Consciousness and Cognition Consciousness and Cognition 12 (2003) 708–716 www.elsevier.com/locate/concog
Intentionality as a constituting condition for the own self—and other selves Andreas Wohlschl€ ager,a,* Kai Engbert,a and Patrick Haggardb a
Max-Planck-Institut f€ur psychologische Forschung, Cognition and Action, Amalienstraße 33, M€unchen D-80799, Germany b University College London, UK Received 11 March 2003
Abstract Introspectively, the awareness of actions includes the awareness of the intentions accompanying them. Therefore, the awareness of self-generated actions might be expected to differ from the awareness of othergenerated actions to the extent that access to oneÕs own and to otherÕs intentions differs. However, we recently showed that the perceived onset times of self- vs. other-generated actions are similar, yet both are different from comparable events that are conceived as being generated by a machine. This similarity raises two interesting possibilities. First we could infer the intentions of others from their actions. Second and more radically, we could equally infer our own intentions from the actions we perform rather than sense them. We present two new experiments which investigate the role of action effects in the awareness of selfand other-generated actions by means of measuring the estimated onset time. The results show that the presence of action effects is necessary for the similarity of awareness of self- and other-generated actions. Ó 2003 Elsevier Inc. All rights reserved.
1. Introduction Self-generated actions are thought to take an exceptional position in conscious experience. Proprioception,1 i.e., tactile and kinesthetic sensations as well as efference copies may offer private *
Corresponding author. Fax: +49-89-38-602-199. E-mail address:
[email protected] (A. Wohlschl€ ager). 1 We use the term proprioception in its strict etymological sense to mean perceptual information associated with the self. This sense of the word includes the classic physiological meaning of the sensory sources about body configuration within the muscles and joints but it also includes many additional sources and kinds of information, including efference copy. In our use, any information that individuals can logically have about themselves, but which another cannot directly have, would count as proprioception. 1053-8100/$ - see front matter Ó 2003 Elsevier Inc. All rights reserved. doi:10.1016/S1053-8100(03)00083-7
A. Wohlschla¨ger et al. / Consciousness and Cognition 12 (2003) 708–716
709
and privileged access. Hence, the awareness of self-generated actions is usually contrasted to the awareness of otherÕs actions and to the awareness of allochthonic events in general. Neuroscientific studies have accumulated evidence about anatomical structures specifically involved in action generation (Goldberg, Kwan, Borrett, & Murphy, 1984; Kornhuber & Deecke, 1965; Tanji, 2001) and in the awareness of self-generated actions (Fried et al., 1991; Haggard, Aschersleben, Gehrke, & Prinz, 2002a; Haggard, Clark, & Kalogeras, 2002b). On the other hand, there is increasing evidence about similarities in the neuronal processing of self- and other-generated actions. Gallese, Fadiga, Fogassi, and Rizzolatti (1996) found evidence that a common brain circuit is used in monkeys to both control object-orientated actions and to represent homologous actions of others. These ‘‘mirror neurons’’ in the pre-motor cortex fire both during grasping and when the monkey observes a human making a similar grip. Recent evidence from both human performance (Craighero, Fadiga, Rizzolatti, & Umilta, 1999, 1996) and neurophysiological studies (Fadiga, Fogassi, Pavesi, & Rizzolatti, 1995; Rizzolatti, Fadiga, Gallese, & Fogassi, 1996) supports a common processing of self- and other-generated actions. Given these similarities at the level of neuronal processing one may have to reconsider assumptions concerning conscious experience of actions. For instance, it seems to be possible that there is no elementary difference in the awareness of self- and other-generated actions. However, no proprioceptive information is available in other-agentsÕ actions, so a potential common construction-mechanism (i.e., an identical mechanism for generating conscious awareness of self- and of other-generated actions) has to consider multimodal information. Action related cues such as action-effects or the conception of intentionality could play a crucial role in this respect. The psychological control of our actions necessarily relies on processes completely within an individual. However, our conscious experience of actions may be more linked to the actionsÕ effects in the world. In that case, the perception of oneÕs own actions and those of others could be quite similar. In order to explore the relationship between the awareness of self-generated actions and the integration across agents, Wohlschl€ ager, Haggard, and Gesierich (in press) performed a series of three experiments. Awareness of actions was quantified by measuring the perceived onset time of actions using the method devised by Libet, Gleason, Wright, and Pearl (1983). This method provides a common metric for describing diverse events, yet it is also sensitive to the properties underlying psychological representations of those events (Haggard et al., 2002a, 2002b). In all experiments subjects viewed a response lever and a clock projected onto the table next to the lever. At the beginning of each trial the clock hand started rotating at a random position. It rotated until the response lever was moved, continued for a random period thereafter, and then stopped. The subjectÕs task was to verbally report the position of the clock hand when the response lever was depressed. In order to create actions with different characteristics the agent moving the response lever was varied across the experimental conditions: In the Self condition, subjects themselves pressed the lever at a time of their own choice. They were instructed to avoid pressing at ‘‘obvious’’ clock positions, or pressing in response to the onset of clock rotation. In the Other condition, subjects observed the experimenter pressing the response lever. The experimenter followed the same constraints as the subjects in the Self condition. In the Machine condition the response lever was moved by a solenoid (which in turn was operated by the experimenter using a hidden switch) invisibly located at the lever fulcrum. Subjects were told that the lever would now move automatically. In all conditions the subjects judged the perceived onset time of the action
710
A. Wohlschla¨ger et al. / Consciousness and Cognition 12 (2003) 708–716
using the clock. In the first two experiments every lever-press was followed by a pure tone after 250 ms, whereas Experiment 3 repeated the design of the first experiment without any ensuing beep. In the Machine conditions of the first and third experiment, subjects viewed just the moving response lever without any hand, whereas in Experiment 2 the subjects saw a rubber hand operating the lever. By comparing the position of the clock hand at lever closure with the subjectÕs judgment of when the lever moved, the judgment error as a dependent measure was calculated. The results of all three experiments clearly showed that the perceived onset time of oneÕs own actions is comparable to the perceived onset time of other peopleÕs actions. Both in turn are substantially later than the perceived onset time of a physically comparable machine (or rubber hand) movement. Varying the anthropomorphic nature of the viewed hand, from something clearly inanimate to something humanoid reduced but did not eliminate this difference. The results also did not change substantially, when the auditory action effect was removed. In summary, one could say that conditions involving a voluntary agent (Self or Other) were similar to each other but different from conditions that did not involve a voluntary agent. Note that the involvement of a voluntary agent in these experiments solely depended on the subjectsÕ conception of the situation. We will use the term ‘‘intentionality’’ for conditions in which an observer conceives him/herself or others as voluntary and intending an action and the term ‘‘non-intentionality’’ for the contrary. Using these terms, our question can be formulated as how and why intentional actions are perceived differently from non-intentional events. One important factor regarding the awareness of actions could be the subjectÕs concept of how and why the action is generated. Wohlschl€ager et al. (in press) speculated that self- and othergenerated actions may be perceived with reference to the effects they cause, while machine actions are not. In this context binding actions to their effects may provide an explanation for the different time course of awareness of intentional actions versus other events. According to Haggard et al. (2002a) the perceived time of actions is shifted towards the effects they cause whenever intentions can be attributed to an acting agent. While previous experiments showed that awareness of the times of oneÕs own and othersÕ actions were similar, they did not show why this was the case. Possibly, subjects attribute agency to others on the same basis that they deploy it themselves. However, a complete test of this account requires comparing the perceived time of self- and other-generated actions made with and without agency. Therefore the present experiments include passive as well as active movements. In addition, the attribution of agency to others necessarily depends on appreciating the effects of othersÕ actions in the external world. Therefore, we studied actions which either did or did not produce an additional external effect, to investigate similarity between judgments of self- and other-generated actions in each case.
2. Experiment 1 Experiment 1 aimed to explore the respective contribution of proprioception and intentionality to the awareness of actions. These factors were put into a 2 2 design, which resulted in four different experimental conditions (see Table 1): Similar to the above experiments in Wohlschl€ager et al. (in press), the subjects pressed the lever at a time of their own choice in the Self condition.
A. Wohlschla¨ger et al. / Consciousness and Cognition 12 (2003) 708–716
711
Table 1 The two experimental factors and the four experimental conditions Factor: Intentionality Factor: Proprioception
Condition
No
Yes
No Yes
Machine Passive
Other Active
In the Other condition, the subject observed the experimenter pressing the response lever so that no proprioceptive information was available. In the Machine condition, a rubber hand was used and the response lever again was moved automatically so that neither proprioceptive information was available nor should intentionality be attributed. In the Passive condition the subjectÕs finger was moved automatically so that proprioceptive information was available but the actions could not be perceived as intentionally caused. In each condition a lever-press was followed by a pure tone after 250 ms. This tone served as an additional external effect of the action. 2.1. Method Twenty-four subjects, 14 women and 10 men, all right handed were tested in an apparatus similar to the one used in the experiments described above, but modified in the following ways (see Fig. 1A): First the subjectÕs (or experimenter or rubber) hand and the response lever (distance between upper open and lower closed position was 2.2 mm) were placed inside a closed box with a cutout at the top above the position of the index finger. Inside the box a hollow wooden shaft in which the subjects inserted their index finger was fixed to the lever (see Fig. 1B). Hence, the subjects could just see the moving shaft when they pressed the lever; all other cues, which could give any hints about the origin of the movement, were eliminated and thus all conditions were visually identical.
Fig. 1. Apparatus. (A) Subjects viewed the response lever through a semi-silvered mirror. The semi-silvered mirror served to superimpose the clock on the wooden shaft in which the index fingers were inserted (B).
712
A. Wohlschla¨ger et al. / Consciousness and Cognition 12 (2003) 708–716
The second modification concerned the clock position. It was now superimposed upon the wooden shaft, i.e., on the position of the subjectÕs finger, rather than projected next to it. The aim was to avoid difficulties with a spatial disunion of clock and lever. Thus a spatial shift of attention was not necessary and at the same time subjects inevitably watched the moving finger. Size and all other attributes of the clock were analogous to the above experiments. A single clock-hand of radius 1.2 cm rotated with a period of 2560 ms within a face marked at conventional 5 ‘‘minute’’ intervals. The sequence of a particular trial also remained unchanged. The experimenter initiated the rotation of the clock at the start of each trial. The initial position of the clock hand was random. The clock hand then rotated until the response lever was moved (see below), continued to rotate for a random period between 1.5 and 2.5 s thereafter, and then stopped. The subject then verbally reported the position of the clock hand when the response lever was depressed. Except for the Other condition, the experimenter sat at a table that was placed about 2 m away from the experimental apparatus. In these conditions, subjects could only see the experimenter if they raised and turned their head to the left. Subjects were encouraged to make their verbal reports as precise as possible and to avoid confining themselves to the 5 ‘‘minute’’ intervals marked on the clock face. Subjects performed 40 such trials in each of four conditions. The experimenter verified that the subject was looking at the clock face during each trial. Third, in all but the Self condition, the movement of the shaft was controlled by the solenoid programmed by the computer. The computer program simulated a random distribution of the onset times relative to clock start that was similar to the one observed in the Self condition. Thus, even though subjects were instructed in the Other condition that now the experimenter will operate the lever, in fact the experimenters finger was also moved passively. The duration and the voltage of the rectangular pulse that was sent to the solenoid was calibrated such that the movement duration was held as constant as possible across Other, Machine, and Passive condition.2 2.2. Results and discussion Mean judgment errors of the 24 subjects are shown in the upper panel of Fig. 2. A repeated measures ANOVA of the judgment errors yielded a highly significant influence of the factor ‘‘intentionality’’ (F ð1; 23Þ ¼ 11:38, p ¼ :0026), but only a marginally significant influence of the factor ‘‘proprioception’’ (F ð1; 23Þ ¼ 4:43, p ¼ :0465). There was no significant interaction between these two factors (F ð1; 23Þ ¼ 0:46, p ¼ :5044). The results show that the perceived onset time of lever presses mainly depends on the attribution of intentionality. If the agent is conceived as an intentional one, the onset times are perceived later than in the case where the subjectÕs finger is moved passively or a rubber hand operates the lever. Although there is a numerical difference between the perceived time of events with and those events without proprioceptive information, this difference was only marginally significant. In addition, there was also no interaction between proprioception and intentionality thus that proprioception does not even modulate the intentionality effect. 2
The movement duration for the Other, Machine, and Passive condition were 40 3, 36 4, and 41 1 ms (mean standard deviation). The movement duration for the Self condition was 66 50 ms, reflecting the fact that it could not be controlled and was therefore more variable.
A. Wohlschla¨ger et al. / Consciousness and Cognition 12 (2003) 708–716
713
Fig. 2. Mean onset time judgment errors of the four conditions of Experiment 1 (top panel) and Experiment 2 (lower panel). Vertical lines indicate standard errors.
Experiment 1 replicated the results of the Wohlschl€ager et al. (in press) study: The awareness of actions that are considered as being intended by someone (oneself or others) is delayed with respect to physically comparable non-intentional events. In addition we showed, that concerning the temporal awareness of events, passive movements of oneself belong to these ‘‘unintended’’ events and are no different to any other arbitrary event.
3. Experiment 2 The event that had to be judged (the key-press) was followed by a short tone 250 ms in all conditions of Experiment 1. As mentioned above, the tone as an effect of the action of pressing the lever might be the crucial cue and the necessary condition for making inferences about the intentions of others (and perhaps oneself). We therefore replicated Experiment 1, leaving out the action effect, i.e., the tone. In this sense, the actions in Experiment 2 did not produce external effects to the same extent as those of Experiment 1.
714
A. Wohlschla¨ger et al. / Consciousness and Cognition 12 (2003) 708–716
3.1. Method Again, 24 subjects, 16 women and 8 men, all right handed were tested. Apparatus, procedure and instructions were identical to Experiment 1, with a single exception. In Experiment 2, no auditory tones occurred in any condition.3 3.2. Results and discussion Mean judgment errors of the 24 subjects are shown in the lower panel of Fig. 2. A repeated measures ANOVA of the judgment errors yielded no significant influence of the factor ‘‘intentionality’’ (F ð1; 23Þ ¼ 1:63, p ¼ :2146). However, now the factor ‘‘proprioception’’ had a highly significant influence (F ð1; 23Þ ¼ 18:22, p ¼ :0003). Again, there was no significant interaction between these two factors (F ð1; 23Þ ¼ 2:56, p ¼ :1229, b ¼ :03). Despite the obvious numerical difference in judgment errors between the Self condition ()15 ms) and the Passive condition (+6 ms), a direct comparison revealed no significant difference (tð23Þ ¼ 1:69; p ¼ :1053). The results show that in the absence of an action effect, the perceived onset time of lever presses solely depends on whether proprioceptive information is available or not. Experiment 2 contradicted the basic finding of Wohlschl€ ager et al. (in press): The awareness of self-generated actions (including passive movements of the subjects finger) is dissimilar (i.e., about 23 ms earlier) to the awareness of other-generated actions. Moreover, the attribution of intentionality does not influence the judgment errors, neither in the case of other-generated actions, nor in the case of selfgenerated ones.
4. General discussion Our results point to the conclusion that the attribution of intentionality determines the time course of our awareness of actions, regardless of whether they are executed by oneself or someone else. Moreover, the attribution of intentionality depends on the presence of an action effect that can be regarded as the realization of the content of an intention. In the lack of such an ‘‘intended’’ effect, there are clear differences in the time course of awareness of events with and without proprioceptive information and the difference between intended and non-intended actions disappears. Surprisingly, the latter is not only true for other-generated actions, but also for selfgenerated ones (see also Haggard et al., 2002b). On the other hand, the presence of an ‘‘intended’’, distal effect seems to weaken the influence proprioception has on the awareness of action. In a series of similar experiments with four levels of beep intensity (including no beep, Engbert & Wohlschl€ager, 2003), we recently showed that the influence of proprioception on the awareness of action clearly depends on the presence of distal action effects. The difference between Self and Other across the four intensity levels were: Nobeep: )33.3 ms; 63 dB (same level as in Experiment 1): )13.2 ms, 71 dB: +3.3 ms, and 76 dB: +7.4 ms. Note that the numerical difference (none of the differences in the 3 beep conditions was 3 The movement duration for the Other, Machine, and Passive condition were 35 4, 41 4, and 46 2 ms (mean standard deviation). The movement duration for the Self condition was 54 50 ms.
A. Wohlschla¨ger et al. / Consciousness and Cognition 12 (2003) 708–716
715
significant, not even at a ¼ :05) even reverses its sign with higher beep intensities. Hence, in these experiments, the effect of proprioception clearly disappeared with the introduction of a beep of arbitrary intensity. In the light of these findings, we prefer to not include the marginally significant effect of proprioception in Experiment 1 in our further interpretation.4 Together, the above findings suggest that our conscious awareness of intentions and related mental states does not arise from a private source within our own minds. Traditionally, the prototype of private access to mental states has been somatic proprioception (Melzack & Wall, 1982; Wittgenstein, 1953). On this view, conscious experience is necessarily restricted to the subject because subjects receive somatic information only from their own body and not from the bodies of others. Our results reflect this differential access only for non-operant actions. The awareness of actions having goals is similar for self- and other-generated actions. Moreover, for non-operant actions, the attribution of intentionality does not influence the awareness anymore, neither for oneself nor for others. We conclude that proprioceptive information only plays a role in the aspects of action awareness studied here, if there is no goal or intention that could be potentially shared between agents. Instead, our results suggest awareness of action distributes successfully across different agents, given that the action has an effect. In this respect, action awareness resembles the ‘‘mirror neuron’’ system reported in monkeys (Gallese et al., 1996) and its apparent analogue in humans (Fadiga et al., 1995; Rizzolatti et al., 1996), since mirror-neurons also require the action to be operant (i.e., being directed towards a goal). Recently, it was shown that goals also play a critical role in human imitation behavior (Wohlschl€ ager & Bekkering, 2002a, 2002b) and that goal directed actions activate the human homologue of the monkeys mirror-neuron area more than non-goal directed actions (Koski et al., 2002). Our data suggests that this neural circuit underlying action generation and action understanding may also participate in constructing conscious experience. Interestingly, the same neural network has also been implicated in other high-level functions, which typically require consciousness, such as communication (Rizzolatti & Arbib, 1998). Our results also raise the issue of whether people represent the actions of others by analogy with their own. On this interpretation, one has private access to oneÕs own intentions, and these intentions generate conscious awareness of action (Haggard & Eimer, 1999). One can then infer the intentions of others from their actions, by analogy with oneÕs own private case. This account rescues the privacy view, but at the same time softens the concept of privacy, since other people can now infer oneÕs ‘‘private’’ intentions, from oneÕs behavior. Moreover, if awareness of action could arise from mental states inferred in other people, rather than directly experienced, then oneÕs awareness of oneÕs own action might likewise be inferential, and therefore not especially privileged. In the above experiments, we have studied the perceived onset times of actions with and without effects. Our results show that the differential activation of the proprioceptive system in self-generated and observed actions yields differential awareness of the onset times of the actions. However, if the action has an effect, the differential activation of the proprioceptive system seems 4
The probability p that the conjunction of n propositions is true is the product of the probability of each individual proposition being true. Choosing the same a level for each individual proposition, p is a function of a and n : pða; nÞ ¼ ð1 aÞn . Choosing a ¼ :01, the conjunct a-level for the final conclusion would be .19 as opposed to .04 when choosing a ¼ :01. This is another reason to exclude the marginally significant effect of proprioception in Experiment 1 for the final conclusion.
716
A. Wohlschla¨ger et al. / Consciousness and Cognition 12 (2003) 708–716
to be overridden by mechanisms that attribute intentions to others (and perhaps also to oneself) and that leads to a common representation of self- and other-generated actions. In short: it is goals that unify people.
References Craighero, L., Fadiga, L., Rizzolatti, G., & Umilta, C. (1999). Action for perception: A motor-visual attentional effect. Journal of Experimental Psychology—Human Perception and Performance, 25(6), 1673–1692. Craighero, L., Fadiga, L., Umilta, C. A., & Rizzolatti, G. (1996). Evidence for visuomotor priming effect. NeuroReport, 8(1), 347–349. Engbert, K., & Wohlschl€ager, A. (2003). The perceived onset time of self- and other-generated actions. Poster presented at the Tutorials in Behavioural and Brain Sciences 2003: The Visible Mind: Recording Cognitive Functions, Leipzig. Fadiga, L., Fogassi, L., Pavesi, G., & Rizzolatti, G. (1995). Motor facilitation during action observation—a magnetic stimulation study. Journal of Neurophysiology, 73(6), 2608–2611. Fried, I., Katz, A., McCarthy, G., Sass, K. J., Williamson, P., Spencer, S. S., & Spencer, D. D. (1991). Functionalorganisation of human supplementary motor cortex studied by electrical-stimulation. Journal of Neuroscience, 11, 3656–3666. Gallese, V., Fadiga, L., Fogassi, L., & Rizzolatti, G. (1996). Action recognition in the premotor cortex. Brain, 119, 593– 609. Goldberg, G., Kwan, H. C., Borrett, D., & Murphy, J. T. (1984). Topography of the movement-associated scalp potential (masp) suggests initiation of spontaneous movement by the supplementary motor area. Archives of Physical Medicine and Rehabilitation, 65(10), 662–663. 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. Haggard, P., Clark, S., & Kalogeras, J. (2002b). Voluntary actions and conscious awareness. Nature Neuroscience, 5(4), 382–385. Haggard, P., & Eimer, M. (1999). On the relation between brain potentials and the awareness of voluntary movements. Experimental Brain Research, 126(1), 128–133. Kornhuber, H. H., & Deecke, L. (1965). Hirnpotential€ anderungen bei Willk€ urbewegungen und passiven Bewegungen des Menschen: Bereitschaftspotential und reafferente Potentiale. Pfl€ ugers Archives, 284, 1–17. Koski, L., Wohlschl€ ager, A., Bekkering, H., Woods, R. P., Dubeau, M.-C., Mazziotta, J. C., & Iacoboni, M. (2002). Modulation of motor and premotor activity during imitation of target-directed actions. Cerebral Cortex, 12(8), 847– 855. Libet, B., Gleason, C. A., Wright, E. W., & Pearl, D. K. (1983). Time of conscious intention to act in relation to onset of cerebral activity (readiness-potential)—the unconscious initiation of a freely voluntary act. Brain, 106, 623–642. Melzack, R., & Wall, P. (1982). The challenge of pain. Harmondsworth, UK: Penguin. Rizzolatti, G., & Arbib, M. A. (1998). Language within our grasp. Trends in Neurosciences, 21(5), 188–194. Rizzolatti, G., Fadiga, L., Gallese, V., & Fogassi, L. (1996). Premotor cortex and the recognition of motor actions. Cognitive Brain Research, 3(2), 131–141. Tanji, J. (2001). Sequential organization of multiple movements: Involvement of cortical motor areas. Annual Review of Neuroscience, 24, 631–651. Wittgenstein, L. (1953). Philosophical investigations. Oxford: Blackwell. Wohlschl€ager, A., & Bekkering, H. (2002a). Is human imitation based on a miror-neurone system? Some behavioural evidence. Experimental Brain Research, 143, 335–341. Wohlschl€ager, A., & Bekkering, A. (2002b). The role of objects in imitation. In M. Stamenov & V. Gallese (Eds.), Mirror Neurons and the Evolution of Brain and Language (pp. S.101–S.114). Amsterdam: John Benjamins. Wohlschl€ager, A., Haggard, P., Gesierich, B., & Prinz W. (in press) The perceived onset time of self- and othergenerated actions. Psychological Science.
