In Defence of the Anscombe Thesis

The Anscombe thesis can be expressed clearly as the claim that that if a person As by Bing, then her act of Aing is her act of Bing (Davidson 1963, p.14–15). It will be argued here that this thesis is true when formulated along the lines Anscombe expressed. To defend this thesis, first the outline of the argument will be presented, followed by responses to two critiques held against this view: the nature of a “by” relation, and the temporal nature of action. Through responding to these critiques, a more robust defence of the Anscombe thesis will be shown to hold.

Anscombe’s formulation of the identity of certain actions stems from her theory of intentional action. In her view, intentional action is understood by their being a response to the question of “why?” one is doing what one is doing. The question of “why” regarding intentions is the search for reasons for actions, with reasons being distinct from causes (Speaks 2004, p.4–5). Reasons act as evidence for an actors intentions, while causes are logically connected to the preceding event. In Anscombe’s metric, this explains the means by which different descriptions can apply to the same action. If one were to ask someone what they were doing, and they responded that they did not that was what they were doing, what has occurred is the framing of that action under a description foreign to the actor. It is impossible for someone to be acting without knowing what they are doing, but it is possible to describe what one is doing differently (Speaks 2004, p.7). With this, one can evaluate the descriptions with which the same action can be described.

Anscombe gave the example of a man moving his arm to operate a pump which is replenishing water to poison a house of Nazis. In this example a means-ends chain is formed through the question of “why”, which resembles: “‘Why are you moving your arm up and down’ ‘To operate the pump’ and he is operating the pump. … ‘Why are you replenishing the water-supply’ ‘To poison the inhabitants, and he is poisoning the inhabitants’” (Ancombe 1957, §23). It is clear from this that the question of “why” being adequately answered makes for all parts of this chain to be intentional action currently being done. It is then noted that a break occurs, for if one were to ask the man why he is poisoning the inhabitants, which in this parable is about resistance to the Nazis, his further responses, such as saving the Jews or to bring about a new high command, are not things that the man is currently doing. For the poisoning of the inhabitants does not in normal circumstances entail any of these things the way that replenishing a home with poisoned water entails the poisoning of its inhabitants, for in normal circumstances, one expects individual’s to drink water from their own home’s water supply.

This example contains four parts: moving the arm up and down, operating the pump, replenishing the water supply, and poisoning the inhabitants, however the relationship between these things remains unaddressed. If the chain holds; however, “The answer we imagined to the question ‘Why?’ brings it out that the four descriptions form a series, A-B-C-D, in which each description is…dependent on the previous one, though independent of the following one” (1957 §26). All actions are therefore dependent on the movement of the arm, and the actor contributes nothing more for the chain to exist. Since one operates a pump by moving one’s arm up and down, and there is not distinct action other than moving the arm, operating the pump in that circumstance is one and the same action as moving one’s arm up and down. This similarly applies to the relationship between replenishing the water-supply by operating the pump, and poisoning the inhabitants by replenishing the water supply. At no point does the individual act any distinctly from the motion as one’s arm, but the relation between that motion and external circumstances simply influences different descriptions. The ordering of this is crucial, however, as a means-ends chain is formed for this one action-sequence, with each description including in it a wider set of external circumstances for the same action. From this the compelling case is put forward that there is only one action occurring, but with four descriptions. If asked what one is doing, one can respond with any of the four claims: moving my arm up and down, operating the pump, replenishing the water supply, or poisoning the inhabitant; all with equal validity, with the only difference between these descriptions being the scope they express of the individual action relating to their external, real-world effects.

The argument presented above provides the foundational claims for the Anscombe thesis, and due to the brevity of her writing style, expresses the whole of her discussion on the topic. As such, responding to the challenges against this thesis provides the best means to making the argument more robust. One of the challenges to this thesis is the use of the term “by” and it’s meaning. The refutation on this grounds rests on the claim that the word “by” is asymmetric and extensional, which makes its use in identity claims unfounded (Goldman 1971, p.764). By asymmetric what is meant is that if the claim X by Y is true, Y by X cannot be true, and by extenstionality what is meant that if X by Y is true, and X is Z, then Z by Y is also true (Ginet 1990, p.53). These two points together are merged with the Anscombe thesis to prove its absurdity with the following chain:

  1. X BY Y
  2. X IS Y (Anscombe Thesis)
  3. Y BY Y (Extensionality and 1 and 2)
  4. Y BY X (Extensionality and 2 and 3)
  5. Not (Y BY X). (Asymmetry of 1 and 4)

This argument, which apparently leads to contradiction by the use of the word “by”, however, can be easily shown to not apply in real-world situations. Anscombe uses the example that since “the U.S. President is the U.S. Commander-in-Chief by being President, while the Commander-in-Chief is not the President by being Commander-in-Chief, the President and the Commander-in-Chief can’t be the same man” (1979, p.224). That this statement is implied by the argument above, but that the U.S. President is the U.S Commander-in-Chief, and that he is Commander-in-Chief by being the President, means that the problem lies with the argument, not with the Anscombe thesis. The problematic occurrence then is regarding statements (3) and (4), both of which do not stand up to scrutiny. Since these steps rest on the presence of extensionality in the word “by”, some issue with that clause must be addressed. When “by” is use in these contexts, one could substitute the term “consisted in” for “by”, such that operating the pump by moving one’s arm, is substantially the same as saying that the operation of the pump consisted in the movements of one’s arm (Ginet 1990, p.54). The argument from the Anscombe thesis is then that something is whatever it is wholly constituted by, as in the U.S. Commander-in-Chief is the U.S President, since being Commander-In-Chief consists wholly in being President. This lack of the applicability for the extensionality claim is stressed by the ungrammatical nature of many of the inverse statements, which serves not as a refutation to the identity, but rather the misuse of the word “by” (Hornsby 1980, p.7). As a result, the formulation of “by” above does not hold.

Schneider, admits this failure in extensionality, but brings about a corollary, that the asymmetry of “by” is a result of the word’s explanatory semantic role, which is incompatible with identity (2008, p.663). The role of “by” in his view is that when one does X by Y, the Y is the explanation of how X was done. Following this, the explanation of X given can be more or less complete, and if so, they cannot constitute identity as they will be partial. This, however, neglects the nuance of the Anscombe thesis in its formulation, which as put forward requires that if one X’s by Y’ing, the whole of the X’ing must occur in the Y’ing for the X’ing to be the Y’ing. If the partial explanation left out features that were additional constitutive means of the X’ing, then the argument does not satisfy the Anscombe thesis to begin with, however, if the difference is simply a matter of descriptiveness, then it presents no issue. Schneider’s example of “NN estimated the number of persons by first counting the number of names in a column, and then multiplying the number of names in a column by the number of columns” being distinct from “NN estimated the number of persons by multiplying the number of names in a column by the number of columns” that the identity of the Anscombe thesis would give, belongs to the latter category (Schneider 2008, p.665–666). That the estimation is the multiplication does not give a partial explanation that changes the validity of the claim, since the counting the number of names in a column is an extension of the multiplication, for the act of multiplication entails knowing the relevant numbers. This refutation does not provide adequate grounds for assuming the lack of identity, only that the Anscombe thesis allows for different descriptions of the same action, some varying in detail, but not altering the content of what was done. The Schneider corollary, as a result, fails to improve upon the argument from “by”, and rather than refuting the Anscombe thesis, serves to highlight its strength.

Another challenge to the Anscombe thesis is concerned with the temporal nature of actions, such that when one acts in the means-end sequence that forms identity, one does certain things prior to the occurrence of other things, meaning their identity is not plausible. The act of killing is an example of such a scenario, for if one were to shoot someone, and they died 12 hours later, the identification of shooting with killing would imply that the individual died after they had been killed. Since killing, however, is associated with a death, the period in which the victim has not yet died by the Anscombe thesis suggests “that a certain act X is identical with a certain act Y — where X has occurred, but Y has not” (Thomson 1971, p.117). This highlights the temporal argument against the identity of these actions, that two events occurring at distinct times could not constitute the same action, for once a shooting has occurred, the action has been completed, though the killing, which is to be equated, has not even occurred. The argument continues, that the instance upon which events occur are distinct even with reflection, that if one turned on an electric heater to melt chocolate, one flips the switch on at time T, but the chocolate melts at time T’, meaning that even though the melting were over a stretch of time, the action that instigated it, flipping the switch, occurred at a distinct time from the completion of the melting (Thomson 1971, p.121). The identification then of killing with shooting, or of melting with turning on the heater, would lead to seemingly absurd statements such as he died hours after he was killed, or the chocolate melted a few minutes after it was melted. As a result, the identity the Anscombe thesis claims between these actions is false, making the Anscombe thesis false.

This argument fails to refute the Anscombe thesis effectively on the grounds that it fails to adequately describe the nature of action, and by extension, misrepresents the means-end reasoning of the Anscombe thesis. Defenders of the temporal argument admit the difficulty in identifying the time at which an action has occurred, for example, is the time of shooting to be identified with the time of pulling the trigger or the time of the bullet’s lodging in the victim. An issue then arises with the reference to when a death occurs, for while it is uncontroversial to say that a killing cannot occur unless there is a death, this does not entail that a killing does not occur until there is a death. This line of reasoning implies “the act which was to become, at the time of [Y’s] death, the killing of [Y] was performed wholly before it became that; it was performed when [X] shot [Y]; indeed it was [X’s] shooting of [Y] (Lombard 1989, p.140). This stems from the fact that if a shooting was to lead to a death, insofar as the death occurs at some point, no action outside of the shooting could be responsible for the death, meaning the shooting was the killing. As such, the act of a killing is distinct from the time of death; for the agent who killed the individual is not necessarily present at the time of the death, but is nonetheless responsible for the killing. According to Davidson in his example of a Queen poisoning her husband; “Is it not absurd to suppose that, after the Queen has moved her hand in such a way as to cause the King’s death, any deed remains for her to do or complete? She has done her work; it only remains for the poison to do its” (Davidson 1971, p. 57–58). This line of reasoning suggests that to assume the killing was a distinct act from the poisoning is to assume that the agent acted in killing at a point at which the agent was no longer acting at all, which is an unnecessary inflation of the ontology of action.

Returning to the shooting example, that no bodily movement occurs beyond pulling the trigger, and that the shooting of the victim and their subsequent killing are wholly constituted by the pulling of the trigger means that they cannot constitute distinct acts simply based on time of their occurrence. For a shooting occurs over a stretch from the pulling of the trigger to the lodging of the bullet, not at a distinct time, and the killing occurs over a stretch including the shooting up until the death, again not a distinct time. These separate descriptions: killing, shooting, pulling the trigger; are all related to the same action, namely A kills B by shooting B by pulling the trigger of his gun. This becomes evident by the means-end sequencing of the Anscombe thesis. That the action-sequence as a whole consists in only one physical movement, pulling the trigger, while the intention of the action-sequence, to kill B, can only be realised over the course of the whole killing, means that the differences between the varying descriptions regards only the scope, not separate actions. All descriptions are equally valid in describing the action, if the action-sequence has occurred, all are equally true, with no description being the action (Anscombe 1979, p.220). From this it is clear that the Anscombe thesis resists the temporal challenge, as action-sequences do not have distinct times, the attaching of times is itself arbitrary, and the desire to isolate times futile. To allow the sequence to entail the times of all constituent acts, therefore, provides it with greater explanatory power.

The challenges to the Anscombe thesis, it has been shown, serve only to highlight the strength of its formulation. That refutations based on semantic or temporal grounds fail to address the nature of the thesis regarding the identity of actions, adds to the resilience of the thesis as a whole. Beyond this, the intuitiveness of not attaching multiple actions, be they a separate killing, shooting, and pulling the trigger of a gun; or a separate operation of a pump from the replenishing of a water supply, has great intuitive merit. On these grounds, that the formulation of the Anscombe thesis simplifies the understanding of actions, and deflates the ontology of actions associated with bodily movements, while simultaneously resisting challenge, the claim that the Anscombe thesis is true has been readily defended.


Anscombe, G. (1957). Intention. 1st ed. Cambridge, Mass.: Harvard University Press.

Anscombe, G. (1979). Under a Description. Noûs, [online] 13(2), pp.219–233. Available at:

Davidson, D. (1963). Actions, Reasons, and Causes. Essays on Actions and Events, pp.3–20.

Davidson, D. (1971). Agency. Essays on Actions and Events, pp.43–62.

Ginet, C. (1990). On Action. 1st ed. Cambridge: Cambridge University Press.

Goldman, A. (1971). The Individuation of Action. The Journal of Philosophy, 68(21), pp.761–774.

Hornsby, J. (1980). Actions. 1st ed. London [etc.]: Routledge & Kegan Paul.

Lombard, L. (1989). ‘Unless’, ‘Until’, and the Time of a Killing. Pacific Philosophical Quarterly, [online] 70(2), pp.135–154. Available at:

Schneider, B. (2008). ‘By’: A Refutation of the Anscombe Thesis. Linguistics and Philosophy, [online] 31(6), pp.649–669. Available at:

Speaks, J. (2004). A guide to Anscombe’s Intention, §§1–31. [online] Notre Dame, Indiana: University of Notre Dame. Available at:

Thomson, J. (1971). The Time of a Killing. Journal of Philosophy, [online] 68(5), pp.115–132. Available at:



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Ryan Khurana

Executive Director of the Institute for Advancing Prosperity. Axiology, Horology, and Technology.