Review of Soames (2018)

A New Vision is the sequel to Soames’ The Analytic Tradition in Philosophy, Volume I: The Founding Giants (Princeton UP, 2014). Founding Giants covered Frege, Moore and Russell. New Vision covers Wittgenstein’s Tractatus, the rise of logical empiricism and its downfall, the advances in logic due to Gödel, Tarski, Church and Turing, Tarski’s theory of truth, and contrasting approaches to ethics and meta-ethics in the 1930s. Soames describes his goal as being to identify major insights and achievements, distinguishing them from major errors or disappointments. His declared focus is explication and evaluation of arguments in the texts of Wittgenstein, Carnap et al. Thereby Soames conceives of himself as “arguing with the greats” rather than historians of analytic philosophy. He thereby seeks to avoid the perils of antiquarianism which besets history of philosophy when it is bowed down by too much attention to historical-textual detail, whilst his engagement with the secondary literature is sparse.

I do believe that it is possible to do insightful history of philosophy by interrogating dead philosophers as though they were walking amongst us—possible because it’s actually been done. Exemplars of this kind of work are Jonathan Bennett’s Kant’s Analytic (1966) and Kant’s Dialectic (1974), volumes which have stood the test of time, proving fruitful for philosophers and historians of philosophy alike. But I don’t think that there’s a simple equation which determines that more history, more textual detail means less philosophy—because sometimes more of that is just what’s needed to channel the philosophy of our forebears. It’s because Soames hasn’t done enough to get the history and the texts right that I think he quite often gets their philosophy wrong.

Soames’ story in New Vision is, as he says, a “complicated” one— understandably so because his aim is to engage directly with the arguments of the greats and they gave a lot of arguments. As a consequence, New Vision might better be characterised as a collection of interrogative episodes rather than as an extended dialogue. To provide an impression of the whole, I’m going to evaluate one such episode in which Soames attempts to strike up an argument with Wittgenstein.

In New Vision Soames takes Wittgenstein to task for what he describes as “among the darkest and most implausible aspects of the Tractatus”, Wittgenstein’s metaphysics of simples and atomic facts configured from them, ideas which Soames does not consider to have had much interest or influence anyway (Soames 2018, 23). Where does Soames think Wittgenstein went wrong? To be blunt: because Wittgenstein had the ill-fortune to come before Kripke. Soames credits Kripke with the land mark discovery that metaphysical and epistemic modalities needn’t march in step but have the potential to diverge, so propositions might be necessary whilst being a posteriori and a priori though contingent. For Soames this discovery was one of the most remarkable achievements of analytic philosophy in the 20^th century. But coming before Kripke, Soames claims, Wittgenstein mistakenly identified necessarily true propositions with propositions knowable a priori. According to Soames it’s this very mistake, “the notorious tractarian collapse of the modalities”, that led Wittgenstein down the false path to his misbegotten metaphysics of simples and atomic facts (Soames 2018, 14).

Wittgenstein famously advanced his atomism by arguing that if there were only complexes all the way down, “then whether a proposition had sense would depend on whether another proposition was true” (2.0211). This would be an intolerable consequence because, Wittgenstein continued, “[i]t would be impossible to form a picture of the world (true or false)” (2.0212). Since it is possible for us to form a true or false picture of the world, Wittgenstein concluded that the analysis of complexes must terminate in absolute simples. Soames reconstructs Wittgenstein’s argument along the following lines.

Suppose \(S_1\) is a statement affirming the existence of a complex designated by the logically proper name “\(O\)”. In order for \(S_1\) to “have sense”, by which Wittgenstein means be true or false, \(S_1\)’s constituent expressions, including “\(O\)”, must have meaning. In order for “\(O\)” to have meaning, \(O\) must exist. Because \(O\) is a complex, \(O\) exists if and only if its parts (\(a\), \(b\), \(c\)) are arranged a certain way. Let \(S_2\) be the statement whose constituent expressions include logically proper names for \(O\)’s parts and which says that \(O\)’s parts are so arranged. Then whether \(S_1\) has sense depends upon whether \(S_2\) is true. But in order for \(S_2\) to have sense its constituent expressions must have meaning too,which they do only if \(O\)’s parts exist. Since \(O\)’s parts are complexes too, they exist if and only if their parts are arranged a certain way. Hence whether \(S_2\) has sense depends upon whether another statement \(S_3\) which says that the parts of \(O\)’s parts are so arranged is true, and so on without end. Represent this chain of meaning-truth dependencies as an unending sequence:

Now the key interpretative question is why does Wittgenstein take this regress of one sentence’s meaningfulness presupposing the truth of another to be vicious? For Soames it’s vital to appreciate that this regress presupposes a chain of necessary connections whereby the existence of a complex is analysed in terms of the existence and arrangement of its parts: necessarily \(O\) exists if and only if \(O\)’s parts exist and they’re arranged a certain way, necessarily \(O\)’s parts exist iff the parts of \(O\)’s parts exist and they’re arranged a certain way, and so on without end. We can represent this chain as an unending sequence of necessary conditionals:

According to Soames, we have seen, Wittgenstein presupposes that necessity and a priori knowability coincide. Hence, for Soames’ Wittgenstein, (\(S_\Box\)) is equivalent to another non-terminating sequence of a priori knowable conditional:

Soames now reasons that if there were no simples “it would follow that knowing that [“\(O\)”] means what it does” and hence knowing the meaning of the sentences in which “\(O\)” occurs, “would require knowing the proposition that \(a\), \(b\) and \(c\) are composed in the right way” (p. 13). But the same reasoning can be repeated for its parts: “knowing that they exist and that propositions about them are meaningful, and have the senses that they do, would require knowing the existence of still further objects, as well as the meaningfulness of still further names for those objects and so on without end” (pp. 13–14). Soames concludes: “Thus, if there were no metaphysically simple objects, then one couldn’t know the meaning of any sentence or perhaps whether it even had a meaning” (p. 14).

Soames’ reconstruction of Wittgenstein’s argument isn’t plausible. Even supposing that (\(S_\Box\)) and (\(S\)_apriori) are equivalent it doesn’t follow that this imposes a requirement upon what must be actually known by a speaker who grasps “\(O\)”. A proposition’s being knowable (a priori or otherwise) is quite different from its being known – possibility doesn’t entail actuality. So even if it is a priori knowable that \(S_1\)\(\rightarrow\)\(S_2\), it doesn’t follow that anyone actually knows this, much less that a speaker has to actually know \(S_2\) in order to actually know \(S_1\). Soames supposes that (\(S\)_apriori) imposes an unending, therefore unsatisfiable set of necessary conditions upon actually knowing that \(O\) exists. But because (\(S\)_apriori) covers only the weaker modality of what is knowable, it remains open that a speaker might know \(S_1\) and not know \(S_2\) even if \(S_1\)\(\rightarrow\)\(S_2\) is a priori knowable.

The upshot is that Soames fails to explain how the a priori knowability of \(S_1\)\(\rightarrow\)\(S_2\) etc. imposes a requirement upon what must be known by someone who understands “\(O\)”. All that Soames establishes is that if there is complexity all the way down, then there is an indefinite potential for unpacking \(O\)’s complexity, a potential that can be realised by actually coming to know a priori \(S_1\)\(\rightarrow\)\(S_2\), \(S_2\)\(\rightarrow\)\(S_3\) etc. This might be a surprising view to hold. But since Soames hasn’t shown that speakers would have to actually exhaust (per impossibile) the potential for unpacking \(O\)’s complexity in order to grasp “\(O\)”’s meaning, Soames sheds no light upon Wittgenstein’s claim that if there was complexity all the way down, it would be impossible to say something about \(O\) (or any other object). So it’s hard to see that Soames succeeds in striking up a conversation with Wittgenstein rather than talking past him.

Where Soames has gone adrift is failing to factor in Wittgenstein’s own insistence that a non-terminating sequence of meaning-truth dependencies would make it impossible to “form”, or more literally “draw up” [“entwerfen”], “a picture of the world (true or false)” (2.0212). By “picture of the world (true or false)”, Wittgenstein doesn’t simply mean “bearer of truth or falsity” but points us further into the interior of the Tractatus where a more demanding notion of a proposition and what it is to grasp a proposition awaits us – Wittgenstein’s picture theory. It’s because a non-terminating sequence of meaning-truth dependencies is incompatible with the possibility of a proposition in this more demanding sense that Wittgenstein concludes that there cannot be complexity all the way down (as I argue in 2018, 188–190).

Let me elaborate briefly upon this alternative interpretation. When we read further into the Tractatus we find that a proposition is a complete picture of reality in the sense that when a speaker understands a proposition, they have an exact knowledge of how objects must be arranged for that statement to be true or false and which arrangements of them are thereby left open. And this is information a speaker can uptake with effortless facility: “The proposition is a picture of reality, for I know the state of affairs presented by it, if I understand the proposition. And I understand the proposition, without its sense having been explained to me” (4.021). So a speaker must already actually know everything she/he needs to know to understand how things must be arranged for a proposition to be true even if the proposition isn’t one she/he has heard before. But a speaker couldn’t have knowledge of what it takes for a proposition to be true (or false) and what is thereby left open if she/he had per impossibile to check and see whether a non-terminating sequence of meaning-truth dependencies was satisfied for every expression of their language. A speaker wouldn’t be in a position to know straightaway that the expressions of their language were meaningful but only have a supertask ahead of them. As finite agents, speakers could never confirm that more than an initial segment of the sequence was satisfied, so never be in a position to exercise the consummate facility with language with which Wittgenstein credits speakers.

By contrast to Soames’ account, this interpretation has the merit of making immediate contact with what speakers are required to know to understand a language and it makes sense of Wittgenstein’s argument at 2.0211-2.0212 in the wider context of Wittgenstein’s commitment to the picture theory. It’s a further consequence of this interpretation that what Soames describes as the “notorious Tractarian collapse of the modalities” plays no significant role in Wittgenstein’s argument – Soames’ original mistake was to read the Tractatus through “Kripke goggles.”

I have concentrated upon one interrogative episode of New Vision to give a representative impression, but I might have taken issue with other episodes where, it seems to me, Soames’ arguments falter for lack of engagement with the historical texts. Consider, for example, his dismissal of the Tractarian conception of a proposition as a propositional sign in its projective relation to the world in favour of his own cognitive act type theory. Or his criticism of the Aufbau that Carnap failed to realise that statements expressed in purely logical vocabulary have no empirical content when, Soames has forgotten, “\(\exists x \exists y (x \neq y)\)” consists of purely logical vocabulary but remains verifiable or falsifiable depending on how many things there are.