Guide to Chapter Four of Gareth Evans'
The Varieties of Reference

Rick Grush (Department of Philosophy, UC San Diego)
Copyright © Rick Grush 1999, 2000, 2001, 2002
Copyright holder grants permission to distribute, quote, and reproduce for any non-commercial purposes provided authorship and copyright are clearly indicated.

Chapter 4: Russell's Principle

In this section Evans motivates the need for an interpretation and defense of Russell's Principle. He begins with a fairly minimal interpretation — that it requires discriminating knowledge, or in other words, it requires the subject to be able to distinguish the object of thought form all other objects. He then provides two examples involving steel balls, one of which presumably provides intuitive support for RP, and the other of which presumably yields results that run counter to RP. Evans uses these conflicting examples to motivate the need for a more thorough understanding of RP, so that its ability to provide theoretically satisfactory analyses of such cases is secured. Evans closes by remarking briefly on why an analysis of thought is important for his ultimate project of understanding reference.

The first hurdle in defending Russell's Principle (RP) has been to get clear about its meaning. Anyone can claim to endorse it if it is given a weak enough interpretation: the adherent to the photograph model might claim that the mere possession of states causally related to the object suffices for discriminating knowledge. Evans proposes to start off with the following rough formulation of RP, which, while still imprecise, is at least substantive enough to get the ball rolling: in order for a subject to think about an object, she must have discriminating knowledge of that object, which means she must be able to distinguish the object of her thought-attempt from all other things. Three ways that this can be discharged are i) when one can perceive it, ii) when one can recognize it, and iii) when one knows uniquely distinguishing features it possesses.

Evans then produces two thought experiments. The first concerns a subject S who one day sees two identical steel balls suspended by string from a single point and rotating about it. The subject has, let us suppose, no ability (descriptive knowledge or recognitional capacity or current perceptual content) to distinguish one of the balls from the other. Evans takes the upshot of this example to be in accord with Russell's Principle: we would not say that the subject could have a thought about just one of the balls.

[Note: Many people find this quick argument unconvincing, but I think this is because it is not always understood. In order to go against RP, the example would have to be one in which the subject cannot distinguish between them, and yet can have a thought about just one of them. So any attempt to exploit discriminating knowledge, such as 'the ball that I saw to the left when I first saw the two balls', while perhaps successful in allowing one to think about just one of the balls, does so in a way that makes the case no longer a counter-example to RP, as the subject has in fact invoked descriptive resources that are sufficient to discriminate them. Evans' claim is not that one couldn't have a thought about one of the balls, but that one could not have a thought about one of the balls without invoking some sort of distinguishing considerations of the sort the defender of RP claims is needed. The other sort of objection people have takes a bit longer to explain. The phenomenon is that they agree that they cannot distinguish them, but they then imagine the two balls rotating, and then focus attention on one of these imagined balls, and (attempt to) issue a thought like 'that ball was manufactured before 1940'. Since they were, supposedly, thinking about the two balls in imagination, and since they, supposedly, managed a thought about just one of these imagined balls, and since, supposedly, they have no means to discriminate one of the balls from the other, it is believed that a counter-example to RP has been produced. But upon closer inspection this doesn't hold up. First, let us just agree off the bat that a (genuine) thought that 'a is F' is true if a is in fact F, and false if a is in fact not F. Now, suppose that one ball, B1, was manufactured in 1938, while the other, B2, was manufactured in 1942. Now suppose that our subject goes through the maneuvers above, imagining the two previously encountered balls, selecting one of the imagined balls out, and attempting to think 'that ball was manufactured before 1940.' Has the subject thought something true, or something false? If the subject was thinking about B1, then the subject was thinking something true, and if B2, something false, of course. So which was the subject thinking about? There is no answer to this. What the subject has managed to do is to single out one of the imagined balls in a manner congenial to RP: by focusing attention on one of them when thinking 'that ball was manufactured before 1940' in a manner relevantly parallel to the way perceptual content would allow one to issue a thought about one of two perceived balls. But there is nothing that connects this imagined ball with either of the balls actually perceived in the past. And as far as I can see, any attempt to make such a connection immediately falls back on some actual discriminating feature (the first one I saw; the one that was on the left when I first noticed them, etc.), and hence is no longer a counter-example to RP.]

The second thought experiment is a variant on the first. In this case, the subject sees two balls, one on one day and the second on a later day (I suppose that in this case, there is on the first day only B1 hanging by a string, and the second day only B2, where B1 and B2 are distinct). Furthermore, the subject, because of focal amnesia, retains no memory of the first encounter. Such a subject would be unable to produce any facts which would discriminate between the two, without relying on descriptions like 'the one from which my memories derive' (but as we saw, such reliance immediately remove the case from contentions as a counter-example to RP).

Evans claims that even the subject would, upon learning about both balls and his amnesia, no longer claim to be thinking about 'that ball' (which would again be in accord with RP). But, he admits that it would nonetheless be natural to say that the subject was thinking of the second ball, or had the second ball in mind. And this is in prima facie conflict with RP. The photograph model, according to which the thought is about the object causally responsible for it (so to speak) would say that in fact the subject is thinking about the second ball. Given that there is nothing the subject can do to discriminate them (we are supposing, of course, that the subject does not exploit a description such as 'the ball that is the cause of the information I retain'), we appear to have a counter-example to RP.

Note that the second sentence on page 91 says that we have 'some apparent counter examples'. This seems to be a slip, as only the second is intended as a candidate counter-example. Perhaps Evans is referring to earlier discussions of the photograph model and Donnellan's views in particular (e.g. p. 81).

Evans then discusses the importance of a theoretical defense of RP. Such a defense will make clear exactly what RP is: in defending the truth of RP one will be defending the necessity of certain kinds of discriminating capacities for thought about objects, and with a better grasp on the kind and workings of these kinds of discrimination, one will be better able to address the adequacy of examples like those presented earlier in the section.

The last paragraph on page 92, which closes this section, Evans explains why the interest in thought. This is rehash of material Evans discussed in the last chapter (3.2) concerning one of the two strategies for establishing the Russellian status of a category of expression.

4.2 Verificationism and ideal verificationism

In this section Evans introduces one way of cashing out the requirement of discriminating knowledge: verificationism (perhaps supplemented in such a way as to make it ideal verificationism). Though Evans will argue (toward the end of this section) that verificationism is inadequate, the form of the verificationist position has some features that Evans will exploit in his own proposal, which will be made in Section 4.4.

Evans gives a brief overview of Dummettian verificationism. A distinction is drawn between two sorts of thought one can have about an object.

First, one can have thoughts about objects — objects such that one is in a position to verify whether or not they have certain properties — because the object thought about is demonstratively identified, and whether or not this object has or lacks the property the thought attributes to it can be immediately assessed, prototypically perceptually. This is the one-step verification procedure. For example, you and I are in the same location, you point to a large object and say 'That is big.' I simply assess the applicability of the predicate ' ... is big' to the object that I am currently perceiving (the one that has been demonstrated to me).

The second sort of thought is one for which a two-step verification procedure is needed. Such thoughts will be of the form 'a is F', where 'a' is a name. The first step will consist in deciding, for some demonstratively identified object, whether it is in fact a, and then once this has been answered in the affirmative, using the one-step procedure outlined above for assessing the truth of thoughts involving demonstratively identified objects.

The first step of the two-step procedure, whereby the thinker determines of some given object whether it is a, can take one of two forms: the subject might be able to recognize a, or might know (and be able to decide the applicability of) some distinguishing features of a. For example, someone tells me that my mother is ill. I first come into the vicinity of some object, and employing my recognitional capacity I determine that that object (demonstratively identified) is my mother. I then use the one step procedure outlined above to determine whether or not the object I am perceiving is such that the predicate ' ... is ill' is truly predicated of it. This example involves my use of a recognitional capacity in the first stage. A different sort of example would involve my ability to see if some object uniquely satisfies some description — e.g. 'the tallest person in the room is sunburnt'.

Thus the verificationist not only makes good on what discriminating knowledge amounts to, but does so in a way which is in harmony with the trichotomy (acquaintance, description, recognition) Evans mentioned earlier. But note that both kinds of thought are such that the one-step procedure is necessarily employed.

From the bottom of page 94 to the bottom of page 95 Evans provides quotes from Dummett (the verificationist whose views are under consideration) to show how Dummett applies these verificationist themes to his interpretation of Frege.

After this, Evans brings up a number of objections against verificationism. There are two sorts of cases which give the model trouble: i) statements involving small, large, distant, or otherwise imperceptible objects (electrons, the Milky Way Galaxy, black holes), and ii) statements involving abstract objects. Evans discusses Dummett's attempts to deal with these cases, and claims that these attempts fail.

In the first sort of case Dummett claims that we can extend the demonstrative model to include i) detection with instruments, ii) the notions of the demonstrative verification procedures being carried out by beings with very different perceptual and cognitive abilities. Evans' objection to this line of extension is that it effectively undercuts the spirit of the verificationist model (this is the first full paragraph on page 97). Why? Because such extended procedures are no longer able to play the same sort of role that their non-extended counterparts play in the deciding of the truth value of propositions about the objects so identified. For example, if the thought is 'Quine is bald', then one can demonstratively identify some object, decide that it is Quine via one's recognitional capacities, and then decide of this demonstratively identified object if it is or is not bald. But a thought such as 'The Milky Way is spinning', does not proceed by singling out something demonstratively in this extended way, and then deciding whether or not this demonstratively identified object is spinning. Even if one's thought to that effect includes some imagined experience, such as imagining seeing the Milky Way from above, and then noticing that it is spinning (which would involve a lot of spatial as well as temporal shifts of scale from our own experience, but is more or less along the lines of what the ideal verificationist has in mind), this has nothing to do with how we actually verify such propositions, and in particular it has nothing to do with how we would actually demonstratively identify the object in question. And this is what undercuts the spirit of the verificationists' position.

In the case of abstract objects, Dummett's suggestion is that we adopt some preferred signs or names for such objects, and the verification procedure is then run not on the objects, but on the preferred set of names. For example, to decide if [(5 x 61) - 33] + (51 x 23) is evenly divisible by 2, we first determine the number according to its canonical name (in whatever notation we use), and then decide, of the number represented in this way (1445) whether it is evenly divisible by 2, and it is not.

Evans' objection is that even if this works for numbers, it is not at all clear that it will work for all abstract objects, because there may not in general be anything like a canonical notation for such objects. Numbers appear to be a rather special case in this regard.

Evans also points out that there are problems when it comes to assessing the truth of thoughts about entities from the past, which cannot be identified at all any more.

[Note for clarification: Near the top of page 99, Evans starts the paragraph with "As for the second of the two points..." He is referring to the top of page 96. What is a bit confusing is at the top of page 96, he says "And, second, those aspects of our thought which it does take account of can be accommodated by other models." So here at the top of p. 99 he first discusses what it is that the verificationist model seems to do an intuitively good job of taking account of, and then he goes on to point in the direction of a different model that does a better job. This alternate model begins in 4.4 and continues throughout later parts of the book, especially chapter 6.]

The final point Evans makes is that the fact that, e.g., imagined experiences or images of various sorts play a role in our thinking should not be interpreted as vindicating verificationism, as it might seem initially to do. Rather, he says that it is an index of the fundamental role that our conception of the spatial world plays in our thinking. (These last few paragraphs are a bit obscure. They are perhaps best read as his way of flagging the fact that he does not want to throw out the spatial representation baby with the verificationist bathwater, as Evans' own proposals make heavy use of spatial representation. Spatial representation of various sorts plays roles in both theories, but as it will turn out, rather different roles.)

4.3 The generality constraint

The basic idea of this section is quite straight-forward. It is the idea that thoughts are structured, in that an ability to think that a is F results from the joint exercise of two distinct abilities, the ability to exercise the Idea of a and the ability to exercise the concept of F. Evans here uses 'Idea' to refer to an ability to think of objects, and 'concept' to refer to an ability to think of properties. Given this, it follows that if a subject can think that a is F, and can also think that b is G, then that subject will have the ability to think that b is F and a is G. This constraint is used to produce an objection to the photograph model.

First, Evans contrasts the Generality Constraint and a language of thought. While a LOT is typically taken to employ symbols, which Evans glosses as entities with both semantic and non-semantic properties, Evans claims that he prefers to think of the thought constituents not as 'elements', but as 'abilities'. Though Evans is getting at an important distinction here, his way of explaining it is inadequate. Certainly skills, such as the ability to think of some object, will have semantic as well as nonsemantic properties (skills are implemented in hardware or wetware, for example), and so appealing to this is insufficient to distinguish Evans' notion from a LOT. Rather, it seems better to think of the difference as one between the nature of the vehicles, and the means by which they carry a content. A symbol has a dedicated vehicle which gets acted upon by processes which are nonsemantic (that is, are sensitive only to the symbols' nonsemantic properties), and the symbol has the semantic value it has independently of whether or not it is being operated on by these processes. A skill, by contrast, may share its 'vehicle' (implementation base) with many other skills, it is essentially active, and its semantic value is a function of the activity it is engaged in, or can be engaged in. On this reading, possessing a LOT, with an inner set of symbols and processes comprising a system characterized by a combinatorial syntax and semantics might be one way to meet the Generality Constraint, but not the only way, and not Evans' preferred way.

The second detail is Evans' contrast of the Generality Constraint, which is a constraint on thought, with the systematicity one can find in language (the topic of discussion right now is public, overt language, not a putative language of thought). Though Evans, on pages 101-102 uses the systematicity of language as an example to illustrate the Generality constraint, he makes it clear that this is just an analogy for illustrative purposes. This is because thoughts are essentially structured, whereas linguistic expressions may or may not be so. It is entirely possible to have one word sentences which are really sentences, though not structured.

Suppressed in this discussion of the Generality Constraint (but required for the conclusions Evans wants to draw) is the idea that these constituent capacities to have thoughts belong to one of some number of categories (analogous to the grammatical categories described in Chapter One), and that combination is constrained by categorial appropriateness. So while it is possible to combine an Idea with a concept to arrive at a thought, such as the thought that John is happy, merely combining two Ideas, for instance of John and Harry, does not yield a thought at all. This sort of categorial constraint, a mental analogue of a grammatical constraint, is not the same as the sort of constraint mentioned in footnote 17. There the issue is one of semantic appropriateness of the property and the object, thus ruling out thoughts such as that the square root of 36 is happy.

Evans closes the section by detailing how the photograph model can be in conflict with the Generality Constraint. His target is Donnellan's claim that when a subject has a belief which would be expressed with the words 'a is F', that belief is about object x iff x is causally responsible for the belief. [For example, when I, at a party, come to have a belief I would express with the words 'The man in the corner drinking a martini is a spy', my belief is about the woman in the corner drinking water if she was the object responsible for my belief, regardless of how misinformed I am about the object.]

Evans expresses two misgivings. First, that a program such as Donnellan's, like any of the Photograph Model family, would violate Russell's Principle. Second, and more to the point in this chapter, Donnellan's apparatus, even if it works, seems capable of treating only beliefs, and thus is in principle applicable only to a small range of possible thoughts, and this is in violation of the Generality Constraint. By putatively circumventing the need to have an Idea of an object (conceived of as an ability essentially playing a role in structured thought), one can, with Donnellan's apparatus, have a subject that could putatively have a belief about an object to the effect that it is F, but not necessarily be able to even think that that object is G or H.

4.4 The fundamental level of thought

This is one of the most important sections of the book, as well as one of the most difficult. In it, Evans offers a minimal account for what is involved in thinking of an object in a manner that satisfies RP (knowing which object it is one is thinking about). The basic idea is that for any particular object of a certain sort, there is a fundamental ground of difference which distinguishes that object from all others of that sort. In order to think of an object, one must think of it as possessing the ground of difference that in fact distinguishes it from all others. For spatio-temporal objects, this ground of difference is spatial location at a time.

Before getting into the details, note that the last sentence of the first paragraph of page 106 (which starts "My strategy...") can be misleading. He is discussing his strategy in the final sections of this chapter here, and not his strategy in this section. Contrary to what is said in this sentence, he is in fact (in section 4.4) trying to give some substance to the notions of what it means to know what one is thinking about. Section 4.6 is where he runs through the consequences of denying Russell's Principle.

His starting point is that in order to think that P, one must know what would make P true. In the case where P is 'a is F' this involves knowledge that requires an Idea of an object (the Idea of a), and an Idea of a property (the Idea of Fness). This follows from the Generality Constraint. Of course, more is required than this, but these will be necessary conditions. [Note that Evans is here using his technical term 'Idea' to cover both Ideas of objects (as in the last section), as well as what would normally be called 'concepts'. I will follow the latter convention and use 'Idea' as a blanket expression.] So now he will go into what it means to have an Idea of an object.

For Evans, an Idea of an object is based on a 'conception of a world of such objects'. Here he has in mind kinds of entities like colors, sounds, chess positions, sets, physical objects, etc. And by a 'world' of such objects, he means something like all the objects of such a type. For every type of object, there will be individuation conditions — conditions that determine how many such objects there are, via determining when two objects are really the same or not. For example, numbers are differentiated by their position along an infinite ordering. Two numbers at the same position are the same number. Two numbers at different positions are not the same. These criterial features for object identity Evans calls their fundamental ground of difference. For temporal objects, objects that exist in time (like statues and unlike numbers), the fundamental ground of difference must be relativized to a time; that is, it must be a fundamental ground of difference that distinguishes that object, from all others of that kind, at a time. For physical objects, their identity conditions consist of i) spatiotemporal location, and ii) the type of object it is.

One will be employing what Evans calls a fundamental Idea of an object if one thinks of that object as the possessor of the specific fundamental ground of difference that it in fact possesses. For example, if I think of the number 3 as being the third natural number, or as the number occupying the third position on an infinite ordering, etc., I will be employing a fundamental Idea of that object. Alternately, I might think of it as Dummett's favorite abstract object and be thinking of the same number in a non-fundamental way, using a non-fundamental Idea. It follows immediately that if one is thinking about an object in virtue of employing a fundamental Idea of that object, then one is satisfying Russell's Principle, since one will be able to distinguish that object from all others (by definition of a fundamental Idea and of fundamental grounds of difference).

Evans now wants to argue that fundamental Ideas and grounds of difference are in play even when one is not employing a fundamental Idea of an object in one's thinking on a given occasion. The first example will be existential statements, such as 'some number is the largest prime less than 2,000,000'. The wherewithal for such a thought does not involve a particular fundamental idea of the object, since we don't know what number is in question. But what is involved is the notion of some object that has a fundamental ground of difference of the kind appropriate to numbers. To think that a G is F is to think that there is an F thing that is distinguished from all other things by a fundamental ground of difference appropriate to Gs. To think that someone stole the evidence is to think some object individuated by fundamental grounds of difference appropriate to people stole the evidence. (Note that we can now think of the 'world' of such objects mentioned by Evans at the beginning of this section as just the totality of objects amenable to a given kind of fundamental ground of difference. So the 'world' of numbers is just the realm of all entities amenable to being distinguished by position along an infinite linear ordering.)

Now for some typographical conventions. Evans uses the Greek letter delta to stand for a fundamental idea of a G. In these notes (since Greek characters are still dicey on web pages) I will simply use a lower case roman d. This would be a general notion of a fundamental idea, like position along an infinite ordering; one's general knowledge of the kinds of fundamental grounds of difference appropriate to a given kind of object. For a specific fundamental Idea, like that for the number 3 as being individuated by being in the third position, Evans uses d*. Evans uses corner-quotes to express propositions that could be formed with such Ideas (In these notes, I will use angle-brackets < >). So in the case where Gs are numbers, the thought that some number is F would be <d is F>, while the thought that a particular number, say 3, is F, would be <d* is F>.

The proposal is that these existential thoughts combine a concept with a fundamental idea 'd' of a type of object, where the d is not any particular fundamental idea d*, but is just one's knowledge of the sorts of fundamental grounds of difference applicable to things of that sort.

Evans then briefly turns to concepts, or Ideas of properties. He says that possession of the concept of being F is having knowledge which, when conjoined with any fundamental Idea d* of a particular object, yields knowledge of what it would be for a proposition of the form <d is F> to be true. So having the concept of pain is just to have knowledge of what it is for some being (human or animal) to be in pain. If one has no idea of what it would be for a person or animal to be in pain, then that person does not have the concept of pain; if one has no idea about what it is that would make it true of some number that it is prime, then one does not have the concept prime number. Notice that in so defining concept possession, Evans has guaranteed that concepts will satisfy the Generality Constraint — concepts are defined as knowledge that, when combined with any arbitrary fundamental Idea (d*) of an object of the G type (again, provided that objects of the G type are not semantically anomalous with F), is sufficient to provide the subject with knowledge of the truth conditions of the thought.

The final full paragraph on page 109 is enlightening. The topic is the relation between specific fundamental Ideas, and what might be called an unspecified fundamental Idea. Reiterating the point of the previous 2 paragraphs, Evans says that knowledge of an existential proposition involving a G is conceived of as being true, if it is true, in virtue of the truth of some proposition involving an unspecified fundamental Idea d. The same holds for non-fundamental Ideas. When one thinks that Dummett's favorite number is even, one conceives of it as being true, if it is true, in virtue of the truth of some proposition of the form <d is F> (where in this case, F is the concept even).

The difference between the two cases is that in the instance of existential statements, there might be more than one proposition of the form <d if F> that can make the existential proposition true. In the case of the particular-proposition involving a non-fundamental Idea, it is understood so that, if true, at most one such proposition <d is F> is capable of making the proposition true. Furthermore, which fundamental Idea this is is determined by the non-fundamental Idea a. (Evans has a remark here about the fact that so expressed, knowledge of what makes some non-fundamental Idea a an Idea of a G is formally the same as knowledge of what makes something a property (the example used F) applicable to Gs. In both cases, the knowledge consists in knowing what it would be for some proposition of the form <d is F> (or <d = a>) to be true. But he remarks that this does not assimilate non-fundamental Ideas to properties.)

Evans then turns to generalizing this account to cover Ideas of temporal objects, and non-fundamental Ideas of (temporal) objects. For temporal objects, the Fundamental Ideas (as well as properties) must be relativized to a time. Evans indicates these with subscripts, but here the notation will be d[t] and F[t], where the characters in square brackets [] are understood to be subscripted. First, Evans relativizes his account of concept possession, so that possession of the concept of property F[t] is knowledge of what would make arbitrary propositions of the form <d[t] [is] F[t]> true. (The square brackets around 'is' indicate that tense is not a factor — the knowledge might need to be expressed with 'was' or 'will be' depending on the time involved.)

Now consider a proposition of the form <d[t'] [is] F[t]> where t' is not the same as t. In such cases, the proposition is understood via an understanding of two propositions: one is just what has been discussed, <d[t] [is] F[t]>. The other involved one's knowledge of identity conditions for objects over time, which allow one to entertain propositions of the form <d[t] = d'[t']>. For example, when I look at an aged and feeble man in a hospital bed, and say that this man was a great linebacker, this proposition, in order to be understood, must be articulated. Suppose that t is now, and t' is some time in the past when the man in question played football. Then the proposition 'This man was a great linebacker' is of the form <d[t'] [is] F[t]>. I must know what it would be for something like <d[t] [is] F[t]> to be true, that is, what would make it the case that someone is a great linebacker, and furthermore, I must have a grasp of identity conditions on objects (in this case persons) which allows me to understand <d'[t'] = d[t]>, that is, this man I see in the hospital bed is the same man as the one who played football in the past.

For non-fundamental Ideas, three propositions are involved. A proposition of the form <a [is] F[t]> involves

i) knowledge of what it would be for a proposition of the form <d[t] = a> to be true.

ii) knowledge of what it would be for a proposition of the form <d[t] = d'[t']>.

iii) knowledge of what it would be for a proposition of the form <d[t] [is] F[t]> to be true.

(i) is just what is required for one to have a non-Fundamental Idea of a G, as was discussed earlier in the section; (ii) is an expression of one's knowledge of identity conditions for objects over time; (iii) is no more than possession condition for the concept of F applicable to Gs. Conditions (ii) and (iii) were just the two conditions required for thoughts involving fundamental ideas of temporal objects.

Evans closes the section by explaining how this account of what is involved in understanding certain sorts of propositions guarantees that subjects who understand propositions in this way conform to the Generality Constraint. In effect, since Evans has defined concept possession (of the concept F) in terms of knowledge of what it would be for an arbitrary proposition of the form <d is F> to be true, he has guaranteed both facets of the GC. For first, it follows immediately that if a subject can think 'a is F', she will be able to think 'b is F', where 'a' and 'b' are stand-ins for different fundamental Ideas of some type G (this man, that man, for example). Furthermore, the ability to think 'a is F' will entail the ability to think 'a is H' for some concept H applicable to G's, for exactly the same reason — the subject will only have the concept H if she knows what it would be for arbitrary propositions of the form <a is H> to be true, and hence must be able to think 'a is H'.

4.5 Comparison with verificationism

Evans draws attention to the formal similarity between his account and that of the ideal verificationist.

In both cases (Evans' account as articulated in 4.4, and the ideal verificationism of Dummett) there are two sorts of thoughts, which we might call basic and non-basic. For the verificationist, basic thoughts involve the immediate assessment of the applicability of some predicate to a demonstratively identified object as decisive concerning the truth of the thought about the object. Non-basic thoughts (employing names rather than demonstrative identification) involve a two step procedure, the ability to determine of a demonstratively identified object whether it is the bearer of the name, and then a determination of the basic type explained above.

Evans' account retains much of the same structure. Instead of understanding a basic proposition as resting on an ability to verify applicability of a predicate to an object given demonstratively, it is understood as consisting in the use of knowledge of the fundamental ground of difference of such an object, and what it means for an object picked out by such a fundamental ground of difference to have this property. Non-basic thoughts involve a two step procedure: first, the identification of the right basic thought (linking the name to the correct Idea); and then the basic thought-procedure as above.

The differences are that Evans has not taken demonstrative identification, or any other kind of identification, as basic. And because actual encounters such as would be required to support demonstrative identification are not basic, abstract, small, distant, and past/future objects are not the problem for his account that they are for the verificationist.

Finally, he mentions that his program has the virtue of making sense of the spatial imaginings discussed in 4.2 (imagining the man who made the table, or imagining the solar system from above). In thinking of spatiotemporal objects, one thinks of them as being in space and time, as spatial differences constitute the fundamental grounds of difference of objects.

When we represent material objects in the imagination, we ipso facto represent them as located and differentiated in space. We imagine the carpenter, as in the example of 4.2, as located in a particular position in space, though, of course, there is no particular position we imagine him as having. Such representations of objects in the imagination are just like arbitrary fundamental Ideas (to be understood on the model of the arbitrary names of certain formal systems). [p.114]

4.6 The counter-examples

In this section Evans returns to the two examples involving the steel balls, and argues that the claim that the subject could issue a thought about just one of the balls — which would constitute a violation of Russell's Principle — is an unattractive one for a number of reasons.

The first paragraph is an abstract for how Evans' way of conforming to Russell's Principle (via fundamental Ideas) will apply to the three ways of 'knowing which': demonstratives will be discussed in Chapter 6, and involve fundamental Ideas of material objects. The two types of non-fundamental identification to be addressed are i) recognitional capacities in Chapter 8; and ii) descriptions, which are covered by his account of concept possession (provided that the description employs a property of which the subject has a concept), and will not receive separate treatment.

Evans returns to the case of the steel balls. The case, recall, was one in which a subject encountered two identical rotating balls, and at a later time, attempts to think about just one of them. The example was supposed to target Russell's Principle via questioning whether it would be possible for the subject to think about one of the balls even though he had no discriminating knowledge of that ball. Evans now argues that in fact the subject cannot think about one of the balls (an argument similar to the one given in 4.1). The part of the argument given on this page is that, supposing he has a fundamental Idea of a steel ball (one sufficient for an existential thought, a 'd' in the terminology of the last section), can he understand what it would be for 'that ball' (the one he is attempting to think of) to be the ball putatively identified by his Idea? The proposition in question is the one I labeled (i) above: does the subject know what it would be for a proposition of the form <d[t] = a> to be true (where the 'a' is 'that ball')? Since there is nothing in the subject's 'conceptual repertoire' that could possibly make a difference between the subject's Idea being of 'that ball' versus the other ball, Evans holds that he cannot. In order for the counter-example to work, it will be necessary to suppose that the subject's thought does manage to home in on one of the balls.

Evans compares such a supposition to Chisholm's contention that if a brain were divided such as to yield two persons, the original person would survive only in one, even though there would be no behavioral or other empirically accessible difference between them. Furthermore, he claims that "Chisholm takes himself to be able to grasp the supposition 'This one, not that one, is the original person'." Evans claims that such a claim is nonsensical. That in absence of any possible empirical means of telling the difference between the two hypotheses, the supposition that there is a difference is illusory.

Evans anticipates an objection to his reasoning, pointing out a disanalogy between the steel ball case and the Chisholm case. In the case of the steel balls, the proponent of the photograph model can suppose that the subject's ability to launch a thought at exactly one of the balls is underwritten by his causal contact with one of the balls, or more accurately from the fact that one of the balls is causally responsible for the information that the subject has about the balls. Evans' rebuttal to this is that the very concept of knowledge has to do with a capacity, things that the subject can or cannot do, and by claiming that causal antecedents are sufficient for knowledge, the PM theorist is subverting this, and in effect turning the useful notion of knowledge into a powerless sham. Specifically, Evans says "The concept [of knowledge] is one of a capacity, and the proof of its being possessed at a given time must surely reside in facts about what the subject can or cannot do at that time." What this means is something like this. Suppose we have two subjects, and we want to discover which of them knows something. Surely, this line proceeds, we are seeking an ability that one of them has and the other does not. It would be a serious shift in the meaning of the word 'know' if according to some theory the two subjects had exactly the same abilities, there was nothing that one could do that the other could not, and have one subject credited with knowledge of some sort and the other not so credited.

A further difficulty is that the PM credits the subject with knowledge to the effect that something is true (the identification of his Idea of the object and one of the two balls), when he has 'no inkling' of what would make it true. Evans reiterates an important point, which is that we are assuming that the subject does not think of using the notion of the causal antecedents of his idea for individuative purposes. If the subject does this, then the case ceases to be a counter-example to Russell's Principle, because the subject will have individuating knowledge, this time by description ('the ball that caused my current Idea').

Evans' argument here is unsatisfactory. The opponent he seems to be — and should be — addressing here is the proponent of the Photograph Model, who will say that the subject in fact manages to think of exactly one of the balls in absence of discriminating knowledge because of the fact that one of the balls is the causal source of the information the subject has. The proponent of PM need not and should not claim that the subject has any knowledge of how or why this is the case. So even if Evans is right that knowledge must be somehow manifestible, the PM proponent is not fazed, for she will simply say that the subject's capacity to think about one of the balls is not based on any knowledge he has. Furthermore, Evans will himself argue that it is possible for subjects to think that they are having a thought of a particular kind when in fact they are not. If this is true, then it will surely be possible, because of the presence or absence of external conditions, for two subjects to be in identical dispositional states and yet for one of them to know that P and the other not to know that P, simply because one of them is in fact truly thinking that P and the other is not thinking that P, even though he supposes himself to be. I am not saying that either of these points is fatal to Evans' position, but at least it is not clear that his argument as presented here shows a) why the proponent of PM need be crediting the subject with any knowledge at all in supposing him able to have a thought about just one of the balls; and second it is not clear that what Evans says here about knowledge is consistent with claims he makes at other parts of this book about thought externalism.

Evans now turns (middle of p. 117) to a potential counter to his argument as presented above. The rebuttal is that one can know that 'This is water', when what makes it true that the stuff is water is the fact that that stuff is H2O, even though the subject has no inkling of what H2O is. Thus, it seems that in fact a subject can know something to be true when he is not in a position to know what makes it true. (This is the line that I claimed the adherent to PM should take.)

Evans' answer to this is that, in effect, though the subject does not know what specifically makes the proposition true, he does, and must, know what sort of thing would make it true — namely that there is some characterization or essence which applies both to the stuff he is identifying, and the stuff in rivers and lakes. He does not know what this characterization is, but he knows there must be one. He knows that the proposition 'this is water' is true in virtue of some other proposition 'this is H', where H is schematic for some suitable characterization or other of water.

Evans then points out that there is no parallel between this case and the case of the steel balls, because ex hypothesi the notion of causal antecedents does not enter into the subject's thought at all, but in the case of water, there must be a sense in which the subject realizes that there is some characterization in virtue of which this stuff and that stuff are the same kind of stuff. That is, as soon as we allow the subject in the steel ball case to exploit anything that would distinguish the balls, the case is no longer a counter-example to RP. The point here is that if the ball example is altered so as to make it analogous to the water case, by allowing some notion of the causal antecedents to enter into the subject's thinking about the ball, then the case no longer is a counter-example to RP. (But as I mentioned above, it is not clear that the proponent of PM needs to claim that the ball example needs to be analogous to the water case, for she is presumably happy to say that the subject is in fact thinking of just one of the balls even though this fact is not underwritten by anything the subject has knowledge of or can even articulate — the mere presence of the correct causal genesis of the information.)

Finally Evans addresses the following adversarial line of thought (from the top of p. 118): The verificationists have given an analysis of what it is to know the truth conditions of a proposition. Their answer seems to be about the best one going. Nonetheless, they have difficulties with certain things, such as space, time, theoretical entities like neutrinos, etc. Given that there are things like time and neutrinos that we have to admit that we have knowledge of in spite of the fact that we cannot give an account of what it is to know the truth conditions of propositions involving reference to such entities, we may as well also drop the requirement that understanding a sentence with suitable singular terms requires the audience to know which object is in question. In short, since we have to give up the idea that we have knowledge of truth conditions anyway, why not just drop Russell's Principle?

Evans' response is that just because there are some things which are difficult for the verificationist, it does not follow that we can just give up and drop all empirical constraints on thought. Our move away from verificationism need not require moving this far. He points out that the case of knowing which object is in question is disanalogous to the cases that trouble verificationism. These cases fall into two classes: on the one hand our concepts of space, time and matter; and on the other, concepts supported by specific empirical theories. He claims that the case of the steel balls cannot be assimilated to either. It cannot be assimilated to the case of theoretical knowledge, because there is no theory. And it cannot be likened to the case of the verification transcendence of space, time, and matter (our conception of an objective world). Evans concludes that the case is not like the others whose existence is undeniable. Furthermore, as he has shown, treating the PM answer as a case of knowing which object is in question is inconsistent with the correct notion that knowledge must consist in capacities that the subject has, and thus would involve an overthrow of any reasonable account of concept possession.

In the next chapter, Evans will explain why it seems natural to say that the subject was thinking of just one of the balls, even though he was not. This explanation will take much of the motivation from the photograph model's sails.

[End of Guide to Chapter Four]

Copyright holder grants permission to distribute, quote, and reproduce for any non-commercial purposes provided authorship and copyright are clearly indicated.

Chapter 4: Russell's Principle

Table of Contents: