, Volume 37, Issue 1, pp 21–30

Agency, Consciousness, and Executive Control


  • Jesús H. Aguilar
    • Department of PhilosophyRochester Institute of Technology
    • Department of Philosophy and Religious StudiesMarist College

DOI: 10.1007/s11406-008-9123-y

Cite this article as:
Aguilar, J.H. & Buckareff, A.A. Philosophia (2009) 37: 21. doi:10.1007/s11406-008-9123-y


On the Causal Theory of Action (CTA), internal proper parts of an agent such as desires and intentions are causally responsible for actions. CTA has increasingly come under attack for its alleged failure to account for agency. A recent version of this criticism due to François Schroeter proposes that CTA cannot provide an adequate account of either the executive control or the autonomous control involved in full-fledged agency. Schroeter offers as an alternative a revised understanding of the proper role of consciousness in agency. In this paper we criticize Schroeter’s analysis of the type of consciousness involved in executive control and examine the way in which the conscious self allegedly intervenes in action. We argue that Schroeter’s proposal should not be preferred over recent versions of CTA.


AgencyIntentional actionCausal theory of actionConsciousnessExecutive control


On the causal theory of action (CTA), some behavior A of an agent S is an action if and only if (1) S’s A-ing is non-deviantly caused by some non-actional mental items R and (2) R constitutes S’s reasons for A-ing.1 A common complaint leveled against CTA is that it commits its proponents to a mechanistic view of intentional behavior that renders any role for the agent otiose. This objection to CTA has a distinguished pedigree. It goes back at least to the work of Roderick Chisholm (1966) and Richard Taylor (1966) who criticized this theory because of its alleged failure to account for the agent’s active participation in the production of an action.2

Recently, there have been renewed criticisms leveled against CTA. Like Chisholm and Taylor in the 1960s, a new wave of critics of CTA believe it is a theory that “should be rejected, not embellished” (Hornsby 2004b, p. 173). The main concern of these critics of CTA is the same concern as earlier critics, viz., that by reducing agency to the realization of the “right type” of causal relation between some mental items and behavior, CTA effectively eliminates the role of the agent as the source of action.3 The agent is reduced to being the mere locus of causal activity from which purposeful behavior is generated. The upshot is that CTA leaves agents out and “in any of its versions…is not a story of agency at all” (Hornsby 2004a, p. 2). Call this the problem of the absent agent.

François Schroeter’s (2004) work on action is representative of the sentiments expressed by the growing chorus of critics of CTA. He echoes the traditional worries, proposing that CTA continues to exhibit fundamental shortcomings despite the efforts to enrich this theory by some defenders of endorsement models of autonomy.4 While there are important differences among the several endorsement theories of autonomous agency, they all regard the formation of higher-order conative states, policies, or intentions as necessary for autonomous agency. The differences notwithstanding, according to Schroeter, all endorsement models of autonomy exhibit the same shortcomings of CTA by failing to adequately explain the participation of the agent in two fundamental types of agential control: executive control and autonomous control. The first type of control takes place at the executive stage of the action while the second type of control occurs whenever the agent exhibits her autonomy. Schroeter believes the agent participates in these two types of agential control in the guise of what he calls “the conscious self”, and, furthermore, that it is precisely this agential participation that has eluded CTA and endorsement models of autonomy.

In this paper we defend CTA by going on the offensive. We concentrate our attention on the nature of the conscious self and its alleged contribution to executive control in Schroeter’s theory of action. For independently of the difficulties that an enriched CTA may face in accounting for autonomy, Schroeter’s suggestion that this theory is essentially mistaken in its account of executive control strikes us as a very serious charge. By examining the shortfalls of Schroeter’s proposed theory of action, it should become apparent that CTA is in no worse shape than his alternative suggestion and is in fact better positioned to be the best candidate theory for a naturalized theory of human action and agency.

The first section of this paper deals with Schroeter’s analysis of the type of consciousness involved in executive control, namely, “access consciousness.” The second section deals with the way in which the conscious self allegedly intervenes in action.

The Conscious Self and Access Consciousness

The conscious self is characterized by Schroeter as a system or set of systems capable of entertaining and manipulating “conscious thoughts, including action plans and the thoughts involved in the conscious monitoring of their realization” (2004, p. 642). It is also capable of deliberating and reflectively endorsing or rejecting mental states such as desires. Thus, the conscious self comes across as a very rich entity intervening in all kinds of agential scenarios from bodily movements such as hitting a vein with a syringe to more complicated cases involving the endorsement of first-order desires and resistance to the force of cravings.5 This richness of properties and functions is something to be expected, for according to Schroeter the agent is the conscious self (2004, p. 642).

Concerning consciousness, which is the key agential feature in Schroeter’s analysis, he adopts Ned Block’s (1995) well known although contentious distinction between “phenomenal-consciousness” and “access-consciousness.”6 Block distinguishes these two types of consciousness as follows. First, phenomenal-conscious content is phenomenal and only sometimes transitive (i.e., involving the subject being conscious of some x) while access-conscious content is representational and necessarily transitive. Second, for every token phenomenal-conscious state there is a phenomenal-conscious type of state that can be instantiated at some other time; but any token access-conscious state could fail to be accessible at some other time. Third, access consciousness is a functional notion and its content is system-relative, but phenomenal consciousness is neither a functional notion nor is it system-relative (Block 1995). Schroeter proposes that access-consciousness is the type of consciousness that characterizes the conscious self (2004, p. 640 footnote 11 and p. 642 footnote 13).

One virtue of Schroeter’s proposal is to remind us of the often ignored role that consciousness plays in the production of an action. Although consciousness is seldom required to be a condition in the production of an intentional action, there is an intuitive sense in which consciousness seems to be involved in the process of executing most actions.7 Even relatively simple intentional actions like extending one’s arms to stop a fall or the avoidance of sudden obstacles when driving a car—to use one of Schroeter’s examples—seem to engage the agent’s consciousness. For in all such cases agents know what they are doing and whether they have accomplished their intended goal.

According to Schroeter, and despite the intuitive pervasive presence of phenomenal-consciousness in the production of actions, one has to be careful in attributing anything like executive control to a subject on the sole basis of phenomenal-consciousness. It may be that the agent’s brain is engaged “in a fair amount of central monitoring” that accounts for the complexity and sophistication of many of these behaviors, but unless they are guided by the conscious self the agent is completely passive. In fact, in cases like these where the agent is a passive bystander of what her brain does (e.g., when an epileptic experiencing a focal point seizure in her left temporal lobe successfully drives a car), Schroeter resists identifying such bodily movements as actions preferring to call them “quasi-reflexive behaviors” (2004, pp. 639–40).

A first difficulty in Schroeter’s way of accounting for the correct attribution of executive control to agents arises out of this way of characterizing such complex “quasi-reflexive” behaviors. Not only does this proposal run against basic intuitions that consider these behaviors to be fully intentional and under the guidance of the agent in the form of her brain’s monitoring their execution, but they are explained in terms of some mysterious capacity possessed by the brain that for all purposes amounts to a parallel center of control. After all, to use Schroeter’s example again, the avoidance of sudden obstacles when driving a car exhibits a great deal of subtle control that guides the fast and precise performance of many actions.

However, even if one grants that there are quasi-reflexive behaviors controlled by the brain sitting halfway between simple reflexes and real actions, there is another difficulty emerging from Schroeter’s proposal: the sole evidence available to the agent that she is indeed in control of her action comes from the introspective presence of phenomenal-consciousness and not, as Schroeter proposes, from access-consciousness.

Recall that in contrast to phenomenal-consciousness, which is always dependent on the subject’s recognition of its presence, access-consciousness works “silently” and can function undetected by the agent herself. Consider one of Block’s favorite examples of access-consciousness. In the middle of a conversation one suddenly notices a disturbing noise (coming from a jackhammer or a refrigerator) that has always been there without one’s attending to it (Block 1995). Block highlights the contrasting nature of these two types of consciousness with this example: while access-consciousness was there all the time providing information about the ongoing noise, phenomenal-consciousness was absent and only later present. Thus, far from engaging the agent at the personal and experiential level in which phenomenal consciousness finds its home, access-consciousness involves paradigmatic sub-personal information processing which often occurs without the subject noticing its presence. But, if this is how access-consciousness typically functions, it is very hard to see how Schroeter can make much use of this type of consciousness as a source of executive control given his way of understanding it. For if the agent is to be in a position to recognize her controlling guidance over an action, it had better be the case that for every correct introspective self-attribution of agential control the relevant information is consciously available to her. Clearly, a type of consciousness that usually works “silently” in the background is not going to satisfy this requirement.

To stress this final point a bit further, imagine what the consequences for Schroeter’s proposal concerning agential control would be if an agent denies having had any phenomenal experience at the moment of producing a complex action like parking a car. It seems that the executive control involved in such an incredible feat would be nothing short of miraculous. So, much in the same way in which blind-sighted patients give rise to puzzles about theories of perceptual consciousness, agential-blindness of this type would give rise to similar puzzles about the precise contribution of consciousness in the production and control of actions.

At this stage, it is evident that an agent’s role in the production of an action is not enhanced by the sort of role Schroeter affords access consciousness. Moreover, some of the functions he attributes to the conscious self seem to be duplicated by intentions. We turn to the role of intentions for CTA and the role of agents in the etiology of action in the next section.

Intentions, Agents, and Controlling Behavior

Recent versions of CTA ascribe intentions a central role in providing an account of executive control over intentional behavior.8 Intentions are seen as internal proper parts of an agent, sometimes passively acquired in response to reasons for action and sometimes actively formed by deciding to act. Intentions are considered as mediating plans and reasons for acting, being formed in response to reasons for action and planning. Furthermore, intentions are taken to be the executive states that coordinate, cause, sustain, guide, and monitor intentional behavior. Intentions enable agents to be responsive to afferent feedback in the performance of actions. In short, we exercise agency by controlling our behavior via our intentions.

Such a reductive account of agency strikes new wave critics of CTA as untenable. Thomas Nagel aptly represents this sentiment. Nagel notes that if you think of yourself as one part in a series of causal relations between events, you will find yourself alienated from your actions: “Everything I do or that anyone else does is part of a larger course of events that no one ‘does’ but that happens” (1986, p. 113). Such a putative external perspective on the production of action undermines agency. The difficulty this poses leads some to declare that, “we need to be released from the straitjacket of [CTA] to give a realistic account of human agency” (Hornsby 2004a, p. 22). The result has been that some have endorsed stories about human agency that introduce opaque causal relations with dubious naturalistic credentials. For instance, some philosophers have promoted versions of agent-causation as at least a proper part of their theory of action.9 Defenders of this program have modified the agent causal theory of action (ACTA) in a variety of ways.10 Others, such as Hornsby, offer less radical proposals.11 And, as noted earlier, some advocates of CTA, such as Bratman (2001) and Velleman (2000), recommend enriching CTA by incorporating the features of endorsement models of autonomy.

Schroeter’s proposal is offered as a remedy for what ails theorizing about action using CTA. His criticism specifically takes aim at the executive role afforded to intentions on these recent versions of CTA. His main concern is the same as other critics of CTA. Specifically, versions of CTA ultimately reduce agents to mere patients due to their commitment to a so-called “Hobbesian” explanatory framework. According to such an explanatory framework, the agent is merely “the theater in which intentions crystallize as a result of the causal interactions of beliefs and desires” (2004, p. 647). This, in turn, leads to a fundamentally mistaken view where “a full explanation of the action does not need mention the agent but only particular mental states such as desires” (2004, p. 647).12 That is, in such a Hobbesian explanatory framework the agent effectively disappears from the picture and causal chains of states and events take over.13

If this recurrent type of criticism against CTA is correct, then it seems the problem of the absent agent will generalize to cases of executing intentions in action. However, this problem also arises even when the agent is afforded the sort of role that a Schroeter-like criticism of CTA demands as a necessary condition for full-blooded agency. Compare the following:
  1. 1.

    Suguro moves his leg at t bringing his foot into contact with a ball, thereby kicking the ball.

  2. 2.

    Suguro moves his leg at t as a causal response to his forming a present-directed intention to kick a ball, his activity aimed at kicking a ball is causally sustained and guided by his intention to kick a ball by which Suguro monitors his activity, and he remains responsive to afferent feedback via his intention throughout the performance of the action.14


If we assume that any agent at a time t is constituted by her body and the mental states and events realized by some neural activity at t, then both (1) and (2) share the problem of not involving Suguro as exercising executive control over his intentional behavior. In (1) some proper part of Suguro (his leg) is what moves and comes into contact with the ball. But it would seem odd if we were to claim that Suguro did not kick the ball. In (2) some proper part of Suguro at the time he kicks the ball (his proximal intention) is what plays an executive role in the etiology and direction of his intentional behavior. But it should seem no less odd to say that Suguro does not exercise executive control over his behavior. Just as a proper part of Suguro is responsible for the end result of the execution of his intention when he kicks the ball and we correctly say “Suguro kicked the ball,” the fact that his intention is the means whereby he exercises executive control does not eliminate Suguro’s role in exercising agency and control over his behavior. Therefore, merely on this basis it is wrong to consider Suguro as passive. He would be wholly passive if something external to him was to control his behavior and he could do nothing to stop it. But, this is not implied by the present picture.

So, if proper parts of an agent’s body proximally cause events that we hold the agent responsible for, then it should seem no less strange that a proper part of an agent causes the bodily movements we describe as intentional and yet we still hold the agent responsible. Moreover, if we are committed to a view of ourselves as part of the causal order, then taking intentions which arise in response to reasons for acting as the means whereby we exercise control should not seem odd. The story we can tell about our agency and the role of our intentions in its etiology and control should only seem strange and inadequate if we are committed to a view of ourselves as substances that can directly cause behavior, transcending the causal order. Such a view, despite its apparent anti-naturalist credentials, is supported by philosophers that defend the agent-causal theory of action (ACTA).

Returning to Schroeter’s particular proposal, he explicitly rejects ACTA. For instance, he claims that he is not committed to “taking agency as a primitive phenomenon” (2004, p. 650). However, given his account of the causal role played by the conscious self and his rejection of intentions as sources of executive control, Schroeter appears to be committed to a theory of executive control that shares the liabilities of ACTA ascribed to it by proponents of naturalism in the philosophy of action.

On Schroeter’s theory of action, the agent or conscious self is the causal source of the mental items that figure in the control of intentional behavior. However, this proposal raises the following dilemma: either the mental items in question are proper parts of the agent, and so it is really some unspecified part of the agent that controls other parts of the agent, or the mental items in question are distinct from the agent.15 If the former, then Schroeter owes us an explanation as to why this proper part of the agent (which is left unspecified) is any better than an executive state like an intention as the means whereby executive control is exercised. If the latter, then Schroeter must explain how he can avoid the charge of merely offering a dressed-up version of ACTA which posits mysterious causal relations between substances and events.

Our belief that indeed Schroeter favors a version of ACTA is greatly increased not just by this uncomfortable dilemma faced by his theory but by passages that seem to support such a view. For instance, compare the following statements made by Chisholm, perhaps the most articulate defender of ACTA, to statements made by Schroeter in his essay. Here is Chisholm:

We may say that the hand was moved by the man, but we may also say that the motion of the hand was caused by the motion of certain muscles; and we may say that the motion of the muscles was caused by certain events that took place within the brain. But some event, and presumably one of those that took place within the brain, was caused by the agent and not by any other events. (Chisholm 1966, p. 18) 16

And, here is Schroeter:

I have claimed that, in monitoring the execution of an action plan, the conscious self can send central commands to the motor system in order to initiate, stop or reschedule the different steps required for the realization of the plan. (Schroeter 2004, p. 649)17

The fundamental similarities between these proposals reveal that Schroeter, like Chisholm before him, is committed to a theory of action with less than stellar naturalist credentials. This feature of Schroeter’s theory does not count decisively against it. But it does seem to undermine its initial appeal if one is already disposed to reject any theory of action that resorts to mystery and posits causal relationships of dubious pedigree. Hence, his comments notwithstanding, there are strong reasons to believe that Schroeter actually embraces a version of ACTA with all its traditional problems.


CTA should be preferred over Schroeter’s theory of action. Although we will not attempt to make the positive case for CTA here in any detail, consider the following assets of CTA and how they compare to Schroeter’s main proposal. First, regarding the role of the conscious self, not only does Schroeter’s theory allow for redundant centers of agential control; but, more seriously, his ascription of access consciousness to the conscious self in exercising control hardly seems adequate for the role he assigns it in controlling behavior. CTA offers a simpler picture of executive control. On most recent proposals, agents exercise executive control via their intentions. Moreover, CTA does not commit its adherents to a particular theory of consciousness. Theorizing about knowledge and self-awareness of our actions is kept separate from theorizing about executive control.

Second, if some proper parts of agents are responsible for exercising executive control over action, it is not obvious that we should prefer alternative parts to those endorsed by recent versions of CTA as the means whereby executive control is exercised. Consciousness, much less access consciousness, does not afford us executive control over our actions.

Finally, if it is the agent and not some proper part of the agent that is causally responsible for the production and control of action, then Schroeter needs to explain why we should not take him as endorsing a version of ACTA. Schroeter’s theory gives us a robust theory of the role of agents in agency and intentional action. But it threatens to introduce mysterious causal relations between agents and actions.

CTA should be preferred over theories like Schroeter’s if we desire simpler explanations and ontological parsimony. CTA provides a simpler explanation of human action and does not needlessly multiply entities. For these reasons, CTA better meets a third desideratum in theorizing about human action. A theory of action that is more amenable to a naturalistic research program about the nature of mind and agency is to be preferred if we regard our philosophical theorizing about human action to be on a continuum with work in the sciences.


This is offered as a schema that represents in rough form the basic elements of CTA typically agreed upon by its defenders. For recent book-length defenses of CTA see Bishop (1989), Brand (1984), Mele (1992 and 2003), and Enç (2003).


The same worry can be traced back as far as T. Reid’s (1788/1969) criticism of proto-causal theories of action such as Hume’s and Hobbes’s. For instance, here is Reid articulating his criticism of Hobbes’s views: “to say that what depends upon the will is in a man’s power, but the will is not in his power, is to say that the end is in his power, but the means necessary to that end are not in his power, which is a contradiction” (1788/1969, p. 266).


See Alvarez and Hyman (1998), Hornsby (2004a and 2004b), Melden (1961), Nagel (1986), and Schroeter (2004) for representative critiques of CTA along these lines.


Schroeter focuses on the work of two prominent defenders of such theories, viz., Harry Frankfurt (1988 and 1999) and J. David Velleman (2000). Strictly speaking, Velleman is the only one of these two authors who is in the business of trying to enrich CTA. Frankfurt explicitly rejects CTA in “The Problem of Action” (in 1988, p. 69). This is not to say that a Frankfurtian enrichment of CTA is impossible as precisely the work of Velleman (1992) and Bratman (2001) have shown.


Essentially, all behavior involving an action plan and the “anticipatory awareness of their unfolding” would be under the control of the conscious self, and this includes pretty much all our intentional actions. Given Schroeter’s analysis of intentional behavior it is surprising that he is not ready to generalize to include every single intentional action. In fact, he is careful to avoid proposing that he is offering anything like a set of necessary and sufficient conditions of intentional action where one would find the presence of executive control. Nonetheless, the fact remains that intentional action requires in his opinion the presence of executive control and thus of the conscious self. His misgivings about making such a definitional move are rather that “intentional action [is] a complex phenomenon which resists analysis in a set of simple necessary and sufficient conditions.” (Schroeter 2004, p. 646 footnote 20).


Two important criticisms of Block’s distinction are found in Dennett (1995) and Burge (1997).


There are important exceptions to this tendency to ignore the role of consciousness in action, for instance Carl Ginet has given a lot of weight to the “actish phenomenal quality” present when acting (Ginet 1990).


See, e.g., Bishop (1989), Brand (1984), Enç (2003), and Mele (1992 and 2003).


This is different from current defenders of agent-causation in the debate over free will, such as Timothy O’Connor (2000). O’Connor is not obviously interested in offering an agent-causal theory of action simpliciter.


For instance, Maria Alvarez and John Hyman reject the thesis that actions are events and suggest the following. “What appears to be unsustainable is the doctrine that actions are events caused by agents. But it is possible to detach the concept of agent-causation from this doctrine, for the claim that there is a defensible conception of agent causation implies only that an action is a causing of an event by an agent: there is no need to suppose, in addition, that this event is the agent’s action, or that an action is itself an event” (1998, p. 224). See also E.J. Lowe (2001).


For reasons we will not explore here, it is not obvious that Hornsby’s proposal in 2004a and 2004b is really much of an improvement over CTA, for it seems to us that her proposal is as mysterious as ACTA is. Furthermore, as we argue below and notwithstanding his denials, Schroeter’s alternative proposal to CTA is little more than a species of ACTA.


Compare Schroeter’s remarks to some offered by Chisholm. “For if what we say he did was really something that was brought about by his own beliefs and desires, if these beliefs and desires in the particular situation in which he happened to have found himself caused him to do just what it was we say he did do, then, since they caused it, he was unable to do anything other than just what he did do. It makes no difference whether the cause of the deed was internal or external: if the cause was some state or event for which the man himself was not responsible, then he was not responsible for what we have been mistakenly calling his act” (Chisholm 1966, p. 13).


Alfred Mele (2003, chapter 10) addresses similar concerns raised by Velleman. Velleman (1992) expresses doubts about whether the CTA can account for “full-blooded” action or “human action par excellence” (Velleman 1992, 462). Velleman is concerned about what Schroeter calls “autonomous control” (what Mele refers to as “self control” in 2003, 222ff). The concern is that CTA affords agents a diminished role in exercising autonomous control. But, contra Schroeter, Velleman does not reject a reductive CTA type of story about executive control. He embellishes it in the interest of providing an account of autonomous agency. So while Mele’s reply to Velleman’s concerns resembles our present effort, our reply is to a more basic critique of the CTA, namely, CTA’s alleged inability to provide us with an adequate conception of the executive control that is the hallmark of all intentional action (whether autonomous or not). Thanks to a referee for pressing us on this point.


See Bishop (1989) for the most developed and rigorous philosophical defense of the role of feedback loops in intentional action.


The following dilemma is similar to one posed by Donald Davidson 1980, pp. 52–53.


Compare these remarks by Chisholm on the production of action to similar comments by him about making something happen directly in 1964, p. 620 and 1966, p. 17. Nonetheless, Chisholm dispensed with ACTA in what is probably his last statement on the topic of agency (see Chisholm 1995). Richard Taylor (1982) also dispensed with ACTA.


Earlier remarks in Schroeter’s paper on how executive control can be exercised are also quite telling: “Once an agent has decided to perform a given option, the action can normally be initiated at his command when the circumstances are appropriate. That is to say, the motor system can be directly activated by a command issued by the conscious self (‘Start going to the gym right now!’)” (Schroeter 2004, pp. 644–45).


Copyright information

© Springer Science+Business Media B.V. 2008