- Email: [email protected]
Outside influence: The sense of agency takes into account what is in our surroundings
Acta Psychologica 186 (2018) 104–109
Contents lists available at ScienceDirect
Acta Psychologica journal homepage: www.elsevier.com/locate/actpsy
Outside influence: The sense of agency takes into account what is in our surroundings
T
⁎
Nicholas Hon , Yin-Yi Seow, Don Pereira Department of Psychology, National University of Singapore, 9 Arts Link, 117570, Singapore
A B S T R A C T
We are quite capable of distinguishing those outcomes we cause from those we do not. This ability to sense selfagency is thought to be produced by a comparison between a predictive representation of an outcome and the actual outcome that occurs. It is unclear, though, specifically what types of information can be entered into agency computations. Here, we demonstrate that information from non-target stimuli (stimuli that are not directly acted upon) incidentally present in our surroundings can influence predictions of outcomes, consequently modulating the sense of agency over clearly-defined target outcomes (those that occur to acted-upon stimuli). This provides the first evidence that our sense of agency is contextualized with respect to what is in our immediate visual environment. Furthermore, our data suggest that agency computations, instead of just a single comparison, may involve comparisons performed in stages, with different stages involving different types/ classes of information. A model of such multi-stage comparisons is described.
1. Introduction In general, we are able to effectively distinguish the outcomes we cause from those we do not. This ability to sense our agency in actions and outcomes is a central one because it grounds our “sense of self” (Knoblich & Flach, 2003) and enables us to distinguish ourselves from others and the operations of the natural world. Additionally, in our social interactions, we generally assign credit and apportion blame on the assumption that individuals can sense their own responsibility in given outcomes (Bandura, 2001; Haggard & Tsakiris, 2009). A sense of agency (SoA) is thought to be produced by a comparison process in which a mental representation of a predicted outcome is compared to the actual outcome that occurs (Frith, Blakemore, & Wolpert, 2000; Moore & Haggard, 2008; Sato & Yasuda, 2005; Wegner & Wheatley, 1999). A match between the predictive representation and the actual outcome produces a strong sense of agency. On the other hand, a mismatch between the two produces feelings of non-agency (or lack of agency). For instance, a ball that travels in a direction and with a speed consistent with the force with which it is thrown will likely produce a strong sense of agency in the thrower. A ball that travels backwards towards the thrower (a mismatch with what would be predicted regarding direction of movement), however, will more than likely produce a strong sense of non-agency. Supporting this, experimental manipulations that affect the relationship between predicted and actual outcomes have been found to reliably influence SoA,
⁎
producing agency distortions. On one hand, manipulations that act to minimize matching between predicted and actual outcomes result in a reduced sense of agency (Sato, 2009; Sato & Yasuda, 2005; Wenke, Fleming, & Haggard, 2010). For example, in one study, the application of mild spatial noise to an outcome such that spatial matches between predictions and actual outcomes were slightly reduced produced lower levels of SoA (Farrer & Frith, 2002). (It should be noted that outright spatial incongruence or mismatches (e.g., a stimulus moving left when the prediction is that it will go right) between predictions and actual outcomes will produce a clear sense of non-agency.) On the other, situations in which an actual outcome appears to match a predicted one result in an elevated sense of agency, and this is so even when an individual may not have had any real control over the outcome in question (Hon & Poh, 2016; Pronin, Wegner, McCarthy, & Rodriguez, 2006; Wegner, Sparrow, & Winerman, 2004; Wegner & Wheatley, 1999). In a seminal study (Pronin et al., 2006), participants who undertook a faux magical ritual aimed at harming another individual reported greater levels of SoA when the targeted person subsequently reported feeling ill. In reality, the ritual did not have any actual influence and the targeted individual was, in fact, a confederate. In this case, “vicarious agency” was produced because the outcome matched the participants' predictions regarding the outcome of the ritual. While the comparison model described above has been very successful in explaining its production, it is nonetheless the case that SoA is generally studied under highly constrained circumstances. In some
Corresponding author at: Department of Psychology, National University of Singapore, 9 Arts Link, 117570, Singapore. E-mail address: [email protected] (N. Hon).
https://doi.org/10.1016/j.actpsy.2018.03.004 Received 27 November 2017; Received in revised form 6 February 2018; Accepted 15 March 2018 Available online 28 March 2018 0001-6918/ © 2018 Elsevier B.V. All rights reserved.
Acta Psychologica 186 (2018) 104–109
N. Hon et al.
case. It is also worth pointing out that this paradigm allowed us to probe the matching of predictions while keeping target signals constant. Notice that target signals, in any given target condition, were the same regardless of what the distractors did. It is worth pointing out that what is being studied here is different from earlier studies that have reported that stimuli like primes can sometimes affect agency ratings over a target action/outcome (Aarts, Custers, & Marien, 2009; Aarts, Custers, & Wegner, 2005; Chambon & Haggard, 2012; Linser & Goschke, 2007; Wenke et al., 2010). While it is possible to characterize such primes as non-target stimuli, their purpose was to affect the fluency with which agentic actions could be performed (Chambon & Haggard, 2012; Sidarus et al., 2013). Thus, while they were non-targets, they were not extraneous to the task. The non-targets in the current study are different in that they neither hinder nor aid action choice or the actual movement of the target. Thus, the non-target stimuli of this study were designed to function more genuinely as extraneous environmental stimuli.
typical experiments, participants are introduced to a single stimulus that they try to control via some behavior (e.g., a button press or joystick movement), with the actual stimulus outcome being consistent with their actions or not (Dewey, Seiffert, & Carr, 2010; Ebert & Wegner, 2010; Hon, Poh, & Soon, 2013). In others, participants perform some motor action and are presented with somatosensory outcomes that are linked with the action or not, judging agency over the production of these outcomes (Chambon, Moore, & Haggard, 2015; Sidarus, Chambon, & Haggard, 2013; Wenke et al., 2010). In these paradigms, it is common for target stimuli/outcomes to be presented in isolation, being the only information presented at a given time. One concern is that this may give the impression that SoA is driven solely by signals from target stimuli/outcomes and nothing else. In everyday life, however, the stimulus one acts upon may not be the sole stimulus present in the environment. An open question, then, pertains to whether or not non-target stimuli (stimuli that are not acted upon) incidentally present in the environment can influence SoA over outcomes relating to an acted-upon target stimulus. To date, the literature has been silent on this issue. In this study, we investigated the possibility that information regarding extraneous, non-target stimuli can be factored into agency computations; in particular, the predictive representations of outcomes. Consider, for example, a standard table setting in which a dinner plate is flanked by a knife and fork. If one were to push the plate, one would clearly predict that it should move forward in a manner consistent with the push applied. If it does this when pushed, one should feel a strong sense of agency. In this case, though, the prediction involves only information from the target stimulus (i.e., the plate). However, it could also be predicted that, in addition to moving in a direction consistent with the push, the plate should end up further away from the knife and fork. Notice that, in this latter scenario, information about the knife and fork is extraneous to the acted-upon stimulus (i.e., the plate) but is built into the prediction nonetheless. In the current study, we investigated this issue - whether information about extraneous, non-target stimuli can be built into predictive representations. Here, participants viewed displays that comprised a central rectangle (target) flanked on the left and right by two other rectangles (non-targets). The target rectangle was always a different colour from the non-target rectangles (Fig. 1). Participants made self-decided and self-initiated presses of the up- or down-arrows on a standard computer keyboard when the display was presented. Following a button press, the target rectangle would move in a direction consistent with that indicated by the button press or not. The task was to assess agency over the movement of the target rectangle. Consistent with the literature, we would expect high agency ratings when the target's movement matched the direction indicated by the button press and low ratings when these were incongruent. Of greater relevance here, at the same time as the target moved, the adjacent non-targets could do one of three things: They could move in the same direction as the target rectangle, move in the opposite direction or remain stationary. If, as in the plate and cutlery example above, predictions include the (spatial) relationship between target and simultaneously-present non-target stimuli, then, in the current experiment, the prediction associated with a given buttonpress would be that the target rectangle should (a) move in a direction consistent with that indicated by the button press and (b) end up further away from the non-targets than before the action was performed. Accordingly, the best matches should occur when the target moves in a direction consistent with the button press and when it ends up clearly separated from the non-targets. Thus, in this study, when there is congruence between button press and target movement, one would expect the greatest SoA to be reported when the target and non-targets move in opposite directions. This would produce the greatest final distance between the two, offering the clearest evidence of a match with the prediction. Similarly, the lowest agency ratings should be observed when target and non-targets move in the same direction, as the final spatial separation between the two would be smallest in this
2. Experiment 1 2.1. Methods 2.1.1. Participants 29 undergraduate students from the National University of Singapore participated in this experiment.1 All participants had normal or corrected-to-normal vision. 2.1.2. Procedure In the experiment proper, participants completed a single block comprising 240 trials of a simple agency judgement task. On each trial, three coloured rectangles were presented side-by-side in the centre of a white screen (Fig. 1). The middle rectangle (designated the target) was of one colour, while the two flanking rectangles (non-targets) were of another. The two colours used for the rectangles were red and blue, and the assignment of the colours to target or non-targets was counterbalanced across participants. In response to the presentation of the rectangles, participants made self-initiated and self-decided up- or down-arrow key presses. A hundred milliseconds (100 ms) after the key-press, the target rectangle moved in a direction that was either congruent (e.g., moving up after an up-arrow key-press) or incongruent (e.g., moving down after an up-arrow key-press) with the direction indicated by the key-press. These were designated Target Congruent and Target Incongruent trials respectively. The distance travelled by the target rectangle was 5.14o of visual angle, regardless of direction. The movement of the target lasted 120 ms. Target Congruent and Target Incongruent trials each accounted for 50% of all trials. When the target moved, one of three things could happen with respect to the flanking non-targets. These could move in the same direction as the target (e.g., upwards when the target moved up), in an opposite direction to the target (e.g., downward when the target moved up), or they could remain in their original position. These are termed Non-target Same, Non-target Different and Non-target Neutral trials respectively. Non-target Neutral trials, in which the non-targets remained stationary accounted for 50% of all trials, with the remaining 50% of the trials being split equally between Non-target Same and Different conditions. Non-target movements began at the same time and were of the same speed as the target. However, they only moved one-third the distance travelled by the target (~1.74o of visual angle). Correspondingly, their movement lasted a shorter time (~ 40 ms) than that of the target. The difference in amount of movement was, like the use of different colours, meant to promote disambiguation between targets and non-targets. 1 Sample size was determined on the basis of other studies form our lab using the same base paradigm (e.g., Hon et al., 2013).
105
Acta Psychologica 186 (2018) 104–109
N. Hon et al.
Fig. 1. Illustration of a trial. Depicted is an example of a Target Congruent - Non-target Same trial. On such trials, the target (i.e. centre) rectangle moved in the direction indicated by the participant's key press, with the non-targets moving in the same direction as the target. Although greyscale was used in this figure, in the actual experiment, red and blue were used for the rectangles.
After the target rectangle had stopped moving, participants rated how much they felt their key-press caused its movement using a 7-point Likert-type scale (1: Not at all; 4: Unsure; 7: A lot). The stimuli were presented on a 21″ LCD monitor, with the experiment controlled by a PC running the E-Prime software. 2.2. Results The agency rating data from this experiment are depicted in Fig. 2. To start, a fully-within 2 (target: congruent, incongruent) × 3 (nontarget: same, different, neutral) ANOVA was performed on the agency ratings. This revealed a main effect of target congruence [F(1, 28) = 153.35, p < 0.001, ɳp2 = 0.84], a main effect of non-target movement [F(2, 56) = 6.43, p = 0.005, ɳp2 = 0.19] and, most importantly, a significant interaction between the two [F(2, 56) = 8.44, p = 0.005, ɳp2 = 0.23]. To elucidate this interaction, Target Congruent and target incongruent trials were examined separately. For Target Congruent trials (Fig. 2A), a significant effect of nontarget movement on agency ratings was found, F(2, 56) = 11.28, p < 0.001, ɳp2 = 0.29. A test of simple effects revealed that agency ratings in Non-target Different trials were significantly higher than in Non-target Same [t(28) = 3.97, p < 0.001] and Non-target Neutral trials [t(28) = 2.07, p = 0.048]. Non-target Neutral trials produced significantly higher agency ratings than Non-target Same trials, t (28) = 3.21, p = 0.003. This indicates that SoA over the movement of the target rectangle was affected by what the non-targets did. For Target Incongruent trials (Fig. 2B), on the other hand, no effect of non-target movement on agency ratings was found (F < 1, n.s.). This is unsurprising because a clear spatial incongruence between predicted and actual outcomes is known to be an unambiguous “nonagency” signal (Hon et al., 2013; Schmidt & Heumuller, 2010). These results are consistent with the idea that information about the non-targets was built into predictive representations of the outcome. Specifically, that, when acting on the target rectangle (via one's button press), the prediction would be that the target rectangle would both (a) move in a direction consistent with one's button press and (b) end up further away from the non-targets than before the action was performed. With respect to (b), on Target Congruent trials, the actual outcome that best matches the prediction would be when the final position of the target is furthest away from non-targets (i.e., clearest evidence that the two are separated), with the poorest match being when the eventual target-non-target separation is the smallest. In the Target Congruent condition, targets ended up furthest away from the non-targets on the Non-target Different trials (~6.84o of visual angle) and closest on the Same trials (~3.43o of visual angle). Agency ratings in the Target Congruent trials followed this pattern: Ratings were highest in the Non-target Different trials and lowest in the Same trials, with ratings on Neutral trials being intermediate between these.
Fig. 2. Agency ratings from the Target Congruent (A) and Target Incongruent trials (B) of Experiment 1. Error bars indicate 1 SEM.
strategically build these into their predictions (as spatial references)? To assess this possibility, we ran another experiment that was identical in set-up to Experiment 1 with the exception that the three non-target conditions – Same, Different and Neutral – now accounted for an equal number of trials each (one-third each). In this new experiment, the nontargets would, in fact, move more often than they would remain stationary. Would the fact that they now moved so often limit their inclusion in predictions? 3.1. Methods 3.1.1. Participants 30 participants, drawn from the same pool as those in Experiment 1, participated in this experiment. 3.1.2. Procedure The procedure and presentation parameters were identical to those in Experiment 1. The central difference was that, in this experiment, the total number of trials was evenly divided between the three non-target conditions.
3. Experiment 2 In Experiment 1, non-targets only moved 50% of the time. Could this spatial invariance on half the trials have prompted participants to 106
Acta Psychologica 186 (2018) 104–109
N. Hon et al.
3.2. Results
4. Experiment 3
Fig. 3 depicts the agency ratings from this experiment, separated by target congruence (Fig. 3A: Target Congruent trials; Fig. 3B: Target Incongruent trials). As with Experiment 1, a fully-within 2 (target: congruent, incongruent) × 3 (non-target: same, different neutral) ANOVA revealed a main effect of target congruence [F(1, 29) = 303.99, p < 0.001, ɳp2 = 0.91], a main effect of non-target movement [F(2, 58) = 3.58, p < 0.034, ɳp2 = 0.110] and a significant interaction between the two, F(2, 58) = 9.22, p < 0.001, ɳp2 = 0.24. Separate one-way ANOVAs on the two types of target trials revealed that, as before, there was no effect of non-target type on target incongruent trials (F < 1, n.s.). At the same time, there was a significant effect of non-target type on Target Congruent trials, F(2, 58) = 6.57, p = 0.003, ɳp2 = 0.19. Simple effects tests revealed that, for these trials, agency ratings in Non-target Different trials were higher than in Non-target Same (t(29) = 2.80, p = 0.009) and Non-target Neutral trials (t(29) = 2.16, p = 0.039). Non-target Neutral trials produced higher agency ratings than Non-target Same trials (t(29) = 2.30, p = 0.029). This is the exact same pattern observed in Experiment 1. Supporting this, a direct comparison between experiments revealed no evidence of an interaction between experiment (Experiment 1, Experiment 2) and the central target and non-target variables (F < 1, n.s.). The data from this experiment replicate the findings of Experiment 1, with agency ratings in the Target Congruent trials following the spatial separation between targets and non-targets: highest ratings for Non-target Different, followed by Neutral and then Same. Additionally, the fact that these findings were obtained, even though the non-targets moved more often than they stayed still, suggests that the inclusion of non-target information into predictions was unlikely to have been driven by a strategy based on the preponderance of non-targets remaining still (as in Experiment 1).
It is also possible that strategy may have played a role in another way: Participants may have been motivated to include non-targets in predictions because these appeared on all trials (as in Experiments 1 and 2). We ran a subsidiary experiment to assess this possibility. In this experiment, non-targets were only present on 50% of the trials.
Fig. 4 depicts the agency ratings from the critical trials (i.e., those with non-targets present) of this experiment, separated by target congruence (Fig. 4A: Target Congruent trials; Fig. 4B: Target Incongruent trials). As with Experiments 1 and 2, a fully-within 2 (target: congruent, incongruent) × 3 (non-target: same, different neutral) ANOVA was performed on the critical non-target present trials. This revealed a main effect of target congruence [F(1, 32) = 143.28, p < 0.001,
Fig. 3. Agency ratings from the Target Congruent (A) and Target Incongruent trials (B) of Experiment 2. Error bars indicate 1 SEM.
Fig. 4. Agency ratings from the Target Congruent (A) and Target Incongruent trials (B) of Experiment 3. Error bars indicate 1 SEM.
4.1. Methods 4.1.1. Participants 33 participants, drawn from the same pool as those in Experiments 1 and 2, participated in this experiment. 4.1.2. Procedure The procedure and presentation parameters were largely identical to those in Experiments 1 and 2. The key difference was that, in this experiment, non-targets were only present on half the trials, being absent on the other half. When non-targets were present, each non-target condition (Same, Different and Neutral) accounted for an equal number of trials. 4.2. Results
107
Acta Psychologica 186 (2018) 104–109
N. Hon et al.
Fig. 5. Model in which SoA is the result of comparisons involving both primary determinants and secondary modulators.
ɳp2 = 0.82], a main effect of non-target movement [F(2, 64) = 6.92, p = 0.002, ɳp2 = 0.18] and a significant interaction between the two, F (2, 64) = 13.45, p < 0.001, ɳp2 = 0.30. Separate one-way ANOVAs performed on the two types of target trials revealed that there was no effect of non-target movement on target incongruent trials (F < 1, n.s.). There was, however, a significant effect of non-target movement on Target Congruent trials, F(2, 64) = 15.31, p < 0.001, ɳp2 = 0.32. Simple effects tests revealed that, for these, agency ratings in Non-target Different trials were higher than in Non-target Same (t(32) = 6.15, p < 0.001) and Non-target Neutral trials (t(32) = 2.17, p = 0.038). Also, Non-target Neutral trials produced higher agency ratings than Non-target Same trials (t(32) = 3.51, p = 0.001). This is, of course, the same pattern observed in Experiments 1 and 2. Finally, we formally compared the findings from the 3 experiments with a mixed 2 (target: congruent, incongruent) × 3 (non-target: same, different, neutral) × 3 (experiment: 1, 2, 3) analysis. This analysis revealed main effects of target [F(1, 89) = 548.76, p < 0.001, ɳp2 = 0.86] and non-target [F(2, 178) = 16.36, p < 0.001, ɳp2 = 0.16] and a two-way interaction between these [F(2, 178) = 30.32, p < 0.001, ɳp2 = 0.25]. Critically, though, there was no main effect of experiment nor did it interact with the other two variables (all Fs < 1, n.s.). In other words, the same pattern of findings was obtained in all 3 experiments. Thus, even though non-targets may not always appear, when they are presented, they are factored into predictions.
conditions would not have been caused by the actual target signals, as these were held constant. Taken as a whole, the current findings suggest an important point about predictive representations: Predictions can include at least two different classes of information. The first relates to information that serves as the primary determinant of agency. In the current experiments, regardless of what the non-targets did, congruence between the button press and the target's movement produced ratings from the selfagency end of the scale (> 4), while incongruence produced ratings from the non-agency end of the scale (< 4). This suggests that, for the task used here, the primary determinant of agency was spatial congruence between the target's movement and the direction indicated by the button one presses. The second class involves information that merely acts to modulate SoA. In the current study, this would have been information involving the non-targets. Notice that, when there was congruence between target movement and the button press, none of the non-target movement conditions produced non-agency ratings (< 4). That being said, the different non-target conditions were able to modulate the level of SoA, with some conditions producing higher ratings than others. Altogether, this suggests that matching with respect to these two classes of information – primary determinants and secondary modulators – occurs in the following stage-wise fashion (Fig. 5). An initial stage assesses matching on the basis of the primary determinant of agency.2 If there is no match, then a strong sense of non-agency ensues. This is consistent with previous findings that clear mismatches between predictions and actual outcomes provides strong, unambiguous non-agency signals, which appear to be unalterable by additional factors (Schmidt & Heumuller, 2010; Shanks & Dickinson, 1991; Shanks, Pearson, & Dickinson, 1989). If, on the other hand, there is a match at this initial stage, then a sense of agency is produced. This sense of agency can then be up- or down-regulated at the second stage, in which matching is performed on the basis of information that acts as a secondary modulator. A good match on the secondary modulator produces a strong sense of agency, while a poor match produces a less strong SoA. As mentioned previously, a dominant idea is that SoA is produced by a single-stage comparison between predicted and actual outcomes. Some researchers, however, have posited that agency computations may involve more than one stage (Synofzik, Vosgerau, & Newen, 2008). Our data are broadly consistent with this latter idea. The current findings suggest that, rather than being underpinned by a single, uni-level comparison, the SoA one eventually experiences can be the product of comparisons performed at multiple levels, with each involving different types/classes of information. The cognitive literature has reported many examples of behavior
5. General discussion The present study set out to explore if information about non-targets present in the immediate environment can affect the sense of agency over a clearly-defined target outcome. In particular, we investigated the hypothesis that predictive representations can include information about incidentally present non-targets. Our findings from 3 experiments support this idea. The current findings are consistent with the idea that information about non-targets was built into predictive representations. In the current study, the prediction would be that, when the target rectangle is acted upon (by one's button press), it will move in a direction consistent with the button pressed and will end up further away from the nontarget rectangles than at the start of the trial (i.e., before any action was performed). The greatest SoA should, therefore, be produced when the actual outcome matches both parts of this prediction. This is what was found. When the target moved in a direction that was congruent with the button press, the highest agency ratings were reported when evidence of a final spatial separation between target and non-targets was clearest (i.e. Non-target Different trials). As mentioned previously, this difference in matching across the different distractor movement
2 The specific type of information (e.g., spatial, temporal) that functions as the primary determinant would be determined by the nature of the paradigm (or, more specifically, the type of actions and outcomes that are involved in this).
108
Acta Psychologica 186 (2018) 104–109
N. Hon et al.
being affected by the congruence between target and simultaneously presented non-target information. Typically, performance is adversely affected by an incongruence between the two (e.g., Stroop effect, flankers distraction). If such a phenomenon underpinned the current findings, one would have expected the lowest ratings to be observed when targets and non-targets moved in opposite directions (i.e., when target and distractor effects were incongruent). However, the opposite pattern was found here. Thus, we can rule this out as an alternative account. Alternatively, could participants have interpreted non-target movements as causing target movements? For example, perhaps, on congruent trials, participants viewed non-targets as causing target outcomes when the two moved in the same direction (i.e., Target Congruent - Non-target Same trials). This possibility is also unlikely. If participants felt that non-targets caused target movements, then ratings on those trials should have fallen within the non-agency range (< 4). This was not observed. Thus, this alternative can also be ruled out. In conclusion, the current results demonstrate that we contextualize our agentic operations with respect to what is in our (physical) surroundings, allowing for non-target signals (signals from stimuli that are not directly acted upon by us) to be incorporated into self-agency computations. It is interesting to note that other studies have demonstrated that an individual's sense of agency can be contextualized with respect to what others do (Dewey, Pacherie, & Knoblich, 2014; van der Wel, 2015). Taken together, those findings and ours paint a picture of SoA as being flexibly sensitive to both our social and physical settings. The results from the experiments reported here indicate that our predictive representations of an outcome, in particular, can include information about both target and non-target stimuli. This clarifies an important point about such representations. In studies in which actedupon target stimuli and their outcomes are presented in isolation, predictions could quite reasonably be thought to involve only information about said stimuli and outcomes. Here, the data suggest that extraneous, non-target information present in our immediate visual environment can also be built into predictive representations, being able to exert an influence over the resultant SoA.
moving object with another person. Cognition, 132, 383–397. Dewey, J. A., Seiffert, A. E., & Carr, T. H. (2010). Taking credit for success: The phenomenology of control in a goal-directed task. Consciousness and Cognition, 19(1), 48–62 (doi: S1053-8100(09)00144-5 [pii]101016/j.concog.2009.09.007). Ebert, J. P., & Wegner, D. M. (2010). Time warp: Authorship shapes the perceived timing of actions and events. Consciousness and Cognition, 19(1), 481–489 (doi: S10538100(09)00154-8 [pii]10.1016/j.concog.2009.12002. Farrer, C., & Frith, C. D. (2002). Experiencing oneself vs another person as being the cause of an action: The neural correlates of the experience of agency. Neuroimage, 15(3), 596–603. Frith, C. D., Blakemore, S. J., & Wolpert, D. M. (2000). Abnormalities in the awareness and control of action. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 355(1404), 1771–1788. http://dx.doi.org/10.1098/rstb.2000. 0734. Haggard, P., & Tsakiris, M. (2009). The experience of agency: Feelings, judgments and responsibility. Current Directions in Psychological Science, 18, 242–246. Hon, N., & Poh, J. H. (2016). Sleep deprivation produces feelings of vicarious agency. Consciousness and Cognition, 40, 86–92. http://dx.doi.org/10.1016/j.concog.2015.12. 007. Hon, N., Poh, J. H., & Soon, C. S. (2013). Preoccupied minds feel less control: Sense of agency is modulated by cognitive load. Consciousness and Cognition, 22(2), 556–561 (doi: S1053-8100(13)00037-8 [pii]171016/j.concog.2013.03.004. Knoblich, G., & Flach, R. (2003). Action identity: Evidence from self-recognition, prediction, and coordination. Consciousness and Cognition, 12(4), 620–632. Linser, K., & Goschke, T. (2007). Unconscious modulation of the conscious experience of voluntary control. Cognition, 104(3), 459–475 (doi: S0010-0277(06)00160-0 [pii] 201016/j.cognition.2006.07.009). Moore, J., & Haggard, P. (2008). Awareness of action: Inference and prediction. Consciousness and Cognition, 17(1), 136–144 (doi: S1053-8100(07)00004-9 [pii] 221016/j.concog.2006.12.004). Pronin, E., Wegner, D. M., McCarthy, K., & Rodriguez, S. (2006). Everyday magical powers: The role of apparent mental causation in the overestimation of personal influence. Journal of Personality and Social Psychology, 91(2), 218–231 (doi: 200609808-002 [pii]241037/0022-3514.91.2.218). Sato, A. (2009). Both motor prediction and conceptual congruency between preview and action-effect contribute to explicit judgment of agency. Cognition, 110(1), 74–83 (doi: S0010-0277(08)00238-2 [pii]10.1016/j.cognition.2008.26011). Sato, A., & Yasuda, A. (2005). Illusion of sense of self-agency: Discrepancy between the predicted and actual sensory consequences of actions modulates the sense of selfagency, but not the sense of self-ownership. Cognition, 94(3), 241–255 (doi: S001002770400099X [pii]281016/j.cognition.2004.04.003). Schmidt, T., & Heumuller, V. C. (2010). Probability judgments of agency: Rational or irrational? Consciousness and Cognition, 19(1), 1–11 (doi: S1053-8100(10)00006-1 [pii]301016/j.concog.2010.01.004). Shanks, D. R., & Dickinson, A. (1991). Instrumental judgment and performance under variations in action-outcome contingency and contiguity. Memory & Cognition, 19(4), 353–360. Shanks, D. R., Pearson, S. M., & Dickinson, A. (1989). Temporal contiguity and the judgement of causality by human subjects. Quarterly Journal of Experimental Psychology, 41, 139–159. Sidarus, N., Chambon, V., & Haggard, P. (2013). Priming of actions increases sense of control over unexpected outcomes. Consciousness and Cognition, 22(4), 1403–1411. http://dx.doi.org/10.1016/j.concog.2013.09.008. Synofzik, M., Vosgerau, G., & Newen, A. (2008). Beyond the comparator model: A multifactorial two-step account of agency. Consciousness and Cognition, 17(1), 219–239 (doi: S1053-8100(07)00026-8 [pii]361016/j.concog.2007.03.010). Wegner, D. M., Sparrow, B., & Winerman, L. (2004). Vicarious agency: Experiencing control over the movements of others. Journal of Personality and Social Psychology, 86(6), 838–848. http://dx.doi.org/10.1037/0022-3514.86.6.838 (39-14304-004 [pii]). Wegner, D. M., & Wheatley, T. (1999). Apparent mental causation. Sources of the experience of will. The American Psychologist, 54(7), 480–492. van der Wel, R. P. R. D. (2015). Me and we: Metacognition and performance evaluation of joint actions. Cognition, 161, 60–65. Wenke, D., Fleming, S. M., & Haggard, P. (2010). Subliminal priming of actions influences sense of control over effects of action. Cognition, 115(1), 26–38 (doi: S0010-0277(09) 00269-8 [pii]10.1016/j.cognition.2009.42016).
References Aarts, H., Custers, R., & Marien, H. (2009). Priming and authorship ascription: When nonconscious goals turn into conscious experiences of self-agency. Journal of Personality and Social Psychology, 96(5), 967–979. http://dx.doi.org/10.1037/ a0015000. Aarts, H., Custers, R., & Wegner, D. M. (2005). On the inference of personal authorship: Enhancing experienced agency by priming effect information. Consciousness and Cognition, 14(3), 439–458 doi: S1053-8100(04)00120-5 [pii]31016/ j.concog.2004.11.001. Bandura, A. (2001). Social cognitive theory: An agentic perspective. Annual Review of Psychology, 52, 1–26. http://dx.doi.org/10.1146/annurev.psych.52.1.1 (52/1/1 [pii]). Chambon, V., & Haggard, P. (2012). Sense of control depends on fluency of action selection, not motor performance. Cognition, 125(3), 441–451. http://dx.doi.org/10. 1016/j.cognition.2012.07.011. Chambon, V., Moore, J. W., & Haggard, P. (2015). TMS stimulation over the inferior parietal cortex disrupts prospective sense of agency. Brain Structure & Function, 220(6), 3627–3639. http://dx.doi.org/10.1007/s00429-014-0878-6. Dewey, J. A., Pacherie, E. A., & Knoblich, G. (2014). The phenomenology of controlling a
109
Contents lists available at ScienceDirect
Acta Psychologica journal homepage: www.elsevier.com/locate/actpsy
Outside influence: The sense of agency takes into account what is in our surroundings
T
⁎
Nicholas Hon , Yin-Yi Seow, Don Pereira Department of Psychology, National University of Singapore, 9 Arts Link, 117570, Singapore
A B S T R A C T
We are quite capable of distinguishing those outcomes we cause from those we do not. This ability to sense selfagency is thought to be produced by a comparison between a predictive representation of an outcome and the actual outcome that occurs. It is unclear, though, specifically what types of information can be entered into agency computations. Here, we demonstrate that information from non-target stimuli (stimuli that are not directly acted upon) incidentally present in our surroundings can influence predictions of outcomes, consequently modulating the sense of agency over clearly-defined target outcomes (those that occur to acted-upon stimuli). This provides the first evidence that our sense of agency is contextualized with respect to what is in our immediate visual environment. Furthermore, our data suggest that agency computations, instead of just a single comparison, may involve comparisons performed in stages, with different stages involving different types/ classes of information. A model of such multi-stage comparisons is described.
1. Introduction In general, we are able to effectively distinguish the outcomes we cause from those we do not. This ability to sense our agency in actions and outcomes is a central one because it grounds our “sense of self” (Knoblich & Flach, 2003) and enables us to distinguish ourselves from others and the operations of the natural world. Additionally, in our social interactions, we generally assign credit and apportion blame on the assumption that individuals can sense their own responsibility in given outcomes (Bandura, 2001; Haggard & Tsakiris, 2009). A sense of agency (SoA) is thought to be produced by a comparison process in which a mental representation of a predicted outcome is compared to the actual outcome that occurs (Frith, Blakemore, & Wolpert, 2000; Moore & Haggard, 2008; Sato & Yasuda, 2005; Wegner & Wheatley, 1999). A match between the predictive representation and the actual outcome produces a strong sense of agency. On the other hand, a mismatch between the two produces feelings of non-agency (or lack of agency). For instance, a ball that travels in a direction and with a speed consistent with the force with which it is thrown will likely produce a strong sense of agency in the thrower. A ball that travels backwards towards the thrower (a mismatch with what would be predicted regarding direction of movement), however, will more than likely produce a strong sense of non-agency. Supporting this, experimental manipulations that affect the relationship between predicted and actual outcomes have been found to reliably influence SoA,
⁎
producing agency distortions. On one hand, manipulations that act to minimize matching between predicted and actual outcomes result in a reduced sense of agency (Sato, 2009; Sato & Yasuda, 2005; Wenke, Fleming, & Haggard, 2010). For example, in one study, the application of mild spatial noise to an outcome such that spatial matches between predictions and actual outcomes were slightly reduced produced lower levels of SoA (Farrer & Frith, 2002). (It should be noted that outright spatial incongruence or mismatches (e.g., a stimulus moving left when the prediction is that it will go right) between predictions and actual outcomes will produce a clear sense of non-agency.) On the other, situations in which an actual outcome appears to match a predicted one result in an elevated sense of agency, and this is so even when an individual may not have had any real control over the outcome in question (Hon & Poh, 2016; Pronin, Wegner, McCarthy, & Rodriguez, 2006; Wegner, Sparrow, & Winerman, 2004; Wegner & Wheatley, 1999). In a seminal study (Pronin et al., 2006), participants who undertook a faux magical ritual aimed at harming another individual reported greater levels of SoA when the targeted person subsequently reported feeling ill. In reality, the ritual did not have any actual influence and the targeted individual was, in fact, a confederate. In this case, “vicarious agency” was produced because the outcome matched the participants' predictions regarding the outcome of the ritual. While the comparison model described above has been very successful in explaining its production, it is nonetheless the case that SoA is generally studied under highly constrained circumstances. In some
Corresponding author at: Department of Psychology, National University of Singapore, 9 Arts Link, 117570, Singapore. E-mail address: [email protected] (N. Hon).
https://doi.org/10.1016/j.actpsy.2018.03.004 Received 27 November 2017; Received in revised form 6 February 2018; Accepted 15 March 2018 Available online 28 March 2018 0001-6918/ © 2018 Elsevier B.V. All rights reserved.
Acta Psychologica 186 (2018) 104–109
N. Hon et al.
case. It is also worth pointing out that this paradigm allowed us to probe the matching of predictions while keeping target signals constant. Notice that target signals, in any given target condition, were the same regardless of what the distractors did. It is worth pointing out that what is being studied here is different from earlier studies that have reported that stimuli like primes can sometimes affect agency ratings over a target action/outcome (Aarts, Custers, & Marien, 2009; Aarts, Custers, & Wegner, 2005; Chambon & Haggard, 2012; Linser & Goschke, 2007; Wenke et al., 2010). While it is possible to characterize such primes as non-target stimuli, their purpose was to affect the fluency with which agentic actions could be performed (Chambon & Haggard, 2012; Sidarus et al., 2013). Thus, while they were non-targets, they were not extraneous to the task. The non-targets in the current study are different in that they neither hinder nor aid action choice or the actual movement of the target. Thus, the non-target stimuli of this study were designed to function more genuinely as extraneous environmental stimuli.
typical experiments, participants are introduced to a single stimulus that they try to control via some behavior (e.g., a button press or joystick movement), with the actual stimulus outcome being consistent with their actions or not (Dewey, Seiffert, & Carr, 2010; Ebert & Wegner, 2010; Hon, Poh, & Soon, 2013). In others, participants perform some motor action and are presented with somatosensory outcomes that are linked with the action or not, judging agency over the production of these outcomes (Chambon, Moore, & Haggard, 2015; Sidarus, Chambon, & Haggard, 2013; Wenke et al., 2010). In these paradigms, it is common for target stimuli/outcomes to be presented in isolation, being the only information presented at a given time. One concern is that this may give the impression that SoA is driven solely by signals from target stimuli/outcomes and nothing else. In everyday life, however, the stimulus one acts upon may not be the sole stimulus present in the environment. An open question, then, pertains to whether or not non-target stimuli (stimuli that are not acted upon) incidentally present in the environment can influence SoA over outcomes relating to an acted-upon target stimulus. To date, the literature has been silent on this issue. In this study, we investigated the possibility that information regarding extraneous, non-target stimuli can be factored into agency computations; in particular, the predictive representations of outcomes. Consider, for example, a standard table setting in which a dinner plate is flanked by a knife and fork. If one were to push the plate, one would clearly predict that it should move forward in a manner consistent with the push applied. If it does this when pushed, one should feel a strong sense of agency. In this case, though, the prediction involves only information from the target stimulus (i.e., the plate). However, it could also be predicted that, in addition to moving in a direction consistent with the push, the plate should end up further away from the knife and fork. Notice that, in this latter scenario, information about the knife and fork is extraneous to the acted-upon stimulus (i.e., the plate) but is built into the prediction nonetheless. In the current study, we investigated this issue - whether information about extraneous, non-target stimuli can be built into predictive representations. Here, participants viewed displays that comprised a central rectangle (target) flanked on the left and right by two other rectangles (non-targets). The target rectangle was always a different colour from the non-target rectangles (Fig. 1). Participants made self-decided and self-initiated presses of the up- or down-arrows on a standard computer keyboard when the display was presented. Following a button press, the target rectangle would move in a direction consistent with that indicated by the button press or not. The task was to assess agency over the movement of the target rectangle. Consistent with the literature, we would expect high agency ratings when the target's movement matched the direction indicated by the button press and low ratings when these were incongruent. Of greater relevance here, at the same time as the target moved, the adjacent non-targets could do one of three things: They could move in the same direction as the target rectangle, move in the opposite direction or remain stationary. If, as in the plate and cutlery example above, predictions include the (spatial) relationship between target and simultaneously-present non-target stimuli, then, in the current experiment, the prediction associated with a given buttonpress would be that the target rectangle should (a) move in a direction consistent with that indicated by the button press and (b) end up further away from the non-targets than before the action was performed. Accordingly, the best matches should occur when the target moves in a direction consistent with the button press and when it ends up clearly separated from the non-targets. Thus, in this study, when there is congruence between button press and target movement, one would expect the greatest SoA to be reported when the target and non-targets move in opposite directions. This would produce the greatest final distance between the two, offering the clearest evidence of a match with the prediction. Similarly, the lowest agency ratings should be observed when target and non-targets move in the same direction, as the final spatial separation between the two would be smallest in this
2. Experiment 1 2.1. Methods 2.1.1. Participants 29 undergraduate students from the National University of Singapore participated in this experiment.1 All participants had normal or corrected-to-normal vision. 2.1.2. Procedure In the experiment proper, participants completed a single block comprising 240 trials of a simple agency judgement task. On each trial, three coloured rectangles were presented side-by-side in the centre of a white screen (Fig. 1). The middle rectangle (designated the target) was of one colour, while the two flanking rectangles (non-targets) were of another. The two colours used for the rectangles were red and blue, and the assignment of the colours to target or non-targets was counterbalanced across participants. In response to the presentation of the rectangles, participants made self-initiated and self-decided up- or down-arrow key presses. A hundred milliseconds (100 ms) after the key-press, the target rectangle moved in a direction that was either congruent (e.g., moving up after an up-arrow key-press) or incongruent (e.g., moving down after an up-arrow key-press) with the direction indicated by the key-press. These were designated Target Congruent and Target Incongruent trials respectively. The distance travelled by the target rectangle was 5.14o of visual angle, regardless of direction. The movement of the target lasted 120 ms. Target Congruent and Target Incongruent trials each accounted for 50% of all trials. When the target moved, one of three things could happen with respect to the flanking non-targets. These could move in the same direction as the target (e.g., upwards when the target moved up), in an opposite direction to the target (e.g., downward when the target moved up), or they could remain in their original position. These are termed Non-target Same, Non-target Different and Non-target Neutral trials respectively. Non-target Neutral trials, in which the non-targets remained stationary accounted for 50% of all trials, with the remaining 50% of the trials being split equally between Non-target Same and Different conditions. Non-target movements began at the same time and were of the same speed as the target. However, they only moved one-third the distance travelled by the target (~1.74o of visual angle). Correspondingly, their movement lasted a shorter time (~ 40 ms) than that of the target. The difference in amount of movement was, like the use of different colours, meant to promote disambiguation between targets and non-targets. 1 Sample size was determined on the basis of other studies form our lab using the same base paradigm (e.g., Hon et al., 2013).
105
Acta Psychologica 186 (2018) 104–109
N. Hon et al.
Fig. 1. Illustration of a trial. Depicted is an example of a Target Congruent - Non-target Same trial. On such trials, the target (i.e. centre) rectangle moved in the direction indicated by the participant's key press, with the non-targets moving in the same direction as the target. Although greyscale was used in this figure, in the actual experiment, red and blue were used for the rectangles.
After the target rectangle had stopped moving, participants rated how much they felt their key-press caused its movement using a 7-point Likert-type scale (1: Not at all; 4: Unsure; 7: A lot). The stimuli were presented on a 21″ LCD monitor, with the experiment controlled by a PC running the E-Prime software. 2.2. Results The agency rating data from this experiment are depicted in Fig. 2. To start, a fully-within 2 (target: congruent, incongruent) × 3 (nontarget: same, different, neutral) ANOVA was performed on the agency ratings. This revealed a main effect of target congruence [F(1, 28) = 153.35, p < 0.001, ɳp2 = 0.84], a main effect of non-target movement [F(2, 56) = 6.43, p = 0.005, ɳp2 = 0.19] and, most importantly, a significant interaction between the two [F(2, 56) = 8.44, p = 0.005, ɳp2 = 0.23]. To elucidate this interaction, Target Congruent and target incongruent trials were examined separately. For Target Congruent trials (Fig. 2A), a significant effect of nontarget movement on agency ratings was found, F(2, 56) = 11.28, p < 0.001, ɳp2 = 0.29. A test of simple effects revealed that agency ratings in Non-target Different trials were significantly higher than in Non-target Same [t(28) = 3.97, p < 0.001] and Non-target Neutral trials [t(28) = 2.07, p = 0.048]. Non-target Neutral trials produced significantly higher agency ratings than Non-target Same trials, t (28) = 3.21, p = 0.003. This indicates that SoA over the movement of the target rectangle was affected by what the non-targets did. For Target Incongruent trials (Fig. 2B), on the other hand, no effect of non-target movement on agency ratings was found (F < 1, n.s.). This is unsurprising because a clear spatial incongruence between predicted and actual outcomes is known to be an unambiguous “nonagency” signal (Hon et al., 2013; Schmidt & Heumuller, 2010). These results are consistent with the idea that information about the non-targets was built into predictive representations of the outcome. Specifically, that, when acting on the target rectangle (via one's button press), the prediction would be that the target rectangle would both (a) move in a direction consistent with one's button press and (b) end up further away from the non-targets than before the action was performed. With respect to (b), on Target Congruent trials, the actual outcome that best matches the prediction would be when the final position of the target is furthest away from non-targets (i.e., clearest evidence that the two are separated), with the poorest match being when the eventual target-non-target separation is the smallest. In the Target Congruent condition, targets ended up furthest away from the non-targets on the Non-target Different trials (~6.84o of visual angle) and closest on the Same trials (~3.43o of visual angle). Agency ratings in the Target Congruent trials followed this pattern: Ratings were highest in the Non-target Different trials and lowest in the Same trials, with ratings on Neutral trials being intermediate between these.
Fig. 2. Agency ratings from the Target Congruent (A) and Target Incongruent trials (B) of Experiment 1. Error bars indicate 1 SEM.
strategically build these into their predictions (as spatial references)? To assess this possibility, we ran another experiment that was identical in set-up to Experiment 1 with the exception that the three non-target conditions – Same, Different and Neutral – now accounted for an equal number of trials each (one-third each). In this new experiment, the nontargets would, in fact, move more often than they would remain stationary. Would the fact that they now moved so often limit their inclusion in predictions? 3.1. Methods 3.1.1. Participants 30 participants, drawn from the same pool as those in Experiment 1, participated in this experiment. 3.1.2. Procedure The procedure and presentation parameters were identical to those in Experiment 1. The central difference was that, in this experiment, the total number of trials was evenly divided between the three non-target conditions.
3. Experiment 2 In Experiment 1, non-targets only moved 50% of the time. Could this spatial invariance on half the trials have prompted participants to 106
Acta Psychologica 186 (2018) 104–109
N. Hon et al.
3.2. Results
4. Experiment 3
Fig. 3 depicts the agency ratings from this experiment, separated by target congruence (Fig. 3A: Target Congruent trials; Fig. 3B: Target Incongruent trials). As with Experiment 1, a fully-within 2 (target: congruent, incongruent) × 3 (non-target: same, different neutral) ANOVA revealed a main effect of target congruence [F(1, 29) = 303.99, p < 0.001, ɳp2 = 0.91], a main effect of non-target movement [F(2, 58) = 3.58, p < 0.034, ɳp2 = 0.110] and a significant interaction between the two, F(2, 58) = 9.22, p < 0.001, ɳp2 = 0.24. Separate one-way ANOVAs on the two types of target trials revealed that, as before, there was no effect of non-target type on target incongruent trials (F < 1, n.s.). At the same time, there was a significant effect of non-target type on Target Congruent trials, F(2, 58) = 6.57, p = 0.003, ɳp2 = 0.19. Simple effects tests revealed that, for these trials, agency ratings in Non-target Different trials were higher than in Non-target Same (t(29) = 2.80, p = 0.009) and Non-target Neutral trials (t(29) = 2.16, p = 0.039). Non-target Neutral trials produced higher agency ratings than Non-target Same trials (t(29) = 2.30, p = 0.029). This is the exact same pattern observed in Experiment 1. Supporting this, a direct comparison between experiments revealed no evidence of an interaction between experiment (Experiment 1, Experiment 2) and the central target and non-target variables (F < 1, n.s.). The data from this experiment replicate the findings of Experiment 1, with agency ratings in the Target Congruent trials following the spatial separation between targets and non-targets: highest ratings for Non-target Different, followed by Neutral and then Same. Additionally, the fact that these findings were obtained, even though the non-targets moved more often than they stayed still, suggests that the inclusion of non-target information into predictions was unlikely to have been driven by a strategy based on the preponderance of non-targets remaining still (as in Experiment 1).
It is also possible that strategy may have played a role in another way: Participants may have been motivated to include non-targets in predictions because these appeared on all trials (as in Experiments 1 and 2). We ran a subsidiary experiment to assess this possibility. In this experiment, non-targets were only present on 50% of the trials.
Fig. 4 depicts the agency ratings from the critical trials (i.e., those with non-targets present) of this experiment, separated by target congruence (Fig. 4A: Target Congruent trials; Fig. 4B: Target Incongruent trials). As with Experiments 1 and 2, a fully-within 2 (target: congruent, incongruent) × 3 (non-target: same, different neutral) ANOVA was performed on the critical non-target present trials. This revealed a main effect of target congruence [F(1, 32) = 143.28, p < 0.001,
Fig. 3. Agency ratings from the Target Congruent (A) and Target Incongruent trials (B) of Experiment 2. Error bars indicate 1 SEM.
Fig. 4. Agency ratings from the Target Congruent (A) and Target Incongruent trials (B) of Experiment 3. Error bars indicate 1 SEM.
4.1. Methods 4.1.1. Participants 33 participants, drawn from the same pool as those in Experiments 1 and 2, participated in this experiment. 4.1.2. Procedure The procedure and presentation parameters were largely identical to those in Experiments 1 and 2. The key difference was that, in this experiment, non-targets were only present on half the trials, being absent on the other half. When non-targets were present, each non-target condition (Same, Different and Neutral) accounted for an equal number of trials. 4.2. Results
107
Acta Psychologica 186 (2018) 104–109
N. Hon et al.
Fig. 5. Model in which SoA is the result of comparisons involving both primary determinants and secondary modulators.
ɳp2 = 0.82], a main effect of non-target movement [F(2, 64) = 6.92, p = 0.002, ɳp2 = 0.18] and a significant interaction between the two, F (2, 64) = 13.45, p < 0.001, ɳp2 = 0.30. Separate one-way ANOVAs performed on the two types of target trials revealed that there was no effect of non-target movement on target incongruent trials (F < 1, n.s.). There was, however, a significant effect of non-target movement on Target Congruent trials, F(2, 64) = 15.31, p < 0.001, ɳp2 = 0.32. Simple effects tests revealed that, for these, agency ratings in Non-target Different trials were higher than in Non-target Same (t(32) = 6.15, p < 0.001) and Non-target Neutral trials (t(32) = 2.17, p = 0.038). Also, Non-target Neutral trials produced higher agency ratings than Non-target Same trials (t(32) = 3.51, p = 0.001). This is, of course, the same pattern observed in Experiments 1 and 2. Finally, we formally compared the findings from the 3 experiments with a mixed 2 (target: congruent, incongruent) × 3 (non-target: same, different, neutral) × 3 (experiment: 1, 2, 3) analysis. This analysis revealed main effects of target [F(1, 89) = 548.76, p < 0.001, ɳp2 = 0.86] and non-target [F(2, 178) = 16.36, p < 0.001, ɳp2 = 0.16] and a two-way interaction between these [F(2, 178) = 30.32, p < 0.001, ɳp2 = 0.25]. Critically, though, there was no main effect of experiment nor did it interact with the other two variables (all Fs < 1, n.s.). In other words, the same pattern of findings was obtained in all 3 experiments. Thus, even though non-targets may not always appear, when they are presented, they are factored into predictions.
conditions would not have been caused by the actual target signals, as these were held constant. Taken as a whole, the current findings suggest an important point about predictive representations: Predictions can include at least two different classes of information. The first relates to information that serves as the primary determinant of agency. In the current experiments, regardless of what the non-targets did, congruence between the button press and the target's movement produced ratings from the selfagency end of the scale (> 4), while incongruence produced ratings from the non-agency end of the scale (< 4). This suggests that, for the task used here, the primary determinant of agency was spatial congruence between the target's movement and the direction indicated by the button one presses. The second class involves information that merely acts to modulate SoA. In the current study, this would have been information involving the non-targets. Notice that, when there was congruence between target movement and the button press, none of the non-target movement conditions produced non-agency ratings (< 4). That being said, the different non-target conditions were able to modulate the level of SoA, with some conditions producing higher ratings than others. Altogether, this suggests that matching with respect to these two classes of information – primary determinants and secondary modulators – occurs in the following stage-wise fashion (Fig. 5). An initial stage assesses matching on the basis of the primary determinant of agency.2 If there is no match, then a strong sense of non-agency ensues. This is consistent with previous findings that clear mismatches between predictions and actual outcomes provides strong, unambiguous non-agency signals, which appear to be unalterable by additional factors (Schmidt & Heumuller, 2010; Shanks & Dickinson, 1991; Shanks, Pearson, & Dickinson, 1989). If, on the other hand, there is a match at this initial stage, then a sense of agency is produced. This sense of agency can then be up- or down-regulated at the second stage, in which matching is performed on the basis of information that acts as a secondary modulator. A good match on the secondary modulator produces a strong sense of agency, while a poor match produces a less strong SoA. As mentioned previously, a dominant idea is that SoA is produced by a single-stage comparison between predicted and actual outcomes. Some researchers, however, have posited that agency computations may involve more than one stage (Synofzik, Vosgerau, & Newen, 2008). Our data are broadly consistent with this latter idea. The current findings suggest that, rather than being underpinned by a single, uni-level comparison, the SoA one eventually experiences can be the product of comparisons performed at multiple levels, with each involving different types/classes of information. The cognitive literature has reported many examples of behavior
5. General discussion The present study set out to explore if information about non-targets present in the immediate environment can affect the sense of agency over a clearly-defined target outcome. In particular, we investigated the hypothesis that predictive representations can include information about incidentally present non-targets. Our findings from 3 experiments support this idea. The current findings are consistent with the idea that information about non-targets was built into predictive representations. In the current study, the prediction would be that, when the target rectangle is acted upon (by one's button press), it will move in a direction consistent with the button pressed and will end up further away from the nontarget rectangles than at the start of the trial (i.e., before any action was performed). The greatest SoA should, therefore, be produced when the actual outcome matches both parts of this prediction. This is what was found. When the target moved in a direction that was congruent with the button press, the highest agency ratings were reported when evidence of a final spatial separation between target and non-targets was clearest (i.e. Non-target Different trials). As mentioned previously, this difference in matching across the different distractor movement
2 The specific type of information (e.g., spatial, temporal) that functions as the primary determinant would be determined by the nature of the paradigm (or, more specifically, the type of actions and outcomes that are involved in this).
108
Acta Psychologica 186 (2018) 104–109
N. Hon et al.
being affected by the congruence between target and simultaneously presented non-target information. Typically, performance is adversely affected by an incongruence between the two (e.g., Stroop effect, flankers distraction). If such a phenomenon underpinned the current findings, one would have expected the lowest ratings to be observed when targets and non-targets moved in opposite directions (i.e., when target and distractor effects were incongruent). However, the opposite pattern was found here. Thus, we can rule this out as an alternative account. Alternatively, could participants have interpreted non-target movements as causing target movements? For example, perhaps, on congruent trials, participants viewed non-targets as causing target outcomes when the two moved in the same direction (i.e., Target Congruent - Non-target Same trials). This possibility is also unlikely. If participants felt that non-targets caused target movements, then ratings on those trials should have fallen within the non-agency range (< 4). This was not observed. Thus, this alternative can also be ruled out. In conclusion, the current results demonstrate that we contextualize our agentic operations with respect to what is in our (physical) surroundings, allowing for non-target signals (signals from stimuli that are not directly acted upon by us) to be incorporated into self-agency computations. It is interesting to note that other studies have demonstrated that an individual's sense of agency can be contextualized with respect to what others do (Dewey, Pacherie, & Knoblich, 2014; van der Wel, 2015). Taken together, those findings and ours paint a picture of SoA as being flexibly sensitive to both our social and physical settings. The results from the experiments reported here indicate that our predictive representations of an outcome, in particular, can include information about both target and non-target stimuli. This clarifies an important point about such representations. In studies in which actedupon target stimuli and their outcomes are presented in isolation, predictions could quite reasonably be thought to involve only information about said stimuli and outcomes. Here, the data suggest that extraneous, non-target information present in our immediate visual environment can also be built into predictive representations, being able to exert an influence over the resultant SoA.
moving object with another person. Cognition, 132, 383–397. Dewey, J. A., Seiffert, A. E., & Carr, T. H. (2010). Taking credit for success: The phenomenology of control in a goal-directed task. Consciousness and Cognition, 19(1), 48–62 (doi: S1053-8100(09)00144-5 [pii]101016/j.concog.2009.09.007). Ebert, J. P., & Wegner, D. M. (2010). Time warp: Authorship shapes the perceived timing of actions and events. Consciousness and Cognition, 19(1), 481–489 (doi: S10538100(09)00154-8 [pii]10.1016/j.concog.2009.12002. Farrer, C., & Frith, C. D. (2002). Experiencing oneself vs another person as being the cause of an action: The neural correlates of the experience of agency. Neuroimage, 15(3), 596–603. Frith, C. D., Blakemore, S. J., & Wolpert, D. M. (2000). Abnormalities in the awareness and control of action. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 355(1404), 1771–1788. http://dx.doi.org/10.1098/rstb.2000. 0734. Haggard, P., & Tsakiris, M. (2009). The experience of agency: Feelings, judgments and responsibility. Current Directions in Psychological Science, 18, 242–246. Hon, N., & Poh, J. H. (2016). Sleep deprivation produces feelings of vicarious agency. Consciousness and Cognition, 40, 86–92. http://dx.doi.org/10.1016/j.concog.2015.12. 007. Hon, N., Poh, J. H., & Soon, C. S. (2013). Preoccupied minds feel less control: Sense of agency is modulated by cognitive load. Consciousness and Cognition, 22(2), 556–561 (doi: S1053-8100(13)00037-8 [pii]171016/j.concog.2013.03.004. Knoblich, G., & Flach, R. (2003). Action identity: Evidence from self-recognition, prediction, and coordination. Consciousness and Cognition, 12(4), 620–632. Linser, K., & Goschke, T. (2007). Unconscious modulation of the conscious experience of voluntary control. Cognition, 104(3), 459–475 (doi: S0010-0277(06)00160-0 [pii] 201016/j.cognition.2006.07.009). Moore, J., & Haggard, P. (2008). Awareness of action: Inference and prediction. Consciousness and Cognition, 17(1), 136–144 (doi: S1053-8100(07)00004-9 [pii] 221016/j.concog.2006.12.004). Pronin, E., Wegner, D. M., McCarthy, K., & Rodriguez, S. (2006). Everyday magical powers: The role of apparent mental causation in the overestimation of personal influence. Journal of Personality and Social Psychology, 91(2), 218–231 (doi: 200609808-002 [pii]241037/0022-3514.91.2.218). Sato, A. (2009). Both motor prediction and conceptual congruency between preview and action-effect contribute to explicit judgment of agency. Cognition, 110(1), 74–83 (doi: S0010-0277(08)00238-2 [pii]10.1016/j.cognition.2008.26011). Sato, A., & Yasuda, A. (2005). Illusion of sense of self-agency: Discrepancy between the predicted and actual sensory consequences of actions modulates the sense of selfagency, but not the sense of self-ownership. Cognition, 94(3), 241–255 (doi: S001002770400099X [pii]281016/j.cognition.2004.04.003). Schmidt, T., & Heumuller, V. C. (2010). Probability judgments of agency: Rational or irrational? Consciousness and Cognition, 19(1), 1–11 (doi: S1053-8100(10)00006-1 [pii]301016/j.concog.2010.01.004). Shanks, D. R., & Dickinson, A. (1991). Instrumental judgment and performance under variations in action-outcome contingency and contiguity. Memory & Cognition, 19(4), 353–360. Shanks, D. R., Pearson, S. M., & Dickinson, A. (1989). Temporal contiguity and the judgement of causality by human subjects. Quarterly Journal of Experimental Psychology, 41, 139–159. Sidarus, N., Chambon, V., & Haggard, P. (2013). Priming of actions increases sense of control over unexpected outcomes. Consciousness and Cognition, 22(4), 1403–1411. http://dx.doi.org/10.1016/j.concog.2013.09.008. Synofzik, M., Vosgerau, G., & Newen, A. (2008). Beyond the comparator model: A multifactorial two-step account of agency. Consciousness and Cognition, 17(1), 219–239 (doi: S1053-8100(07)00026-8 [pii]361016/j.concog.2007.03.010). Wegner, D. M., Sparrow, B., & Winerman, L. (2004). Vicarious agency: Experiencing control over the movements of others. Journal of Personality and Social Psychology, 86(6), 838–848. http://dx.doi.org/10.1037/0022-3514.86.6.838 (39-14304-004 [pii]). Wegner, D. M., & Wheatley, T. (1999). Apparent mental causation. Sources of the experience of will. The American Psychologist, 54(7), 480–492. van der Wel, R. P. R. D. (2015). Me and we: Metacognition and performance evaluation of joint actions. Cognition, 161, 60–65. Wenke, D., Fleming, S. M., & Haggard, P. (2010). Subliminal priming of actions influences sense of control over effects of action. Cognition, 115(1), 26–38 (doi: S0010-0277(09) 00269-8 [pii]10.1016/j.cognition.2009.42016).
References Aarts, H., Custers, R., & Marien, H. (2009). Priming and authorship ascription: When nonconscious goals turn into conscious experiences of self-agency. Journal of Personality and Social Psychology, 96(5), 967–979. http://dx.doi.org/10.1037/ a0015000. Aarts, H., Custers, R., & Wegner, D. M. (2005). On the inference of personal authorship: Enhancing experienced agency by priming effect information. Consciousness and Cognition, 14(3), 439–458 doi: S1053-8100(04)00120-5 [pii]31016/ j.concog.2004.11.001. Bandura, A. (2001). Social cognitive theory: An agentic perspective. Annual Review of Psychology, 52, 1–26. http://dx.doi.org/10.1146/annurev.psych.52.1.1 (52/1/1 [pii]). Chambon, V., & Haggard, P. (2012). Sense of control depends on fluency of action selection, not motor performance. Cognition, 125(3), 441–451. http://dx.doi.org/10. 1016/j.cognition.2012.07.011. Chambon, V., Moore, J. W., & Haggard, P. (2015). TMS stimulation over the inferior parietal cortex disrupts prospective sense of agency. Brain Structure & Function, 220(6), 3627–3639. http://dx.doi.org/10.1007/s00429-014-0878-6. Dewey, J. A., Pacherie, E. A., & Knoblich, G. (2014). The phenomenology of controlling a
109