PART I

Wittgenstein, Anscombe, and the Activity of Philosophy

INTRODUCTION

1.

Although Anscombe’s Introduction to Wittgenstein’s Tractatus is a wonderful book, I disagree with Anscombe about a good number of things. This introduction is about the disagreements that emerge in Part I, and about the unRussellian character of the Tractatus. It is not a very introductory introduction, and can be read instead as an afterword to the three essays in Part I. The first section is about unRussellianism, and in the following sections I look at my disagreements with Anscombe in Essays 3, 2, and 1.
I start with something from Essay 3—the significance of Frege for reading the Tractatus, and what Anscombe says about it. Here I want to approach in a different way the questions Anscombe raises. This will lead me to a point from which I can address the main ways I disagree with Anscombe in Part I. My reading of the Tractatus is, like hers, unRussellian, but my understanding of the unRussellianism of the Tractatus is different from hers—and that, I think, is the source of my disagreements with her. But in trying to work out, for this introduction, how to present my disagreements with her, I realized that there is a further disagreement, not touched on in any of the essays collected here. Anscombe wrote that “Wittgenstein’s conception of ‘sense’ may be called the same as Frege’s, if we are careful to add that Wittgenstein had different theses about it” (IWT, 17). She does take seriously the importance for Wittgenstein of the connection between his conception of sense and the directionality of sense (about which, more below); but in writing this introduction, I’ve come to think that she underestimates the significance of directionality for Wittgenstein’s conception of sense. She did not see, I think, what a profound difference from Frege’s conception it marks. Wittgenstein did not modify Frege’s conception of sense, nor did he have different theses about it. He started from a Russellian conception of sense (articulated by Russell in his account of asymmetrical relations), and transformed it radically, winding up with something altogether different from anything in Frege or Russell.1 Writing this introduction has made me realize anew how deep the questions are that one gets into as one reads and thinks about Wittgenstein with Anscombe.
In discussing Frege and Wittgenstein here, I generally use the word “proposition” as the translation of “Satz.” In passages that are specifically about Anscombe, I follow as far as possible her use of “sentence” and “proposition.” For more about the use of “Satz” in the Tractatus, see Essay 4.
At the beginning of her book, Anscombe said that “almost all that has been published about [Wittgenstein’s Tractatus] has been wildly irrelevant”; and she added that if this irrelevance has had any single cause, it is “the neglect of Frege and of the new direction that he gave to philosophy.” She then set out what she took to be distinctive in Frege’s sort of approach to philosophy, and how “empiricist and idealist preconceptions” get in the way of understanding such an approach. She also explained how she took Frege’s approach to be significantly different from that of Russell. In Essay 3, I argued that her account is puzzling, especially in her treatment of the contrast between Frege and Russell. But my claim was that she was anyway right in thinking that there was a significant contrast—one which is important in thinking about the Tractatus, and which can be seen when we look at Anscombe’s account of the picture theory. When I wrote about this in Essay 3, I explained the contrast in two ways. I first referred to the contrast drawn by Warren Goldfarb and Peter Hylton between an object-based view of propositions and a judgment-based view. I quoted Goldfarb’s characterization of Russell’s view: the primitive parts of propositions “subsist in and of themselves”; they are put together into propositions, but are recognizable independently of the particular role they may have in this or that proposition (Goldfarb 2002, 190–191). I argued that the Russellian readings of the Tractatus, which Anscombe was criticizing, ascribed to Wittgenstein an object-based view of propositions, while Anscombe’s own reading of the picture theory involved a judgment-based approach. I then looked at the role of the context principle in Anscombe’s account of the picture theory, and contrasted it with (what I took to be) Russellian readings of the picture theory—those of Norman Malcolm and David Pears.
It might be objected to my approach in Essay 3 that it depends on the contrast between a supposedly Fregean judgment-based view of propositions and a supposedly Russellian object-based view, and that that contrast doesn’t hold up. Although I think such an objection doesn’t work, I won’t here lay out the putative objection or defend my formulation of the contrast, but will instead try to get at the contrast in a different way, from within Russell’s own thinking. That is, I want to contrast Russell’s Russellian approach to propositions and their constituents with an unRussellian approach to which Russell is driven in one sort of case. Focusing on Russell’s general Russellianism will bring out how different it is from the unRussellian approach he very reluctantly takes in the case of propositional functions. The contrast as we can see it in Russell’s own thought can bring into clearer view what is at stake in Anscombe’s insistence on unRussellianism. (Russell’s own unRussellianism is in fact picked out by Frege as something he agrees with. I’ll get back to this.) My aim is not just to give an alternative explanation of the contrast that Anscombe had introduced between Fregean and Russellian readings of the Tractatus. Thinking about unRussellianism, and thinking unRussellianly, lead, I think, into the most fundamental issues with which the Tractatus is concerned; and this was what Anscombe saw. I am trying here, not to defend unRussellianism, but to present the significant contrast between the kinds of approach I had set out originally as “object-based” and “judgment-based.” It turns out to be more complicated than I had realized.
In chapter 4 of The Principles of Mathematics, there is a good statement of Russell’s Russellianism. He says there that every object of thought, everything we can think of, everything that can occur in a proposition, counts as what he calls a term, an expression that he treats as synonymous with the word “entity.” (Here “proposition” is used to mean nonlinguistic propositions.) Every term is a logical subject, and Russell argues that any attempt to treat anything as not a logical subject leads to contradiction. He does allow for a possible exception in the case of some denoted complexes of terms, but he does not make any exception for concepts. Thus, for example, he holds that the concept human, when it occurs as concept in the (nonlinguistic) proposition “Socrates is human,” is intrinsically the same as the concept when it occurs as logical subject—for example, in the proposition “Humanity is a term.” The concept as concept is no less self-subsistent in its occurrence-as-concept than in its occurrence-as-logical-subject. Although it can be a part of a proposition, it is an independent self-subsistent entity. It is clear in Russell’s discussion of such examples as human and humanity that he believes that what it is that is being thought of can be separated from how it occurs in a proposition. He continued to hold versions of this view even while much else in his thought shifted. In 1913, for example, he wrote that the relation precedes can occur in the two different ways, in “A precedes B” and in “Preceding is the converse of succeeding” (Russell 1992, 80).
Russell’s Russellianism comes under strain when he discusses propositional functions, in chapter 7 of The Principles of Mathematics. He had introduced the notion of propositional functions in chapter 2, and had there explained it this way: “ϕx is a propositional function if, for every value of x, ϕx is a proposition, determinate when x is given” (Russell 1937, 19). If this or that term occurs in a proposition, we can imagine replacing it by other terms. Thus, in the case of the (nonlinguistic) proposition “Socrates is a man,” we can imagine replacing the term Socrates by other terms; and in that way we get such other propositions as “Plato is a man,” “Aristotle is a man,” and so on. Because Russell was there simply introducing the notion of propositional function, he did not deal with complications. These come up in chapter 7, when he tries to give an account of how we can distinguish in a proposition the subject and what is asserted about the subject. The background idea, as Russell begins the discussion, comes from the earlier treatment of propositional functions in chapter 2—the idea that the way to get hold of what is asserted about Socrates by the proposition “Socrates is a man” is to omit the term Socrates from the proposition. In that way, we get what is also asserted about Plato by “Plato is a man,” what is asserted about Aristotle by “Aristotle is a man,” and so on. While this appears to work for “Socrates is a man,” it emphatically does not work for “Socrates is a man implies Socrates is mortal.” We may indeed take that to be asserting of Socrates what “Plato is a man implies Plato is mortal” asserts of Plato, but we cannot get hold of what that is by removing Socrates from “Socrates is a man implies Socrates is mortal.” For the result of omitting Socrates is this: “… is a man implies … is mortal,” which does not include any indication that the same term must be included in both places, if we are to get a proposition asserting about the term in question what “Socrates is a man implies Socrates is mortal” asserts about Socrates. Here we seem to have something that we cannot pull out of the proposition in which it occurs, the what-is-asserted-about-the-term. This is what then drives Russell to conclude, reluctantly, that “the ϕ in ϕx is not a separate and distinguishable entity: it lives in propositions of the form ϕx and cannot survive analysis.” When I say that Russell takes this view reluctantly, I mean that he believes he has no choice, and that the view may indeed lead to contradiction (although he also thinks that the opposite view leads to contradiction). The unRussellianism of the view that Russell has wound up with is plain. In contrast with the Russellian account of how the concept human can occur in a proposition as concept or as term, and is an entity independent of its occurrence in this or that way in this or that proposition, Russell is led to a “non-entity” account of what is in common to “Socrates is a man implies Socrates is mortal” and “Plato is a man implies Plato is mortal.” Those propositions are values of the propositional function “x is a man implies x is mortal”—but what is in common to the propositions isn’t something that can occur independently of the propositional contexts within which it is recognizable, within which it “lives.” One might well ask how close Russell has come, in this sort of case, to what Frege speaks of as a function, and Russell himself recognized the Fregean-ness of the view to which he had been driven, when in his exposition of Frege, he said (in appendix A of Principles) that, if his conclusion in chapter 7 is right (that is, the conclusion that the ϕ in ϕx is not an entity), then what Frege calls a function is not an entity.
When Frege comments on The Principles of Mathematics, he singles out the passage in chapter 4, which I used to explain Russell’s Russellianism, in explaining his disagreement with Russell. And he then mentions that Russell “appears to incline” toward the Fregean position in the passage cited in my last paragraph, where Russell is explaining Frege’s views and says of the unRussellian conclusion that he reached in chapter 7, that if that is right, then a Fregean function is in general not an entity. Frege then summarizes his view:
It is clear that we cannot present a concept as independent, like an object: rather it can occur only in connection. One may say that it can be distinguished within, but that it cannot be separated from the context within which it occurs. (Frege 1984a, 282)
Frege’s way of putting the point there—in terms of what is distinguishable within but not separable from the context—is close to Russell’s “the ϕ in ϕx is not a separate and distinguishable entity: it lives in propositions of the form ϕx and cannot survive analysis.”
In discussing the influence of Frege on Wittgenstein, Warren Goldfarb (2002) has argued that the resemblances between features of Frege’s view and ideas that are significant in the Tractatus should not be assumed to have come from Wittgenstein’s reading of Frege. It may be, as he suggests, that Wittgenstein started off with a Russellian view, and, in thinking it through, arrived at an understanding of propositions and ontology that was close in various ways to Frege’s unRussellian understanding. I want to think about the issues here in a somewhat different way: there is Russell’s unRussellianism, Frege’s unRussellianism, and the quite distinctive unRussellianism of the Tractatus. I want to get into view some of the differences between the unRussellianisms. What I came to see, in thinking about these different unRussellianisms, is that one of the starting points for Wittgenstein’s unRussellianism may have been Russell’s Russellian account (in The Principles of Mathematics) of the directionality of asymmetrical relations. Another likely starting point is Wittgenstein’s dissatisfaction with the theory of types. This is not to discount influences directly from Frege, some of which are indeed spelled out by Goldfarb.

i. Russell’s UnRussellianism and That of Wittgenstein

Russell’s Russellianism, as I have explained it, involves nonlinguistic propositions and what Russell speaks of as terms—that is, whatever can occur in propositions, where Socrates is his most frequent example. (The word “occur” there is meant in a logical sense; so Parkinson, for example, does not occur in most of the nonlinguistic propositions in which Parkinson’s disease occurs as term.) Russell’s unRussellianism involves what is common to the values of a propositional function, where these values themselves are nonlinguistic propositions. What is common to the values of a propositional function (the ϕ in ϕx) is some nonseparable feature of propositions. These nonseparable features of propositions, then, do not fit what Russell had said about “everything that can be an object of thought.” To see the relation to Wittgenstein, think of two changes from Russell’s unRussellianism. First, switch from talking about nonlinguistic propositions to talking about propositions in the Tractatus sense, as propositional signs in use. And, secondly, switch from talking about what is in common to the propositions that are values of a particular propositional function (where this is what Russell says “lives in” the proposition) and talk instead about whatever is in common to some propositions and is a mark, in them, of a shared form and content; and say of this that it lives in the proposition. What meaning this proposition-feature has, it has in its occurrences in propositions. These two changes from Russell’s unRussellianism move us closer to ...