Beyond the comparator model: A multifactorial two-step account of agency

https://doi.org/10.1016/j.concog.2007.03.010Get rights and content

Abstract

There is an increasing amount of empirical work investigating the sense of agency, i.e. the registration that we are the initiators of our own actions. Many studies try to relate the sense of agency to an internal feed-forward mechanism, called the “comparator model”. In this paper, we draw a sharp distinction between a non-conceptual level of feeling of agency and a conceptual level of judgement of agency. By analyzing recent empirical studies, we show that the comparator model is not able to explain either. Rather, we argue for a two-step account: a multifactorial weighting process of different agency indicators accounts for the feeling of agency, which is, in a second step, further processed by conceptual modules to form an attribution judgement. This new framework is then applied to disruptions of agency in schizophrenia, for which the comparator model also fails. Two further extensions are discussed: We show that the comparator model can neither be extended to account for the sense of ownership (which also has to be differentiated into a feeling and a judgement of ownership) nor for the sense of agency for thoughts. Our framework, however, is able to provide a unified account for the sense of agency for both actions and thoughts.

Introduction

Throughout the last years, computational and experimental evidence has largely extended our understanding of the mechanisms that govern the control and perception of self-action. It largely elaborates on theoretical groundwork that was laid many decades ago by the notion of internal comparison processes (cf. Helmholtz, 1866, Held and Freedman, 1963, Sperry, 1950, von Holst and Mittelstaedt, 1950). According to the most recent versions of this notion (Kawato, 1999, Wolpert et al., 1995, Wolpert et al., 1998, Wolpert and Ghahramani, 2000, Blakemore et al., 2002, Frith et al., 2000, Lindner et al., 2006, Wolpert and Flanagan, 2001), the motor system can be considered a control system with the input being a “desired state” and the output being the “estimated actual state” (see Fig. 1). On the basis of these two representations, the system specifies a sequence of motor commands in order to reach a certain goal. This process is controlled by different comparator mechanisms. In a feedback control loop, a comparison between the desired and the estimated actual state allows to calculate a motor error, which is fed back to the system to improve its functioning (comparator 1). On the basis of a given motor command, the system predicts the outcome of one’s behaviour (“predicted state”). When being compared with the initial desired state, such predictions can be used for feed-forward motor control (“central error monitoring”) of the movement which operates prior to any sensory feedback (comparator 2). Moreover, the predictions can be compared to the incoming sensory flow (ex- and reafference) and thus be used to cancel out or attenuate sensory feedback that is self-produced (reafference; comparator 3).

With respect to sensorimotor control, the “comparator model” (CM) is well-supported by empirical evidence. Especially the last comparator—which compares predicted and estimated actual state—has received wide support from elegant experimental work with all kinds of different sensorimotor sub-systems in both non-human species (e.g. weak electric fish (Bell, 2001) or crickets (Poulet & Hedwig, 2002)) and humans (Blakemore et al., 1998a, Haarmeier et al., 2001, Lindner et al., 2006, Shergill et al., 2003, Synofzik et al., 2006, Voss et al., 2006). Recently, however, the explanatory claim of the CM has been largely extended in that it would not only explain sensorimotor control, but also hold as a neurocognitive model of self-agency1 (Feinberg, 1978, Frith, 1992; for a recent version and modification see Frith et al., 2000, Frith, 2004, Frith, 2005). According to this model, the comparators not only serve a pragmatic, executive function on the level of sensorimotor control, but also assume an evaluation function on the level of action awareness: They may directly underlie our sense of agency (SoA), i.e. the registration that we are the initiators of our own actions (cf. Gallagher, 2000, Gallagher, 2004), in that they “label movements as generated by oneself or an external source” (Shergill, Samson, Bays, Frith, & Wolpert, 2005, p. 2384).

According to the CM, mainly two feed-forward comparators2 underlie our sense of agency: The comparison between predicted state and desired state (comparator 2, see Fig. 1) evokes a sense of being in control (Frith, 2005), while a comparison between predicted state and estimated actual state allows to self-attribute sensory events (comparator 3, see Fig. 1). In the following critique we will exemplarily focus on the latter comparator (comparator 3) since it is this comparator that is (i) responsible for attributing self-agency to sensory events and that is (ii) currently held responsible for disruptions of agency in schizophrenia patients (Frith, 2005, Lindner et al., 2005, Shergill et al., 2005). In contrast, the former comparator (comparator 2) (i) cannot explain self-agency for sensory events (since no kind of sensory signal is entered in the comparison process), (ii) is more speculative in character and less well empirically established and (iii) is not primarily used anymore to explain delusions of control in schizophrenia (Frith, 2005). However, most of the following critique regarding comparator 3 applies to comparator 2 as well.

The argumentation for comparator 3 as underlying our SoA seems convincing: Since there is no intrinsic difference between sensory signals arising as consequences of our own actions and sensory signals arising as results from events in the outside world, we need to resort to an internal central signal, i.e. the internal prediction, and compare it with the actual sensory afference in order to distinguish between externally produced and self-produced events (Frith et al., 2000, von Holst and Mittelstaedt, 1950).3 If predicted and estimated actual state are congruent, the sensory event is attributed to one’s own agency. If the sensory feedback is incongruent with sensory prediction, an external attribution of the causation of the sensory stimuli occurs (see Fig. 1). This agency registration process usually runs outside of consciousness: Rather than being explicitly aware of the motor representations involved in the comparator process, we experience self-agency by a rather diffuse sense of a coherent, harmonious ongoing flow of anticipations and sensory feedback (Pacherie, 2001). But, for example, in case of a striking mismatch between anticipations and sensory feedback, the comparator output becomes conscious and awareness of one’s motor predictions, one’s actual movements and the sensory consequences arises (Slachevsky et al., 2001).4

The idea of the comparator as a subpersonal mechanism underlying a subject’s registration of self-agency has many theoretical advantages: (1) The information necessary for agency is not added to the perception of an action, e.g. by a higher-order cognitive reasoning process, but is an immediate part of the perceptual registration itself. This registration is not necessarily conscious or explicit, but in fact occurs mostly in an implicit, pre-conscious way. However, it always frames our phenomenal content: Depending on the context requirements, it cancels (Haarmeier et al., 2001, Lindner et al., 2006), attenuates (Blakemore et al., 1999, Shergill et al., 2003) or highlights (Synofzik, Schlotterbeck, Leube, Thier, & Lindner, 2005) the experience of our action consequences. (2) The experience of agency is not dependent on conceptual capacities, but functions on a pre-conceptual, subpersonal level. (3) The self-relation involved in agency is not assumed to be represented as independent of action-processing, as suggested by many philosophical conceptions (e.g. Descartes’ dualism which presupposes an independent self that is able to initiate an act of will unrelated and prior to any bodily movement or Kant’s transcendental philosophy which presupposes a transcendental I as an independent unknowable entity which constitutes the SoA by unifying all our sense-experiences). The CM is able to explain the SoA as an intrinsic property of the action processing itself (Gallagher, 2004, Haggard, 2005, Legrand, 2006, Tsakiris and Haggard, 2005a, Tsakiris and Haggard, 2005b): Via an efference copy, motor processes directly inform the comparator, which in turn evokes our SoA. Thus, by directly connecting the SoA to action perception and motor processes, the CM seems to be a parsimonious explanation of the SoA.

To test the contribution of the CM to explaining the SoA, it has first to be clarified what is meant by “sense of agency”. In fact, the aforementioned definition of the “sense of agency”, i.e. the registration that we are the initiators of our actions (and not merely the initiators of our movements which would not comprise the effects), differs from many current definitions of the SoA and holds several advantages. For example, it does not require any direct phenomenal awareness (as the case in Gallagher’s notion of the SoA, Gallagher, 2007) or any meta-representations of self-agency (as one might suspect in the notions of the SoA proposed by Frith (2005) or by Stephens & Graham (2000), for a discussion see Gallagher (2004)). Nor does it require the conceptual ability to register self-agency (as would be the case when defining the SoA as “the ability to refer to oneself as the author of one’s actions”, de Vignemont & Fourneret, 2004, p. 2). Yet, it does entail the representation of oneself as causally responsible for the action and its direct effects (as opposed to a “compound sense of agency” which entails a sense of action initiation and a sense of action execution, i.e. “the sense of intending and executing an action”, Tsakiris & Haggard, 2005a, p. 387). The representational content of the SoA is not only determined by the action initiation per se (and thus not only by action antecedents), but also refers to the guidance and consequences of one’s actions and the causal relation between action intention, action performance and action consequences (for a broadly similar notion with respect to the notion of the content of an “intention-in-action”, see Pacherie, 2000).

However, not only the conceptual understanding of the SoA is highly heterogeneous, but also its experimental implementation and testing. This can be best shown by analyzing the SoA in terms of a distinction between feeling of agency and judgement of agency: Whereas feeling of agency (FoA) represents the non-conceptual, low-level feeling of being the agent of an action, judgement of agency (JoA) refers to the conceptual, interpretative judgement of being an agent. Both have to be distinguished from the sense of ownership (SoO), i.e. the sense that my body is moving regardless of whether the movement is voluntary or involuntary (Gallagher, 2000). Experiments claiming to investigate the SoA sometimes test the FoA (e.g. when testing the direct visual result of an internal comparison between predicted and actual sensory action-outcome; Lindner et al., 2005), sometimes the JoA (e.g. by requiring subjects “to indicate whether it was they themselves or the computer” that caused an event to happen; Aarts, Custers, & Wegner, 2005, p. 443), sometimes the SoO (e.g. by asking “You have just seen the image of a moving hand. Was it your hand?”; Daprati et al., 1997, p. 77) or sometimes a mixture of JoA and SoO (e.g. when requiring subjects to judge whether it was “their own movement, their own movement distorted, or the movement of another agent”; Farrer et al., 2003a, p. 325). Hence, the diversity of behavioural and neuroimaging results of these studies seems to be largely due to conceptual differences and confusions.

Relying on these distinctions we are able to investigate the contribution of the CM to the explanation of the SoA. In particular, we will show (1) that the CM does in fact contribute largely in explaining the FoA, yet is neither sufficient nor necessary, and that it fails completely to explain the JoA.5 Correspondingly, (2) it cannot sufficiently account for disruptions of the FoA or the JoA in schizophrenia. Finally, it also fails completely to explain (3) the SoO and (4) the SoA of thoughts. We put forward an empirical and conceptual critique and, moreover, outline a two-step multifactorial account of agency as an alternative framework which keeps the advantages of the CM.

Section snippets

Registration of self-agency despite a mismatch at the comparator?

Certainly, the CM should not be taken in a too strict version: Not any mismatch between predicted and actual state should lead to an evaluation of the event as being externally produced, but only mismatches beyond a certain sensitivity range. Accordingly, experimental evidence has shown that—within a certain range—even distorted sensory action consequences are experienced as self-produced, despite the underlying mismatch at the comparator. For example, in action recognition studies based on

The umbrella term “self-monitoring”

It was recently claimed that the CM of agency straightforwardly explains delusions of control in schizophrenia (e.g., Frith, 1992, Frith et al., 2000, Gallagher, 2004, Knoblich et al., 2004, Lindner et al., 2005). As articulated by Frith, in schizophrenia patients “something is wrong with the generation of a forward model and the representation of the predicted state of the system” (Frith et al., 2000). Even the most recent versions of the CM account of delusions is not more precise, contending

Can the comparator model explain the sense of ownership?

In an attempt to elaborate on Frith’s model, it was proposed that a version of the comparator would not only underlie our sense of agency but also our sense of ownership (SoO), i.e. the sense that part of my body is moving regardless of whether the movement is voluntary or involuntary.For example, Shaun Gallagher contends that “a match at the feedback comparator provides a sense of ownership” (Gallagher, 2000, p.16).

Can the comparator model explain the agency of thoughts?

The sense of agency in thoughts recently became a lively debated topic mainly because of the investigation of the positive symptom “thought insertion” in schizophrenia. Patients suffering from thought insertion experience thoughts which they claim to be inserted into their minds by other people or alien forces. The best description of this phenomenon is that the patients experience a thought in their stream of consciousness while they do not experience themselves as the author of this thought.

Conclusion

The CM is well suited to provide a parsimonious, elegant way of understanding the brain’s and body’s interaction with the world: In case of self-produced somatosensory stimulation—for example when we touch ourselves (Blakemore et al., 1999, Shergill et al., 2003, Voss et al., 2006) or when we perform eye movements (Haarmeier et al., 2001, Lindner et al., 2006, von Holst and Mittelstaedt, 1950)—the actual sensory consequences match the prediction and thus are specifically attenuated. Conversely,

References (122)

  • C. Farrer et al.

    The role of proprioception in action recognition

    Conscious Cognition

    (2003)
  • J.R. Flanagan et al.

    Prediction precedes control in motor learning

    Current Biology

    (2003)
  • C. Frith

    The self in action: Lessons from delusions of control

    Conscious Cognition

    (2005)
  • C.D. Frith et al.

    Explaining the symptoms of schizophrenia: Abnormalities in the awareness of action

    Brain Research Brain Research Review

    (2000)
  • S. Gallagher

    Philosophical conceptions of the self: Implications for cognitive science

    Trends in Cognitive Sciences

    (2000)
  • T. Haarmeier et al.

    Optimizing visual motion perception during eye movements

    Neuron

    (2001)
  • P. Haggard

    Conscious intention and motor cognition

    Trends in Cognitive Sciences

    (2005)
  • C. Henquet et al.

    Confusing thoughts and speech: Source monitoring and psychosis

    Psychiatry Research

    (2005)
  • M. Kawato

    Internal models for motor control and trajectory planning

    Current Opinion in Neurobiology

    (1999)
  • R.S. Keefe et al.

    Source-monitoring deficits for self-generated stimuli in schizophrenia: Multinomial modeling of data from three sources

    Schizophrenia Research

    (2002)
  • T.T. Kircher et al.

    Self-consciousness, self-agency, and schizophrenia

    Conscious Cognition

    (2003)
  • G. Knoblich et al.

    Grounding the self in action

    Conscious and Cognition

    (2003)
  • D.T. Leube et al.

    The neural correlates of perceiving one’s own movements

    Neuroimage

    (2003)
  • A. Lindner et al.

    Disorders of agency in schizophrenia correlate with an inability to compensate for the sensory consequences of actions

    Current Biology

    (2005)
  • H. Mushiake et al.

    Activity in the lateral prefrontal cortex reflects multiple steps of future events in action plans

    Neuron

    (2006)
  • A. Newen et al.

    Self-representation: Searching for a neural signature of self-consciousness

    Conscious Cognition

    (2003)
  • E. Pacherie et al.

    Phenomenology and delusions: Who put the ’alien’ in alien control?

    Consciousness and Cognition

    (2006)
  • K.L. Subotnik et al.

    Neurocognitive and social cognitive correlates of formal thought disorder in schizophrenia patients

    Schizophrenia Research

    (2006)
  • G.E.M. Anscombe

    Intention

    (1957)
  • T. Bayne et al.

    In defence of a doxastic conception of delusions

    Mind & Language

    (2005)
  • E. Bisiach et al.

    Anosognosia related to hemiplegia and hemianopia

  • S.J. Blakemore et al.

    The detection of contingency and animacy from simple animations in the human brain

    Cerebral Cortex

    (2003)
  • S.J. Blakemore et al.

    Spatio-temporal prediction modulates the perception of self-produced stimuli

    Journal of Cognitive Neuroscience

    (1999)
  • S.J. Blakemore et al.

    Predicting the consequences of our own actions: the role of sensorimotor context estimation

    Journal of Neuroscience

    (1998)
  • S.J. Blakemore et al.

    The detection of intentional contingencies in simple animations in patients with delusions of persecution

    Psychological Medicine

    (2003)
  • S.J. Blakemore et al.

    The perception of self-produced sensory stimuli in patients with auditory hallucinations and passivity experiences: Evidence for a breakdown in self-monitoring

    Psychological Medicine

    (2000)
  • S.J. Blakemore et al.

    Central cancellation of self-produced tickle sensation

    Nature Neuroscience

    (1998)
  • M. Botvinick

    Probing the neural basis of body ownership

    Science

    (2004)
  • M. Botvinick et al.

    Rubber hands ’feel’ touch that eyes see

    Nature

    (1998)
  • C. Cahill et al.

    Psychotic experiences induced in deluded patients using distorted auditory feedback

    Cognitive Neuropsychiatry

    (1996)
  • J. Campbell

    Schizophrenia, the space of reasons, and thinking as motor process

    The Monist

    (1999)
  • J. Campbell

    The ownership of thoughts

    Philosophy Psychiatry Psychology

    (2002)
  • J. Campbell

    The first person, embodiment, and the certainty that one exists

    The Monist

    (2004)
  • A. Coliva

    Thought insertion and immunity to error through misidentification

    Philosophy Psychiatry Psychology

    (2002)
  • M. Davies et al.

    Monothematic delusions: Towards a two-factor account

    Philosophy Psychiatry Psychology

    (2001)
  • Y. Delevoye-Turrell et al.

    A deficit in the adjustment of grip force responses in schizophrenia

    Neuroreport

    (2002)
  • H.H. Ehrsson et al.

    Touching a rubber hand: Feeling of body ownership is associated with activity in multisensory brain areas

    Journal of Neuroscience

    (2005)
  • H.H. Ehrsson et al.

    That’s my hand! Activity in premotor cortex reflects feeling of ownership of a limb

    Science

    (2004)
  • N. Eilan et al.

    Agency and self-awareness: Mechanisms and epistemology

  • F. Erichsen

    Die Bedeutung von Werthaltungen für die Schizophrenie [the significance of valence in schizophrenia]

    Psychiatria Clinica

    (1973)
  • Cited by (649)

    • The sense of agency in near and far space

      2024, Consciousness and Cognition
    View all citing articles on Scopus
    View full text