Constitution and Causation in Pietroski 's
attack on Neuralism in Causing Actions
Because Paul Pietroski’s book, Causing Actions is such a rich and suggestive work, I will limit my remarks on one of the central theses it espouses: that neuralism --the claim that mental events just are neural events -- is false (p. ). In particular, I will show how behind many of the arguments presented in favor of it there is a particular view of (a) whole/part relationships, and (b) causation tout court. Insofar as these assumptions are correct Pietroski is right: neuralism is false. If, however, these assumptions are mistaken, the thesis Pietroski sets up to demolish is in fact something of a red herring.
From the outset, Pietroski warns that he will be not distinguish between the following two claims: (1) that "mental events are identical with certain neural events; and (2) that mental events are constituted by neural events, where `constitute’ is understood so that constituted events are not causes distinct from their constituting events" (p. 6). Choosing to limit "constitute" in this restrictive sense has the following significant consequences: If global neural processes are mere "fusions" of their components -- that is, causally indistinguishable from the micro neural events which so constitute them -- a fortiori, neuralism is true. "In the context of debates about mental causation," Pietroski adds by way of explanation, "this is to grant that if a part p of an event fusion F has an effect e, then F is not a cause of e distinct from p. That is, there is nothing puzzling in the fact that a fusion and some part(s) of the fusion caused a given effect" (ibid).
Are other interpretations of "constitution" possible? Fully aware that his project rests in large part on a specific understanding of that term, Pietroski is not blind to other possible interpretations, He warns that "the mere fact that mental events are constituted by neural events in some way of other does not guarantee that mental events are not causes distinct from neural events (p. 7). If, for example, mental events turned out to be "causes distinct from" the neural events on which they "supervene," then event dualism such as Pietroski advocates would be vindicated, as supervenience does not entail neuralism (p. 9).
"Perhaps there is an intermediately strong notion, [Pietroski suggests] such that constituted events are neither fusions of their constituting events nor causes distinct from their constituting events" (p. 7). But what about the possibility that constituted events might be causally distinct from their constituting events and yet not fusions of them? Pietroski never considers this possibility. And yet, he realizes that "one cannot just assume that mental events bear some metaphysical relation to events that satisfy untendentiously neural predicates. Absence of an account of how constituted events can be causes that are not distinct from their constituting events, the main virtue of neuralism... has been lost. By way of comparison, suppose an event dualist appealed to an intermediately weak notion of being distinct, such that mental causes can be distinct from neural causes without overdetermining any effects of neural causes. One would want to hear more about this notion of distinctness; for therein lies the account of mental causation" (p. 7). It seems clear that throughout this section Pietroski means by "distinct" something akin to spatio-temporally "separate" in the sense that if both were simultaneously causally efficacious (note question-begging manner in which problem is formulated) then the effect would be "overdetermined."
"In short," Pietroski concludes, "historically and theoretically, the following choice has been particularly important: should one adopt the simple account of mental causation, and embrace the consequences of saying that each human mental event is (or at least fails to be distinct from) some biochemical event; or should one eschew these consequences, and defend another conception of mental causation?... One is free to characterize other notions of constitutions, given which constituted events are not causes distinct from their constituting events... But the word `constitution’ is not a talisman that protects against objections to both neuralism and dualism; it labels work that needs doing" (p. 7). Instead of taking on that task, Pietroski assumes a "simple account" of causation as well as on an understanding of "constitution" borrowed from a mechanistic physics that is no longer plausible" (p. 150), according to which wholes are no different from aggregates reducible to the sum (fusion) of their parts.
In part, the source of the problem is the following: whereas Pietroski recognizes the load he places on the concept of "constitution,’ I believe that despite the comment that "therein lies the account of mental causation" he fails to recognize that implicit assumptions also weigh down the concept "distinct" -- and the relation between "constitution" and "distinctiveness." Throughout the book Pietroski appears to take a position that can be summarized in the slogan: "if X is constituted by Y, then X is ipso facto not distinct from Y." When read as part of a pointilistic event ontology such as Pietroski’s, this conclusion is reinforced by his statement that "event dualists will say that mental causes fail to be biochemical causes, in any sense of `be’ that avoids the need for a substantive account of how an event can have mental causes and neural causes" (p. 8). Throughout the book (cf p. 148), he repeatedly questions whether bodily motions can have "neural causes and distinct mental causes," hereby begging the question as to whether mental events can be constituted by neural ones while at the same time, qua mental, operating as distinct causal powers. (I will return to this theme below.)
Pietroski describes Descartes’ conception of mental causation as follows:
S –> N1 -> ... Nk -> M -> T -> Np -> ...Nz -> B
A stimulus S causes neural event N1, which in turn causes a chain of other neural events, until one neural event Nk causes a mental event, which in turn causes a trying, which then causes another neural event Np, which subsequently causes a chain of other neural events that terminate in bodily motion B.
The problem with this account of causation, Pietroski maintains, is that whereas ordinary neural causal processes such as those spanning N1 to Nk, and those spanning Np to Nz, can be conceptualized as a causal chain akin to a line of dominoes, "Causation involving mental events [such as those engaging M and T] is not to be understood in terms of a mechanical model" (p. 150). Whereas domino toppling is determined, he states, mental events are merely influenced (p. 149) by outside events. Although Pietroski recognizes that this assertion requires that he spell out the difference between "determination" and "influence," he does not. Instead he abruptly changes the terminology in which the discussion is framed and begins to speak of the difference between explaining behavior in terms of "impersonal" causes versus explaining actions in terms of "rationalizing" causes.
How are we to understand causation involving mental events? Why do "instances of causation involving mental events seem different from instances of causation?" (P. 150). His reply is similar that given by Socrates when he worried that earlier philosophers made air, ether, and water the only causes. What about an agent’s reasons for not escaping from prison, Socrates asks in the Phaedo. Are not these the true causes of his behavior? Likewise, Pietroski notes that whereas some events can be explained by appeals to "impersonal" causes (ibid), others – of which human actions are paradigmatic – are better explained by reference to the agent’s reasons. Moreover, the idea that reasons can be causes is associated with the intuition that our actions are in some sense free (ibid) – not in the sense of uncaused, but in the sense that because they are caused by my reasons they are mine and within my control. My actions are free because they issue from my reasons, which are changes in me, in my person, not just in my biochemistry. The difference between M and T and any of the neural events Nk and Nz in the above chain, then, is that while M and T can serve as "rationalizing" causes, Nk and Nz cannot. If mental causation operated like the chain of dominoes that Descartes proposed, then "[t]he occurrence of Nk and then Np should be explicable from the impersonal point of view" – but it can’t. Only by postulating a mental event "between these neural events" (p. 151) can intentional actions become explicable. "But there is no reason to think, and ample reason to deny, that the impersonal order exhibits such causal/explanatory gaps" (ibid). Pietroski’s conclusions here would be correct if the impersonal order where a linear, mechanistic order. But it isn’t. If as I believe complex systems allow a reconceptualization of how wholes can be constituted by parts which they in turn affect, then phrasing the dynamics as "Nk and then Np" would constitute a category mistake. If Nk is a global dynamical system and Np one of its components, there is no "and then" in the sense of "distinct" which Pietroski takes for granted throughout the book. Nor is the suggestion that a mental event needs to be postulated between neural ones a plausible way out of the difficulty. As much neurological research has shown in the last few years, brain dynamics is simply not a matter of one atomic event following another in atomistic, linear fashion. But I anticipate.
Neuralists such as Fodor, Pietroski recognizes, would counsel a different alternative: that some neural events are not mere neural events – that "some neural events are tryings and their mental causes" (p. 152). Pietroski dismisses this proposal on the grounds that since there is to date no inductive evidence for cashing out mental events in the language of neuroscience, there is no reason at present to accept neuralism. The two sentences immediately following this facile dismissal of neuralism are particularly revealing of their underlying assumptions concerning the nature of causation: "Certain abstract (metaphysical) considerations about causation can make it difficult to see alternatives [to neuralism]," Pietroski states (pp. 152-153). Moreover, "[n]euralism also makes it hard to see why the domino analogy is at all inadequate" (p. 153): some dominoes might be red-spotted [mental events] and others not, but each toppling would be just like any other (ibid). Pietroski’s response to the latter possibility is once again, Socratic: an agent’s reasons that are the cause of his actions seem to be a toppling unlike any other, [he asserts]. Given neuralism, "it is hard to see how such [biochemical] events can be mine in the sense that my mental events are mine" (p. 155). Again, Pietroski would be correct if the only dynamics available were mechanistic dynamics; if so, it would be hard to see alternatives. Unfortunately, a different dynamics is readily available, but the opportunity to reconsider those "abstract metaphysical considerations about causation" is allowed to slip away.
Consider, therefore, the following picture of mental causation (p. 151), which appears often in the book:
I note at once that Pietroski’s position is not, as may appear from the diagram, that a person houses "two independent but coordinated causal processes – one in the mind, one in the brain... Rather, there are two ways of explaining bodily motions by citing and describing inner causes [one rationalizing and one impersonal]" (p. 152). "[C]ontrary to neuralism, however, M and T cannot be characterized from the N point of view (p. 156). In describing what happens inside persons in a distinctively personal way, one is speaking about events that cannot be described in impersonal terms" (p. 156). The two chains do not represent two independent but parallel causal processes; nor do the two chains "talk about the same causes in different ways" (that would be neuralism). Rather, the two chains are "two ways of explaining bodily motions by citing and describing inner causes" (p. 152). The same, or distinct inner causes. In light of Pietroski’s earlier assumptions concerning "constitution" and "distinctiveness" (i.e., M and T are not "constituted" out of N’s) how are the personal, mental events on the upper chain related to those impersonal neural biochemical events on the lower chain? Does a person have both mental [personal] -- and, distinctly – neural, biochemical [impersonal] causes? If so, as Pietroski recognizes, the problem of overdetermination looms large. If not, Pietroski must explain "how tryings can be causes of the muscle contractions that cause bodily motions" (p. 152). Since Pietroski eschews substance dualism, T cannot cause B in the absence of Nz;. But if T is not constituted by Nz the way neuralists proposed (as a fusion), and yet T is indeed causally efficacious and is distinct from Nz, then what is T and how does it cause? The same considerations, mutatis mutandi, apply to the question of how M is caused
If (1) wholes are reducible to the fusion of their parts, (2) if the only model of causation available for abstract metaphysical reasons is a chain of "topplings," and (3) if mental events are "inner" events distinct from –not merely different descriptions of N1...Nz (and are in some sense inner to the agent or person – to me – not just the body), then, ontologically, what are mental events? Where and when are they? Both persons and their mental events. Pietroski believes, are "spatio-temporal particulars," and mental events are changes that persons undergo and which take place "inside" the person (p. 168). Since, to repeat, Pietroski explicitly rules out substance dualism while insisting that a person is not just (the fusion of) his body, the ball is in his court to provide some account of mental events and personhood.
Unlike merely neural, biochemical changes, Pietroski claims, mental events are global changes in the brain. Since Pietroski has only mechanistic physics to go on, however (whereby wholes are identical to the aggregation of their parts), the distinctiveness of the mental, he suggests, is due to the fact that mental events are more diffuse and vague than just biochemical ones; that is, they extend over a larger spatio-temporal area of the brain than do neural events that embody impersonal phenomena. The two are distinct in that "there is ... no independent reason for thinking that any biochemical event has boundaries that are vague in just this way" (p. 174, emphasis mine). Phrased somewhat differently, mental events like tryings are different from biochemical events insofar as the former have a different "vagueness graph"; their "grain" is different from that of biochemical events" (p. 173). Mental events are vague in the sense that they have a less determinate location than impersonal neural events; the former extend over or span a larger spatio-temporal region than the latter." If each mental event is [therefore] vague in a way that no neural event is – and presumably, no neural event is vaguely constituted by neural events – then each mental event is distinct from every neural event" (p. 177). With these two sentences, Pietroski’s reliance on inappropriate dynamics becomes apparent, which is regrettable, because he’s on the right track: mental events are global properties. And if, indeed, the only dynamics available were Newtonian, Pietroski’s criticisms would be right on target. However, this is not the case. global properties become suddenly tractable when one adopts a different dynamics than does Newtonian mechanics, one that operate in a qualitatively different state space. As Hinton and Shallice (1991) note on a related matter, intuitions that rest on the assumption of a much simpler state space must be discarded when the multidimensional state space of complex dynamical systems is involved.
A different conceptual framework based on complex dynamical systems theory, I submit, is readily available for handling the relation between the global and the local – and has been for some time now yielding empirical neurological results. Formulating questions concerning mental causation in terms of self-organizing complex dynamical systems can dissolve many difficulties concerning the ontology and manner of causation of actions with which Pietroski and earlier action theorists have struggled. The recent concept of a dynamical attractor whose constraints cascade across dimensions is particularly useful in this regard. Part II consists in a brief sketch of this different conceptual framework for interlevel causation.
When under favorable conditions of non-equilibrium open systems which exchange matter and energy with their environment are driven to a critical instability threshold (called a bifurcation point), a random, naturally occurring fluctuation can become amplified, serving as the nucleation that drives the initial regime into a new organization with emergent properties. A qualitatively different, higher level, distributed dynamical (not just structural) organization emerges when uncorrelated components suddenly become correlated into a new, global, dynamical system, whose specific properties cannot be inferred from the sum or fusion of the properties at the earlier stage. All self-organizing complex dynamical systems including "dissipative structures" like Benard cells and the Belousov-Zhabotinsky (BZ) reaction – as well as slime molds, tornadoes and organisms – exhibit homologous, far from equilibrium dynamics.
Research in this new science (and without appealing to any mysterious elan vital) shows that systematic, self-organized wholes (which are not just fusions or aggregates) are created bottom-up from interacting components which the global system then, in turn, controls. As a result of this strange loop, circular relationship -- which Kant noted in his third Critique is a type of causality "unknown to us" -- complex dynamical systems are not mere epiphenomena; they actively exercise causal power over their components. As structures of processes (not pointilistic events), autocatalytic webs and other dissipative structures are distributed patterns of dynamic relations. (Here is the diffuseness Pietroski was searching for.) As such, dissipative structures are not mechanical processes; in autocatalysis no one molecule pushes the others around. The central nervous system, too, has no localized grandmother unit that, in the manner of a Newtonian force, activates and determines the others by bumping into them. Pietroski is correct: domino-like causation is inapplicable here. But that doesn’t mean causation tout court is. Rather, the dynamics of self-organized complex systems embody mereological causality in the sense that the global system is both constituted by its components and is also distinct from them the higher level exhibits emergent properties irreducible to those found at the lower levels. And it is in virtue of the emergent properties of the global level that the system as a whole is causally effective (see Hinton and Shallice’s research on the emergence of a semantic attractor in recurrent neural networks).
How to conceptualized the distributed and interlevel causation exhibited by these dynamic processes? Complex dynamical systems are relational and property, not concrete systems. It is therefore a category mistake to think of them as exercising causal power by impressing a force on their components. And yet, far from being inert epiphenomena, as the "simple account" of causation of traditional physics – as well as action theory – would have them, complex systems, as global, distributed wholes, do indeed affect their components top-down. This mereological causality can be conceptualized from the point of view of complex dynamical systems theory as the operation of contextual constraints (Juarrero, 1999).
From the combined effects of context-free and first-order contextual constraints, dynamical structures and patterns at a higher level of complexity self-organize (Juarrero 1999, 142). Through the imposition of the second-order contextual constraints that are embodied in the self-organized global level, a complex adaptive system’s global level then limits its components’ degrees of freedom such that its overall dynamical organization (order, pattern) is maintained and enhanced. A complex system’s contextual constraints effect this top-down control by altering the prior probability of the behavior of the components. The higher level constrains the lower, for example, by regulating the rate at which its component processes occur, that is, by altering their natural frequency. Given its embedded ness in a systematic whole, that is, the prior likelihood that a lower level process will be activated automatically alters. That conditional probability is what embeddedness in a systematic whole (not an aggregate or a fusion) is (ibid).
Unlike Pietroski’s smeared out but still traditionally conceptualized mental events, complex adaptive systems are best characterized as multidimensional mathematical attractors describing characteristic behaviors or states that tend to draw the system towards themselves. I speculate that in the brain, coherent wave patterns of a far greater dimensionality than those of the individual neural events that constitute them, and embodying the emergent properties of consciousness and meaning (Pietroski’s personal, rationalizing features) appear as neurons entrain and self-organize into global structures. These higher levels of dynamical organization (not merely spatio-temporal organization, as Pietroski suggests), impose second-order, context-sensitive constraints on their components (Pietroski’s impersonal, merely neuronal features). They do so by carving out strange attractors that reset the natural intrinsic firing frequency of the individual neurons, including those controlling motor processes. In so doing the global level entrains and constrains the lower levels to the former attractors’ dynamics. And that is how all interlevel, top-down causes cause, including conscious states such as intentions and tryings.
Although this proposal is speculative, evidence is plentiful and increasing that organisms are self-organizing dissipative structures (Collier 1986; Brooks and Wiley,1988) that exhibit dynamics homologous to the Benard cells of thermodynamics and to the BZ chemical reaction. IN particular as it pertains to the mind-brain relationship, brain research into the role of non-linear, complex dynamics in the appearance of emotions, intentions, and even consciousness has come a long way since the time during which the so-called Identity Theory and classical, rule-based artificial intelligence were in vogue. In particular, the neuroscientific data summarized in book form in Edelman and Tononi (2000) on the role of non-linear dynamics in the neural basis of consciousness; in Freeman (1999) on the role of dynamical attractors in intentional action; and in Damasio (2000) on the role of dynamics in emotion, suggest that the proper physics and the proper dynamics will provide renewed insights into the way wholes are "constituted" from interacting parts, as well as novel "abstract, metaphysical considerations" concerning interlevel causation. In particular, I believe neurological research informed by the conceptual framework of the theory of self-organizing systems will yield fruitful solutions to the problem of intentional behavior – in particular, into the way actions are caused by personal, rationalizing intentions.
Damasio, A. 1999. The Feeling of What Happens. New York: Harcourt Brace.
Edelman, G.M and G. Tononi. 2000. A Universe of Consciousness. New York: Basic Books.
Freeman, W. J. 1999. How Brains Make Up their Minds. London: Weidenfeld and Nicholson.
Hinton, G. and T. Shallice. 1991. Lesioning an attractor network: investigations of acquired dyslexia. Psychological Review 98 (January): 74-95.
Juarrero, A. 1999. Dynamics in Action: Intentional Behavior as a Complex System. Cambridge, MA: MIT Press.
Pietroski, P. 2000, Causing Actions, Oxford: Oxford University Press.