SAINSBURY ON THINKING ABOUT AN OBJECT

R.M. Sainsbury’s account of reference has many compelling and attractive features. But it has the undesirable consequence that sentences of the form “x is thinking about y” can never be true when y is replaced by a non-referring term. Of the two obvious ways to deal with this problem within Sainsbury’s framework, I reject one (the analysis of thinking about as a propositional attitude) and endorse the other (treating “thinks about” as akin to an intensional transitive verb). This endorsement is also within the spirit of Sainsbury’s account of reference.


. Introduction
One of the many distinctive and plausible features of R.M. Sainsbury's Reference without Referents (2005) is the way it treats all singular referring expressions uniformly, whether or not they have referents, without assigning them all descriptive content.Moreover, Sainsbury's system (RWR) does this while preserving a conservative, orthodox ontology: non-empty terms refer to ordinary things, but empty terms do not refer to non-ordinary things, like non-existent objects, abstract objects, concrete possibilia or other "exotic" things.Sainsbury (rightly, in my view) holds the orthodox opinion that there are no such things.Empty terms do not refer to anything, but they are semantically significant.
I believe that RWR solves many of the outstanding puzzles about empty terms, fiction and non-existence, and gives a very plausible account of names and their role in thought.However, RWR in its 2005 version seems to have the consequence that it can never be literally true that someone can think about a non-existent object.In Reference without Referents, Sainsbury explicitly accepts this consequence.But given the centrality of the notion of "thinking about" to the study of intentionality, this seems somewhat perverse.
In some recent work, Sainsbury agrees, and has given a different account of "thinking about" consistent with the main theses of RWR (see Sainsbury forthcoming).My aim in this note is threefold: to explain the problem with the 2005 account; to criticise Sainsbury's proposed solution; and to sketch an alternative solution which is within the spirit of RWR (even if it is not quite within the letter of the 2005 book).

. Negative Free Logic and Empty Terms
Sainsbury follows Burge (1974) and others in endorsing a negative free logic (NFL), which holds that all simple sentences containing empty names are false.A simple sentence Sainsbury defines as "one constructed by inserting n referring expressions into an n-place predicate" (Sainsbury 2005, p. 66).So sentences which are embedded in non-truth-functional or intensional operators do not count as simple.
According to NFL, "Sherlock Holmes is a detective" is false.Not (as Nathan Salmon (1998) says) because Sherlock Holmes is an abstract object, and no abstract object is a detective, but simply because Sherlock Holmes does not exist.Likewise, "Sherlock Holmes exists" is false, and its negation "Sherlock Holmes does not exist" is true.
This account of negative existential sentences has a pleasing simplicity compared to a view like Evans's (1982), which holds that "Sherlock Holmes does not really exist" (rather than "Sherlock Holmes does not exist") is the negation of "Sherlock Holmes exists".NFL also can maintain an intuitive conception of predication and truth: where "a" is a name and "is F" is a predicate, a simple sentence of the form "a is F" is true iff there is an object denoted by "a" and the object denoted by "a" has the property denoted by "F", and false otherwise (Sainsbury 2005, p. 46). 1 When simple sentences containing empty names are embedded within intensional operators like "believes that", then of course there is no mechanical way to calculate the truth-value of the resulting sentence, which is just as things should be.
One obvious question for the view is how exactly we should demarcate the "simple" sentences.Mere morphology or surface form will not do it: "Holmes is unreal" or "Holmes is non-existent" may look simple, but treating them as such under NFL would give the wrong truth-value: false instead of true.So we have to treat (surely not implausibly) the sentence "Holmes is non-existent" as meaning "It's not the case that Holmes is existent" to get the right result.Sainsbury addresses the question of what simplicity is head-on, and frankly concedes that the notion of a simple sentence is itself partly a product of the theory (2005, p. 69).This seems right: the detection of more hidden structure in a sentence is not ad hoc but rather is a result of getting the best balance of semantic, syntactic and logical requirements.
A deeper problem arises, however, when we consider apparently simple sentences which themselves explicitly concern thought about the non-existent.NFL treats "Sherlock Holmes is a detective" as false, as we know.So it should do the same with (S): (S) Vladimir is thinking about Sherlock Holmes (S) looks like a simple sentence because it seems to be constructed out of a two-place predicate and two names.But since one of these names is empty, then (S) is false.So it is not true that Vladimir is thinking about Sherlock Holmes.Yet surely there can be such a thing as thought about the non-existent?
In his 2005 book, Sainsbury bites the bullet and denies that there is such a thing: How could Le Verrier have used an individual concept to think about Vulcan if there is no such object?On the face of it, thinking about is a relation, and "x thinks about y" entails that y exists.The obvious conclusion is that we speak incorrectly when we say that Le Verrier thought about Vulcan [ . . .]. Strictly speaking, there is no state of thinking about something which does not exist.We use this loose description when the thinker is trying but failing to think about something, which typically means (setting aside pretence, fiction, etc.) that she is in a state which she cannot discriminate from one in which she is actually thinking about something.(2005, pp. 237-238) I find this hard to accept.A thought about the planet Vulcan is as much a genuine case of thinking about something as a thought about the planet Neptune.I think it is literally true that the Greeks thought about Zeus, that Le Verrier thought about Vulcan, and that Vladimir thought about Sherlock Holmes.Moreover, the idea that these claims are not literally true is somewhat out of tune with the spirit of RWR.After all, RWR's aim is to treat all referring expressions in the same way, regardless of whether they have a referent, and surely one of the roles for empty terms is to pick out episodes where someone is thinking about the non-existent?2 If we reject Sainsbury's response, then it seems that there are only two possibilities.One is to insist that (S) expresses a relation after all, and that Sherlock Holmes is one of its relata.Sherlock Holmes would have to be what Sainsbury (forthcoming) calls an "exotic" relatum: abstract, non-existent, merely possible or some such.I agree with Sainsbury that we should not take this route.
The other is to deny that (S) is a simple sentence after all.As noted above, something can seem like a simple sentence even when it is not.But filling out what exactly this means in the case of "thinking about" is a little more complicated.For it means that the inferential role of sentences containing "thinking about" cannot be derived on the basis purely of its apparent form as a two-place predicate.Any attempt to show that sentences like (S) are not simple will require a substantial analysis of "thinking about".Such analyses are inevitably controversial, but I think the outlines of one can be provided.But first I must say something about the way I think an analysis should not be given.

. Propositionalism
It is natural to say "(S) is not a simple sentence because the context 'x is thinking about y' is an intensional context".This is true, I think; but it does not get us very far.It is sufficient for a context to be intensional that classical existential generalisation is not a valid rule of inference for such a context.In other words, something like "There exists something, namely Sherlock Holmes, such that Vladimir is thinking about it" cannot be validly inferred from (S).But this is something we already know, since we already know that Vladimir is thinking about the non-existent.So calling (S) an intensional context does not solve our problem; it names it.
The question is not whether (S) is an intensional context, but why?One answer is that (S) might be shorthand or elliptical for a description of a propositional attitude, and propositional attitude contexts are one of the paradigms of intensional contexts.If sentences like (S) should be analysed in terms of some propositional attitude construction, then this would explain why in its underlying structure it is not a simple sentence.Suppose, for example, that (S) could be analysed as something of the form: (S*) Vladimir is in some propositional attitude state with the content that Sherlock Holmes is F (for some value of F) (S*) clearly isn't a simple sentence, since the name "Sherlock Holmes" is embedded within a non-truth-functional construction, its truth is compatible with the falsity of the embedded sentence.Whether or not (S*) is true depends on Vladimir's state of mind and not simply on the semantic facts.
The view that all reports of intentional states can be analysed as propositional attitude reports is what I call "propositionalism" (some call it "sententialism"). 3Propositionalism says that (S) is really a report of a propositional attitude state.But is propositionalism true?
Certainly in a case where someone is truly describable as thinking about something, this may because they are in a propositional attitude state.My conscious judgement that Wagner wrote Parsifal in Ravello, for example, is a case of thinking about Wagner.But the question is whether all cases of thinking about an object can be analysed in terms of the propositional attitudes.Or to put it more directly, are all intentional states propositional attitudes?
Philosophers often use "intentional state" and "propositional attitude" as names for the same subject matter.But there seem to be many apparently intentional states whose content is not given by a complete sentence.On the face of it seeing the Pope, noticing the guests arriving, admiring the Dalai Lama, are all intentional states.They all involve the direction of the mind upon objects, and represent their objects in particular ways: these are some marks of intentionality.
However, seeing, noticing and admiring, like loving, hating and knowing (someone) are all relations to their objects.And it might be thought that since one of the marks of intentionality is the possible absence of the object, then these relations are not really properly called intentional states.Maybe they are some kind of hybrid of a propositional attitude and a non-mental relation -as, for example, in those theories which analyse seeing X in terms of having a visual experience with the propositional content that X is F (for some visible property F), plus some causal relationship to X itself.So perhaps the existence of these relational states does not threaten propositionalism.
But even once we put these relational states to one side, there seems to be a large class of intentional states which are neither relations nor propositional attitudes.These are the states ascribed by verbs which have become known as intensional transitives: transitive verbs (verbs which take direct objects) which exhibit the characteristics of intensionality (that is, non-extensionality).Here are some categories of intensional transitive verbs (drawn from the taxonomy in Forbes 2006, p. 37) which seem to describe (or to depend upon) intentional states: 1. Verbs of depiction or representation: imagine, portray, visualize, write (about), belief (in).
And there are others, some of which have to do with actions (seeking, hunting for, etc.) and some more remotely connected to intentionality properly so-called (avoiding, lacking, omitting).In all these cases, it is plausible that we have linguistic constructions which bear all the marks of intensionality (with an "s") -resistance to substitution of co-referring terms, and to classical existential generalisation.If these constructions describe genuine intentional mental states, then we seem to have clear examples of non-relational intentional states which are not propositional attitudes.If Vladimir fears the ghost under the bed, then it does not follow that there exists such a ghost.And nor does it follow that he fears the gorgon under the bed even though (on some popular semantic treatments) "the gorgon" and "the ghost" have the same extension, namely: nothing (Forbes 2006, p. 37).Propositionalists believe that intensional transitive verbs like "think about", "hope for", "fear", "want" (etc.) have underlying forms or truth-conditions of propositional attitude attributions.Antipropositionalists (like Zoltán Szabó 2003 andGraeme Forbes 2006) reject this analysis: some intensional transitive constructions have propositional attitude truth-conditions and some do not.
In his recent paper, Sainsbury (forthcoming) explicitly addresses the problem that intensional transitives pose for RWR.He rejects his previous analysis which led him to say that sentences like (S) are false.Maintaining his ontological conservativeness, he argues that intensional transitives encode a hidden structure, and he proposes a propositionalist account of certain central cases of intensional transitives.The essence of his analysis is that whenever someone thinks about X, they are in some propositional attitude state with a propositional content that concerns X.And there is a one-way entailment from these propositional attitude claims to the troublesome intensional transitive sentences.This proposal enables him to maintain the central theses of RWR by relegating all intensionality into the propositional attitude constructions.However, although I think Sainsbury is right to reject his previous analysis, he is wrong to embrace the essence of propositionalism, as I shall now explain.

. The Reduction of Intensional Transitives
The usual way to defend propositionalism is to say that ascriptions of intentional states by intensional transitives can be analysed to reveal covert material which shows their true propositional structure.The psychological implication of this semantic proposal is to identify a propositional attitude the existence of which makes true the ascription of any intentional state by means of an intensional transitive.I will refer to this idea (with an abbreviation hiding a harmless usemention conflation) as the reduction of intensional transitives.
Philosophers of mind will be familiar with this proposal, since they will be familiar with the usual reduction of desiring something (or wanting something) to desiring/wanting that p.Although we talk in an everyday way about wanting a bottle of burgundy, etc., what this really means, on reflection, is that we want to have a bottle of burgundy."To have a bottle of burgundy" is an infinitival complement which is normally taken to have the underlying form of a propositional construction.
As Sainsbury (forthcoming) notes, the linguistic evidence for this kind of "reduction of desire" is strong (see also Larson 2002, andLarson et al. 1997).For example, the ways in which verbs of desire interact with adverbs seems to support it: if I say "Vladimir wanted a Crítica, vol.40, no.120 (diciembre 2008) drink before everyone else" what does "before everyone else" modify?It could modify his wanting -i.e.his want occurs before everyone else's wants-or it could modify his actually having (i.e.drinking) a drink -he wanted to start drinking before everyone else.Both these readings can be heard in "Vladimir wanted to have a drink before everyone else" and this is evidence for the evidence of a covert "have" in the structure.
But although this might work for desire, this is only one example, and there are many others for which it doesn't work as well, as Graeme Forbes has shown (Forbes 2006).To take a couple of examples: "looking for an X" cannot mean the same as "looking to find an X" because "Vladimir will look for a bottle of burgundy tomorrow" does not mean the same as "Vladimir will look to find a bottle of burgundy tomorrow".The first refers to when his looking will get started, but the second has an extra meaning, referring to when the object of the search should be found (Forbes 2006, p. 58; the same point applies to "looking to have").And attempts to fill out the covert material in the case of verbs of depiction have met with similarly little success.For example, it has been proposed (by Larson 2002) that "Vladimir imagines a unicorn" would amount to "Vladimir imagines a unicorn in front of him".But as Forbes points out, this does not deliver the right result when negated: "Vladimir does not imagine a unicorn" does not mean "Vladimir does not imagine a unicorn in front of him".The second could be true even when Vladimir is still imagining a unicorn.
These examples suggest, I think, what an uphill task it would be to defend the propositionalist strategy for all the intensional transitives.So it looks as if there will still be intractable intensional transitives for RWR to deal with.What, then, should we say about "thinking about" itself?

. Thinking about an Object
On the face of it, the problem we encountered for RWR suggests that "thinking about" is an intensional transitive.But as noted above, Sainsbury proposes a specific propositionalist reduction of thinking about an object, which I paraphrase as follows: whenever someone is in the state of thinking about something X, they are in a propositional attitude state whose propositional content concerns X.What should we make of this?
It is true that thinking about something always entails having some beliefs about it.I can hardly imagine winning the lottery -a form of thinking about winning the lottery -unless I have some beliefs about what the lottery is.But this is just a necessary condition for imagining; the beliefs do not constitute imagining it.The same is true of "entertaining the proposition that": whenever anyone thinks about X, they are ipso facto entertaining some proposition about X.This may be plausible as a necessary condition for thinking about X; but, as just noted, Sainsbury wants more than a necessary condition.It would not be easy to establish that every case of thinking about something can simply be identified with entertaining the proposition, or with believing that, or with holding some other propositional attitude.
I doubt that thinking about should be identified with a propositional attitude at all.It seems that whenever someone thinks about X, they have to be thinking about it in some way.That is, they might be imagining X, visualising X, speculating about X, contemplating X, judging something about X, fearing X, hoping for X and so on.And for only a few of these is a propositionalist reduction available.My alternative suggestion is that rather than being a specific or determinate kind of mental state, "thinking about" picks out a determinable kind.Some of the determinates of this kind will be propositional attitudes (judging); but some will not (fearing, visualising).
What all these determinates have in common is that they all involve consciously mentally representing X. Unconscious intentional states do not count as cases of thinking about something.This suggests the following analysis of "thinking about X": (T) S thinks about X = d f S is in some conscious state of mind which represents X.
Of course, this is not a reductive analysis, since the right hand side of (T) introduces the ideas of consciousness and representation.But this is the whole point, since the defence of the claim that (S) is not a simple sentence requires that it has "hidden" semantic structure.I am proposing that the hidden structure makes reference to conscious states of mind and mental representation: in fact, it contains a quantification over conscious representational states of mind.Because of this, we should say that the predicate "is in a conscious state of mind which represents" is not semantically or syntactically simple.If this is true of some thinker, then it is also true that they are in some state of mind, that they are conscious, and that their state represents something.This kind of complexity is of just the sort to create an intensional context.

TIM CRANE
My speculative hypothesis is that the same is true for all intensional transitives: they all involve a reference to a representation in their semantic analysis (see Forbes 2006, chapter 5).Verbs like "think about" create intensional contexts precisely because they contain a reference to something other than their extension.We might say, roughly speaking, reference is made to their intensions.But because of the rather specific meaning that has now come to be attached to the term "intension", I would rather say that these verbs make reference to a mental representation of their objects, by quantifying over such representational states.
These tentative suggestions are, it seems to me, entirely in harmony with the basic principles of RWR.As Sainsbury says, according to an "exotic" view, "intensional verbs emerge as semantically ordinary, and only metaphysically strange.I prefer a more conservative view according to which they are metaphysically ordinary and semantically (slightly) strange" (Sainsbury forthcoming).I agree.The semantic strangeness of intensional transitives resides in the fact that they are not semantically what they seem: analysis reveals that their semantic structure is not that of a two-place predicate, but makes reference to a representation.

. Conclusion
Sainsbury's RWR gives a very convincing account of empty names in general, but (in its 2005 version at least) it assigns the wrong truth-value to sentences like (S).I have argued that the way out of this is not to give a propositionalist reduction of thinking about an object (as Sainsbury does) but rather to treat "thinking about" as an intensional transitive.As such, it needs to have a semantic analysis to display its hidden semantic structure.But such a semantic analysis is quite consistent with the basic principles of RWR. 4