² The main contention of this piece – that actions are primitive and not reducible to other psychological and bodily phenomenon – is also argued for in Chapter 8 of my Self-Knowing Agents. (L. O'Brien, Self-Knowing Agents, Oxford: OUP.) Douglas Lavin in his ‘Action as a form of temporal unity: on Anscombe's intention’ (Canadian Journal of Philosophy, 54) also sets out a non-reductionist view.

³ I use the – admittedly clumsy – expression ‘self-change’ deliberately so as to avoid locutions like ‘I change my self’ because the latter brings connotations of my taking myself as an object to act on, when what actually happens is that I take myself to be that I act with. The main point is that when I act, I change. I can only change the non-me world by changing. So whenever I change any thing in action, I change.

⁴ J. Bennett (ed.), Objections to the Meditations and Descartes’ Replies: ‘Fifth Objections (Gassendi) and Descartes’ replies: Objections to Second Meditation; Objection 1’. www.earlymoderntexts.com/pdfs/descartes1642/pdf, page 86.

⁵ For discussion of this see O'Brien, , ‘Ambulo Ergo Sum’, Royal Institute of Philosophy Supplement 76 (2015)CrossRefGoogle Scholar.

⁶ Evans, G., The Varieties of Reference (Oxford: Clarendon Press, 1982), 207Google Scholar.

⁷ Wittgenstein, L., Philosophical Investigations (1953), §621 Google Scholar.

⁸ Boyle, M. and Lavin, D., ‘Goodness and Desire’ in Tenenbaum, S., (ed.), Desire, Practical Reason, and the Good (Oxford: Oxford University Press, 2010), 170Google Scholar.

⁹ Davidson, D., Essays on Actions and Events (Oxford: Clarendon Press, 1980), 79Google Scholar.

¹⁰ Williamson in fact intends something a bit narrower, construing primeness as non-factorisability in to specifically a conjunction of a purely internal condition and a purely external condition. Thanks to Aidan McGlynn for reminding me of this.

¹¹ O'Brien, Self-Knowing Agents, 137.

¹² Williamson., T. Knowledge and its Limits (Oxford: OUP, 2000)Google ScholarPubMed, 1 and 6–8.

¹³ Williamson, T. in Carter, J.A., Gordon, E. and Jarvis, B. (eds), Knowledge-First (Oxford: OUP, forthcoming). See 22–23 of online draft: http://www.philosophy.ox.ac.uk/__data/assets/pdf_file/0005/35834/KfirstCarter.pdf

¹⁴ Perhaps, we are wrong to think that the kinds of account being offered – of knowledge, or action – are supposed to be applied in this way to the particular case? Perhaps they are accounts only of the general kinds knowledge and action – not directly applicable to particular knowings and actions? (Thanks to Jen Hornsby for this question.) If that is so then the question we started with cannot be understood to engage these forms of non-reductionism. Our starting question was whether, when a particular agent – me – acts at a particular time, we can understand the agent's action in terms of more basic psychological and worldly conditions. My non-reductionism was intended to answer that question in the negative. Correlatively, my understanding of the non-reductionist about knowledge is intended to be a non-reductionist about particular acts of knowing – and as such permitting this assumption. Perhaps Williamson himself is not this kind of reductionist.

¹⁵ If we thought propositions that were true were different in kind from propositions that are false this may not seem as plausible – but why would be think that, unless we took true propositions to involve facts?

¹⁶ Williamson notes that ‘An apparent asymmetry between the two columns is that the contents on the knowledge side were just treated as propositional while those on the action side were not’ but he considers ‘this asymmetry is largely an artefact of presentation’. In one way this is true – the decision to represent the contents of actions and intentions as incomplete but the decision to present the objects of knowledge and beliefs as complete is a presentational decision. However, the decision is not a superficial one: the contents of actions – and the contents of intentions – are usefully presented as non-propositional to capture the fact that the subject must stand in a reflexive relation to herself in acting. This relation is distinct from the relation she stands in when she acts on another – or to indeed her herself when she acts on herself as she might act on another. The object of an act or intention needs to be reflexively bound to the agent. That is economically effected by removing the specification of the agent from the object of the action or intention altogether – and thus removing both the suggestion that the subject needs to single herself out as the thing to be changed when she acts, and the suggestion that someone other than the agent could be in the subject place. But to argue for this is a job for another occasion.

¹⁷ It is more plausible that there are success conditions, not of the form Aφs, for transactional actions – actions in which an agent acts on an object distinct from herself. If an agent A acts on a patient P, P having a certain property may be a candidate for a success condition independent of A's action: If Polly puts the kettle on the stove, the kettle being on the stove is plausibly a success condition which is not just Polly's action – but is the result of Polly's action.

Exploring this is a job for another occasion.