What are self-generated actions?

Abstract

The concept of self-generated action is controversial, despite extensive study of its neural basis. Why is this concept so troublesome? We analyse the concept of self-generated action as employed by Passingham et al., 2010a, Passingham et al., 2010b. There are two definitions of self-generated action; as operant action and as underdetermined action. The latter draws on subjective experience. Experiments on action awareness suggest that experience may not be a good guide for defining self-generated action. Nevertheless, we agree with Passingham and colleagues that self-generated actions exist distinct from operant actions. But defining ‘self-generated’ by the degree of involvement of an endogenous process risks regress. We sketch an alternative account of self-generated action that tries to avoid these problems to make self-generated actions accessible to experimental science.

Introduction

The scientific study of self-generated action, and its relation to volition and conscious experience, has long been controversial. For example, Libet’s finding that the onset of the readiness-potential precedes reports of conscious intention (Libet, Gleason, Wright, & Pearl, 1983) continues to provoke discussions about ‘conscious free will’. While the neural basis of self-generated actions has been studied extensively (Eccles, 1982, Fleming et al., 2009, Okano and Tanji, 1987, Romo and Schultz, 1987, Shima et al., 1991, Thaler et al., 1995), the concept of self-generated action remains difficult. Why is this concept of self-generated action so troublesome?

In a recent exchange between Passingham et al., 2010a, Passingham et al., 2010b and Nachev and Husain (2010), the difficulties with the concept of self-generated actions have become apparent. This exchange exemplifies differences in opinion in the field of study of self-generated actions. As such it is informative about why it is that the study of self-generated action has been plagued by controversies and debate.

In this article, we first review the debate between Passingham et al., 2010a, Passingham et al., 2010b and Nachev and Husain (2010). Then, we take a closer look at the literature on self-generated actions that Passingham et al., 2010a, Passingham et al., 2010b refer to for support. There are two definitions of self-generated action: self-generated actions can be defined either as operant actions, or as underdetermined action. We accept the first definition as a scientifically viable conception of self-generated action. By contrast, we argue that the definition of self-generated action as underdetermined actions draws from subjective experience, in particular regarding the (often implicit) idea of the involvement of a self in self-generated action. We review literature on action awareness to show that subjective experience is not a good guide for defining self-generated action. Then, we assess whether brain anatomy can provide arguments for relating the self to self-generated action. After exploring the concept of ‘endogenous process’ (Passingham, Bengtsson, & Lau, 2010b) in relation to self-generated action as underdetermined action, we sketch a novel definition of self-generated action as complex integration. This approach suggests new directions for the future study of self-generated action.

In the article that sparked off the controversy, Passingham, Bengtsson, and Lau, review and endorse the claim that the medial frontal cortex in general, and the supplementary motor cortex in particular, are crucially involved in self-generated action (2010a). Actions are self-generated, according to the authors, if these actions are not uniquely specified by external cues (Passingham et al., 2010b, Passingham et al., 2010a). Passingham et al. (2010a) discuss and dismiss challenges to their view, for example that differences in neural activity for self-generated compared to externally triggered action may be explained solely by differential demands on working memory, increased conflict processing, and increased task-complexity (p. 17). The authors speculate that internal as opposed to external cues could give rise to self-generated action. Internal cues could are, for example, one’s own preceding actions (e.g., Halsband and Passingham, 1982, Passingham, 1987), elapsed time (e.g., Mita, Mushiake, Shima, Matsuzaka, & Tanji, 2009), or (changing) behavioural goals (e.g., Kennerley, Dahmubed, Lara, & Wallis, 2009).

Then, Passingham and colleagues expand their claim: They propose that the medial frontal cortex underlies self-reflection, episodic memory, and self-referential thoughts during ‘mind-wandering’. More generally, they propose that information related to the self is processed on the medial frontal surface of the brain. By contrast, information about the external environment is processed on the lateral surface of the brain. This proposal is consistent with what is currently known about the anatomical connections of the medial and lateral surfaces of the brain (Passingham et al., 2010a, p. 19).

Nachev and Husain question the possibility of investigating self-generated action scientifically on purely conceptual grounds: “… the contrast between internally- and externally-guided actions is empirically intractable.” (Nachev & Husain, 2010). According to Nachev and Husain, internal cues are by definition hidden from any experimenter’s view (p. 192). The invisibility of cues, however, does not rule out their existence or importance for behaviour. The effects of external cues on action can be investigated using conventional experimental methods. However, because these methods cannot or are not used to investigate the effects internal cues on action, there are a myriad of possibly relevant internal cues and we cannot determine what they are. Therefore, comparison between self-generated and externally triggered action cannot yield scientifically meaningful data (Nachev & Husain, 2010).

Passingham and colleagues (2010b) reply to this criticism is twofold. First, they state that the number and nature of cues, internal and external, can be controlled for by carefully balancing task features such as complexity and level of conflict, between self-generated and externally-triggered action. Second, they state that the distinction between externally-trigged and self-generated action is a matter of degree, rather than a sharp divide: “… it would be fallacious to claim that there are no external components in self-generated actions and no internal components in externally generated actions.” (Passingham et al., 2010b). Rather: “It is the difference between two conditions [i.e. self-generated and externally-trigged] that can be said to reflect a difference in the degree of involvement of an endogenous processes.” (Passingham et al., 2010b, p. 194).

Passingham’s et al. answer (2010b) may not fully address Nachev and Husain’s criticism. For example, if removing external cues were, by itself, to increase task complexity, then one could not modulate the self-generatedness of an action while at the same time equating for complexity. In this article, instead of taking sides in this debate, we take a closer look at the literature of self-generated action that Passingham et al., 2010a, Passingham et al., 2010b refer to for support. In particular, we focus on how it is that self-generated actions are defined and operationalised in these studies.