Physical and mental effort disrupts the implicit sense of agency

Abstract

We investigated the effect of effort on implicit agency ascription for actions performed under varying levels of physical effort or cognitive load. People are able to estimate the interval between two events accurately, but they underestimate the interval between their own actions and their outcomes. This effect is known as ‘intentional binding’, and may provide feedback regarding the consequences of our actions. Concurrently with the interval reproduction task, our participants pulled sports resistance bands at high and low resistance levels (Experiments 1 and 2), or performed a working memory task with high and low set-sizes (Experiment 3). Intentional binding was greater under low than high effort. When the effort was task-related (Experiment 1), this effect depended on the individual’s explicit appraisal of exertion, while the effect of effort was evident at the group level when the effort was task-unrelated (physical, Experiment 2; mental, Experiment 3). These findings imply that the process of intentional binding is compromised when cognitive resources are depleted, either through physical or mental strain. We discuss this notion in relation to the integration of direct sensorimotor feedback with signals of agency and other instances of cognitive resource depletion and action control during strain.

Introduction

It is important that the human motor system can efficiently process events which are the result of its own actions, and to discriminate these from events in the world for which it is not responsible. For instance, if I kick a ball and it knocks over and smashes a vase of flowers, I know my action of kicking the ball was responsible for the vase smashing. Self-authored events like this tend to be easy to identify and this feeling that ‘I did it’ is known as a sense of agency. Some actions are more effortful than others; kicking a ball as hard as one can might break a vase, but so might brushing one’s arm against it as one walks past it precariously positioned near the edge of a table. Both these actions have the same outcome, but might require the action monitoring system to respond differently in order to correctly ascribe agency. Here, we tested the role of physical and mental effort on the ascription of the sense of agency, using an implicit measure.

Self-agency is detected where there are cues relating to intentionality, volition, predictability and contiguity. An interesting phenomena occurring under these conditions is that actions and their effects are perceived as occurring closer together in time than they did, an effect known as temporal or intentional binding (Barlas and Obhi, 2013, Engbert and Wohlschläger, 2007, Engbert et al., 2007, Haggard et al., 2002; see Moore & Obhi, 2012, for a review). One theory of subjective time perception suggests that ‘ticks’ of an ‘internal clock’ give rise to our sense of time passing. The pace of this clock varies with arousal and motor activity (Gibbon et al., 1984, Treisman, 1963, Wearden et al., 1999). When the pace of neural ‘ticks’ slows, durations appear shorter due to the accumulation of fewer pacing ‘ticks’. Conversely, when the pace of the ‘ticks’ quickens, durations appear longer. Contexts characteristic of self-agency are believed to slow the pace of the internal clock as a consequence of motor prediction. This results in the shortening of subjective time and temporal binding (Wenke & Haggard, 2009). This may be an adaptive process to help to create a sense of agency, and in a general sense this process could assist the sensorimotor system to identify and monitor its effects and optimise performance (Buhrmann and Di Paolo, 2015, Wenke and Haggard, 2009).

Several accounts as to how agency is attributed to the self have been suggested, including the forward predictive comparator model (Blakemore et al., 2001, Blakemore et al., 2002, Wegner et al., 2004), the post-hoc inference account (Wegner & Wheatley, 1999), and the optimal cue integration account (Moore and Fletcher, 2012, Synofzik et al., 2010, Synofzik et al., 2009, Synofzik et al., 2013). The comparator model provides a predictive account of agency attribution, characterised by the comparison between predicted action effects with actual action effects. Congruence between these effects results in perceived self-authorship and a sense of agency, whereas incongruence between them results in a diminished sense of agency.

An alternative model, the post-hoc inference account, provides a post-dictive re-constructionist account of agency attribution. Here, sense of agency self-attribution is dependent upon reflection on the action-effect relationship after the outcome has occurred. For instance, when there is an intention to act, when the perceived effects can be explained by the intended action, and when there is no other plausible cause for the effect, sense of agency is then experienced and retrospectively introduced into consciousness. Finally, the optimal cue integration account recognises the importance of both pre- and post-dictive cues. These cues are then weighted for their reliability for agency attribution depending on the context and then used to determine self-authorship. The ability to construct these cues and make comparisons between expected and actual effects of actions may depend on the availability of cognitive resources. Indeed, diminished attentional resources have been shown to impair explicit ratings of agency (Hon, Poh, & Soon, 2013). Moreover, studies concerning cognitive load, kinematics and motor awareness offer support to this prediction. Dual task paradigms employing mental arithmetic, memory tasks and fine motor movements during balance, gait, posture and walking tasks have shown reductions in motor control and motor awareness (Kannape et al., 2014, Lindenberger et al., 2000, Woollacott and Shumway-Cook, 2002). Dual task performance models suggest that motor control and cognitive activity compete for cognitive resources (Huxhold et al., 2006, Lacour et al., 2008). In such cases motor control and awareness become less efficient due to cross-domain resource competition. The deficits in motor awareness caused by limited cognitive resources are especially of relevance to the sense of agency as, in accordance with the forward models of motor control, it is a crucial factor in the ability to monitor self-initiated actions. This cognitive resource limitation notion therefore has interesting implications regarding the role cognitive resource availability may have on constructing the attribution of agency.

We are concerned primarily with how effort might influence the implicit sense of agency. It is important then to note that despite appearing to be independent concepts, physical and mental effort similarly put strain on the cognitive system by expending cognitive resources (Dietrich, 2003, Dietrich and Sparling, 2004, Franconeri et al., 2013). Mental and physical effort therefore draw from and deplete a common cognitive resource. It is also important to note that exertion influences perception in other domains. For instance, perceived distance increases and hills seem steeper under conditions requiring more physical exertion (e.g. when carrying a heavy load; Bhalla and Proffitt, 1999, Sugovic and Witt, 2013, Witt et al., 2004). These apparent spatial distortions as a function of required effort are also mirrored for the perception of time.

A recent meta-analysis investigating the effect of physical load on duration judgements revealed that physical workload results in longer perceived durations (Block, Hancock, & Zakay, 2016; seven studies spanning from 1963–2011). However, the impact of effort on perceived duration need not be experiential, as stimuli that allude to motion, action, or exertion also elongate perceived durations. For example, faster moving non-biological stimuli are perceived to last longer than slower moving stimuli (Brown, 1995, Kaneko and Murakami, 2009), and the perceived duration of images of ballet dancer statues are lengthened when the poses reflected greater levels of exertion (Nather, Bueno, Bigand, & Droit-Volet, 2011). The elongation of subjective time as a result of effort are also found for mental activity. Depleted attentional resources and increased cognitive load result in longer retrospective subjective time judgements (for a review see Block, Hancock, & Zakay, 2010). Given these findings highlighting the similarities between the effects of physical and mental effort on time perception, one can hypothesise that physical and mental exertion could have similar disruptive effects on temporal measures of the sense of agency, driven by the depletion of cognitive resources (Dietrich, 2003, Dietrich and Sparling, 2004, Franconeri et al., 2013, Hon et al., 2013).

Given what we know about the effects of physical and mental exertion on performance and perception, and assuming that ascribing agency is a costly cognitive process, the hypothesis follows that agency should be reduced under conditions of mental or physical effort. There is some support for this hypothesis from a study using an explicit measure, where participants reported the degree to which they felt agency over an event (Hon et al., 2013). These authors found that explicit ratings of agency over a dot that moved following an arrow key press were reduced under conditions of high cognitive load, which was manipulated using a working memory task. This is an interesting finding but converging evidence using an implicit task would be valuable in understanding the processes involved in agency attribution during strain. This is especially important given that explicit and implicit measures are sometimes found to be dissociated (Dewey and Knoblich, 2014, Obhi and Hall, 2011).

Some studies that have used implicit measures lend indirect support to the notion that mental effort disrupts the implicit sense of agency, though their research questions did not directly address the role of cognitive load. Specifically, temporal binding for self-actions has been shown to be weaker when the outcome of the action is socially negative (Yoshie & Haggard, 2013), or when the actor performs the action under coercion (Caspar, Christensen, Cleeremans, & Haggard, 2016). Individual ratings of agency over outcomes to actions have also shown to be lower when there is conflict in action selection caused by distractor stimuli (Sidarus & Haggard, 2016). Indeed, each of these acts imply a significant degree of cognitive conflict (see Greene, Nystrom, Engell, Darley, & Cohen, 2004), which would make their findings appear broadly in line with our hypothesis.

A general explanation of these effects could be drawn from the effect of resource depletion on time perception. As noted above, subjective time lengthens under load (Block et al., 2010, Block et al., 2016), which would result in less temporal binding (i.e. smaller underestimation errors) in agency conditions due to the resources required to bind the action and its effect together already being committed to the cognitively effortful primary task. However, the only research to have directly addressed cognitive effort during the task itself used an explicit measure (Hon et al., 2013), as did the research inducing conflict in action selection (Sidarus & Haggard, 2016). The work by Caspar et al. (2016) and Yoshie and Haggard (2013), although employing implicit measures, manipulated the outcome of, or the motivation for, the action, rather than the effort context under which the action itself was performed. Therefore, a direct test of how effort influences agency using an implicit measure would be valuable in understanding the processes involved in agency attribution during strain, and determining if effort affects interval estimation measures of sense of agency in a comparable way to explicit ratings previous found (Hon et al., 2013).

One recent study has more directly investigated the effect of physical effort exerted during the action, using an implicit measure (Demanet, Muhle-Karbe, Lynn, Blotenberg, & Brass, 2013). Levels of physical task-unrelated effort were manipulated so that the arm contralateral to the active hand used to carry out the time perception measure held an exercise resistance band. The authors used the ‘Libet clock’ method in which participants attempted to state where the hand on an analogue clock face had been both when they performed their action (a key press) and also when they heard the consequent tone. They compared these trials to baseline trials in which either only a key press occurred or only a tone sounded. In contrast to predictions one could derive from the above discussed research, Demanet et al. found that greater effort increased temporal binding, therefore implying a stronger implicit sense of agency under conditions of high task-unrelated physical effort. These findings were attributed to the notion that effortful actions boost the interoceptive sensory-motor information of willed effort, which may act as a cue to self-agency (Haggard et al., 2002, Lafargue and Franck, 2009, Vierkant, 2014).

The pattern of data observed by Demanet et al. (2013) and the conclusion drawn accord with Maine de Biran’s hypothesis (1805, as cited in Demanet et al.) that effort is a cue to identify self-agency. This explanation is convincing and a good account of their finding of enhanced temporal binding under effort. Nevertheless, their finding is contrary to what one might predict given that it is known that physical effort depletes cognitive resources, which is associated with time expansion, not compression, and that the depletion of cognitive resources reduces motor control and awareness - elements fundamental for agency ascription (Block et al., 2010, Block et al., 2016, Huxhold et al., 2006, Kannape et al., 2014). Hence, it might be expected that the opposite effect would be observed; that exertion would result in disrupted – not enhanced - intentional binding. If the role of effort in reducing cognitive resources specifically disrupts the ascription of agency – rather than a general disruption of time perception - then it can be expected that effort will not have a similar disruptive effect during a passive control task in which agency is absent. To provide evidence that can help to establish the role of effort in implicit agency, we present three experiments that for the first time to our knowledge investigate the effect of task-related physical effort, and in further experiments, task-unrelated physical and mental effort, on implicit agency attributions.

To determine whether physical and mental effort modulate agency over actions and their outcomes, a temporal binding paradigm measuring implicit agency was used. In particular, an interval reproduction paradigm, previously demonstrated to successfully show temporal binding in agentic tasks (Buehner and Humphreys, 2009, Engbert et al., 2008, Humphreys and Buehner, 2010, Poonian and Cunnington, 2013), was used here to measure implicit sense of agency. Experiments 1 and 2 investigate the role of task-related and task-unrelated physical effort on agency, respectively. In this way, Experiment 2 is a close replication of Demanet et al. (2013), except that we use an interval reproduction method of intentional binding and they used the ‘Libet clock’ method. Experiment 3 extends the examination of the effects of effort for sense of agency to the cognitive domain, investigating the role of mental effort in the form of cognitive load, for the ascription of agency. Experiment 3 can be compared with Hon et al. (2013) who investigated explicit measures of agency under varying load.

We believe these studies are the first to investigate the role of task-related physical effort on implicit agency. Moreover, these experiments are the first to investigate the influence of physical and mental effort on sense of agency using the interval reproduction paradigm. Mental effort has also not before been directly investigated for its role in modulating implicit agency. Therefore, these features of our approach and design afford a more direct appraisal of different sources of effort on the implicit sense of agency than has been undertaken previously.

Given the common effects of physical and mental effort depleting cognitive resources, we can predict that high effort will result in a decrease in temporal binding across physical and mental effort in agentic conditions. However, the only previous study to investigate physical effort and sense of agency, found the converse effect, that high effort increased implicit agency (Demanet et al., 2013). Our investigation therefore also aims to clarify whether there is a general effect of effort on agency, or whether the specific context and source of the effort plays a critical role in the extent to which an implcit sense of agency is generated.