Whether for simple or more complex tasks, accurate perception of one’s own motion is crucial. It is now widely accepted that accurate perception of self-motion requires integrating a variety of information provided by both environment-centred (e.g., optic flow) and body-centred cues (e.g., vestibular, proprioceptive inputs and motor output; Cheng and Gu, 2018; Cullen, 2019). During the last decades, the human brain has been credited with developing adaptive mechanisms that contribute to an optimal integration of multisensory cues by combining redundant and complementary inputs accounting for stimulus characteristics (Ernst and Banks, 2002; Alais and Burr, 2004; Ernst and Bülthoff, 2004; Fetsch et al., 2013). Seminal works stressed the importance of sensory reliability of inputs in multisensory integration for self-motion perception (Gu et al., 2008; Morgan et al., 2008; ter Horst et al., 2015).

As promoted by the concept of Active sensing (Kveraga et al., 2007; Schroeder et al., 2010; van Atteveldt et al., 2014), it has been recently suggested that multisensory processing does not depend solely on the nature of sensory inputs, but also on the motor and attentional contexts of an action (Donohue et al., 2015). A growing body of studies recently investigated perceptual responses in situations where an external event (stimulus) is the result of an intentional action (van Kemenade et al., 2016; Arikan et al., 2017; Straube et al., 2017). Sensory integration during self-generated actions has been found to be modulated at both physiological (Hughes et al., 2013) and behavioural (Bays et al., 2006) level compared to the processing of the same sensory inputs generated by an external system. Yet most studies exploring self-motion perception are limited to “passive” stimulations, i.e., where motion is not self-generated (Cheng and Gu, 2018). So far, visuo-vestibular integration for motion perception has not been investigated through the prism of intentional action.

However, the link between intention and perception has been widely studied within the theoretical framework of agency. Indeed, the sense of agency describes the subjective feeling associated with controlling one’s own actions and, through these actions, events in the outside world (Haggard and Tsakiris, 2009; Chambon and Haggard, 2012). Agency is largely explained through a comparator model (CM) that describes internal computational predictive mechanisms of human action control (Frith et al., 2000; Blakemore et al., 2002). Interestingly, previous work has highlighted the dependence of the vestibular system on this model. Indeed, a decrease in response of the VO (vestibular only) neurons was specifically observed when the efference copy due to active motion was in agreement with current sensory inputs, both for rotation (Roy and Cullen, 2004) and translation (Carriot et al., 2013). Specifically, it has recently been reported that the active vs. passive internal state distinction at an early stage of integration may be a source of modulation in the computation of motion (Gu, 2018; Brooks and Cullen, 2019; Cullen and Wang, 2020; Cullen and Zobeiri, 2021). There is, however, strikingly little information on the consequences of active versus passive states on self-motion perception. In this context, our study explored the impact of the intentional nature of an action triggering a visuo-vestibular stimulation on self-motion perception, as seen through the prism of agency.

Yon and Frith (2021) recently argued that intentional action comes with prior knowledge (e.g., prediction) that could be used to optimise perception in an uncertain world Yon and Frith (2021). In addition, recent work demonstrates that being active (in terms of motor control of the action) potentiates the integration of relevant cues at the audio-visual level (van Kemenade et al., 2016; Arikan et al., 2017). It can therefore reasonably be hypothesised that being in control of an action may optimise the integration of the different sensory inputs relevant to the task at hand. Here, we speculate that the agentive context of an action may help reduce uncertainty by promoting multisensory integration of relevant inputs. Our hypothesis is consistent with the fact that action can be considered as a powerful way to reduce uncertainty, since it allows better prediction of outcomes (Yon et al., 2018). We therefore speculate that this reduction of uncertainty during sensory integration would help refine the distinction of motions whose characteristics would slightly vary. Thus, being intentionally active during a perceptual task may be particularly valuable in situations with a high level of sensory ambiguity.

The present study sought to explore how being active might impact the perception of one’s own motion at different levels of sensory ambiguity. Specifically, we aimed at investigating to what extent having control over one’s own motion help distinguish it from an externally generated motion under uncertainty. To that purpose, we used a two-alternative forced-choice (2AFC) discrimination task adapted from previous studies on audition (Reznik et al., 2015; Paraskevoudi and SanMiguel, 2021). In this task, the participants had to compare two displacements and identify which was larger. To compare active versus passive motion, two different conditions were presented. In the active condition, the first displacement was intentionally triggered by the participant whilst the second was externally generated (that is, without any participant-intentional action). In the passive condition, both displacements were externally generated. The first displacement had a fixed acceleration value, and the second displacement varied around this value. Two levels of ambiguity were introduced into the discrimination task by manipulating acceleration ratios between the two movements. We expected better perceptual discrimination between movements when participants intentionally triggered the first displacement than when both displacements were externally generated, particularly under a high level of ambiguity between movements.

According to Bergson’s thinking (1896), voluntary action is linked to perception. Thus, understanding what governs being in control could help to elucidate human perception. Exploring the intentional nature of action may shed light on the mechanisms underlying perception. Recently, it has been proposed that action control could be a way to sharpen representation of expected outcomes (Yon et al., 2018; Jagini, 2021). Based on these premises, the present study explored the role of action control in the perception of one’s own motion at different levels of sensory ambiguity. Indeed, we designed a discrimination task enabling us to manipulate both the ambiguity of pairs of displacements and the voluntary nature of the action. Performance on the discrimination task as well as participants’ reported confidence in their perception of the displacements were considered for each trial. We expected better perceptual discrimination between displacements when participants intentionally triggered the first movement, especially under a high level of motion ambiguity.

Analysis of the psychometric curves and of confidence levels highlighted three main results. First, we found better sensitivity (lower JND) in the active condition than in the passive condition, specific to a high level of ambiguity. Second, we observed an effect of agentive condition (higher PSE in the manual condition) but not of the level of ambiguity on PSE (perceptual bias). Third, we observed a general effect of agentive condition, with a higher level of confidence under the agentive condition but no effect of interaction with level of ambiguity.

The two main parameters (PSE and JND) obtained from participants’ psychometric curves indicate the way they performed the task. The PSE represents the value to be added to the standard displacement for the comparison displacement to be perceived as equal. It indicates how that standard displacement is perceived since a difference in perceptual bias (PSE) between conditions represents either a perceptual attenuation or an enhancement of the first displacement relative to its physical reality. We did not observe any effect of level of ambiguity. In other words, the first displacement was perceived identically under both levels of ambiguity. Also, we did not find any interaction effect between ambiguity and agentive conditions. In contrast, we did find a general effect of the agentive condition with higher PSE in manual condition. Overall, participants tended to enhance the first displacement in the agentive condition. In recent years, the control of action has been associated with both phenomena of attenuation and enhancement of the related sensory processing. It seems here that the context of action and its sensory consequences themselves play indeed a role in the underlying mechanisms. For example, previous studies suggest that this may depend on the intensity of the sensory consequences (Reznik et al., 2015; Paraskevoudi and SanMiguel, 2021). In our study, two levels of intensity have been manipulated (“High” and “Low”). However, they can be both considered as rather low displacement intensities (as reported by the participants). We can assume here that having control over the first displacement increases in both cases its perception to better perceive it and compare it to the second one. The second parameter (JND) is particularly informative on sensitivity of discrimination between displacements, representing the value at which a difference between the two displacements is perceived: the lower the JND, the better the discrimination performance. Better sensitivity (lower JND) was observed in the active condition with a high level of ambiguity than in the passive condition. Taken together, these results suggest that having control over the first displacement lead to higher stability of mechanisms enabling its discrimination from the second when ambiguity vary. Therefore, being active leads to better discrimination of controlled information (first displacement) from uncontrolled information (second displacement) when ambiguity increases.

It should be noted that previous studies using the same type of discrimination task observed different effects of the active condition on these two parameters (Reznik et al., 2015; Paraskevoudi and SanMiguel, 2021). For example, Paraskevoudi and SanMiguel (2021) found reduced perceived intensity (i.e., perceptual bias) for self-generated sounds when presented at supra-threshold intensities (high level), but which increased when presented at near-threshold intensities (low level). Also, the authors found no difference in discrimination sensitivity (i.e., JND). The fact that we extended the aforementioned paradigm to cover several new and important aspects could explain such differences.

First, in our study, participants had control over both the level of intensity they wanted to work at and the timing of the first displacement (the pair of displacements starting as soon as they confirmed their choice). Therefore, the agentive condition strongly mobilises the participant’s intention in the forthcoming comparison pair, rather than in continuous control of the displacements. We decided to add this intentional selection because the choice between different possible actions strongly influences the experience of control and agency (Barlas and Obhi, 2013; Barlas et al., 2017; Barlas and Kopp, 2018). One could hypothesise here that it comes from a higher predictability of event following the choice (i.e., regarding both timing and intensity of the first displacement). In addition, we speculate that the predictive mechanisms involved in this agentive situation lie more in the predictability and attentional commitment related to the participant’s intention. Indeed, it is now established that agency is associated with a better commitment to the task (Caspar et al., 2016), with the mobilisation of attentional mechanisms (Wen and Haggard, 2018). Since the participant’s intention was always respected, we speculate that being in intentional control of the first displacement allowed for a better integration of the related information. More precisely, according to the comparator model, the agentive situation refines the comparative action-consequence mechanisms linking action to its consequences. We can speculate that this serves a better integration of sensory inputs related to the action. Besides, it has recently been suggested that action may help sharpen sensory integration of relevant sensory outcomes (Schroeder et al., 2010; Yon et al., 2018). In addition, we did observe a general stability of both parameters (i.e., perceptual bias and sensitivity) when ambiguity vary. Indeed, we note that sensitivity of discrimination between the two levels of ambiguity remained constant in the active condition whereas it decreased in the passive condition (higher JND) when ambiguity increased. Thus, we speculate that intentional control led to smaller prediction error on this first displacement, allowing better detection of the prediction error gap on a second uncontrolled displacement, even when ambiguity increased. In our high ambiguity condition, the differences in the second displacement were more difficult to perceive (in terms of ratio from the standard displacement). Rather than demonstrating better discrimination in ambiguous situations under the agentive condition, our results tend to show a reduction in performance under the passive condition when the level of ambiguity increases. Therefore, a lower error signal in the agentive condition appears to promote stability and consistency in integrative performance when ambiguity increases. In this case, the prediction error difference between the first controlled displacement and the second uncontrolled displacement remains sufficient to maintain discrimination. In contrast, a fully passive situation generates higher prediction error, preventing the detection of smaller deviations from the first displacement during the second. Thus, our results highlight the impact of the decrease in prediction error in an agentive situation when there is a match between intentional control and the consequences. The fact that the difference from the passive situation is observed under more ambiguous conditions strengthens this hypothesis.

Second, we considered a far more complex multisensory stimulation than in previous studies, which mainly considered unimodal input (mainly bip sound). Recent studies suggest that predictive mechanisms engaged during an action promote the binding of sensory inputs relevant to the task at hand (van Kemenade et al., 2016; Arikan et al., 2017; Straube et al., 2017). Since our results differ from those of previous studies conducted under a specific single sensory stimulation, we also speculate that action may shape sensory consequences differently depending on the amount of task-relevant information available. Also, it is undeniable that future studies would clarify the part of each of underlying mechanisms in such agentive condition (i.e., attention, prediction, predictability, choice).

In addition, we performed an analysis of the confidence levels and of the metacognition of the participants. Having observed a general effect of being in control on confidence levels, but no interaction effect with the level of ambiguity, we decided to complement our analysis by considering metacognition variables. The meta-d’ variable was used to evaluate the correlation between the participant’s confidence in his/her performance and the reality of the performance itself (Maniscalco and Lau, 2012; Fleming and Lau, 2014). Therefore, when this score decreases, the participant’s metacognitive performance decreases. An effect of the level of ambiguity on metacognitive performance is observed: as the level of ambiguity increases, metacognitive performance decreases. Interestingly, we observe a trend towards interaction effects with the agentive condition. Associated with the graphical representation of these results (Figure 5B), the metacognitive performance can be seen to strongly decrease in the passive condition, compared to the active condition.

One limitation here is the number of trials compared to other studies using much shorter stimulations (increasing the number of trials using our multisensory stimulations would have made the study too long and onerous for the participants). A greater number of trials might highlight this trend towards maintaining metacognitive performance in the agentive condition. However, it is interesting to note that the agentive nature of the stimulation could promote not only the mechanisms of multisensory integration but also the mechanisms underlying metacognitive performance.

To our knowledge, our study is the first to extend the notions of both agency and ambiguity management to human motion perception. Its main contribution lies in showing that being in control of one’s motion is beneficial when faced with ambiguous situations. Our conclusions are strengthened by the fact that participants did not perceive themselves as performing better in the high-ambiguity active condition. Such a difference in the management of ambiguous situations should be further explored to better understand the perception of an observer versus an operator, particularly for critical situations. This would provide answers in key areas such as aeronautics and the automotive industry. Thus, there is clearly a need to better understand the underlying integrative mechanisms involved according to the operator’s level of control.

The raw data supporting the conclusions of this article will be made available by the authors under request and without undue reservation.

The studies involving human participants were reviewed and approved by The French CERSTAPS ethics committee (IRB00012476-2021-23-06-119). The patients/participants provided their written informed consent to participate in this study.

A-LR, BB, J-CS, and LB: study conception and design, interpretation of results, and draft manuscript. A-LR: data collection and analysis preparation. All authors contributed to the article and approved the submitted version.

This work was supported by APR Grants (DAR 1107) from the French National Space Research Centre (CNES).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.