«James W. Moore1,2 & Sukhvinder S. Obhi3,4 1. Department of Psychology, Goldsmiths, University of London, London, UK 2. Brain Mapping Unit, Department ...»
Intentional Binding and the Sense of Agency: A review
James W. Moore1,2 & Sukhvinder S. Obhi3,4
1. Department of Psychology, Goldsmiths, University of London, London, UK
2. Brain Mapping Unit, Department of Psychiatry, University of Cambridge, Cambridge,
3. Centre for Cognitive Neuroscience & Department of Psychology, Wilfrid Laurier
University, Waterloo, Ontario, Canada
4. Institute of Cognitive Neuroscience, University College London, London, U.K.
Correspondence: email@example.com PrePrint (uncorrected proof) – Consciousness & Cognition (2012) 1 Abstract It is nearly 10 years since Patrick Haggard and colleagues first reported the ‘intentional binding’ effect (Haggard et al, 2002). The intentional binding effect refers to the subjective compression of the temporal interval between a voluntary action and its external sensory consequence. Since the first report, considerable interest has been generated and a fascinating array of studies has accumulated. Much of the interest in intentional binding comes from the promise to shed light on human agency. In this review we survey studies on intentional binding, focusing, in particular, on the link between intentional binding and the sense of agency (the experience of controlling action to influence events in the environment). We suggest that, whilst it is yet to be fully explicated, the link between intentional binding and the sense of agency is compelling. We conclude by considering outstanding questions and future directions for research on intentional binding.
1. Introduction Humans are agents. That is, they have the capacity to bring about change in the external world through their own goal-directed behaviour. Often, humans also have a corresponding conscious experience of this capacity, which is referred to as the ‘sense of agency’. A key challenge for scientific investigations of the sense of agency is the discovery and use of appropriate measures. In 2002, Haggard and colleagues introduced a novel measure of the sense of agency based on an intriguing relationship between voluntary action and subjective time (Haggard, Clark, & Kalogeras, 2002; Haggard, Aschersleben, et al., 2002). This so- called ‘intentional binding’ measure has generated considerable interest and has been used in a number of experiments on the sense of agency. However, as a measure of the sense of agency it is not without its detractors (e.g. Buehner & Humphreys, 2009). We therefore feel that the time is right for a review of experiments using intentional binding to study the sense of agency. We first provide a brief historical background to the use of subjective time in experimental psychology. We begin by introducing Benjamin Libet’s seminal work and the ‘clock methodology’ he used (which forms the basis of the original intentional binding paradigm). Having set the scene we then review experiments using the intentional binding paradigm. We conclude this review by considering a) the validity of intentional binding as a measure of the sense of agency, and b) future directions for research.
The use of subjective time in Experimental Psychology: Libet’s work on volition 1.1.
Measures based on the subjective experience of time have a long history in Experimental Psychology. In the 1880s Wilhelm Wundt developed his complication-clock apparatus to explore the time course of attention (Figure 1). Participants had to orient to a clock (or pendulum) when presented with a certain stimulus (such as an auditory click) and report the onset of that stimulus by noting the position of the clock hand (or pendulum) when the stimulus occurred. Wundt observed systematic differences in the perceived onset of the auditory stimulus: People either perceived the auditory event earlier or later relative to the position of the clock hand (or pendulum). Moreover, this difference was attributed to whether participants were attending to the clock hand or or the auditory stimulus. Wundt’s chronometric methodology thus provides an invaluable tool for comparing subjective and objective stimulus onset timings.
3Figure 1. The complication-clock apparatus developed by Wundt.
In the 1980s Benjamin Libet adopted Wundt’s classic methodology to explore human volition. In Libet et al’s (1983) seminal study, participants sat in front of a clock face marked at regular intervals. During each trial a spot rotated around the clock-face at a speed of 1 revolution every 2.56s. Participants used the clock to judge the onset of certain events. In one condition participants flexed their wrist when they felt the urge, and judged the time they became aware of raising their hand. In a second condition participants again flexed their wrist when they felt the urge, and this time they judged the time they became aware of their conscious intention to raise their hand. In a third condition, a somatosensory stimulus was applied at an unpredictable time during the trial and participants judged the time that they felt this stimulus. To make these timing judgements participants reported the position of the spot on the clock face when they were perceived the event (intention, action or somatosensory stimulus). Libet and colleagues concurrently used electroencephalography (EEG) to record the readiness potential, a cascade of neural activity that reliably precedes the onset of voluntary movement. Libet found that, judgements of movements were slightly early compared to actual movement onset, whereas judgements of somatosensory stimuli were slightly delayed compared to their actual onset. Famously, although intentions were perceived as occurring prior to movements, they lagged behind the onset of the readiness potential. This suggests that the intention to act may arise after the brain has initiated an action. The implications of this work have been extensively discussed elsewhere. What is most relevant to the current paper is the ‘clock methodology’ that Libet used as it formed the basis of the original work on intentional binding, which is discussed in the following section.
1.2. Original experiment and some basic conditions for intentional binding
4 In the first journal article to report the intentional binding effect, Haggard, Clark & Kalogeras (2002) used the Libet clock method (see earlier description of this approach) to study the perceived time of actions and their consequent effects (see also Haggard, Aschersleben, Gehrke, & Prinz, 2002). There were four critical conditions in their experiment. In baseline conditions, participants either made voluntary actions or listened for the occurrence of an auditory tone (in the absence of action) while they watched a rotating clock hand on a computer screen. They were asked to report the position of the clock hand when they moved or when the tone occurred. In operant conditions, participants made a voluntary key press on every trial, but this time it was followed 250 ms later by an auditory tone. In specific blocks, participants were asked to judge either the time of their action or the time of the tone. The trial structure of a typical operant condition is shown in figure 2. The key comparison was the perceived times of actions and tones in baseline conditions with the perceived time of actions and tones in operant conditions. The authors found that, in operant conditions, the perceived time of their actions was later than in baseline conditions and the perceived time of the tone was earlier than in baseline conditions. Hence, the derived interval between the action and tone in operant conditions was compressed compared to the derived interval between these two events in baseline conditions. Critically, in an identical set of conditions involving involuntary movements induced via transcranial magnetic stimulation over the primary motor cortex, the binding effect was reversed such that the interval between action and effect actually increased in ‘operant’ conditions compared to baseline conditions. Figure 3 shows the classic pattern of intentional binding found in this seminal experiment.
In a second experiment the authors varied the delay and the predictability of when the tone occurred after voluntary action. The results from this study indicated that temporal contiguity and predictability are important determinants for binding. On the basis of their findings the authors speculated that a specific cognitive function of the central nervous system is to bind together critical sensorimotor events that surround voluntary action, and that this function may be crucial for the normal experience of agency. There are two things to highlight about this early hypothesis. The first is the assumed link with sense of agency. The aim of this review is to scrutinise this assumption, and, while the evidence reviewed provides compelling support for this link, it should be noted that alternative proposals have been advanced. For example, intentional binding may reflect the ‘unity of perception’ (Yarrow et al., 2001) or be more generally linked to causality rather than agency (Stetson, Montague and Eagleman, 2006). The second thing to highlight is that this early hypothesis suggested that binding may help support inferences of agency. However, as we shall see in this review, more recent evidence suggests that inferences of agency can also support binding (e.g. Moore & Haggard, 2008). This serves to emphasise a recurring theme in this review; namely, that although there 6 is compelling evidence supporting a link between intentional binding and sense of agency, the exact nature of that relationship is yet to be fully understood Figure 3. The classic pattern of intentional binding described by Haggard, Clark and Kalogeras (2002). The perception of onset of voluntary actions was shifted later in time and the perception of tone onset was shifted earlier. This was not the case for involuntary actions induced by TMS, which showed the opposite effects.
Original figure used with permission.
Since the seminal journal article on intentional binding, many studies have been performed using the Libet clock method. However, another, more direct method has also been employed, yielding confirmatory results. Specifically, citing some of the methodological criticisms surrounding the Libet clock method, several authors have used a direct interval estimation procedure in which participants are instructed to simply report the perceived interval (in milliseconds) between an action and effect (see Figure 4). These studies, which we describe in more detail later, have shown that mean interval estimates are lower (indicating a shorter perceived interval) in voluntary vs. Passive conditions. This temporal illusion is just one example of perceptual distortion surrounding voluntary action. Another well-known example is the spatial compression effect in which targets flashed around the onset of a saccade are systematically mislocalised toward the saccade landing position (interested readers are referred to Ross, Morrone, & Burr, 1997 for more on this fascinating perceptual effect). In the context of this review, we shall refer to the direct estimation of temporal intervals as the (direct) interval estimation approach.
The original demonstration of intentional binding and the suggestion of a link to agency sparked considerable interest from researchers interested in the phenomenology of human action. The remaining sections of our review highlight some of the most significant work that followed the original journal article. As will be evident, intentional binding and its potential relationship with the sense of agency is far from fully understood, although with each additional experiment, a slightly clearer view of the significance of intentional binding in human action contexts and the factors that modulate intentional binding, is emerging.
2. Prediction and retrospective inference
The purported link between intentional binding and the sense of agency is perhaps the most tantalising aspect of the original article. To appreciate why the link was suggested, consideration must be given to current ideas about how the sense of agency might arise. In this regard, it is possible to distinguish between two theoretical positions on the 8 neurocognitive origins of the sense of agency. On the one hand there is the ‘predictive’ position. On this view the sense of agency is generated by processes dedicated to the control of voluntary action. Optimal motor control and learning require predictions of both the future states of the motor system and the sensory consequences of movement (Wolpert & Ghahramani, 2000). These predictions are derived from internal forward models, of which there are two classes: forward dynamic and forward sensory. The forward dynamic model captures the dynamics of bodily movement. The forward sensory model captures the causal relation between movements and their sensory consequences, generating predictions of the likely sensory consequences of movement based on efference copy of motor commands.
According to the so-called, ‘comparator model’ of the sense of agency, this plays a key role in the sense of agency (Blakemore, Wolpert, & Frith, 2002): the sense of agency is produced when there is a match between predicted and actual sensory consequences of movement.
On the other hand there are those who downplay the specific contribution of the motor system. Instead, it is suggested that the sense of agency is generated by a process of retrospective inference. On this view a general-purpose inferential mechanism uses sensory information to establish the causal origins of actions and their effects. One influential version of this position is Wegner’s ‘theory of apparent mental causation’ (Wegner & Wheatley, 1999; Wegner, 2002; Wegner, 2003), which takes an explicitly Humean approach. According to this theory, the experience of willing an action arises if a thought (intention) 1) occurs prior to action, 2) is consistent with the action, and 3) is the most plausible cause of the action.
This same debate, prediction vs. retrospective inference, concerns the intentional binding effect (Haggard & Clark, 2003). Intentional binding could be produced by dedicated motor control mechanisms predicting the sensory consequences of an action. Alternatively, intentional binding could be the result of retrospective inference triggered by the actual presence of the sensory consequences of movement. The apparent specificity of the binding effect to voluntary movement strongly implies the involvement of predictive motor processes.