Why might one think that propositions are act-types?
1 Introduction
Generally, propositions are taken to be entities that fulfil certain roles; propositions are the semantic contents of sentences, the contents of intentional attitudes like believing and doubting, and bearers of truth and falsity. In this paper, I shall explore reasons to think that propositions are act-types, which is to say, that propositions are types which are tokened by mental acts, in particular, those mental acts that predicate properties of objects, as well as considering one strong reason to reject the act-type hypothesis. First, I shall outline what the act-type hypothesis is taken to be, and then compare it with other available theories of propositions to find reasons to prefer it.
A brief overview of the act-type hypothesis
The act-type hypothesis, as defended by Soames (2014a) and Hanks (2015), states that propositions are a type that encapsulates certain sorts of cognitive acts. In particular, those cognitive acts that involve predication. So, according to this thesis, the proposition instantiated by the sentence a is F is the type made up of all the cognitive acts that predicate F-ness of a. In so doing, these types are composed of cognitive acts that have truth conditions; the predication a is F is true iff a instantiates/participates in/exemplifies F-ness. Soames argues (2014a, pp. 96-97) that propositions, as types of the such tokens, inherit this quality from their tokens; since each act token enters into a truth- or falsity- bearing relationship with the world, so too does the type. Moreover, these types can act as semantic and intentional contents in a similar way. If I believe that a is F, then my cognitive act is one of predicating F-ness of a, and then entering into a believing relation with that predication, whatever we might take the believing relation to be (pp. 97-98). Similarly, for the sentences ‘gold is heavy’ and ‘Gold ist schwer,’ each of these attributes the property of heaviness to the substance gold, and thereby can be taken to be arising from the same mental act, one of predicating heaviness of gold, which in turn makes them tokens of the same act-type. This brief overview naturally fails to encapsulate the full complexity of this view; however, it shall suffice to explain the positive reasons for accepting such an account as explored in section 2. Then, in section 3 I shall address a very significant objection to this view of propositions.
2 Reasons for act-type hypothesis
Reasons for accepting the act-type hypothesis shall come in the form of preferring it to other accounts available. Therefore, I shall give a cursory glance to these other accounts below:
Possible worlds (PW) propositions: Propositions are sets composed of possible worlds in which the proposition in question is true. So, the proposition expressed by the sentence ‘gold is heavy’ is the set of possible worlds in which gold is in fact heavy. (For a defence of this, see (Stalnaker, 1984)).
Russellian-Fregean (RF) propositions: Propositions are abstract entities that are composed of either the entities that enter into them (Russellian) or conceptualisations of the objects that enter into them (Fregean). So, the proposition expressed by ‘gold is heavy’ is made up of the relation of exemplification, the objects that instantiate gold and the property of heaviness, for Russellians, or concepts of gold and heavy with the relation of exemplification for Fregeans. Importantly, for both these views, propositions exist prior to any thought about them, and exist independent of any agent’s ability to express or cognize them; they exist independently of minds. The intentionality of thought and language is to be explained in terms of the intentionality of propositions (for a critical discussion of these theories see (Soames, 2014b)). A thorough-going exploration of these related theories is not required here, it is their abstract, mind-independent nature, and that they take propositions to be prior to thought, that is important for my purposes. Since both Russellian and Fregean propositions share these traits, I shall treat them as a single entity, since any argument that is raised against these traits can be taken equally against both Russellian and Fregean propositions.
In the following subsections, I shall outline three features of the act-type hypothesis which make it preferable to these accounts and thereby provide good reasons for accepting the hypothesis.
Naturalism
According to the act-type hypothesis, propositions are simply categories that agents collect the components shared by certain sorts of mental and linguistic events into; ontologically speaking, nothing is being stipulated to exist beyond those mental events with which we are already familiar, i.e., acts of predication. Propositions have the qualities they do in virtue of the tokens that the type categories. This stands in stark contrast to the other major accounts of propositions, which posit further ontological categories. Taking RF propositions first, since these propositions are agent- and mind-independent, and contain objects, relations and properties, accepting RF propositions entails accepting a ‘third realm’ (after the mental and physical) in which these abstract propositions exist. To do otherwise would be to lose their mind-independent status. PW propositions may be more ontologically extravagant, since on the most naturalistic reading, PW propositions require abstract, hypothetical possible worlds exist and that we group them into sets. Now, in the case of both PW and RF propositions, we have to postulate a plurality of entities to make these theories function; either sets of hypothetical possible worlds, or propositions that exist ‘out there’ in a third realm which is independent of the mind.
Though such speculation is not without philosophical merit (such a ‘third realm’ is sometimes postulated to explain the apparent mind independence of numbers and arithmetic, for example, and possible worlds often invoked to explain how modal statements come to have truth values), the principle of parsimony suggests that if an otherwise equally acceptable account exists that is more parsimonious, then the more parsimonious account is to be preferred. In the following parts of section 2 I shall seek to establish that, on at least two counts, the act-type hypothesis has superior explanatory power compared to PW and RF propositions. Therefore, assuming that I am successful, the principle of parsimony gives us a reason to prefer the naturalistic account given by the act-type hypothesis to PW and RF accounts of propositions.
Unity of propositions
RF & PW propositions also produce a strange phenomenon which they struggled to account for. According to the RF account, propositions gain their intentionality by their relationship to the objects and property they refer to, or conceptualisations thereof. However, if that were the case, why wouldn’t a simple list suffice? For example, ‘gold is heavy’ is a proposition, but ‘exemplify, gold, heaviness’ is not, nor is ‘exemplify, Cgold, Cheaviness’ (Soames, 2014b, p. 28). Yet, according to RF theories, the lists given here contain all the components of a proposition. Propositions, it seems, must have some unifying feature that concatenates the list components into a proposition, but RF theories remain mysterious as to what. Similarly, advocates of PW propositions must explain how a set of possible worlds ‘hangs together’ to create the sort of thing that a proposition is; a set is equivalent to the members of the set, so how do this group of possible worlds ‘come together’ to represent gold as heavy?
This problem is usually called ‘the unity of propositions’ and the act-type hypothesis has a plausible account of how propositions are unified. In short, propositions are unified because they are types, and are unified in virtue of the tokens that instantiate them. Consider the proposition instantiated by ‘gold is heavy’ once more. The type that the sentence ‘gold is heavy’ instantiates is common to ‘Gold ist schwer,’ since the cognitive act of predication that is required to create these sentences, predicating heaviness of gold, is a common to all instances thereof. This, however, raises the question as to whether all such cognitive acts are in fact the same, and what this sameness is in virtue of. It might be argued by detractors of the act-type hypothesis that, in fact, all such cognitive acts in fact do not share anything in common. For example, what part of Alfred thinking ‘gold is heavy’ is shared by Suzanne thinking ‘Gold ist schwer?’ For surely, their neurochemical processes are significantly different, so we cannot appeal directly to their brain states. This leaves us only with the mental, and in the case of the mental, we have no good grounds one way or the other to suggest that the mental states involved are in fact similar. This puts the act-type hypothesis on shaky ground, our detracting interlocutor might assert.
This objection has merit; however, I believe that the philosophy of psychology can provide us with some measure of a response. Carruthers (2004) argues that the mind should be thought of as a selection of mental, though not necessarily neurochemical, modules shaped by natural selection, and as I argued in a previous submission (O'Keefe, 2018), we should accept this view with some modifications. In short, if we accept the mind is computational in nature, and I think we should (see (Haugeland, 2000) for a defence of this view) then those mental functions that evolved in our pre-agrarian history should be processed by a series of mentally simulated modules, which have evolved to performed specific functions, that operate with a proprietary database and on a small set of inputs to generate a small set of outputs.
If this view is correct, then the mental process of predication should also be modular; it seems reasonable to assume that humans have been predicating properties of objects since the emergence of the species, or perhaps shortly thereafter. Therefore, I would appeal to these mental modules as the basis for commonality between the cognitive actions undertaken by individuals. Since these modules would be similar in all individuals, as part of our evolutionary heritage, we have good grounds to think that the events they engender would also be similar. Naturally, this defence may be open to objection on grounds beyond the scope of this paper, and I have dealt with some of them in my previous submission. Accepting that this may weaken the argument somewhat, I shall move on to one more positive reason one might accept the act-type hypothesis.
De se vs de re
This reason, and various responses, are adapted from Soames (2014a, pp. 106-109), and although it does not represent a complete answer to the problem as outlined here, the act-type hypothesis provides new insight on this particular problem.
To illustrate the problem, here’s a brief scenario. I am walking along a shopping street and in the windows that I am passing I catch sight of a person with their shirt untucked from their trousers and form the belief ‘some person here has his or her shirt untucked.’ I am passing at speed and don’t make out clearly any other features of the person I caught sight of. Then, further down the road, I pass mirrored window, and then come to realise that it is in fact I who has his shirt untucked and come to form the belief ‘I have my shirt untucked.’ The mystery is as follows: our intuition tells us that I have come to learn something new, that it is in fact I who have my shirt untucked. Yet, according to RF theories of propositions, all that is in propositions is in its representational content, and there is no new representational content to be found. I don’t see or represent anything new in the latter case at all, and yet somehow, I come to know something new. There are several ways in which the RF theorist can respond to this intuition, such as suggesting we come to know the same thing in some new way (attributed to Perry in (2014a)) or attributing a new property to the proposition (attributed to Lewis in (ibid.)). These are, as Soames puts it, ‘explaining away, rather than preserving, our pre-theoretic thoughts on the matter,’ (p. 108), and it seems like instead act-type theory permits us to preserve our intuitions in this case in a way that RF theory does not. I endorse Soames’ view, itself adapted from Frege, that every agent has a type of cognitive act that relates themselves to the world that is not shared by anyone else. If this is the case, then when a third-party cognizes the de re proposition ‘he has his shirt untucked,’ that mental act does not in fact resemble the mental act of me thinking ‘I have my shirt untucked.’ Similarly, in the first instance, when I come to entertain the proposition ‘someone here has his shirt untucked,’ that mental event is unlike ‘I have my shirt untucked,’ because the latter involves a cognitive act which includes myself. This, I would argue, seems fairly intuitive in that how I relate myself to the world is different in a fundamental way to how I relate others to the world. Thinking of myself, given my psychological ‘closeness’ to myself seems fundamentally different to thinking of others. Relating this back to the modular theory of mind mentioned earlier, it seems plausible that I employ a different modular process when I am relating myself to the world than when I am relating others to the world. So, because propositions are act-types, then in fact a new proposition is in fact being entertained, even though the representational content remains the same, because the cognitive act is different in its composition.
It must be stressed that this is simply a glimpse at a potential solution to this problem that retains our pre-theoretic intuitions. Much must be said to flesh out the details on how this difference in cognizance operates, however, that must be for others to determine. Given this, I shall move on to reasons that stand against act-type theory.
3 Reasons against act-type hypothesis
Propositions about Vitali sets
Potentially the most fatal objection to the act-type hypothesis provided by Keller (2016). For the sake of brevity, I shall summarise what is an exceptionally complicated argument. According to the most widely accepted version of set theory, the Axiom of Choice is true. One consequence of the Axiom of Choice is the existence of Vitali sets. Vitali sets are sets composed by drawing, without any systematicity, from other sets, and they can be uncountably large. This entails the existence of sets that are wholly unsystematic and uncountably large in their composition.
Keller then argues that for a set to be cognizable it must meet one of the following conditions, where A is some agent and S is some set:
‘(i) A cognizes all of the members of S.
(ii) A cognizes a rule that generates all and only the members of S recursively from a base of elements cognized by A.
(iii) A cognizes a predicate under which all and only members of S fall.’ (p. 10).
Keller then argues that the proposition ‘Vitali sets are sets’ and other propositions that feature Vitali sets are therefore un-cognizable, since they contain an element that is similarly un-cognizable by a finite agent. This presents a problem for the act-type hypothesis, since this entails that no actual agent could possibly have cognized a Vitali set. This means that there is at least one proposition instantiated by, ‘Vitali sets are sets,’ that could never be a type whose tokens are mental events, since one constituent is un-cognizable. Since there exists propositions which act-type hypothesis cannot capture, then act-type hypothesis must be false.
Two possible responses to this problem must be ruled out immediately:
rejecting the Axiom of Choice would be a strange response to a problem in metaphysics, given that mathematicians seem to have accepted it as part of set theory and,
positing an infinite intelligence makes any claim to naturalism laughable.
Keller also roundly responds to the possibility of creating languages or agents that could cognize Vitali sets, and in a manner that I find to be unarguable, so I must avoid these also in answering the point. Therefore, I’m going to suggest a response to the Vitali set objection that accepts Keller’s argument, and may rescue the act-type hypothesis at the cost of a small number of counter-intuitive outcomes.
I’d like to posit that, since they are un-cognizable, when we use the term ‘Vitali set’, we do not in fact take actual Vitali sets are the referent for that term. What we take as the referent for that term is a linguistic approximation of the mathematical entity that is a Vitali set. This linguistic approximation attempts to grasp the nature of such a thing, being an uncountably large set of unenumerable items that cannot be cognized, as Keller insists. Since it cannot be cognized, it seems reasonable to think that the term ‘Vitali set’ cannot pick out the things that mathematicians have proven to exist, other than by custom. It seems no loss to our language to admit that we only speak about approximations of an entity we can prove to exist, since we cannot ever grasp the actual entity itself. A similar argument might be deployed about God in the philosophy of religion: that the term God only approximates the entity we are discussing, and we can never capture actual content about him. The cost of this is to say that, counterintuitively, the proposition ‘Vitali sets are sets’ is false, because the term Vitali sets doesn’t pick out the mathematical entity it claims to pick out, it picks out our mental approximation of such entities. I suggest that for such an otherwise powerful theory, this trade-off is acceptable. Propositions that discuss uncognizable entities come out looking strange since what that proposition captures is an approximation of a thing that, by definition, we cannot think about. That said, a proposition like ‘Vitali sets exist’ would come out as true, since the proposition simply points to the existence of our mental approximation of the mathematical entities. This approach may have unseemly outcomes for mathematical propositions, particularly those that prove the existence of Vitali sets; further research would have to be done to explain how these propositions can be captured while accepting that un-cognizable entities cannot be referred to.
This solution involves an unintuitive outcome in certain special cases, i.e., when trying to express the un-cognizable, and this makes my case vulnerable to objection from those unwilling to make such a trade-off, nonetheless, in the service of preserving an otherwise powerful hypothesis that explains much about propositions, I am willing to accept this trade-off.
4 Conclusion
In this essay I have aimed to provide three positive reasons for accepting the act-type hypothesis by demonstrating how is answers two major problems for competing hypothesis: the problem of naturalism, the problems of unity of propositions, and how it provides the potential for a more intuitive answer for de se versus de re cases. I have also endeavoured to provide a response to a very strong candidate for a reason to reject the act-type hypothesis, by means of suggesting that any proposition that endeavours to capture an un-cognizable entity must come out as false, because of it’s uncognizable nature, but that this is a small price to pay to preserve an otherwise powerful hypothesis. Naturally, these responses, especially the last two, leave open much for future research, however, the arguments so presented set out reasons why one might come to believe that propositions are act-types.
Bibliography
Carruther, P. (2004). The mind as a system of modules shaped by natural selection. Retrieved 11 11, 2017, from University of Maryland: http://faculty.philosophy.umd.edu/pcarruther/Shaped-modules.htm
Hanks, P. (2015). Predication and Unity. In Propositional Content (pp. 87-121). Oxford: Oxford University Press.
Haugeland, J. (2000). Semantic Engines: An Introduction to Mind Design. In D. D. Cummins, & R. Cummins (Eds.), Minds, Brains and Computers (pp. 34-50). Oxford: Blackwell Publishing.
Keller, L. J. (2016). Against Naturalized Coggnitive Propositions. Erkenntnis, 82(4). Retrieved from https://www.academia.edu/31329823/AGAINST_NATURALIZED_COGNITIVE_PROPOSITIONS
O'Keefe, W. (2018, 01 08). Do evolutionary considerations provide good reason to think the mind is a collection of special-purpose modules. Submitted to Philosophy of Psychology module. Birkbeck College, University of London.
Soames, S. (2014a). Cognitive Propositions. In J. C. King, S. Soame, & J. Speaks, New Thinking About Propositions (pp. 91-124). Oxford: Oxford University Press.
Soames, S. (2014b). Why the Traditional Conceptions of Propositions Can't be Correct. In J. C. King, S. Soames, & J. Speaks, New Thinking About Propositions (pp. 25-44). Oxford: Oxford University Press.
Stalnaker, R. (1984). Chapter 1. In Inquiry (pp. 1-25). Cambridge, MA: MIT Press.