. doi: 10.1016/j.neuroimage.2018.02.049 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref002] 2.Shenhav A, Botvinick M, Cohen J. The expected value of control: An integrative theory of anterior cingulate cortex function. Neuron. 2013;79(2):217–40. doi: 10.1016/j.neuron.2013.07.007 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref003] 3.Silvestrini N, Musslick S, Berry AS, Vassena E. An integrative effort: Bridging motivational intensity theory and recent neurocomputational and neuronal models of effort and control allocation. Psychol Rev. 2022; doi: 10.1037/rev0000372 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref004] 4.Hull C. Principles of Behavior. New York, USA: Appleton-Century-Crofts; 1943. [Google Scholar]

[pone.0287954.ref005] 5.Kool W, McGuire JT, Rosen ZB, Botvinick MM. Decision Making and the Avoidance of Cogntive Demand. J Exp Psychol Gen. 2010;139(4):665–82. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref006] 6.Bustos B, Colvett JS, Bugg JM, Kool W. Humans do not avoid reactively implementing cognitive control [Internet]. 2023. [cited 2023 Jun 13]. Available from: doi: 10.31234/osf.io/x5vgj [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref007] 7.Schouppe N, Ridderinkhof KR, Verguts T, Notebaert W. Context-specific control and context selection in conflict tasks. Acta Psychol (Amst). 2014. Feb 1;146:63–6. doi: 10.1016/j.actpsy.2013.11.010 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref008] 8.Sayalı C, Badre D. Neural systems of cognitive demand avoidance. Neuropsychologia. 2019;123(October 2017):41–54. doi: 10.1016/j.neuropsychologia.2018.06.016 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref009] 9.Sayali C, Rubin-McGregor J, Badre D. Policy abstraction as a predictor of cognitive effort [Internet]. PsyArXiv; 2022. Jun [cited 2023 Jun 13]. Available from: https://osf.io/by7gk [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref010] 10.Vogel TA, Savelson ZM, Otto AR, Roy M. Forced choices reveal a trade-off between cognitive effort and physical pain. eLife. 2020. Nov 17;9:e59410. doi: 10.7554/eLife.59410 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref011] 11.Cohen JD. Cognitive Control. In: The Wiley Handbook of Cognitive Control [Internet]. John Wiley & Sons, Ltd; 2017. [cited 2023 Jun 12]. p. 1–28. Available from: https://onlinelibrary.wiley.com/doi/abs/10.1002/9781118920497.ch1 [Google Scholar]

[pone.0287954.ref012] 12.Miller EK, Cohen JD. An integrative theory of prefrontal cortex function. 2001;167–202. [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref013] 13.Shenhav A, Musslick S, Lieder F, Kool W, Griffiths TL, Cohen JD, et al. Toward a Rational and Mechanistic Account of Mental Effort. Annu Rev Neurosci. 2017;40(1):annurev-neuro-072116-031526. doi: 10.1146/annurev-neuro-072116-031526 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref014] 14.Botvinick MM, Huffstetler S, McGuire JT. Effort discounting in human nucleus accumbens. Cogn Affect Behav Neurosci. 2009; doi: 10.3758/CABN.9.1.16 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref015] 15.Massar SAA, Libedinsky C, Weiyan C, Huettel SA, Chee MWL. Separate and overlapping brain areas encode subjective value during delay and effort discounting. NeuroImage. 2015. Oct 15;120:104–13. doi: 10.1016/j.neuroimage.2015.06.080 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref016] 16.Salamone JD, Correa M. The Mysterious Motivational Functions of Mesolimbic Dopamine. Neuron. 2012;76(3):470–85. doi: 10.1016/j.neuron.2012.10.021 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref017] 17.Leknes S, Lee M, Berna C, Andersson J, Tracey I. Relief as a Reward: Hedonic and Neural Responses to Safety from Pain. PLOS ONE. 2011. Apr 7;6(4):e17870. doi: 10.1371/journal.pone.0017870 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref018] 18.Porreca F, Navratilova E. Reward, motivation, and emotion of pain and its relief. Pain. 2017. Apr;158 Suppl 1(Suppl 1):S43–9. doi: 10.1097/j.pain.0000000000000798 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref019] 19.Hajcak G, Holroyd CB, Moser JS, Simons RF. Brain potentials associated with expected and unexpected good and bad outcomes. Psychophysiology. 2005;42(2):161–70. doi: 10.1111/j.1469-8986.2005.00278.x [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref020] 20.Gheza D, Paul K, Pourtois G. Dissociable effects of reward and expectancy during evaluative feedback processing revealed by topographic ERP mapping analysis. Int J Psychophysiol. 2018;132:213–25. doi: 10.1016/j.ijpsycho.2017.11.013 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref021] 21.Klein ED, Bhatt RS, Zentall TR. Contrast and the justification of effort. Psychon Bull Rev. 2005;12(2):335–9. doi: 10.3758/bf03196381 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref022] 22.Proudfit GH. The reward positivity: From basic research on reward to a biomarker for depression. Psychophysiology. 2015;52(4):449–59. doi: 10.1111/psyp.12370 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref023] 23.Webb CA, Auerbach RP, Bondy E, Stanton CH, Foti D, Pizzagalli DA. Abnormal neural responses to feedback in depressed adolescents. J Abnorm Psychol. 2017;126(1):19–31. doi: 10.1037/abn0000228 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref024] 24.Cohen MX, Elger CE, Ranganath C. Reward expectation modulates feedback-related negativity and EEG spectra. NeuroImage. 2007;35(2):968–78. doi: 10.1016/j.neuroimage.2006.11.056 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref025] 25.Marco-Pallares J, Cucurell D, Cunillera T, García R, Andrés-Pueyo A, Münte TF, et al. Human oscillatory activity associated to reward processing in a gambling task. Neuropsychologia. 2008;46(1):241–8. doi: 10.1016/j.neuropsychologia.2007.07.016 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref026] 26.Mas-herrero E, Ripollés P, Hajihosseini A, Rodríguez-fornells A, Marco-pallarés J. NeuroImage Beta oscillations and reward processing: Coupling oscillatory activity and hemodynamic responses. NeuroImage. 2015;119:13–9. [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref027] 27.Botvinick MM, Braver TS. Motivation and Cognitive Control: From Behavior to Neural Mechanism. Annu Rev Psychol. 2015;66(1):83–113. doi: 10.1146/annurev-psych-010814-015044 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref028] 28.Westbrook A, Braver TS. Dopamine Does Double Duty in Motivating Cognitive Effort. Neuron. 2016;89(4):695–710. doi: 10.1016/j.neuron.2015.12.029 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref029] 29.Kurzban R, Duckworth A, Kable JW, Myers J. An opportunity cost model of subjective effort and task performance. Behav Brain Sci. 2013;36(06):661–79. doi: 10.1017/S0140525X12003196 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref030] 30.Vassena E, Silvetti M, Boehler CN, Achten E, Fias W, Verguts T. Overlapping neural systems represent cognitive effort and reward anticipation. PLoS ONE. 2014;9(3). doi: 10.1371/journal.pone.0091008 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref031] 31.Bogdanov M, Nitschke JP, LoParco S, Bartz JA, Otto AR. Acute Psychosocial Stress Increases Cognitive-Effort Avoidance. Psychol Sci. 2021. Sep 1;32(9):1463–75. doi: 10.1177/09567976211005465 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref032] 32.Pizzagalli DA. Depression, stress, and anhedonia: toward a synthesis and integrated model. Annu Rev Clin Psychol. 2014;10(January):393–423. doi: 10.1146/annurev-clinpsy-050212-185606 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref033] 33.Cacioppo JT, Petty RE. The need for cognition. J Pers Soc Psychol. 1982;42:116–31. [Google Scholar]

[pone.0287954.ref034] 34.Zerna J, Strobel A, Strobel A. The role of Need for Cognition in well-being–Review and meta-analyses of associations and potentially underlying mechanisms [Internet]. PsyArXiv; 2021. [cited 2023 Jun 4]. Available from: https://psyarxiv.com/p6gwh/ [Google Scholar]

[pone.0287954.ref035] 35.Cacioppo JT, Petty RE, Feinstein JA, Jarvis WBG. Dispositional differences in cognitive motivation: The life and times of individuals varying in need for cognition. Psychol Bull. 1996;119(2):197–253. [Google Scholar]

[pone.0287954.ref036] 36.Yee DM, Leng X, Shenhav A, Braver TS. Aversive motivation and cognitive control. Neurosci Biobehav Rev. 2022. Feb;133:104493. doi: 10.1016/j.neubiorev.2021.12.016 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref037] 37.Hajcak G, Moser JS, Holroyd CB, Simons RF. It’s worse than you thought: The feedback negativity and violations of reward prediction in gambling tasks. Psychophysiology. 2007;44(6):905–12. doi: 10.1111/j.1469-8986.2007.00567.x [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref038] 38.Vassena E, Cobbaert S, Andres M, Fias W, Verguts T. Unsigned value prediction-error modulates the motor system in absence of choice. NeuroImage. 2015;122:73–9. doi: 10.1016/j.neuroimage.2015.07.081 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref039] 39.Vassena E, Deraeve J, Alexander WH. Task-specific prioritization of reward and effort information: Novel insights from behavior and computational modeling. Cogn Affect Behav Neurosci. 2019; doi: 10.3758/s13415-018-00685-w [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref040] 40.Vassena E, Gerrits R, Demanet J, Verguts T, Siugzdaite R. Anticipation of a mentally effortful task recruits Dorsolateral Prefrontal Cortex: An fNIRS validation study. Neuropsychologia. 2018;123(October 2017):106–15. doi: 10.1016/j.neuropsychologia.2018.04.033 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref041] 41.Russell JA. A circumplex model of affect. J Pers Soc Psychol. 1980;39(6):1161–78. [Google Scholar]

[pone.0287954.ref042] 42.Cacioppo J. T., Petty R. E., Kao C. F. Need for Cognition Scale. Measurement Instrument Database for the Social Science. [Internet]. 2013. [cited 2023 May 23]. Available from: Retrieved from www.midss.org [Google Scholar]

[pone.0287954.ref043] 43.Psychology Software Tools, Inc. [E-Prime 2.0] [Internet]. Sharpsburg, PA: Psychology Software Tools, Inc.; 2008. Available from: https://support.pstnet.com/.

[pone.0287954.ref044] 44.Rouder JN, Morey RD, Verhagen J, Swagman AR, Wagenmakers EJ. Bayesian analysis of factorial designs. Psychol Methods. 2017;22(2):304–21. doi: 10.1037/met0000057 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref045] 45.Schindler S, Schettino A, Pourtois G. Electrophysiological correlates of the interplay between low-level visual features and emotional content during word reading. Sci Rep. 2018; doi: 10.1038/s41598-018-30701-5 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref046] 46.R Core Team. R. R Core Team. 2017.

[pone.0287954.ref047] 47.Morey RD, Rouder JN. Bayesfactor: Computation of Bayes factors for common designs. R package version 0.9.12–2. BayesFactor Comput Bayes Factors Common Des. 2015; [Google Scholar]

[pone.0287954.ref048] 48.Bürkner PC. brms: An R Package for Bayesian Multilevel Models Using Stan. J Stat Softw. 2017;80(1):1–28. [Google Scholar]

[pone.0287954.ref049] 49.Morey RD. Confidence Intervals from Normalized Data: A correction to Cousineau (2005). Tutor Quant Methods Psychol. 2008. Sep 1;4(2):61–4. [Google Scholar]

[pone.0287954.ref050] 50.Kool W, Botvinick M. Mental labour. Nat Hum Behav. 2018;2(12):899–908. doi: 10.1038/s41562-018-0401-9 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref051] 51.Devine S, Vassena E, Otto AR. More than a feeling: physiological measures of affect index the integration of effort costs and rewards during anticipatory effort evaluation. Cogn Affect Behav Neurosci [Internet]. 2023. Apr 14 [cited 2023 Jun 13]; Available from: https://link.springer.com/10.3758/s13415-023-01095-3 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref052] 52.Kool W, Gershman SJ, Cushman FA. Cost-Benefit Arbitration Between Multiple Reinforcement-Learning Systems. Psychol Sci. 2017;28(9):1321–33. doi: 10.1177/0956797617708288 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref053] 53.Kool W, Shenhav A. Cognitive Control as Cost‐Benefit Decision Making. 2017; [Google Scholar]

[pone.0287954.ref054] 54.Padmala S, Pessoa L. Reward Reduces Conflict by Enhancing Attentional Control and Biasing Visual Cortical Processing. J Cogn Neurosci. 2011. Nov 1;23(11):3419–32. doi: 10.1162/jocn_a_00011 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref055] 55.Treadway MT, Buckholtz JW, Schwartzman AN, Lambert WE, Zald DH. Worth the “EEfRT”? The effort expenditure for rewards task as an objective measure of motivation and anhedonia. PLoS ONE. 2009;4(8):1–9. doi: 10.1371/journal.pone.0006598 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref056] 56.Westbrook A, Kester D, Braver TS. What Is the Subjective Cost of Cognitive Effort? Load, Trait, and Aging Effects Revealed by Economic Preference. PLoS ONE. 2013;8(7):1–8. doi: 10.1371/journal.pone.0068210 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref057] 57.Westbrook A, Van Den Bosch R, Määttä JI, Hofmans L, Papadopetraki D, Cools R, et al. Dopamine promotes cognitive effort by biasing the benefits versus costs of cognitive work. Science. 2020. Mar 20;367(6484):1362–6. doi: 10.1126/science.aaz5891 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref058] 58.Inzlicht M, Shenhav A, Olivola CY. The Effort Paradox: Effort Is Both Costly and Valued. Trends Cogn Sci. 2018;22(4):337–49. doi: 10.1016/j.tics.2018.01.007 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref059] 59.McGuire JT, Botvinick MM. Prefrontal cortex, cognitive control, and the registration of decision costs. Proc Natl Acad Sci. 2010;107(17):7922–6. doi: 10.1073/pnas.0910662107 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref060] 60.Vassena E, Deraeve J, Alexander WH. Predicting Motivation: Computational Models of PFC Can Explain Neural Coding of Motivation and Effort-based Decision-making in Health and Disease. J Cogn Neurosci. 2017;29(10):1633–45. doi: 10.1162/jocn_a_01160 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref061] 61.Thompson EP, Chaiken S, Hazlewood JD. Need for cognition and desire for control as moderators of extrinsic reward effects: a person x situation approach to the study of intrinsic motivation. J Pers Soc Psychol. 1993. Jun;64(6):987–99. doi: 10.1037//0022-3514.64.6.987 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref062] 62.Clay G, Mlynski C, Korb FM, Goschke T, Job V. Rewarding cognitive effort increases the intrinsic value of mental labor. Proc Natl Acad Sci. 2022. Feb;119(5):e2111785119. doi: 10.1073/pnas.2111785119 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287954.ref063] 63.Lin H, Westbrook A, Fan F, Inzlicht M. An experimental manipulation increases the value of effort [Internet]. PsyArXiv; 2021. Feb [cited 2023 Jun 13]. Available from: https://osf.io/gnk4m [Google Scholar]

[pone.0287954.ref064] 64.Kool W, Botvinick M. The intrinsic cost of cognitive control. Behav Brain Sci. 2013. Dec;36(6):697–8. doi: 10.1017/S0140525X1300109X [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref065] 65.Gärtner A, Grass J, Wolff M, Goschke T, Strobel A, Strobel A. No relation of Need for Cognition to basic executive functions. J Pers. 2021. Dec;89(6):1113–25. doi: 10.1111/jopy.12639 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref066] 66.Bogdanov M, Renault H, LoParco S, Weinberg A, Otto AR. Cognitive effort exertion enhances electrophysiological responses to rewarding outcomes. Cereb Cortex. 2022. Sep 19;32(19):4255–70. doi: 10.1093/cercor/bhab480 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref067] 67.Ma Q, Meng L, Wang L, Shen Q. I endeavor to make it: Effort increases valuation of subsequent monetary reward. Behav Brain Res. 2013;261:1–7. doi: 10.1016/j.bbr.2013.11.045 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref068] 68.Norton MI, Mochon D, Ariely D. The IKEA effect: When labor leads to love. J Consum Psychol. 2012. Jul 1;22(3):453–60. [Google Scholar]

[pone.0287954.ref069] 69.Bandura A. Self-efficacy: toward a unifying theory of behavioral change. Psychol Rev. 1977. Mar;84(2):191–215. doi: 10.1037//0033-295x.84.2.191 [DOI] [PubMed] [Google Scholar]

[pone.0287954.ref070] 70.Ryan RM, Deci EL. Intrinsic and Extrinsic Motivations: Classic Definitions and New Directions. 2000;67:54–67. [DOI] [PubMed] [Google Scholar]

PERMALINK

Need for cognition moderates the relief of avoiding cognitive effort

Davide Gheza

Wouter Kool

Gilles Pourtois

Roles

Abstract

Introduction

Methods

Participants

Stimuli and task

Fig 1. Overview of the task and trial structure.

Data analysis

Statistical analyses

Results

Experiment 1

Fig 2. Experiment 1.

Table 1. Bayesian model comparison (Experiment 1).

Experiment 2

Fig 3. Experiment 2.

Table 2. Bayesian model comparison (Experiment 2).

Table 3. Posterior parameter estimation summary (Experiment 2).

Fig 4. Results of the Bayesian parameter estimation for model 5.

Fig 5. Ratings of reward-related feedback that followed the gambling task.

Fig 6. Effect of NFC on arithmetic performance.

Fig 7. Ratings of performance feedback that followed the effort task, by difficulty level (easy or hard), and affective dimension (pleasure, reverse-scored frustration, and relief).

Discussion

Supporting information

Data Availability

Funding Statement

References

Decision Letter 0

Poppy Watson

Roles

Author response to Decision Letter 0

Decision Letter 1

Poppy Watson

Roles

Author response to Decision Letter 1

Decision Letter 2

Poppy Watson

Roles

Acceptance letter

Poppy Watson

Roles

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases