Natural Agency, by John Bishop. Cambridge: Cambridge University Press, 

1990. Pp. xi + 211. £25. 

Natural Agency is a very good book. It is tightly constructed and argued, insightful, and lucidly written. In addition to summarising and contributing to a number of fairly technical and difficult issues, Bishop is always keen to show the reader why the discussions matter in terms of the great and perennial debates in philosophy. The book is a model of good, indeed, first-class, analytic philosophy. 

Bishop argues that many of the debates about the compatibility of moral 
responsibility, or agency, and a naturalistic view of the world, eventually turn on 
the question of the plausibility of a Causal Theory of Action (CTA). Thus, the 
central topic of the book is the causal theory of action, the idea (roughly) that a 
piece of behaviour is an (intentional) action iff that behaviour is caused in the 
right sort of way by appropriate mental states or events. (Bishop believes that 
every mental event is token identical with some brain event, but does not seek to 
add to this debate in the book.) If a CTA can be defended, then Bishop thinks that 
he will have rebutted one very important source of scepticism about the place of 
agency in the natural world. And since moral responsibility requires agency, such 
a rebuttal would also serve as an important step in the justification of the belief 
that moral agents can be within the natural order of things. The nub of Bishop's 
rebuttal is that the concept of action can be cashed out in CTA terms that involve 
only an event-causal ontology; no irreducible reference to action, or agent-causation, is required. 

However, there are well-known difficulties which any successful CTA must 
overcome. Two of these are the problem of deviant causal chains and the problem 
of akratic or weak-willed actions. Let us take the deviant causal chain problem 
first. Suppose that my intention to move my finger causes me to become nervous, 
or distracted, or whatever, and that my finger's moving is caused by my nervousness or state of distraction. If so, then a behavioural outcome being brought about 
by an agent's mental states is insufficient to yield the idea of intentional action. 

Bishop thinks that he can state a CTA in a form that will handle all the deviant 
chain counterexamples that have been proposed. His subtle and perceptive discussion of this point, extending over two chapters, takes one carefully through the 
relevant literature. Bishop argues that the cases which involve nonbasic action 
can be handled by adding restrictions to a CTA which require a match of sorts 
between the causal chain leading to the behaviour and the agent's plan of action, 
as represented in his practical reasoning. Can a CTA handle deviant causal chains 
involving basic actions? After a long and persuasive discussion of the many variants of the sensitivity strategy (Lewis, Peacocke, and Davies), according to which behavioural outcomes must be sensitive to alterations in mental states in order to count as actions, Bishop concentrates on cases of heteromesy, that is, cases in which the causal chain runs through an additional agent. 

Consider two cases. In case A, one of my natural but diseased nerves is 
replaced by an artificial one, which snaps. A friendly neurophysiologist holds 
them together until they can be resoldered in a more permanent way, thereby 
allowing me to act on my intentions. Bishop judges that case A of heteromesy does not produce deviance. The causal chain leading from mental state to behaviour is abnormal and heteromesic, but not deviant. I still intentionally act, I am exercising control, albeit with help. 

Now consider case B: the direct causal route from my mental states to behavioural outcomes involving parts of my body is blocked, but I am so wired up that a neurophysiologist can glean my intentions from his control panel, and initiate the very same behavioural outcomes that I had intended in the first place Bishop judges that case B of heteromesy does produce a deviant causal chain, so that I do not act intentionally. Not I, but he, is in control. Notice that in both A and B the behavioural outcomes are sensitive to alterations in mental states, so sensitivity alone cannot distinguish the two cases. 

The trick now is to distinguish the deviant from non-deviant cases of heteromesy in event-causal terms for cases of basic action, and it is on this question that Bishop is willing to have the fate of the project of producing an adequate CTA decided. Briefly, his solution is this: the sensitivity displayed by such causal chains which lead from mental states to outcome behaviour does not always involve a feedback mechanism to central mental processes, but if it does, the causal chain is non-deviant and results in intentional action iff the feedback goes to the mental processes of the agent rather than to that of the interfering neurophysiologist (p. 172). 

This proposal seems to me to be open to further counterexample. Return to the non-deviant case, A. Bishop's friendly neurophysiologist has nothing to do but to hold the two ends of the artificial nerve together, while he waits for the solderers to arrive. But suppose a slightly different case, A*. In A*, there is an established correlation between what goes on inside of me (or how I move) and how the friendly neurophysiologist must continue to hold the nerves together. As I move, the friendly neurophysiologist must take account of my movements and adjust the level at which he must hold the nerve ends. The friendly neurophysiologist receives "feedback information about orientation and muscular states" in A*, just as he did in B, perhaps by simple perception (p. 170). 

If A is heteromesial but non-deviant, so is A *, although Bishop's suggestion would imply otherwise. Suppose the solderers had soldered in a device Joining the nerve ends that incorporated a feedback mechanism. If so, the hitherto nondeviant causal chain in A would surely not have become deviant just because a feedback mechanism had been implanted. In case A*, the feedback signals go to the friendly neurophysiologist as well as to (or instead of) me. He acts merely as an external variant on the implanted feedback mechanism. So I cannot see that Bishop has yet provided necessary and sufficient conditions for action in event causal terms, although his must be the best and most sophisticated attempt at so doing in the literature. 

The second difficulty for a CTA is akrasia, or incontinence. Bishop claims that 
his account makes sense of Davidson's more obscure treatment of this same problem. Bishop's idea is that, in akrasia, the agent forms an intention to do an act 
other then the act he believes to be best, all things considered. Thus, the act that 
is done incontinently is rational relative to the intention, but not rational relative 
to the beliefs and desires of the agent. But the act is an intentional act, since it is 
produced by the right intention. 

Davidson's original puzzle was to understand how, if (1) agents always most 
want to do what they believe is best, and (2) agents always act to get what they 
most want (assuming some other conditions not relevant here), incontinent action 
was possible. On Bishop's view, incontinent actions are possible because there 
are (as we might call them) incontinent intentions. Even if Bishop solves the problem of incontinence for action, won't Davidson's problem arise again for intentions? If (2) was true, so would be (2*): Agents always intend to get what they 
most want. But if (1) and (2*) are both true, how are incontinent intentions possible? And if one is inclined merely to deny the truth of (2*), which I think Bishop must, could we not have saved ourselves some trouble and denied (2) initially? 

Finally, Bishop spends some time in trying to convince the reader that intending is not itself a mental act, and hence including intendings in an event-causal analysis of action will not introduce unwanted circularity (pp. 180-2). What relation is there between (actively) intending, and having an intention? A plausible view, that Bishop would have to dismiss, is that the former is either primary in some sense, or in any case is irreducible to the latter. 



There is no doubt that cognitive and appetitive states such as belief and desire 
are states which can arise passively; whether intentions can will crucially depend 
on what sort of animal conative states are. Some theorists (F. Schick) regard intentions as comings-to-want, others (Carl Ginet) as comings-to-believe, and still others (e.g., Myles Brand) as composed of both (I conflate for these comparative purposes intentions, decisions, and choices). I think that there are serious objections to theories of all three types. If Bishop addresses himself further to questions of agency, as I certainly hope he will, I also hope that he will discuss the possibility of a non-agent causal analysis of intending itself, and how intendings relate to other apparently "active" mental items such as decisions and choices.