2026/07/01 by Denis Buehler
Neuroscience · Psychology · #Embodied and Extended Cognition #Phenomenology and Existential Philosophy #Philosophy and Theoretical Science
paper · doi:10.1215/00318108-12455553
In his book Movements of the Mind, Wayne Wu presents a theory of agency. While the book focuses on mental action, the theory is meant to cover all forms of agency. At the center of this theory is attention and its interaction with intention.Wu’s book is a significant contribution to current action theory. While the book can be challenging to read, it raises many hard questions and is chock-full of interesting, provocative theses. Wu’s steadfast endorsement of an empirically based approach deserves praise. And having one of the prominent figures in current action theory draw attention to the connection between working memory, executive control, and human agency does the discipline a major service. The book is bound to engender much discussion.Part 1 contains chapters 1 and 2, which lay out Wu’s accounts of attention and action, respectively. Part 2, containing chapters 3 and 4, presents a theory of intention-in-action. Chapters 5–7 (pt. 3) apply the account to bias and skill, deduction, and introspection. The book’s first four chapters present the “core of [Wu’s] theory of agency” (149). I focus on these in what follows.In chapters 1 and 2, Wu presents his account of attention and action. Attention is selection-for-action, in Wu’s technical terminology, a stimulus’s being coupled to (and guiding) a response. Action, in turn, is a response coupled to a stimulus. The two notions are interdefined.To use Wu’s illustration, if you kick one of two balls and can do so with one of two legs, you must select, and thereby couple, one of the balls to one available response—that is, kicking it with one of two legs.Is all attention/action like that? Isn’t it possible for us to act inattentively? When I walk across the hallway while speaking on the phone, I do not seem to attend to my walking. And can’t we attend without necessarily acting on what we attend to? If my attention is captured by a bright light, might I not attend to the light before going back to my reading, without otherwise responding? Commentators have pointed to examples like these, taken from everyday life, and cognitive science (Watzl 2011; Jennings and Nanay 2016; Buehler 2018b; Lange 2023). Wu suggests that such cases draw on unreliable intuitions. In his technical sense, even these cases feature attention and action. And cognitive science, he claims, supports this view as against intuition. We should go with science in our theorizing.What, then, do his technical terms mean? And how does cognitive science support his claims? Wu seems to have settled on this: a stimulus is coupled to a response just in case they both occur at the individual level, they both are processed nondeviantly, and the coupling might not have occurred. Responses are exercises of action-capacities. That may seem to be tight circle, if the aim is to provide an account of action. But Wu’s action-capacities are not the familiar ones. They include, for instance, the “sharpening of one’s visual experience” (179) and “improved cognitive access” (76). What, then, are these capacities? The notion remains unspecified. This allows Wu to handle challenging cases: paradigm passive episodes like captured attention, for example, turn into “passive actions” (79). Purported cases of inattentive action become attentive, since some action-related information-processing usually occurs. And yet, without substantive constraints on these central notions, one may wonder how they relate to the target-explananda.Wu’s theory, in any case, is revisionary. Why should we adopt it? Wu’s central argument revolves around the claim that cognitive science is committed to:Wu notes that cognitive science studies attention through paradigms like, for example, visual search, which operationalize attention as selection of a search-target for a response. He contends that this implies that all of cognitive science is committed to (ESC). But, for one, these paradigms are carefully designed to investigate attention. From the fact that they study attention, it does not follow that all processing of stimuli in a task-setting involves attention. Many experimental tasks do not investigate attention. For another, even if (ESC) were true, moving from it to a claim about all attention and action outside experimental settings requires argument.Most importantly, however, cognitive science does not appear to be committed to (ESC). There are several avenues of work that explore the possibility of processing information for a task—coupling it to a response—without attending to it. Thus scientists ask whether scene gist, certain object-categories, or statistical learning require attention. They design sophisticated experimental paradigms—the attentional blink, multiple-object-tracking, and others—to probe whether these stimuli can be reported, despite attention’s being allocated elsewhere. While establishing that all attention is absent from some task is practically extremely difficult, scientists treat claims like (ESC) as an empirically open matter (Buehler 2022b). Wu nowhere satisfyingly engages with these points (see also Wu 2019, 2025). But if Wu’s theory conflicts with both common sense and cognitive science, some other, powerful form of support is needed to lay the foundation for the rest of the book.Part 2 discusses the connection between working memory, executive function, and action. Theorists have previously explored this connection (Carruthers 2015; Buehler 2014, 2022a, 2023; Christensen et al. 2016; Watzl 2017), and Wu draws on this work. The precise relationship with that work could be more clearly articulated.Wu’s own take on the connection consists in a theory of intention-in-action as practical memory, presented in chapter 3. His central claim is that there is a “functional equivalence between executive control in working memory and intention[-in-action] (94).” From this he appears to infer that intention ‘is’ practical memory (190).The exact nature of the target claim is not fully specified. Wu writes that intention-in-action is working memory (109); that working memory constitutes (3, 46); underwrites the dynamics of (104); parallels (103); is the basis of (98); realizes, or implements intention-in-action (104, 106). These ideas would benefit from further clarification. Standard theorizing in cognitive science describes occurrent intentions as held in working memory. It is unclear how to square this with some of the claims that Wu seems to make.Be that as it may, Wu’s argument apparently proceeds from the mentioned claim of a functional equivalence. To establish such a functional equivalence, one would have to provide a characterization of the purportedly equivalent items, and then show that they indeed are equivalent.What is the characterization of intention-in-action that Wu is working with? Maybe he takes the observation that intention “sets” (roughly, directs) attention to identify the functional role of intention? If so, then his characterization is nonstandard. Work on intention’s functional role does not emphasize the setting of attention but points to other features as central to intention’s functional role (Searle 1983; Bratman 1987; Mele 1992; Pacherie 2006, 2008; Fridland 2019; Mylopoulos and Pacherie 2017, 2019; Shepherd 2021; Grünbaum and Kyllingsbaek 2020). Wu’s argument here would benefit from more engagement with extant conceptions of intention(-in-action).Regarding executive control and working memory, Wu largely relies on an article by Alan Baddeley (1996). Wu reports as functions of the central executive the “coordinating performance during dual task cases, coordinating search strategies during random generation of words from long-term memory, controlling attention to one stimulus and suppressing distraction from another, manipulating and maintaining items in long-term memory as needed in working memory tasks (e.g. span tasks)” (102). Wu emphasizes that “one function consistently tied to the central executive is that it regulates attention” (102).However, if all the cited functions constitute the executive’s functional role, then it is unclear how the functional equivalence with intention is established: Standard characterizations of intention’s functional role do not mention them. And if only the regulation of attention is relevant, then an argument is needed for focusing on this function over others, in characterizing the executive’s functional role.One might have also hoped for some discussion of why Baddeley’s, and not one of the many other characterizations of executive control, is the right one. Research on executive function, working memory, and cognitive control is vast and complicated. It has rapidly grown and developed since Baddeley wrote thirty years ago (Buehler 2018a, 2025). More engagement with this literature would have been welcome.Overall, one feels that the chapter’s central claim about functional equivalence would have benefited from more discussion and support.Chapter 4 builds on this link between intention and working memory. In it, Wu claims that the agent’s intending—her acting with an intention —itself is “an action of practical reasoning as she acts” (125). More specifically, “intending-in-action is constituted by fine-tuning of practical memory in practical reasoning” (129).Wu’s example is that of eating cake. As I intend to eat cake and see this cake, I fine-tune my intention by forming the further intention to eat this cake. The updating of my intention through working memory-updating in light of incoming visual information constitutes the fine-tuning. But “fine-tuning in turn is basic practical reasoning” (139). “Updating of memory [is] a type of thinking” (133). Thus “the dynamics of working memory reveal the dynamics of intention in practical reasoning” (135).Dynamic aspects of intention-in-action have been discussed in prior work, notably by Elisabeth Pacherie and Myrto Mylopoulos (cf. their dynamic hierarchical theory: Pacherie 2006, 2008; Mylopoulos and Pacherie 2017, 2019). These authors describe sources of dynamics and fine-tuning beyond working memory, such as motor processing. If they are right, how can working memory dynamics constitute fine-tuning? Wu does not discuss this work. Does he reject this idea? Or is his claim rather that working memory-updating sometimes fine-tunes intention? If the latter, how is the proposal different from those available in the literature?Maybe the answer lies in Wu’s broader claim that fine-tuning of intention is a form of practical reasoning. That plausibly is often the case. And attention often plays an important role in such reasoning (Grünbaum 2013). Yet Wu apparently aims for a stronger claim, on which all fine-tuning, certainly all working memory-updating, constitutes practical reasoning. This claim would set Wu’s account apart from others in the literature. But working memory-updating through, for example, encoding perceptual information about the cake does not always appear to be practical reasoning. Philosophical accounts of reasoning tend to impose conditions—that reasoning must be propositional and rule-guided, must involve a taking, or some such—that working memory-updating does not always meet. What recommends the stronger claim?Where, finally, does all this leave attention? The emphasis on working memory and executive control may suggest that we here find the locus of agential control. For Wu, this would be too quick. On his view, “attention is everywhere” (80). Working memory encoding, maintenance, updating, output-gating, all are episodes of selection, hence attention, hence action. This puts attention right back at the heart of the account. Whether or not we find this outcome plausible, it raises the question of whether the notion of attention at play here is too thin to do much explanatory work. The explaining rather appears to be done by appeal to working memory and executive control, as I’ve argued elsewhere (Buehler 2014, 2022a, 2023).Thanks for their feedback to Thor Gruenbaum, Myrto Mylopoulos, Elisabeth Pacherie, and Josh Shepherd.