Who is causing what? The sense of agency is relational and efferent-triggered.

Abstract

The sense of agency ("I did that") is a basic feature of our subjective experience. Experimental studies usually focus on either its attributional aspects (the "I" of "I did that") or on its motoric aspects (the "did" aspect of "I did that"). Here, we combine both aspects and focus on the subjective experience of the time between action and effect. Previous studies [Haggard, P., Aschersleben, G., Gehrke, J., & Prinz, W. (2002a). Action, binding and awareness. In W. Prinz, & B. Hommel (Eds.), Common mechanisms in perception and action: Attention and performance (Vol. XIX, pp. 266-285). Oxford: Oxford University Press] have shown a temporal attraction in the perceived times of actions and effects, but did directly not study the relation between them. In three experiments, time estimates of an interval between an action and its subsequent sensory effect were obtained. The actions were either voluntary key press actions performed by the participant or kinematically identical movements applied passively to the finger. The effects were either auditory or visual events or a passive movement induced to another finger. The results first indicated a shortening of the interval between one's own voluntary action and a subsequent effect, relative to passive movement conditions. Second, intervals initiated by observed movements, either of another person or of an inanimate object, were always perceived like those involving passive movements of one's own body, and never like those involving active movements. Third, this binding effect was comparable for auditory, somatic and visual effects of action. Our results provide the first direct evidence that agency involves a generalisable relation between actions and their consequences, and is triggered by efferent motor commands.

Full text

Page 1

This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright

Page 2

Author's personal copy
Brief article
Who is causing what? The sense of agency
is relational and efferent-triggered
Kai Engberta,*, Andreas Wohlschla ¨gera, Patrick Haggardb
aMax Planck Institute for Human Cognitive and Brain Sciences, Department of Psychology,
Stephanstraße 1A, 04103 Leipzig, Germany
bInstitute of Cognitive Neuroscience and Department of Psychology,
University College London, London, UK
Received 22 March 2007; revised 30 July 2007; accepted 31 July 2007
Abstract
The sense of agency (‘‘I did that’’) is a basic feature of our subjective experience. Experi-
mental studies usually focus on either its attributional aspects (the ‘‘I’’ of ‘‘I did that’’) or
on its motoric aspects (the ‘‘did’’ aspect of ‘‘I did that’’). Here, we combine both aspects
and focus on the subjective experience of the time between action and effect. Previous studies
[Haggard, P., Aschersleben, G., Gehrke, J., & Prinz, W. (2002a). Action, binding and aware-
ness. In W. Prinz, & B. Hommel (Eds.), Common mechanisms in perception and action: Atten-
tion and performance (Vol. XIX, pp. 266–285). Oxford: Oxford University Press] have shown a
temporal attraction in the perceived times of actions and effects, but did directly not study the
relation between them.
In three experiments, time estimates of an interval between an action and its subsequent
sensory effect were obtained. The actions were either voluntary key press actions performed
by the participant or kinematically identical movements applied passively to the finger. The
effects were either auditory or visual events or a passive movement induced to another finger.
The results first indicated a shortening of the interval between one’s own voluntary action
and a subsequent effect, relative to passive movement conditions. Second, intervals initiated by
observed movements, either of another person or of an inanimate object, were always
0010-0277/$ - see front matter ? 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.cognition.2007.07.021
*Corresponding author. Present address: Technical University Munich, Department of Sports Science,
Sports and exercise Psychology, Connollystrasse 32, 80809 Munich, Germany. Tel.: +49 177 34 34 624.
E-mail address: kaiengbert@hotmail.com (K. Engbert).
www.elsevier.com/locate/COGNIT
Cognition 107 (2008) 693–704

Page 3

Author's personal copy
perceived like those involving passive movements of one’s own body, and never like those
involving active movements. Third, this binding effect was comparable for auditory, somatic
and visual effects of action. Our results provide the first direct evidence that agency involves a
generalisable relation between actions and their consequences, and is triggered by efferent
motor commands.
? 2007 Elsevier B.V. All rights reserved.
Keywords: Intentional binding; Agency; Time estimation; Action; Volition; Action; Self and other
1. Introduction
The sense of agency, of controlling one’s own actions, is a basic feature of every-
day life, and indeed of human nature. However, the precise conditions under which
the sense of agency occurs, and its precise phenomenology are poorly understood.
This may be largely because introspection about our own actions is rather uninfor-
mative. In recent years, laboratory studies of the sense of agency have attempted to
shed more light in this area.
These studies generally focus on one of two subcomponents of the sense of agency
that can be described as attribution and association. To clarify this point, we take the
basic content of agency as ‘‘I did that’’. Studies focusing on the ‘‘I’’ component of
this content deal with the attributional side of agency, and are often described as
self-recognition studies. Participants typically make explicit judgements about
whether they or another agent was responsible for a given event, such as a visual
object triggered by a key press (Wegner & Wheatley, 1999), or visual feedback from
a movement (Daprati et al., 1997; Tsakiris, Haggard, Franck, Mainy, & Sirigu, 2005;
van den Bos & Jeannerod, 2002).
A second approach focuses on the ‘‘did’’ aspect of ‘‘I did that’’. This considers
agency as a dyadic association between an action and a consequence or effect. Such
associations have been studied in animal operant learning (Dickinson, 1980; Thorn-
dike, 1911), but few studies have considered how they produce an experience of
agency in humans. Two key findings have placed linkage across time at the heart
of this approach. First, the mental representation of the action predicts the later
effect (Elsner & Hommel, 2001; James, 1890). Second, the strength of association,
and thus the feeling of agency, operates over a limited time window. As the interval
between an action and its sensory effect increases, subjects become less likely to agree
that they caused the sensory effect (Wegner & Wheatley, 1999).
A series of recent studies suggested that temporal attraction between action and
effectispartofthephenomenology ofagency itself, notjust apreconditionforit.Hag-
gard,Aschersleben,Gehrke,andPrinz(2002a)askedparticipantstopressakey,which
producedanauditorystimulusashortintervalafterwards.Participantsusedarotating
clock hand (Libet, Gleason, Wright, & Pearl, 1983) to estimate when they made the
action on each trial, or, in separate blocks, when they heard the tone. The results
showedastrong temporalattractionbetweenactionandtone, relative tobaselinecon-
trol conditions in which actions occurred without tones, or tones occurred without
694
K. Engbert et al. / Cognition 107 (2008) 693–704

Page 4

Author's personal copy
actions. Because no such temporal attraction occurred for involuntary movements
evoked by transcranial magnetic stimulation, the effect was called intentional binding,
and taken as an implicit measure of agency (Haggard, Clark, & Kalogeras, 2002b).
However, subjects in such event-judgement studies focus on just the action or its
effect. The task does not require them to represent both events, and the relation
between them. Therefore, these studies cannot provide direct evidence that sense
agency involves representing the relation between action and effect. Here we investi-
gated the relational aspect of agency directly in three interval estimation experi-
ments. Interval estimates avoid some methodological difficulties associated with
the Libet-clock method (Libet, 1985; Pockett & Miller, 2007) and capture the rela-
tion between action and effect more directly than estimates based on the perceived
time of a single event. Accordingly, we hypothesised that the action-effect interval
would be estimated as shorter when the participant was acting intentionally com-
pared to the control conditions where they were not.
Experimental studies require comparing an agency condition to a non-agency
control condition. In the attribution or self-recognition approach, the control condi-
tion is provided by trials where the subject sees the action of another person, rather
than their own action (Daprati et al., 1997; Sirigu et al., 2004). In psychophysical
studies of binding, the control condition typically involves an involuntary movement
lacking any efferent motor command, though these have generally not been physi-
cally comparable with the movements in the agency condition (e.g., Haggard
et al., 2002b). Thus, one’s own agency can be compared either to the agency of
another person, or to one’s own non-agentic body movements. Interestingly, few
studies have combined both traditions to investigate the attributional and motoric
aspects of agency in parallel (but see Engbert, Wohlschla ¨ger, Thomas, & Haggard,
in press; Tsakiris et al., 2005). Accordingly, the present experiments focus on three
outstanding issues regarding human sense of agency, all of which can appropriately
be addressed using interval estimates. First, we have tested directly whether agency
involves a change in the represented relation between action and effect. Second, we
have investigated the generality of sense of agency, by investigating different effects
of the same class of action. Third we have investigated the conditions necessary for
agency by manipulating both the social attribution of action to agents, and the pres-
ence of an efferent motor command in a factorial design. Taken as a whole, the
results confirm that a distinct chain of phenomenal experience is triggered by efferent
signals in the motor system, relating actions to their subsequent effects, and provid-
ing a basic sense of agency.
2. Experiment 1
2.1. Methods
Experiment 1 compared estimates of intervals between actions and subsequent
auditory effects, using a 2 · 2 factorial repeated measures design. The factors were:
whether the subject actually performed the action, or merely observed an action
K. Engbert et al. / Cognition 107 (2008) 693–704
695

Page 5

Author's personal copy
made by another agent, and whether the interval began with an intentional action or
a passive, involuntary movement. We call these factors and their levels ‘‘Person’’
(self/other) and ‘‘Agency’’ (active/passive), respectively. In the self/active condition,
participants pressed a lever themselves. In the self/passive condition, the lever and
the participant’s finger was moved by a computer-controlled stepper motor. In the
other/active condition, participants merely observed while the experimenter ‘pressed’
the lever. In fact, to increase experimental control and avoid subtle cues, the exper-
imenter’s finger was moved passively by the motor. Debriefing of participants at the
end of the experiments confirmed that they were not aware of this: all thought the
experimenter moved intentionally. Finally, in the other/passive condition, the exper-
imenter’s hand was replaced on the lever by a rubber hand. The lever was then
moved by the motor. We chose to replace the experimenter’s hand by the rubber
hand to ensure that there was clear visual information to distinguish between the
active and passive movements in the other condition. The other passive condition
was obviously passive because no biological agent was involved. The other active
condition allowed attribution of intentional actions to the experimenter (even
though the movements were in fact passive to ensure comparability with the remain-
ing conditions).
These four conditions were tested in different blocks, in counterbalanced order.
The four blocks therefore differed in the information available about the action.
In the self/active condition, participants had both efferent and proprioceptive infor-
mation about their own action. In the self/passive condition, proprioceptive but not
efferent information was available. In the other/active condition, participants had
only visual information about the experimenter’s action, though the instruction
and setup was designed to promote an ‘intentional stance’ (Dennett, 1987). In the
other/passive condition, visual information was identical, but the intentional stance
was discouraged both by instruction and the absence of any biological agent.
Twenty-four paid right-handed volunteers (6 Male) aged between 20 and 36 years
(mean age 26.3 years) participated on the basis of informed consent and with local
ethical approval. The right index finger either of the participant, the experimenter
or a rubber hand was attached to a lever using Velcro. The lever could easily be
depressed by the subject, or could be moved via a computer-controlled stepper
motor. The lever displacement was 10 mm. After each movement of the lever an
auditory tone (100 ms, 1 kHz, 76 dB) was presented via stereo loudspeakers.
Participants made unspeeded verbal judgements of the duration of the interval
between the movement and the tone in milliseconds, and these were recorded by
an experimenter. They were reminded that 1 s would correspond to a judgment of
1000, 0.5 s to 500 etc. Moreover, participants were told that none of the intervals
would be longer than one second. That is, only judgments between ‘‘1’’ and
‘‘1000’’ were counted as a valid answer. Within this range, every judgement was
allowed and participants were encouraged to use a full range. In fact, only three
action-effect intervals occurred: 200, 250 and 300 ms. Each interval was presented
42 times per block, in random order. No reference intervals or training were given,
because we were not interested in the accuracy of interval estimation so much as sys-
tematic biases in interval estimation between conditions.
696
K. Engbert et al. / Cognition 107 (2008) 693–704