- Top of page
Mandik (2010) defends a motor theory of control consciousness according to which nonsensory states, like motor commands, directly contribute to the awareness we have of ourselves as being in control of our actions. Along the way, he argues that his theory is to be preferred over Prinz’s (2007) sensory imagery theory, which denies that nonsensory states play any direct role in the generation of control consciousness. I argue that Mandik’s criticisms of Prinz’s theory fall short, but that nonetheless there are reasons to favor a motor theory of control consciousness over a sensory imagery theory.
In his paper, “Control Consciousness,”Mandik (2010) defends a nonsensory, or motor theory of control consciousness. According to such a view, nonsensory states, for example, motor commands, make a direct contribution to one’s awareness of being in control of one’s actions. Such a view is to be contrasted with a rival theory, which Mandik considers and rejects: the sensory imagery theory of control consciousness defended by Prinz (2007). Prinz’s theory aims to explain control consciousness in terms of sensory imagery and sensory input alone. On Prinz’s view, immediately prior to performing an action we generate an anticipatory sensory image of the sensory consequences of that action. Such sensory images are the outputs of a forward model, which calculates them on the basis of a copy of the motor command. When the actual sensory consequences of a bodily movement match the anticipatory sensory image generated by the forward model, we are aware of ourselves as being in control of that movement. So, by contrast to the motor theory, there is no role left to play here for any nonsensory states.
While several issues, both interesting and important, are raised and addressed by Mandik in his paper, I restrict my attention here to his case for rejecting Prinz’s sensory imagery theory in favor of his own motor theory of control consciousness. I will argue that such a move is indeed warranted, but not for the two main reasons that Mandik offers.
Mandik’s first objection to Prinz’s (2007) account is that he fails to justify his classification of the output of a forward model as a sensory image. According to Prinz (2002), what it is for something to be a sensory state is for its representational format to be specific to a sensory modality. The senses themselves are dedicated input systems, and sensory imagery, on Prinz’s view, is the result of “willful reactivation” of these input systems. Mandik argues that since forward models do not receive sensory inputs, they are not sensory input systems, as conceived by Prinz. Thus, they cannot be willfully reactivated in order to give rise to sensory imagery.
But a proponent of the sensory imagery theory need not be committed to Prinz’s particular characterization of sensory systems as receiving sensory input. Perhaps, like some theorists, one might allow that forward models are sensory systems despite taking as input motor representations, that is, copies of motor commands, rather than sensory input. Accordingly, Bayne (in press) draws a distinction between what he calls basic sensory systems like vision, touch, taste, olfaction, and hearing, and nonbasic sensory systems, claiming that the former take energy as input, whereas the latter take representations as input. On such a view, the forward model constitutes a nonbasic sensory system, but a sensory system no less.
Moreover, even if forward models are not themselves sensory systems by any account, they need not be in order to output sensory images. All that is required, at least on Prinz’s view, is that their outputs result from the willful reactivation of our existing sensory systems, such as vision or proprioception. Mandik’s case would be made stronger were he to offer evidence that the forward model is unable to draw on the capacities of our sensory systems in this way.
It is worth noting, however, that even if forward models do indeed output sensory images, this does not pose a threat to the motor theory of control consciousness. After all, this picture is consistent with both the claim that a purely sensory theory of control consciousness will not suffice, and the claim that control signals directly contribute to control consciousness, which together form the core of Mandik’s position.
Moving on, Mandik’s second objection to the sensory imagery theory of control consciousness also faces some challenges. Mandik argues that his motor theory of control consciousness should be favored over the sensory imagery theory because it can account for the way in which we distinguish sensory imagery from perception, whereas the sensory imagery theory cannot. To elaborate, on Mandik’s view, sensory images are distinguishable from perceptions with similar content in virtue of being accompanied by control consciousness. In other words, we are aware of ourselves as having control over our sensory images, but not over our perceptions, and this allows us to tell them apart. But, argues Mandik, since the sensory imagery theory does not allow for nonsensory states to directly contribute to control consciousness, a proponent of this view is committed to saying that there is no first-person way by which to distinguish sensory imagery from perception. And since we often do distinguish between the two from the first person, Mandik urges, we should reject Prinz’s sensory imagery theory in favor of the motor theory.
There are several worries here. First, the sensory imagery theorist may deny that control consciousness is involved in how we distinguish between sensory imagery and perception, and yet still insist that the two are distinguishable from the first person. There may be other criteria to go on. For instance, the first-person difference between perception and sensory imagery may lie in the higher degree of vivacity with which perceptions manifest themselves, compared to the relative dullness of sensory images. Moreover, perception depends on our sense organs and environment in a way that sensory imagery, of course, does not, and this dependence is discoverable from the first person. If we accept these proposed criteria, then the motor theory does not have any clear advantage over the sensory imagery theory when it comes to distinguishing sensory images from perceptions.
Next, Mandik’s argument would likely not be compelling for a sensory imagery theorist like Prinz (2007) himself, since Prinz seems to accept that there is no qualitative difference between sensory imagery and perception with similar content. About sensory images, Prinz writes:
It is perfectly plausible that we form conscious anticipatory images of our actions before executing them. These images may be generated just a few milliseconds before movement, so they are quickly succeeded by the perception of our bodies in motion. The matching process may consist in the fact that there is no significant change in the image that immediately precedes the movement and the one that follows, so we don’t have any experience of two representations being compared, but only one experience of our bodies in motion, which begins just before acting and continues on after initiating movement. (p. 342)
This passage clearly suggests that Prinz does not view sensory imagery as being first-person distinguishable from perception when content is fixed, at least not in the case of bodily movement. Mandik does not address this point.
Finally, it is not clear that the motor theory of control consciousness itself has the resources to explain control consciousness for sensory imagery. Mandik’s motor theory appeals specifically to motor commands, and although there is some evidence to suggest that the motor system plays a role in generating action-related sensory imagery (see Jeannerod, 2006, for a useful review), what we need is evidence that it plays a role in the production of sensory imagery more generally. And if there is no such evidence, then the motor theory is no better off than the sensory imagery theory in this particular respect.
Still, there may be other reasons for preferring a motor theory of control consciousness. For one, the motor theory seems to win out on phenomenological grounds. The sensory imagery theory predicts that each time we perform a controlled bodily movement, we form a conscious sensory image of the sensory consequences of that movement. As mentioned, Prinz argues that we are not aware of these sensory images as distinct from the perception of our bodies in motion because their content matches that of the perception. But in cases where bodily movements are not as anticipated due, perhaps, to an execution error, Prinz’s theory predicts that we would be aware of these sensory images in addition to the sensory consequences of our bodily movements, since they would not match in content. And there is no evidence that this ever happens.1 A motor theory of control consciousness does not face this worrisome consequence.
Further, as it stands, the sensory imagery theory cannot obviously account for control consciousness that accompanies the movement of an anaesthetized, or even a deafferented limb. After all, control consciousness is supposedly generated as the result of a match between the predicted sensory consequences and the actual sensory consequences of a bodily movement. When there are no bodily sensations resulting from a movement, no match is possible, so no control consciousness would be predicted.
To this objection, Prinz (2007) replies that “[w]e feel that we are commanding our bodies to move even though we can not feel the movement, because we experience somatosensory imagery” (p. 344). In other words, the forward model’s output of sensory imagery in these cases is sufficient for control consciousness to arise. But, if so, then we are left with a puzzling picture. According to Prinz, our anticipatory sensory images are, in cases of controlled movement, qualitatively indistinguishable from the sensations that result from the ensuing bodily movements. And yet, the sensations that result from bodily movements do not by themselves give rise to control consciousness, as is evident in the case of passive movement, while anticipatory sensory images by themselves do. What is the relevant difference between them such that this result might be explained? The only plausible answer is that the sensory image that is the output of a forward model is always generated on the basis of a motor command, or a copy thereof, whereas sensations resulting from bodily movement might have no causal link whatsoever to a motor command. But then, especially in light of the phenomenological issues already raised, we now have no reason to suppose that control consciousness is underpinned by sensory imagery, rather than motor commands themselves. The sensory imagery theory of control consciousness, therefore, needs further motivation and refinement.
In sum, while Mandik’s particular reasons for preferring a motor theory to a sensory imagery theory of control consciousness are not convincing, there are some considerations that weigh in favor of his type of view. But the question of which theory is the best one to adopt is far from settled. For his part, Mandik has sparked an interesting debate and offered a rich and insightful discussion of some of the main issues to be tackled in future work.