Putting our minds together: on Roeloefs’ Combining Minds


It is usually considered impossible for distinct subjects to together constitute a further subject; in particular, it is thought impossible for the conscious experiences of a subject to be composed of the experiences of distinct subjects which make it up. This assumption has recently been challenged by Luke Roelofs, who defends Combinationalism: “The experiential properties of a conscious subject are sometimes mere combinations of the experiential properties of other subjects which compose it” (2018, 6).

I shall present two objections to this thesis. The first questions some of the general metaphysical claims made by Roelofs. In particular, I shall question Experience Inheritance, the thesis that whenever a member of an aggregate undergoes an experience (i.e., instantiates a phenomenal property) the whole aggregate undergoes the same token experience (op. cit., 79).

Roelofs assumes that if a member of an aggregate is F, the aggregate is thereby (at least) partly F. But being partly F is just to have a part (member) which itself is F; e.g., an aggregate which contains me as a member is partly spatially located wherever I am located. Correlatively, it may be that an aggregate of which I am a member is partly undergoing a certain visual experience. But this is just for an aggregate to have a member, i.e., me, which is undergoing that visual experience. And for an aggregate to instantiate the property of having a part which is undergoing a visual experience is very different to that aggregate’s itself instantiating a phenomenal property. Therefore, Roelofs has not provided a robust defence of Experience Inheritance.

The second objection to Combinationalism depends on the idea of phenomenal unity: for two distinct experiences e1 and e2 to be phenomenally unified, there must be something it is like to experience e1 and e2 together (e.g., when one hears the radio together with tasting coffee). Phenomenal unity contributes to the phenomenal character of the experiences which are unified. As Barry Dainton puts it, when token experiences are phenomenally unified, “a full description of the phenomenal features of each experience […] will mention (in some manner) all the other experiences” with which they are unified (2008, 281).

On Roelofs’s view, it is possible for a subject S3 to be made up of distinct subjects S1 and S2; S1 experiences e1 but not e2, S2 experiences e2 but not e1, and S3 experiences both e1 and e2 together. The problem is that the phenomenal character of e1 will differ for S1 and S3; for S3, e1’s character involves being experienced together with e2, but for S1 its character cannot involve this, since S1 does not experience e2. But a single token experience cannot have distinct phenomenal characters. Therefore, the above scenario is not coherent: S1 and S3 cannot each have literally the same experience, if S3 but not S1 experiences it together with a distinct experience.


4 thoughts on “Putting our minds together: on Roeloefs’ Combining Minds

  1. Taillard, Antoine says:

    Hi Donnchadh,

    Thank you for this very interesting talk. I couldn’t ask you my question during the Q&A, so here is it on comment form.

    It wasn’t mentioned explicitly at the beginning of your video, but one of the assumptions you make for this whole discussion is that the component experiences (taken together) are different from the composite experience, isn’t it? If you admit that there is an identity between the composite experience and its components (because you are a friend of composition as identity, say), the debate does not get off the ground, right?

    And thus: could someone sympathetic with Roelofs’ view maintain composition as identity to answer your objections? Or is there something in Combinationism that makes this option unviable?

    1. O Conaill, Donnchadh says:

      Hi Antoine,
      Good comment. Luke certainly seems at least sympathetic to composition as identity (e.g., “aggregates in some sense just are their many parts, taken together” – p. 27). Furthermore, he defines combination partly in terms of grounding, and he allows that in some cases grounded entities can be identical with their grounds. So (as he suggested to me) there is a danger that the issue I describe doesn’t get going, because the composite experiences are (together) identical with the experience they compose.

      I have two quick things to say in response. First, it seems to me that this identity would not in and of itself remove the question of how the parts are unified to form the whole (i.e., whether they are unified in a Leibnizian or a Newtonian manner). What might be at stake here is something like the following: is this entity (the whole, which is identical with the parts taken together) most fundamentally a whole, or most fundamentally a collection of parts? If one thinks this distinction can be drawn, then the issue of how to explain the unity of composite experiences will not be removed by identifying such experiences with their components taken together.

      Second, there is something about combination which, in my opinion, is in tension with composition as identity, and that is the clause that the composite entity be fully explained by the components and the relations between them. The orthodox view of metaphysical explanation is that it is irreflexive: in that case, if the composite entity is identical with the (collective) components, it is not clear to me how one could explain the former by appeal to the latter.
      There are various moves available here (e.g., specifying that explanations relate descriptions or depictions of entities rather than the entities so described), but I think that at any rate there is a problem here for someone wishing to combine combinationism with composition as identity.

  2. O Conaill, Donnchadh says:

    Hi Olivier,
    Thanks for your comments. As you noticed, the argument I outline in the talk is rather different to those I sketched in the abstract. I haven’t changed my mind about those other arguments, but I realised that it would not be feasible to present them properly in the time allocated. I don’t think of the argument I have ended up giving as developing either of the other argments – rather, it complements them, in that it targets a slightly different thesis put forward by Luke.

    On your second point, this idea (that for different phenomenal properties or experiences to be phenomenally unified requires that they be experienced by a single subject) is very widely accepted in the relevant literature. I can’t think of anyone who has questioned it, certainly not in any detail (though many would question my stronger claim, that what it is for experiences to be phenomenally unified is, in part, for them to be had by the same subject).
    In defence of this idea, phenomenal unity is typically understood in terms of conjoint phenomenology, i.e., there being something it is like to experience A and B together, and this ‘something it is like’ talk is usually understood as short for there being something it is like for a certain subject to experience A and B together.

    On the third point, I’m actually sympathetic with your thought that grounding and ‘nothing over and above’ are distinct; at the very least, I don’t think that for A to be grounded in B entails that A is NOAA B (I have a paper with Martine Nida-Ruemelin on grounding, physicalism and consciousness where we defend precisely this claim). The reason I slipped into NOAA-talk is because this is how Luke himself prefers to characterise what is grounded (I had other references to this earlier in the talk but I had taken them out, but that one managed to stay in).
    One other thing: technically I disagree with the gloss on ‘NOAA’ which you offer (that it suggests there is just one thing). I am more open to A being NOAA B, even though A and B are numerically distinct. I understand this as in effect A’s not being an ontologically significiant entity in addition in B. Obviously uncontroversial examples are hard to come by, but I would tentatively sggest that singleton sets are NOAA their members, conjunctive properties (if there are any) are NOAA their conjuncts, and mereological sums (if there are any) are NOAA their members.

  3. Massin, Olivier says:

    hi Donnchadh, excellent videos!

    -have you given up on the worry about phenomenal properties being non-cumulative you raise in the abstract, or is your anti-leibnizian take a development of it ?

    -perhaps I missed a step, but how do your support the claim that for two phenomenal properties to be phenomenally unified, they have to be experienced by a single subject ? I find it very plausible, but couldn’t your leibnizian opponent simply deny this ?

    -A rather side issue about your suggestion ni the end —composite subjects are not fully grounded in their components subjects, but sth over and above them: this assumes that whatever is fully grounded in X in nothing over & above X. I know Fine, Rosen & other seems to accept this, but I’m never get that idea. Grounding is irreflixe, nothing over and above suggest there is just one thing, so I don’t see how both can get together. To put it otherwise, if B is (partly or fully) grounded A, B does not reduce to A.

Leave a Reply