It’s About Time! ⁠The Sense of Agency Increases After Spontaneous vs. Planned Actions but It Decreases After Externally Planned Actions

Imagine two people going to a cinema to watch a movie. One person makes a spontaneous decision about what to watch only when they arrive at the movie theater. The second person made this decision a week earlier, and at this point, they only need to collect the tickets they prebooked. Will these two people have the same agentic experience when the movie begins? If not, will an action (i.e., purchasing a ticket) that follows a preplanned schedule result in a stronger or weaker sense of agency compared to that for a spontaneous action?

Generally, people experience agency when they feel in control of their actions and can influence what happens in the outside world (Haggard & Chambon, 2012). Those actions can range from simple movements, such as taking a step or clicking a button, to complex ones, such as having a conversation or making a life-changing decision. Acknowledgement of individual control creates the sense of agency (Haggard & Tsakiris, 2009), and maintaining that feeling is a core motivation (Burger, 1992; Fiske, 2004; Skinner, 1996). Despite being a common human experience, the sense of agency has been studied and operationalized in a variety of ways in different research areas.

For instance, research in cognitive psychology and neuroscience has predominantly focused on actions with immediate outcomes (e.g., Haering & Kiesel, 2014; Moore et al., 2012; Sidarus et al., 2017). In a typical experiment, participants are asked to perform a task (e.g., clicking a button), which results in a certain consequence (e.g., a sound). They subsequently rate the extent to which they feel they caused the consequence. This simple example shows that the sense of agency is inherently related to goal pursuit because it involves a comparison between the intended and the actual outcome (Blakemore et al., 2002). In other words, it arises when a person believes that they caused what they aimed to cause. Moreover, studies have shown that rewards associated with outcomes play an important role in generating a sense of agency (Di Costa et al., 2018; Majchrowicz et al., 2020; Takahata et al., 2012), which further attests to the motivational nature of this construct.

Some actions are indeed spontaneous, performed shortly after the intention is formed, and have consequences that are experienced nearly immediately. However, other actions are planned in advance, and there is a delay between the intention and the behavior. Past research has demonstrated the important consequences of different phases of goal pursuit on psychological functioning (Gollwitzer, 2012). Currently, it is unclear how, if at all, the temporal dynamics of setting an intention impacts the sense of agency (Pacherie, 2015). The current research aims to address this issue and investigate the role of the delay between forming an intention and experiencing outcomes of one’s actions on the sense of agency. After reviewing the inconsistent and scarce evidence for the role of planning in forming a sense of agency, we describe two studies that addressed these inconsistencies.

One piece of evidence that suggests that planned action might result in a stronger sense of agency is the research of Vinding and colleagues (Vinding et al., 2013, 2015). In their studies, participants were asked to form an intention and either realize it immediately or wait before doing so. Results revealed a greater sense of agency for delayed than for immediate intentions, prompting the authors to conclude that setting an intention in advance plays a prominent role in agentic experiences. Several theories can be used to support this interpretation.

One theoretical explanation focuses on the congruence between expected and actual outcomes, which has been shown to increase the sense of agency (Sato & Yasuda, 2005). When an action is executed according to the plan, it should be easier to detect congruence and infer agency. One reason for that would be that when a person plans an action in advance they should have a clearer representation of what this action will look like. Planning would be therefore related to making more specific predictions and forming specific motor plans aimed at fulfilling such predictions. This expectation is in line with the ideomotor theories of action control, according to which action preparation produces an internal response similar to the one when actual effects are perceived (e.g., Shin et al., 2010). Whereas these theoretical predictions have been primarily tested in the context of simple actions, similar expectations could be made for higher-order goals and actions directed at achieving them. Due to a better elaboration of one’s expectation, as long as an action is executed according to a cognitively accessible plan (Wen et al., 2015), one should be more likely to establish intention-action congruence. In contrast, spontaneous action should come with a less elaborate representation of action, which should result in a less clear-cut comparison, and a lower sense of agency as a consequence.

Moreover, if forming an intention primes a subsequent action, then setting an intention could increase the fluency of the action. Stronger fluency might then account for a higher sense of agency after planned vs. unplanned actions. This reasoning is consistent with several studies that revealed a positive association between priming and the sense of agency (Chambon & Haggard, 2012; Sidarus et al., 2013). If explicit cues such as plans work in a similar way as implicit cues, they could positively affect the fluency of action and therefore increase the sense of agency.

These two mechanisms—congruence and fluency—would then explain why planned actions might be linked to a stronger agentic experience compared to spontaneous actions. However, several theoretical arguments suggest that the sense of agency could in fact diminish when an action is planned in advance compared to taking spontaneous action.

First, being aware of one’s commitment to previous choices—even when they are made by oneself—may imply that the agent is constrained and no longer able to choose their actions. A restricted range of choices, as well as an external source of choice, have been shown to impact agency negatively (e.g., Barlas & Obhi, 2013; Sidarus et al., 2017). By making decisions in advance, one inevitably narrows the potential range of future choices. Past research suggests that, at least in Western cultures, choice is the mark of independence and autonomy (Markus & Schwartz, 2010). Moreover, building on classic experimental studies (Libet et al., 1983), spontaneity has been suggested to be a crucial aspect of volitional actions (Haggard, 2019), serving as a basic cue for the sense of agency. That is, deliberating and choosing an action could be an empowering and agentic experience in itself. In contrast, following through on a past choice— as one does when one executes a previous plan— could induce feelings of being constrained. Thus, when one enacts a planned action, even though it is one’s past self who planned it, one may create a situation that resembles following someone’s orders, thus leading to a decreased sense of agency.

Second, forming a plan may result in the automatization of subsequent actions, which could reduce one’s awareness of the actions. Research on implementation intentions, which represent a specific form of planning, is relevant in this regard (Gollwitzer, 1993). This research shows that forming simple plans helps perform actions more smoothly but nevertheless decreases the awareness associated with the enactment of the behavior. If the sense of agency depends on awareness of the behavior (see Wegner, 2002), reducing awareness through planning and automatization could reduce the sense of agency as well. Consonant with these theoretical considerations are the results obtained by Damen and colleagues (Damen et al., 2015), who showed that prior planning reduced the feeling of agency, and a greater time gap between forming an intention and executing an action was related to a lower sense of agency.

This review shows that direct evidence of the relationship between planning and the sense of agency is scarce and restricted to a few studies. Moreover, the mixed results obtained in past studies could indicate that contradictory mechanisms may be involved. Given that many (if not most) human actions serve intentions formed in advance, the question concerning the consequences of such delays on the sense of agency is important.

Our goal was to address these shortcomings by investigating the role of planning in forming a sense of agency in the motivationally relevant context and examining possible mediators of the primary relationship. Based on the literature, we focused on two factors: 1) expectation-outcome congruence and 2) a feeling of having an unconstrained choice. Regarding the former, we expected the outcomes of planned and spontaneous actions to vary in terms of perceived congruence (and therefore surprising character). We predicted that people would perceive the outcomes of their actions as more congruent with their expectations after planned than spontaneous action. If that is the case they should report less surprise as a reaction to the obtained outcomes and they should attribute more agency to themselves. Regarding the latter, we expected that the more people would perceive their actions as being constrained and their choice as being restricted after planned (vs. spontaneous) action, the less agency they would attribute to themselves. The relative strength of these two processes would determine the total effect of planning on the sense of agency (see Fig. 1)

We tested these hypotheses in two studies designed as a registered report. In both studies, participants were asked to perform a planned (vs. unplanned) action, and their explicit sense of agency, as well as proposed meditators, were measured. To examine the sense of agency in motivationally relevant contexts, actions that participants performed during the experiment had meaningful and gratifying consequences. Specifically, in both studies, participants were asked to choose the images they wanted to watch later. In the planned condition, they chose the categories of pictures in advance. In the unplanned condition, they made this decision spontaneously immediately before seeing the picture. Additionally, we explored the role of fluency for sense of agency by measuring perceived difficulty of making decisions in both conditions.

Study 1 was conducted online using a between-subjects design. Here, we also explored the role of personality traits that could moderate the effects of planning on the sense of agency. Two parts of the study (i.e., action planning and execution) were separated by a few days break. Even though such a long break is rarely seen in cognitive psychology experiments, it fits well into the motivational framework. Moreover, such a delay may better resemble a typical situation of plan-making (e.g., making a to-do list for an upcoming week). In order to address the possible problem of forgetting their past choices, we reminded people of their choices at the beginning of each trial and we also measured how well they remember their decisions. Study 2 was conducted in the lab using a within-subjects design. It additionally included an implicit measure of agency (i.e., a temporal binding task).

Finally, both studies included a theoretically intriguing scenario in which a person was asked to execute an action randomly designated for them by an external source (i.e., a computer). On the surface, such an externally planned condition should evoke a limited sense of agency because a person following instructions has little (if any) choice. However, classic studies on obedience in social psychology (Milgram, 1963) demonstrate that people do take responsibility for following orders and experience relevant emotional reactions. Purely involuntary action has been found to decrease implicit agency (e.g., Moore et al., 2009). The proposed mechanism of choice constraints is in line with past studies on following orders or instructions that led to reduced implicit and explicit agency (Barlas et al., 2018; Caspar et al., 2016; Wenke et al., 2010), although this effect was not always observed (e.g., Sidarus & Haggard, 2016). Regarding the two mechanisms we aimed to test, even though externally planned actions should result in some expectation-outcome congruence, as participants knew what to expect, their feeling of having an unconstrained choice would be substantially reduced, as would be their explicit sense of agency.

Power Analysis. Our model included three experimental conditions as an independent variable and two parallel mediators affecting our dependent variables. We assumed that the smallest effect of interest of all paths in the model (i.e., from experimental manipulation to mediators and from mediators to the outcome variables) is 0.25 in standardized units. Using MedPower (Kenny, 2017) ¹ the estimated sample size given 0.9 power was 206 participants, which meant 103 per experimental condition for the similar model with two conditions. Although our main hypotheses were formulated each time as comparisons between two conditions, we planned to include three such conditions altogether. Given that we were not aware of the tool that would allow us to assess power for such a complex mediation model, to simplify it we decided to recruit at least 310 participants. However, given that it was just a rough estimation we decided to increase the number of participants in this study and recruit 400 Polish-speaking participants from the general population with usable data (i.e., after exclusions) through an online research panel. The company recruited more participants for the first part of the study assuming that some would not participate in the second part.

Exclusions. In line with the preregistered conditions, we excluded participants who failed our attentional checks (N = 17). We also planned to exclude those with a substantial proportion of missing values (>50% of unanswered items in the scales), but as the responses were eventually forced, that was not an issue. Thus, the ultimate sample included 455 participants (220 women, 234 men, one person of other identity; M_age = 45.25, SD = 15.82). In the second part, we also excluded trials in which participants did not adhere to the self-planned and externally planned choices. However, most participants complied with the preselected choices and the median level of obedience was equal to 20 out of 20 possible choices with only two participants who did not comply in more than 50% of their choices. For regression models, we used Cook’s distance > 0.5 as an exclusion criterion, but no outliers were detected on the basis of this criterion.

Ethics Approval. The project was approved by the Ethics Committee of the Institute of Psychology, Jagiellonian University.

This study was run online in two parts separated by a few days, thus creating a delay between planning and action in one of the conditions. All participants were told that the goal of the study was to explore the relationship between personality traits and the evaluation of pictures, and therefore, the study had to be divided into two parts. After completing the questionnaires related to personality traits (described below), participants were randomly assigned to one of three conditions. In all conditions, participants were presented with four subsequent screens. Each screen contained five sets of choices between two categories of pictures (e.g., “Entertainment” vs. “Animals”).

Participants in the planned condition were asked to choose and indicate which category in each set they wanted to see in the second part of the study (e.g., “Entertainment” from the first pair of choices, “Animals” from the second pair, etc.). Participants in the unplanned condition were presented with the same options, but they were told that they would be able to choose the categories in the second part of the study. For participants in the externally planned condition, one category in each set was marked as preselected by the computer for them; they were told that they would see the corresponding selection of pictures in the second part of the study. In total, there were eight categories of pictures in the experiment, and in unplanned and planned conditions, participants eventually made 20 choices (see Fig. 2 for an outline).

Seven days later participants were recontacted to complete the second part of the study. Participants were presented with 20 sets of categories that they had seen in the first part. Each set was displayed on a separate screen. In the planned condition, participants were reminded of their earlier choice (i.e., “For this trial you chose __”). In the unplanned condition, they were given a choice to make a spontaneous decision (i.e., “For this trial you can choose between __”). In the externally planned condition, they were reminded of the selection that had been made for them by the computer in the earlier part (i.e., “For this trial ___ was selected for you.”). Participants were asked to indicate one of two categories. After a participant had indicated one of the categories, they saw a picture from that category for up to 5 seconds (they could have exited earlier if they wanted, but the picture disappeared after that time). Next, participants evaluated how much they liked the picture. After evaluating all 20 pictures, they were asked questions regarding their sense of agency, the feeling of having an unconstrained choice, expectation-outcome congruence, and fluency. To address the issue of possible forgetting, we also asked participants a question about remembering their past decisions or the assigned choices.

Stimuli for this and the following study were taken from the open-access database OASIS (Kurdi et al., 2017). We selected the pictures of moderately positive valence (4 - 6 on the scale 1 - 7). The list of selected pictures and categories can be found in the Supplementary Online Materials (SOM).

Expectation-Outcome Congruence. The perception of congruence (vs. incongruence) between expectations and outcome was measured with the following three items (Cronbach’s α = .43²): “The pictures I saw were consistent with my expectations,” “I was surprised by what pictures appeared (reverse-coded),” and “I had expected to see very different pictures than those that appeared (reverse-coded).”

Feeling of Having an Unconstrained Choice. The perception of having an unconstrained (vs. restricted) choice when making a decision was measured with the following three items (α = .77): “I felt free to choose pictures that appeared in each trial,” “It was my autonomous decision to choose the pictures that I saw,” and “I did not have much choice (reverse-coded).”

Explicit Sense of Agency. The explicit sense of agency was measured with the following four items (α = .84): “It was up to me what pictures appeared,” “I had no impact on what pictures appeared (reverse-coded),” “It was me who controlled what pictures appeared,” and “My actions caused which pictures appeared.” This measure was a modification of a scale used by Majchrowicz and Wierzchoń (2018).

Participants responded to all the questions above on a scale from 1 (Definitely disagree) to 7 (Definitely agree). We calculated means for each scale and used the means as variable indices.

Pictures Evaluation. More/less favorable perception of pictures was measured after every single display, with the question “How much did you like this picture?” Participants responded on a scale from 1 (Not at all) to 7 (Very much). We calculated the mean of all evaluation items (α = .91).

Fluency. One item was used to measure subjective ease in action selection: “Choosing the pictures today was difficult (1) – easy (7)” (Graf et al., 2018) 

Forgetting. Memory of previous choices for planned and externally planned conditions was assessed with one question: “How well do you remember the choices of categories of pictures that you made / were presented with a week ago?” Participants responded on a scale from 1 (Not at all) to 7 (Very well).³

Attention Checks. We included two attention checks, one in each part of the study. Participants were asked to mark a given answer (i.e., “If you’re reading this, please mark number two”).

Individual Differences. To test the moderating role of individual differences, we measured the following variables: self-control, a tendency to initiate new commitments, and the need for cognitive closure. All items are included in SOM.

Brief Self-Control. This 13-item scale (α = .84) measured individuals’ ability to override or change their inner responses when necessary, as well as to interrupt undesired behavioral tendencies and refrain from acting on them (Tangney et al., 2004).

Need for Closure (NFC). This 15-item scale (α = .57) measured individuals’ desire to seek certainty and closure and to avoid ambiguity and uncertainty (Kossowska et al., 2012).

Tendency to Initiate New Goals. This 9-items subscale captured whether a person enjoyed initiating new goals and did so frequently (α = .90; Jasko et al., 2020). Another subscale measuring maintenance tendencies was administered but not utilized in this project.

To test our primary hypotheses (described below in frequentist terms), we ran the mediation models with experimental conditions as a predictor, expectations-outcome congruence and feeling of having an unconstrained choice as parallel mediators, and the explicit sense of agency as an outcome variable.

Mediation via Expectation-Outcome Congruence (Unplanned vs. Planned). We predicted that, compared with planned actions, unplanned actions would be related to the lower expectation-outcome congruence (H1a), which would be positively related to the explicit sense of agency (H1b). Overall, we predicted a significant indirect effect from the experimental condition (unplanned vs. planned) through the expectation-outcome congruence to a lower explicit sense of agency (H1c).

Mediation via Feeling of Having an Unconstrained Choice (Unplanned vs. Planned). We predicted that, compared with planned actions, unplanned actions would be related to the higher feeling of having an unconstrained choice (H2a), which would be positively related to the explicit sense of agency (H2b). Overall, we predicted a significant indirect effect from the experimental condition through the feeling of having an unconstrained choice to a greater explicit sense of agency (H2c).

Mediation via Expectation-Outcome Congruence and Feeling of Having an Unconstrained Choice (Externally Planned vs. Planned or Unplanned). We predicted that externally planned actions, compared to both planned and unplanned actions, would be related to the lowest explicit sense of agency, due to the lowest feeling of having an unconstrained choice (H3a). However, we did not expect expectation-outcome congruence to differ when comparing externally planned actions with the planned condition (H3b).

Pictures Evaluation. Finally, we expected that the sense of agency would be correlated with evaluation of stimuli (because the sense of agency is a positive feeling). Therefore, we expected the effects of the manipulation on evaluation to mirror those on the sense of agency, described above (H4).

To test the above hypotheses, we used two alternative methods of estimation: frequentist and Bayesian, fitting regression models with the explicit sense of agency and evaluation as dependent variables and the described mediators. To obtain frequentist estimates we used the PROCESS macro (Hayes, 2018), version 4.3.1, designed for testing complex mediation and moderation models. We used 95% confidence intervals to draw inferences. We considered CIs that do not contain zero as indicators of presence of the tested effects.

To test the above hypotheses within the Bayesian framework, we fitted the Bayesian multivariate linear model using the brms package (Bürkner, 2018; Carpenter et al., 2017) in R (R Core Team, 2020). We used default priors. In order to assess convergence we verified if the Ȓ was close to 1 and the number of effective samples was at least 10% of the number of post-warmup samples. We used 95% quantile-based Bayesian credible intervals to draw inferences. We considered CIs that do not contain zero as indicators of presence of the tested effects.

Individual Differences. While our primary goal was to examine the effects of planning on the sense of agency, we explored the moderating role of individual differences. We predicted that people who felt a high degree of self-control would feel less constrained by their past choices and might even enjoy being able to execute exactly what they had planned in advance. We made the same prediction for individuals who scored high on NFC, who enjoy maintaining their past choices and prefer predictability and consistency over novelty and surprise. In contrast, those who score high on a tendency to initiate new goals might feel more constrained by their past choices and would experience more negative effects of planning on sense of agency.

We predicted that the effect of planning on the explicit sense of agency would be moderated by individual differences. Namely, the effect of unplanned (vs. planned) actions should be less positive or more negative among those who scored high on self-control and NFC but more positive or less negative among those who had a high tendency to initiate new goals (H5). To test these hypotheses, we ran moderation analyses with the experimental condition as a predictor, individual differences as separate moderators, and the explicit sense of agency as a dependent variable.

Fluency. In addition to measuring the congruence between expectations and outcomes, we included one item to measure processing fluency. We expected this variable to play a similar role to perceived congruence and thus to be highly correlated with it. Therefore, we expected the mediations through fluency to mirror those through congruence (H6). To test these hypotheses, we ran the mediation model with the experimental condition as a predictor, fluency and the feeling of having an unconstrained choice as parallel mediators, and the explicit sense of agency and the pictures evaluation as outcome variables. We also correlated fluency and congruence.

To establish the direct and indirect effects of planning manipulation on the explicit sense of agency, we conducted a parallel mediation analysis using PROCESS (Model 4; Hayes, 2018; see Figure 1). Experimental conditions were dummy-coded. This resulted in two main comparisons (unplanned vs. planned, externally planned vs. planned). The expectation-outcome congruence and the feeling of having an unconstrained choice were parallel mediators. Indirect effects were established with 10,000 bootstrap samples. The same analyses were conducted for the second outcome variable (i.e., pictures evaluation).

Explicit Sense of Agency. The total effect of unplanned (vs. planned) condition on the sense of agency was positive and significant (b = 0.44, 95% CI [0.11, 0.78], p = .009). The analysis of the effects of manipulation on the mediators showed that whereas participants in the unplanned (vs. planned) condition did not experience lower expectation-outcome congruence (b = -0.10, 95% CI [-0.34, 0.15], p = .451), they reported a stronger feeling of having an unconstrained choice (b = 0.56, 95% CI [0.21, 0.91], p = .002). Expectation-outcome congruence was not significantly related to the sense of agency (b = 0.03, 95% CI [-0.08, 0.13], p = .632) but the feeling of having an unconstrained choice was positively related to it (b = 0.53, 95% CI [0.45, 0.60], p < .001). When both mediators were included in the model, the effect of unplanned (vs. planned) condition became nonsignificant (b = 0.15, 95% CI [-0.13, 0.43], p = .298). The indirect effect through expectation-outcome congruence was not significant (b = -0.00, 95% CI [-0.03, 0.01]) but the indirect effect through the feeling of having an unconstrained choice was positive and significant (b = 0.30, 95% CI [0.11, 0.49]). In short, the hypotheses regarding the first mediator were not supported but the hypotheses regarding the second mediator were supported.

The total effect of the externally planned (vs. planned) condition was negative and significant (b = -1.08, 95% CI [-1.42, -0.74], p < .001). Even though participants in the externally planned condition did not experience lower expectation-outcome congruence (b = -0.11, 95% CI [-0.37, 0.14], p = .377), they reported having a more constrained choice (b = -0.97, 95% CI [-1.32, -0.61], p < .001) than those in the planned condition. When both mediators were included in the model, the effect of externally planned (vs. planned) condition remained significant (b = -0.56, 95% CI [-0.85, -0.27], p < .001). The indirect effect through congruence was not significant (b = -0.00, 95% CI [-0.03, 0.01]) but the indirect effect through the feeling of having an unconstrained choice was negative and significant (b = -0.51, 95% CI [-0.75, -0.30]) (see Figure 3).

When we repeated the analyses to obtain estimates for the comparison between externally planned vs. unplanned conditions, the results were even stronger (the details are presented in SOM).

Pictures Evaluation. Neither the total (b = -0.04, 95% CI [-0.26, 0.18], p = .728) nor the direct effect (b = -0.06, 95% CI [-0.28, 0.15], p = .581) of unplanned (vs. planned) condition on pictures evaluation was significant. However, both expectation-outcome congruence (b = 0.25, 95% CI [0.17, 0.33], p < .001) and feeling of having an unconstrained choice (b = 0.08, 95% CI [0.02, 0.14], p = .007) were positively related to pictures evaluation. The indirect effect of planning manipulation through congruence was not significant (b = -0.02, 95% CI [-0.09, 0.04]) but the indirect effect of planning manipulation through the feeling of having an unconstrained choice was positive and significant (b = 0.04, 95% CI [0.01, 0.10]).

When it comes to the externally planned (vs. planned) condition, neither the total effect (b = -0.20, 95% CI [-0.43, 0.03], p = .081) nor the direct effect (b = -0.10, 95% CI [-0.32, 0.13], p = .393) were significant. The indirect effect through congruence was not significant (b = -0.03, 95% CI [-0.10, 0.04]), but the indirect effect through the feeling of having an unconstrained choice was negative and significant (b = -0.08, 95% CI [-0.15, -0.02]) (see Figure 4). Again, when comparing the externally planned condition with the unplanned condition we obtained very similar results (see SOM).

Additionally, pictures evaluation was positively but not strongly correlated with the explicit sense of agency (Pearson’s r = .19, p < .001).

We conducted Bayesian analyses as an alternative test of our primary hypotheses. In short, the results of these analyses were very similar to those obtained using the frequentist analyses: the corresponding effects were significant and in the same direction. Additionally, estimates were close numerically. These results are presented in the SOM.

Individual Differences. To examine whether the impact of the planning manipulations on the explicit sense of agency was moderated by individual differences, we conducted a series of moderation analyses using PROCESS (Model 1; Hayes, 2018). We entered experimental conditions (dummy-coded) as a predictor and self-control, the need for cognitive closure, and the initiation tendencies as moderators (each in a separate model). In short, we did not find any of the variables to significantly moderate the impact of planning manipulation (for interactions, all ps >= .310). The detailed statistics are presented in SOM. For exploratory purposes, we also correlated individual differences with the explicit sense of agency, but the results were all nonsignificant (all ps >= .108).

Fluency. To examine the potential role of processing fluency as a mediator, we performed parallel mediation analyses identical to those reported earlier, except that the expectation-outcome congruence was replaced with processing fluency (the correlation between the two was Pearson’s r = .22, p < .001).

These analyses showed that planning manipulation was not significantly related to processing fluency (unplanned vs. planned: b = 0.04, 95% CI [-0.21, 0.29], p = .768; externally planned vs. planned: b = -0.12, 95% CI [-0.37, 0.14], p = .360). Processing fluency was not significantly related to the explicit sense of agency either (b = -0.04, 95% CI [-0.14, 0.06], p = .450). The indirect effect of planning manipulation through processing fluency was not significant (unplanned vs. planned: b = -0.00, 95% CI [-0.02, 0.02], externally planned vs. planned: b = 0.00, 95% CI [-0.01, 0.03]). The replacement of this mediator did not substantially change either the estimates for the other mediator (feeling of having an unconstrained choice), or the direct effects reported previously. In short, the results did not support the hypothesis that processing fluency would be responsible for the effects of planning on sense of agency.

Even though processing fluency was positively related to pictures evaluation (b = 0.15, 95% CI [0.07, 0.23], p < .001), the indirect effects of planning manipulation through processing fluency were not significant (unplanned vs. planned: b = 0.01, 95% CI [-0.03, 0.05], externally planned vs. planned: b = -0.02, 95% CI [-0.06, 0.02]). Again, the replacement of this mediator did not substantially change either the estimates for the other mediator (feeling of having an unconstrained choice), or the direct effects. The detailed statistics are presented in SOM.

In the first study, we found partial support for our hypotheses. First, we observed the total effect of the planning manipulation on the sense of agency. Specifically, in the condition where participants spontaneously decided about which pictures they wanted to see, they reported a higher sense of agency compared to the condition in which they had planned those decisions a week in advance. In contrast, when they executed actions according to an externally established plan, their sense of agency significantly decreased in comparison to other conditions. The effect of externally planned condition, while less surprising than the effect of spontaneous vs. planned decisions, is important as it validates the logic of our manipulation. Taken together, these results suggest that planning may ultimately decrease the sense of agency, and the degree of this decrease may be related to specific attributes of planning, such as the actor responsible for making the plan (self vs. others).

Second, among the two proposed mediators, only the feeling of having an unconstrained choice played a significant mediating role. Participants felt the least constrained when they chose pictures spontaneously and that resulted in their higher sense of agency, as well as better evaluation of pictures they saw. The opposite was true for participants who had their pictures selected by the computer. These results indicate that planning may restrict perceived freedom and that feeling may explain, at least to some extent, how agentic people feel in a given situation.

The other possible mediator, expectation-outcome congruence, was neither affected by the manipulation nor significantly related to the sense of agency. Intriguingly, expectation-outcome congruence showed a positive correlation with the evaluation of pictures, which suggests that it captured a meaningful psychological concept, though this construct was not influenced by the planning manipulation. It is important to note that the reliability of that measure was not satisfactory, which could have contributed to the lack of significant findings.

Importantly, we examined these processes within a context that held motivational relevance. Firstly, participants were given the opportunity to select pictures they wanted to view, mirroring decisions they make regularly, such as when scrolling through social media platforms. Secondly, unlike many experiments in which intention-action-outcome intervals are measured in mere seconds, we instructed participants to make a plan several days in advance. This time gap between their plans and actions may more closely resemble the typical situation of daily planning.

Regarding our secondary hypotheses, we did not find individual differences (i.e., self-control, the need for cognitive closure, the initiation tendencies) to moderate the impact of the manipulation on the explicit sense of agency. They were not correlated with it either. However, it is important to note that the Need for Closure measure did not reach satisfactory reliability levels, so inferences regarding it should be made with caution. Overall, while it could be the case that these personality characteristics play a larger role in the context of more personally meaningful goals, testing this hypothesis would require further studies. Fluency, used interchangeably as a second mediator instead of congruence, was hypothesized to play a similar role to it. It was not affected by the manipulation and neither was it related to the explicit sense of agency. Nonetheless, similarly to congruence, it positively predicted pictures evaluation.

In the next study, we continued examining our hypotheses in a different context, one in which a delay between planning and actions was reduced. We implemented a within-person design and conducted the experiment in the laboratory setting in order to obtain more precise estimations. In this study we did not measure individual differences, but we extended the measures of sense of agency to include the implicit measure frequently used in this research area, namely the interval estimation task. To enable comparison of the results between the studies, we kept the majority of other measures the same.

Given the sample size estimated in Study 1 and difficulties in performing power analysis for a within-subject mediation, we recruited 106 participants with normal or corrected-to-normal vision. It was one more person than planned in the preregistration, since we were unable to predict exactly the final attendance. They were recruited via the advertisement posted on social media. Participants consented to participate in writing prior to the experiment. Due to unforeseen technical issues (i.e., crashing of the procedure) we had to exclude eight participants.

To ensure good quality of data, we planned to exclude trials with a temporal estimate higher or lower than +/-3SD from each subject’s mean per interval per condition. However, we did not observe such trials. We also planned to exclude trials in which participants did not adhere to the planned choices (i.e., they either pressed a different key or did not press any key at all).⁴ There were 964 such trials in total out of 11760 (8%). Finally, we conducted a linear trend analysis with actual and reported intervals. We applied contrast coefficients -3, -1, 1, and 3 for the increasing delays. We performed a separate analysis for each participant, entering data points from three conditions altogether. We excluded nine participants whose variation in reported intervals did not follow the actual pattern of increase. After those exclusions there were 89 people in the final sample (71 women, 18 men; M_age = 23.54, SD = 5.10). ⁵

The study, main part of which lasted for about 30 minutes (a mean duration of the main task completion was 30.49 min), was run in a laboratory setting; we used MATLAB software, version 2018b. We implemented an additional, implicit measure of the sense of agency, namely the interval estimation task. In various versions of this task, participants are asked to assess temporal distance between the action (e.g., pressing the key) and the outcome of that action (e.g., tone or picture display). In our experiment, participants pressed the key and that action resulted in displaying the picture on the screen, which was accompanied by a tone. Each time, they were asked to report an interval between their action (i.e., pressing a key) and outcome (i.e., seeing the picture and hearing a tone) on the slider 0 – 1000ms. Intervals were randomly drawn from the pool of fixed values of 200, 400, 600, and 800ms (equally presented in each block), but participants were told they may mark any value between 0 and 1000ms, as none of the intervals would last longer than one second. In line with the interpretation used in past studies, differences between participants’ estimates and actual intervals were used as indicators of participants’ implicit sense of agency. Specifically, participants should report a compressed time interval between their action and an outcome when they experienced greater contingency between those two events, which is treated as an indicator of inferred causality related to a greater sense of agency.

At the beginning of each trial, participants saw a screen that displayed two labels (e.g., “Entertainment” vs. “Animals”), one on the left side and the other on the right side of the screen. Participants were instructed to choose one of two possible categories of pictures that they would see (the specific instructions differed between the experimental conditions; see next) and prepare to press the corresponding key. After the screen with two labels disappeared, a screen with a fixation cross appeared, and participants clicked with their dominant hand to indicate their choice of the category. Clicking resulted in the display of a picture from a chosen category, signalized by a neutral tone (1000 Hz), played for 100ms. After the picture disappeared, participants reported the interval between their action (i.e. pressing the key) and outcome (i.e. a picture and tone appearance) using the slider.

Planning Manipulation. At the beginning of each block, participants saw a list of choices to be made in that block, displayed on 4 screens with 10 choices on each screen (e.g., “Entertainment” vs. “Animals”). In the planned condition, participants were instructed to make a choice for each of 40 trials at the beginning of the block. In the unplanned condition, they were presented with the same options but they were not asked to plan anything at the beginning of the block. Instead, they were informed that they would be asked to make a decision at the beginning of each trial. In the externally planned condition, they were informed that they would complete the next 40 trials according to the plan selected for them randomly by the computer and indicated on the list.

After the initial presentation of the screens with choices for the block, the list of categories disappeared. At the beginning of each trial, two labels that represented a choice for that trial were again presented. In the planned condition, participants were reminded of their earlier choice (i.e., “For this trial you chose ___”). In the unplanned condition, they were given a choice to make a spontaneous decision (i.e., “For this trial you can choose between ___”). In the externally planned condition, they were reminded of the choice that had been made for them for that trial (i.e., “For this trial ___ was selected for you.”).

The main procedure consisted of 120 trials grouped in three blocks (conditions) in a randomized order. At the end of each block, participants reported the explicit sense of agency, the feeling of having an unconstrained choice, expectation-outcome congruence, fluency, pictures evaluation, and forgetting (the last measure not included in the unplanned condition).

Additionally, there were 20 practice trials at the beginning of the procedure. At the beginning of each of these trials, participants saw a screen that displayed two labels, one on the left side and the other on the right side of the screen. Participants were told they would see a picture from one of the presented categories after they press a key (i.e., ‘space’). After the screen with two labels disappeared, a screen with a fixation cross appeared, and participants clicked with their dominant hand. Clicking resulted in the display of a picture, signalized by a tone. After the picture disappeared, participants reported the interval between their action (i.e. pressing the key) and outcome (i.e. a picture and tone appearance) on the slider. Each time, feedback informing participants about the actual time interval was provided on a separate screen.

Implicit Sense of Agency. We measured time compression in the temporal binding task described earlier. First, a report error was calculated for each trial separately (i.e., a difference between reported time and actual time). Next, report errors were averaged across all trials in each block. In sum, we had three indices, one for each of the three conditions. We used these scores as our dependent variable.

To measure Explicit Sense of Agency (planned condition: α = .90, unplanned condition: α = .94, externally planned condition: α = .91), Expectation - Outcome Congruence (planned condition: α = .66, unplanned condition: α = .72, externally planned condition: α = .71), and Feeling of Having an Unconstrained Choice (planned condition: α = .78, unplanned condition: α = .71, externally planned condition: α = .79) we used the measures described in Study 1. We calculated means from the responses for each experimental condition (planned, unplanned, externally planned actions).

Fluency. One item was used to measure subjective ease in action selection: “Choosing the pictures in this part was difficult (1) – easy (7)” (Graf et al., 2018).

Pictures Evaluation. More/less favorable perception of pictures was measured with the question “How much did you like the pictures in this part?”. Participants responded on a scale from 1 (Not at all) to 7 (Very much).

Forgetting. Memory of previous choices for planned condition and externally planned condition was assessed with one question: “How well do you remember the choices that you made (externally planned condition: were presented with) at the beginning of this part?” Participants responded on a scale from 1 (Not at all) to 7 (Very well).⁶

Regarding the explicit sense of agency, we tested the same hypotheses as described in Study 1 (H1 - H3). Regarding the implicit sense of agency, the effects of planning should mirror the effects on the explicit sense of agency (H4). Again, we predicted that externally planned actions would be related to the lowest level of implicit sense of agency.

To test the above hypotheses, we again used two alternative methods of estimation: frequentist and Bayesian, fitting regression models with the explicit sense of agency and the implicit sense of agency as dependent variables and the described mediators. We re-ran the analyses described in Study 1, with several changes. First, we accounted for the within-participant design in our models by using different MEMORE macro in frequentist analysis (Montoya & Hayes, 2017) and by establishing variable intercepts for participants in Bayesian analysis. Second, due to the software limitations, in frequentist analysis we tested models each time entering two (out of three) conditions into within-person mediation models. The rest of the analyses remained the same.

For the sake of coherence, we tested the same hypotheses and implemented similar analyses as those proposed in Study 1. We expected that the sense of agency would be correlated with evaluation of stimuli. Therefore, we expected the effects of the manipulation on evaluation to mirror those on the sense of agency (H5)⁷, described in H1-H3. We expected that fluency would be correlated with congruence. Therefore, we expected the mediations through fluency to mirror those through congruence (H6).

Additionally, we examined the impact of condition on the implicit sense of agency with mixed models using 40 measurements of implicit agency (i.e., differences between reported time and actual time in each trial) per condition for each person. This way we accounted for within-person variance more precisely, even though it only allowed us to explore the effect of a condition without testing the mediating mechanisms. To address the effects of block order, we tested additional models with that factor accounted for.

To test the effects of planning manipulation on the explicit sense of agency, we conducted a parallel mediation analysis using MEMORE software, version 2.1.2 (Montoya & Hayes, 2017; see Figure 1). Due to the limitations of the software, which only permitted two conditions at a time as a default, it led to three distinct sets of analyses (unplanned vs. planned, externally planned vs. planned, and externally planned vs. unplanned). As the first two comparisons are the most relevant to our hypotheses we present them here and we included the results for the externally planned vs. unplanned comparison in SOM. The expectation-outcome congruence and the feeling of having an unconstrained choice were again parallel mediators. Indirect effects were established with 5,000 bootstrap samples. The same analyses were conducted for the second outcome variable: the implicit sense of agency.

Explicit Sense of Agency. The total effect of unplanned (vs. planned) condition on the explicit sense of agency was not significant (b = 0.25, 95% CI [-0.06, 0.56], p = .115). However, whereas participants in the unplanned condition did not experience lower expectation-outcome congruence (b = 0.06, 95% CI [-0.23, 0.36], p = .671), they again reported a stronger feeling of having an unconstrained choice (b = 0.40, 95% CI [0.11, 0.68], p = .006). Both the expectation-outcome congruence (b = 0.22, 95% CI [0.03, 0.42], p = .025) and the feeling of having an unconstrained choice (b = 0.52, 95% CI [0.31, 0.74], p < .001) were positively related to the sense of agency. When both mediators were included in the model, the effect of planning manipulation on the explicit sense of agency remained nonsignificant (b = 0.03, 95% CI [-0.25, 0.30], p = .840). The indirect effect of planning manipulation through congruence was not significant (b = 0.01, 95% CI [-0.07, 0.09]), but the indirect effect through the feeling of having an unconstrained choice was positive and significant (b = 0.21, 95% CI [0.04, 0.43]) (see Figure 6).

The total effect of externally planned (vs. planned) condition on the explicit sense of agency was negative and significant (b = -1.36, 95% CI [-1.84, -0.88], p < .001). Even though participants in the externally planned condition did not experience lower expectation-outcome congruence (b = -0.15, 95% CI [-0.50, 0.20], p = .398), they reported having a more constrained choice (b = -1.58, 95% CI [-2.04, -1.13], p < .001). Again, both the expectation-outcome congruence (b = 0.27, 95% CI [0.09, 0.45], p = .005) and the feeling of having an unconstrained choice (b = 0.73, 95% CI [0.59, 0.88], p < .001) were positively related to the sense of agency. When both mediators were included in the model, the effect of planning manipulation on the explicit sense of agency became nonsignificant (b = -0.16, 95% CI [-0.53, 0.21], p = .399). The indirect effect through congruence was not significant (b = -0.04, 95% CI [-0.16, 0.05]), but the indirect effect through the feeling of having an unconstrained choice was negative and significant (b = -1.16, 95% CI [-1.58, -0.78]) (see Figure 7). When we repeated the analyses to obtain estimates for the externally planned vs. unplanned comparison, the results were similar and even stronger, except for the relation between congruence and the explicit sense of agency, which was not significant (the details are presented in SOM).

Implicit Sense of Agency. The total effect of unplanned (vs. planned) condition on the implicit sense of agency was not significant (b = -9.50, 95% CI [-23.23, 4.23], p = .173). Neither expectation-outcome congruence (b = -3.24, 95% CI [-13.63, 7.15], p = .537) nor the feeling of having an unconstrained choice (b = 3.61, 95% CI [-7.66, 14.89], p = .526) was significantly related to the implicit sense of agency. When both mediators were included in the model, the effect of unplanned (vs. planned) condition remained nonsignificant (b = -10.73, 95% CI [-25.39, 3.93], p = .149). The indirect effect of planning manipulation through congruence was not significant (b = -0.21, 95% CI [-2.30, 1.54]) and neither was the indirect effect through the feeling of having an unconstrained choice (b = 1.43, 95% CI [-5.45, 7.01]) (see Figure 8).

Interestingly, the total effect of externally planned (vs. planned) condition was significant but, contrary to the expectations, it was negative (b = -16.81, 95% CI [-29.84, -3.78], p = .012) meaning that participants in the externally planned condition perceived shorter time intervals between their actions and outcomes than those making spontaneous decisions. Neither the expectation-outcome congruence (b = 1.81, 95% CI [-6.07, 9.68], p = .649) nor the feeling of having an unconstrained choice (b = -4.66, 95% CI [-10.79, 1.47], p = .134) was significantly related to the implicit sense of agency. When both mediators were included in the model, the effect of externally planned (vs. planned) condition remained significant (b = -24.00, 95% CI [-40.16, -7.84], p = .004). The indirect effect of planning manipulation through congruence was not significant (b = -0.26, 95% CI [-2.85, 1.49]) and neither was the indirect effect through the feeling of having an unconstrained choice (b = 7.45, 95% CI [-4.26, 17.39]) (see Figure 9). When we repeated the analyses for the externally planned vs. unplanned comparison, neither the total nor the direct effect of condition was significant. The remaining relations were similar (i.e., most of them not significant). The details are presented in SOM.

Pictures Evaluation. We conducted parallel mediation analyses identical to those reported earlier, with pictures evaluation as a dependent variable.

The total effect of unplanned (vs. planned) condition on pictures evaluation was not significant (b = 0.11, 95% CI [-0.13, 0.36], p = .368). Whereas expectation-outcome congruence was not significantly related to pictures evaluation (b = 0.08, 95% CI [-0.09, 0.26], p = .339), the feeling of having an unconstrained choice was positively related to it (b = 0.28, 95% CI [0.09, 0.47], p = .004). When both mediators were included in the model, the effect of planning manipulation on pictures evaluation remained nonsignificant (b = -0.01, 95% CI [-0.25, 0.24], p = .968). The indirect effect of planning manipulation through congruence was not significant (b = 0.01, 95% CI [-0.03, 0.06]), but the indirect effect of planning manipulation through the feeling of having an unconstrained choice was positive and significant (b = 0.11, 95% CI [0.01, 0.25]) (see Figure 10).

With regard to the comparison between externally planned (vs. planned) condition, the total effect on pictures evaluation was not significant (b = -0.29, 95% CI [-0.59, 0.00], p = .051). Whereas expectation-outcome congruence was positively related to pictures evaluation (b = 0.32, 95% CI [0.15, 0.48], p < .001), the feeling of having an unconstrained choice was not significantly related to it (b = 0.10, 95% CI [-0.03, 0.23], p = .116). When both mediators were included in the model, the effect of planning manipulation on pictures evaluation remained nonsignificant (b = -0.08, 95% CI [-0.42, 0.25], p = .627). The indirect effect of planning manipulation through congruence was not significant (b = -0.05, 95% CI [-0.15, 0.07]) and neither was the indirect effect through the feeling of having an unconstrained choice (b = -0.16, 95% CI [-0.35, 0.09]) (see Figure 11). When we repeated the analyses for the externally planned vs. unplanned comparison, we observed some differences. First, there was a significant and negative total effect of condition on the pictures evaluation. Second, the expectation-outcome congruence was not significantly related to the pictures evaluation. The remaining relations were similar and most of them were not significant. The details are presented in SOM.

Additionally, pictures evaluation was positively correlated with the explicit sense of agency for planned, unplanned, and externally planned conditions (Pearson’s r = .47, p < .001, Pearson’s r = .40, p < .001, and Pearson’s r = .22, p = .043, respectively).

We have conducted Bayesian analyses as an alternative approach to test our primary hypotheses and those related to pictures evaluation. In short, the results of these analyses were similar to the models tested in the frequentist analyses: the same effects were significant and in the same direction, with a few exceptions.

A positive relationship between congruence and the explicit sense of agency was significant also when the externally planned condition was compared with the unplanned one; a similar (i.e., positive and significant) result was observed for pictures evaluation in this comparison and when the unplanned condition was compared with the planned one. Importantly, these changes did not affect indirect effects that remained not significant. Additionally, a positive relation between the feeling of having an unconstrained choice and pictures evaluation was significant also when the externally planned condition was compared with the planned or the unplanned. The negative indirect effects through this mediator were significant as well.

It should be noted that in the frequentist analyses due to software limitations we decided to run separate models for different pairs of conditions, and so the relations between mediators and dependent variables in our models were estimated separately. That was not the case in Bayesian models, in which minor changes in the estimates resulted only from repeated sampling when the reference condition was changed. Therefore, some discrepancies between frequentist and Bayesian estimations were expected. Details of these analyses may be found in the SOM.

Fluency. Again, we conducted parallel mediation analyses identical to those reported earlier, but the expectation-outcome congruence was replaced with processing fluency.

These analyses showed that whereas the unplanned (vs. planned) condition was not significantly related to processing fluency, the externally planned (vs. planned) condition was negatively related to it (b = -0.58, 95% CI [-0.97, -0.20], p = .004).

Processing fluency was not significantly related to the implicit sense of agency. The indirect effects of planning manipulation through processing fluency were not significant either.

Finally, processing fluency was positively related to pictures evaluation in the unplanned (vs. planned) condition (b = 0.26, 95% CI [0.12, 0.40], p < .001), but it was not related to it when externally planned vs. planned condition were compared. The indirect effects of planning manipulation on the evaluation of pictures through processing fluency were not significant.

There were minor differences in the remaining estimates when we replaced congruence with fluency: these and other detailed statistics are presented in SOM.

To establish the impact of condition on the implicit sense of agency utilizing multiple datapoints instead of mean indices, we conducted mixed model analyses with the package lme4 (Bates et al., 2015). Condition (contrast-coded) was used as a predictor, while the measures of the implicit sense agency (i.e., errors of time estimation) were treated as a dependent variable. We added random intercepts for participants and used maximum likelihood as a method of estimation. Two models were estimated: one with the planned condition and one with the unplanned condition set a reference point.

We found a significant negative effect of the externally planned condition when contrasted with the planned one (b = -15.57, t(9832.29) = -2.93, p = .003). Again, time estimates were shorter when actions were planned by the external entity instead of oneself. The unplanned condition was neither significantly related to the implicit sense of agency when contrasted with the planned condition (b = -9.35, t(9819.35) = -1.83, p = .067), nor when it was a reference category and the externally planned condition was compared with it (b = -6.22, t(9830.52) = -1.19, p = .233).

The impact of block order was inspected with a similar set of analyses, with the order of block (e.g., 1, 2, 3) added as an additional predictor. We set the first block as our reference category. We found that in the second block temporal estimates were significantly longer than in the first one, while estimates in the third block did not differ (b = 10.33, t(9814.26) = 1.98, p = .047 and b = 8.61, t(9836.60) = 1.65, p = .098, respectively). Importantly, addition of the order did not substantially change the results. Details of these analyses may be found in SOM.

In Study 2, we again found some support for our hypotheses. Similar to Study 1, we observed that the feeling of having an unconstrained choice mediated the relationship between the planning manipulation and the explicit sense of agency. In line with the hypotheses, this mediation path was positive for the unplanned condition and negative for the externally planned condition when compared with the remaining conditions. These results support the idea that planning (both self- and externally-imposed) reduces the perceived freedom of choice, which subsequently has a negative impact on agency. This notion was additionally supported by the significant indirect effect on pictures evaluation, especially in the Bayesian models. It is important to note, though, that we did not find any mediation patterns with respect to the implicit measure of agency, a topic we will revisit in the general discussion.

When inspecting total effects of our manipulation, externally planned actions, compared with both the planned and the unplanned actions, led to a decreased explicit sense of agency, which aligns with the findings of Study 1. However, the same condition, compared with the planned one, resulted in the highest level of agency as measured implicitly. While the first effect might have been anticipated, the second one was unexpected. These results may indicate the need to treat explicit and implicit measures of agency differently, which raises questions about whether they capture the same construct at all. Additionally, externally planned actions, when compared with the unplanned ones, resulted in lower evaluation of pictures. This notion aligns with the results regarding the explicit sense of agency and strengthens evidence on the overall positive effects of spontaneity.

Again, expectation-outcome congruence remained unaffected by the manipulation, although it did show a positive relation with the sense of agency and, to some extent, with picture evaluation. Fluency, used as an alternative measure of congruence, decreased when actions were externally planned. We observed a positive relationship between fluency and pictures evaluation. Importantly, replacing congruence with fluency did not substantially change most of the results.

In summary, despite making several changes to the experimental design, we obtained a similar pattern of results, which speaks to the generalizability of our findings across different experimental contexts.

The main goal of our research was to test the impact of planning on the sense of agency, both explicit and implicit. Prior studies exploring this question have produced mixed results, with some indicating positive effects of planning (Vinding et al., 2013, 2015), while others have shown negative effects (Damen et al., 2015). To address these discrepancies, we proposed two opposing mediating mechanisms: one based on the congruence between expectations and outcomes, and the other via the sense of having a less constrained choice when making spontaneous decisions as opposed to planned ones.

To examine these mechanisms, we conducted two studies. These studies differed in several aspects, including the time gap between decisions and outcomes, the use of between-person versus within-person designs, and the inclusion of additional measures. In Study 1 we verified the theoretical model using a between-person design in the context of a longer delay between the planned decisions and outcomes. In Study 2 we induced a shorter gap between the decisions and outcomes and we used a within-person design, which additionally included a measure intended to capture implicit agency. Together, these studies provided an opportunity to test the generalizability of the hypothesized mechanisms across different experimental contexts while keeping the focal measures and the cover story (i.e., choosing pictures) constant.

Despite differences between studies, both experiments consistently revealed that spontaneous decision-making resulted in the strongest sense of unconstrained choice, which, in turn, was positively related to explicit sense of agency. Whether the delay between the plan and its execution spanned days or mere minutes, and regardless of whether the manipulation was between or within participants, planned decisions consistently led to a diminished sense of agency due to the perception of having more limited choices.

These findings are in line with the notion that spontaneity is a defining characteristic of human volition (Haggard, 2019). They also support the idea that decisions made by one’s past self tend to feel constraining, thus leading to feeling less autonomous and agentic. Despite the fact that participants’ actual choices did not necessarily differ between the conditions, the moment of making commitment has clearly impacted the subjective perception of the available options. This pattern of results may have important practical implications, considering the extent to which many human actions are planned and organized in advance. Future research could explore the circumstances under which planning does not negatively impact the sense of agency. One potentially relevant factor could be the specific attribution for following through on past intentions. It is possible that interpreting such actions, for example, as a display of self-control, could result in a higher, rather than lower, perception of one’s agency. An experimental design could be used to verify this hypothesis.

The comparison between the external plans and the (self-imposed) plans serves as additional validation of our manipulation in both studies. Our hypothesis was that if self-imposed plans reduce the sense of having a choice and subsequently diminish agency, then externally-imposed plans should have an even more pronounced impact. Indeed, the condition where participants’ actions were determined by an external source resulted in the perception of having the most restricted choice and the lowest explicit sense of agency. These effects were robust and statistically significant, suggesting that the results observed in the unplanned condition, which were in the opposite direction, indeed captured the crucial aspect that we primarily focused on. While often overlooked in research on the sense of agency, the externally planned condition provided important insights and should be considered more frequently in future studies on this topic.

In contrast to the mediating role of unconstrained choice, our hypotheses regarding the second mediator were not supported. Despite the fact that in the planned condition participants were able to fully execute their plans, which does not always happen in real life, we did not observe a positive mediating effect based on the congruence between expectations and outcomes. One potential explanation for this could be that, in our studies, participants were only able to plan the general category (e.g., “animals”) rather than the specific image they would see. This form of planning might not be enough to facilitate a comprehensive representation of their intended outcome. In future studies, it might be useful to manipulate the level of plan elaboration as a potential moderating factor to investigate whether this could play a role. Another reason could be that the reliability of the measure was low in Study 1, and although it reached a satisfactory level in Study 2, it was still not particularly high, which may have limited our ability to detect the effect.

Even when the congruence was operationalized through processing fluency, the mediating effects were not significant either. However, there were some significant relationships between perceived congruence and fluency and the positive evaluation of the pictures. As such relationships were expected on the basis of theory, these findings suggest that we did capture meaningful constructs. Given our results, what remains less clear is the relationship between these constructs and the sense of agency.

While not a primary focus of our research, it is worth noting that the relationship between perceived congruence and the explicit sense of agency was inconsistent between the two studies. In particular, in Study 1, these two variables were not significantly related, whereas in Study 2, significant relationships emerged. This prompts the question of whether these differences occurred due to variations in the time delay between planning and execution or if other aspects of the studies contributed to these disparities. For instance, given that participants recalled their planned choices in Study 2 better compared to Study 1, presumably due to the shorter time gap between planning and execution, the measures of congruence could capture a more meaningful construct in that study. Again, a lower reliability of congruence in Study 1 than in Study 2 could be another source of differences between these studies.

While the effects of the manipulation on the explicit sense of agency mostly aligned with our expectations, we did not observe similarly meaningful effects for the implicit sense of agency. The only significant results we obtained were related to the comparison between the externally planned and planned conditions. However, even these results were opposite to what we had expected on the basis of the literature. Surprisingly, individuals in the externally planned condition reported a shorter time interval between their decisions and outcomes, which, according to the typical interpretation of this measure, would indicate a higher sense of agency. The absence of a correlation between explicit and implicit senses of agency in our studies is consistent with some past studies as well (e.g. Dewey & Knoblich, 2014; Lafleur et al., 2020; Majchrowicz & Wierzchoń, 2018; Reis et al., 2022; Wen et al., 2015). Our study thus adds to the literature that underscores the need for a more thorough assessment of the validity and reliability of the measures of implicit agency (see Wen & Imamizu, 2022).

Importantly, whereas we observed a significant total effect of planning (vs. making spontaneous decisions) on the sense of agency in Study 1, in Study 2 only the indirect effect through the feeling of having an unconstrained choice was significant. This difference in results might be attributed to the smaller overall sample size in Study 2.

Another potential explanation could relate to the duration between forming the intention and taking action, which differed between these studies. It is possible that the greater the time gap between planning and execution, the more negative impact of planning on the sense of agency becomes. It would be in line with the results showing that the sense of agency decreases with increasing action-effects intervals (Ruess et al., 2017). Future research could examine this possibility further by testing the moderating role of time delay, a factor previously explored by Damen and colleagues (2015). This possibility also aligns with the construal level theory (Trope & Liberman, 2010), which suggests that when thinking about events closer in time, people focus on the dimension of feasibility, but when they think about more distant events, these concerns become relatively less salient (Lutchyn & Yzer, 2011). It is possible that the sense of agency, a concept closely related to efficacy, is more active when planning occurred relatively recently but it is less salient when the plan is more distant in time and therefore the greatest differences are observed when recent events are contrasted with more distant ones. Finally, lay theories of change might also play a role in this context if people are more inclined to acknowledge changes in their preferences that occurred over the course of a week as opposed to changes that occurred within a few minutes.

The total effect of external plans (compared to any plans made by oneself) on the explicit sense of agency was consistently negative across studies, which attests to the robust impact that autonomy of decisions has, regardless of the circumstances (e.g. the timeline) of given actions.

One limitation of our studies is that they relied on a single decision-making task, which involved the selection of images. It is thus possible that the content of these chosen images influenced not only participants’ evaluations of the pictures but also other aspects of the study such as precision of time estimates in Study 2. Although we took care to select images with relatively neutral content, individual preferences may have still played a role in shaping particular reactions. Future research could consider accounting for individual preferences in the task. Furthermore, while we designed our studies to ensure some motivational relevance to participants, the decisions they made were most likely not of high importance in their daily lives. In future studies, it would be beneficial to use similar planning designs but apply them to activities that individuals freely engage in or find personally relevant.

Despite these limitations, we believe that our studies offer valuable insights into the role of plans in shaping the sense of agency. Planning is a common aspect of daily life, and it carries significant implications for various psychological phenomena. Surprisingly, it has not received much attention in the context of agentic experiences. We hope that our findings will encourage further exploration of these two important and highly relevant constructs.

The approved protocol of Stage 1 Registered Report is available at: https://osf.io/jrzys/

Additional materials related to this paper (including anonymized study data, digital materials, code, etc.) are available at: https://osf.io/mkxjy/

Conceptualization: MM and KJ; Methodology: MM, KJ, BM, ZS, MW, and PMG; Formal analysis: MM, KJ, and ZS; Investigation: MM and KJ; Writing - Original Draft: MM, KJ, and BM; Writing - Review & Editing: MM, KJ, BM, ZS, MW, and PMG; Project administration and funding: MM and KJ; Supervision: KJ.

We have no conflicts of interest to disclose.

The research was supported by a grant from the Faculty of Philosophy under the Strategic Programme Excellence Initiative at Jagiellonian University and by a grant from the National Science Center (2015/19/B/HS6/01253) received by Katarzyna Jasko.

Article Processing Charges were funded by a grant from the Faculty of Philosophy under the Strategic Programme Excellence Initiative at Jagiellonian University.

The data was processed using the PS IMAGO PRO solution, whose analytical engine is IBM SPSS Statistics.