# Wittgenstein on the Chain of Reasons

Author: Matthieu Queloz
Published in: *Wittgenstein-Studien* 7 (1): 105–30. 2016. [doi:10.1515/witt-2016-0108](https://doi.org/10.1515/witt-2016-0108)
DOI: [10.1515/witt-2016-0108](https://doi.org/10.1515/witt-2016-0108)
Canonical entry: https://www.matthieuqueloz.com/entries/wittgenstein-on-the-chain-of-reasons/
Published PDF: https://philarchive.org/archive/QUEWOT

Machine-readable text companion generated from the PDF. Page markers follow the journal pagination.

[p. 106]

rich and nuanced overview of the roles played by reasons in how we act and speak, developing a conception of reason and rationality that went on to form an enduring buttress to his later thought.

The key to Wittgenstein’s conception of reasons lies in understanding a guiding image of this middle period: the image of the chain of reasons. That “the chain of reasons has an end” (PI 2009: 326) is perhaps one of Wittgenstein’s best-known claims; yet it is not in the Investigations, but in his work from the early 1930s that he most explicitly engages with this idea and with its correlate, the expectation that one should always be in a position to provide reasons. In many scattered and apophthegmatic remarks, he provides grounds for thinking that this expectation is misguided, tracing it to the tendency to think of chains of reasons on the model of chains of causes. My aim here is to elucidate Wittgenstein’s own conception of the chain of reasons and to contrast it with rival conceptions by reconstructing Wittgenstein’s answers to three questions: (i) what reasons are, (ii) why chains of reasons must come to an end while chains of causes must not, and (iii) how this relates to the distinction between explanation and justification.

## 2. Giving Reasons

What are reasons? Characteristically, Wittgenstein transposes this Socratic question into a pragmatic key: what do we do when we give reasons? His answer is composed of four interrelated ideas: (i) giving reasons consists in making thought and action intelligible by delineating the reasoning routes that lead to them; (ii) whether something is a reason and what kind of reason it is depends on its role on occasions of utterance; (iii) citing something as a reason does not characterise the item cited psychologically, in terms of the psychological or causal connections it stands in, but rationally, in terms of the rational connections it stands in; (iv) the norms of reasoning delineating what rational connections an item stands in depend on the language game in which the item is adduced as a reason.

First, the concept of a reason is tied for Wittgenstein to the concept of reasoning, the “transition from one proposition to another” (RFM 1978: 39; cf. AWL 1979: 4 f.; BBB 1958: 14 f.; LA 1966: 21 f.). This holds for theoretical as well as for practical reasoning: “An inference is a transition to an assertion; and so also to the behaviour that corresponds to the assertion. ‘I draw the consequences’ not only in words, but also in deeds” (PI 2009: 486). The importance of the seemingly innocuous claim that reasons are tied to reasoning lies in the fact that if reasons are to be capable of being the sort of entities we reason our way from Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 107]

and to, i. e. of being premises and conclusions in theoretical and practical reasoning, they must be capable of being expressed in propositional form. Otherwise, some have argued, the connection between reasons and reasoning is severed.i But for Wittgenstein, the claim that reasons must be capable of being expressed propositionally is implicit in his concern with reasons as the correlates of demands for reasons and explanations, as the kinds of things that can be given in answer to “Why?”-questions. Reasons are considerations expressible in propositional form that serve to guide, justify and explain what we believe, desire and do.

In the Blue Book, Wittgenstein gives two characterisations of what giving a reason consists in: “Giving a reason for something one did or said means showing a way which leads to this action” (BBB 1958: 14); soon thereafter, he writes: “Giving a reason is like giving a calculation by which you have arrived at a certain result” (BBB 1958: 15). Both characterisations owe much to traditional imagery. The first characterisation, which we might call a topological one, is reminiscent of ancient rhetoric in that it links the capacity to give reasons – Plato’s lógon didónai (1930: 510c7) – to a spatial image, much as Aristotle does in the Topics with the tópoi, the commonplaces of reason from which arguments are derived (1989: 100a25 – 27). It also echoes locutions such as “ways of thought”, “deliberative routes”, or “steps in reasoning”, and Wittgenstein himself refers to the giving of reasons in several places as the indication of “the road one walks” (BBB 1958: 14), “the road you went” (LA 1966: 21), “the step in the calculus” (AWL 1979: 4) and “the process by which” (LA 1966: 22) one arrives at an answer to a “Why?”-question. The second characterisation, which compares reasoning to calculating, also harks back to classical notions: the word “reason” shares its etymology with the word “rational”, which has its roots in the latin verb reri, one meaning of which is “to calculate”. Both characterisations indicate that giving a reason serves to make a thought or an action intelligible by describing the train of thoughts which leads to it. Even if it is true that “in the beginning was the deed” (OC 1969: 402), deeds need interpretation. Mere behaviour, in order to be intelligible as a form of action, must be interpretable as an expression of intentions, as a set of movements conducted in the light of reasons. As long as those reasons escape us, the behaviour will appear puzzling. Wittgenstein gives the example of someone who “gets angry, when we see no reason for it” (LW 1982: 192). This behaviour can become intelligible to me, however, when I “see the trains of thought and I know how they lead � An example is Alvarez (����: ��).

Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 108]

to his actions” (LW 1982: 193). When this happens, “I understand his actions”, and he stops “being a riddle to me” (LW 1982: 192– 94).

Second, Wittgenstein’s characterisation of reasons is situated not at the ontological level, but at the semantic level: it focuses not on what reasons are or what their intrinsic nature is, but on their role in language use. Whether something is a reason depends on its role on occasions of utterance. What underpins this approach is the realisation that “what characterizes an order as such, or a description as such, or a question as such, etc., is […] the role which the utterance of these signs plays in the whole practice of the language” (BBB 1958: 103 – 4). Wittgenstein is interested in what he calls the “function of giving reasons” (RPP 1980b: 314). He rejects the idea that reasons have an intrinsic nature to be investigated by philosophy independently of the role they play in language. Reasons are adduced in answer to “Why?”-questions to serve a particular purpose, namely to delineate a path of deliberation or calculation towards what they are reasons for.

At the core of this characterisation of reasons is the idea that reasons are relational: what is characteristic of reasons is not what one mentions in giving them, but the relation between what is mentioned and what it is mentioned for. For example, one might say that the fact that the fridge door has been left open is a reason to think that the ice will melt. As Wittgenstein approaches the subject of reasons, it is not something peculiar about the fact that the fridge door has been left open that makes it a reason for anything – it is a fact like any other. But in adducing that fact as a reason, one characterises it in terms of a relation, namely the relation of being a reason to believe something. In citing something as a reason, one characterises the item cited in terms of its role in a system of rationally related propositions and actions – its role in what Wittgenstein already thought of at the time as an “activity” (BT 2005: 230), a “game” (BT 2005: 231), a “system of communication” (BBB 1958: 81) or a “language-game” (BBB 1958: 17; BT 2005: 4): “If one asks for the reason behind an individual act of thought (act of calculation), the answer one gets is an analysis of a system to which the act belongs” (BT 2005: 231). In the (not fully dependable) lecture notes taken by his students, Wittgenstein gives the example of someone who is asked: “Why did you write 6249 under the line?”, and who answers with: “I arrived at it by this multiplication” (LA 1966: 21). As Wittgenstein points out, this means that “I passed through such and such a process of reasoning” (LA 1966: 21), and “to ask for a reason is to ask how one arrived at the result” (AWL 1979: 5). A reason is given when one “goes back one step in the calculus” (AWL 1979: 4), making the concluding step in the train of reasoning – the result – intelligible by indicating the step that led up to it: “A reason is a step preceding the step of the choice” (BBB 1958: 88). On this account, it is only within and as a Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 109]

part of a “system” (BT 2005: 231) or a “calculus” (BT 2005: 267) that something can have the significance of a reason. To cite something as a reason is to characterise the item cited relationally by redescribing it in terms of its role as a step in reasoning towards some proposition or action.

Third, Wittgenstein spells out this role-oriented view of reasons not in psychological or causal terms, but in normative ones: what is cited as a reason performs its function of indicating the reasoning process leading to an action not in the sense that it specifies the process by which, causally or psychologically, one in fact got there through the psychological or causal connections the item cited stands in, but in the sense that it shows how, rationally, one gets there through the rational connections the item cited stands in. Giving a reason for something one did does not provide a psychological answer to the question of “how one arrived at the result”, but rather “teaches us connections in a calculus” (BT 2005: 267), connections which are determined independently of us: “reasons for accepting [a] proposition are not personal matters, but parts of the calculus to which the proposition belongs” (BT 2005: 267). Only in “some cases” does giving a reason mean “telling the way which one has gone oneself” (BBB 1958: 14); “in others it means describing a way which leads there and is in accordance with certain accepted rules” (BBB 1958: 14). The latter cases Wittgenstein calls justifications post hoc (BBB 1958: 14). Hence, in saying that a reason is a step in reasoning, the reasoning in question is to be understood as including not just the reasoning one actually went through, but also the reasoning one could have indicated ex post actu if asked. This distinction does much to avoid what Ryle called “the intellectualist legend”, according to which “whenever an agent does anything intelligently, his act is preceded and steered by another internal act of considering a regulative proposition appropriate to his practical problem” (2009: 18 – 19). Wittgenstein separates reasons from the psychological processes that led to an action by introducing an element which is precisely not “personal”, but significantly independent of the agent whose reasons are in question. This sense in which giving a reason refers to something independent of the agent is brought out by Wittgenstein’s distinction between ways which one has gone oneself, and ways which are “in accordance with certain accepted rules”, but which one has not gone oneself. Consider the following passages: Giving a reason sometimes means “I actually went this way”, sometimes “I could have gone this way”, i. e. sometimes what we say acts as a justification, not as a report of what was done, e. g. I remember the answer to a question; when asked why I give this answer, I gave a process leading to it, though I didn’t go through this process. (LA 1966: 22) Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 110]

The question “For what reasons do you believe this?” might mean: “From what reasons are you now deriving it (have you just derived it)?” But it might also mean: “With hindsight, what reasons can you give me for this supposition?” (PI 2009: 479) The wording here is revealing in that the possibility of “showing” and “describing” the way one went or “could have gone” presupposes that the ways leading to the action are independently there to be shown or described. As Wittgenstein puts it in the Big Typescript: “We are guided, our step has been marked out ahead of time” (BT 2005: 186). If giving reasons is to be understood as showing a way “which leads there”, this presupposes the idea of an independently obtaining rational connection between two items, of a set path allowing one to move from one item to the other in the reasoning process. Asked to give reasons for what one said or did, citing a string of random considerations and presenting them as the reasoning one went through is, by itself, insufficient. It only constitutes the giving of a reason if the considerations cited link up to form a reasoning route that is both rule-conforming and conducive to the statement or action to be justified. What is brought out here is that “to give a reason” typically functions as what Ryle calls an “achievement” or “success” verb (2009: 114), which refers not just to actions, but to “suitable or correct” (2009: 115) actions: actions whose performance is at least minimally successful as measured by some standard – in this case, the network of permissible reasoning routes determining what counts as a reason for what. It is this normative requirement to accord with an independent standard which Wittgenstein presumably points to when he speaks of “a way which leads there” (BBB 1958: 14, emphasis added), independently of whether one took it or not. As he later puts it, “justification consists in appealing to an independent authority” (PI 2009: 265). To give a reason is to ask for a way which “leads” to what is done not in the mechanical sense in which it causally accounts for it, but in the normative sense in which it rationally accounts for it as what is to be done.

Fourth, the norms of reasoning which delineate what rational connections an item stands in are context-sensitive: they depend on the language game, that is, on the interactive and rule-guided complex of activities and languageuse within which the item is adduced as a reason. It has often been emphasised – for example by Blackburn (2010: 284) – that “X is a reason” must be understood relationally, that is, not as “X is a reason tout court”, but as “X is a reason for Y”. Wittgenstein, by contrast, emphasises the dependence on context, that is, on the language game in which the reason is given. If, having said or done Y, I am asked for my reason, answering X will constitute giving a reason for me to say or do Y only if X does in fact, according the norms of the language game the reason is given in, enable me to make the transition in reasoning from X to Y. We Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 111]

may therefore say that “…is a reason…” is a four-place relational predicate. Let us sharpen this with a definition: DEF: X is a reason for S to perform an action Y iff X is such as to enable the transition in reasoning from X to Y according the norms of reasoning that apply to S in context C.

This means that what functions as a reason for what is a question which, for Wittgenstein, can only be answered in the context of a given language game. Language games play a fundamental role in contexts of reason-giving, because it is within language games that reasons are given, and it is only with language games that they come into being. As Wittgenstein puts it: “Not until there is a language game are there reasons” (RPP 1980b: 689), and a “reason is a reason only inside the game” (AWL 1979: 4). This dependence of reasons on language games derives from the fact that giving reasons requires one to put forward conceptually articulated propositions as reasons. Even seemingly less linguistically demanding ways of giving reasons, such as pointing at a colour, require the concept of colour to provide reasons (LW 1992: 64). Yet our concepts only have meaning in the context of the language games in which they have a use – “a concept is in its element within the language game” (RPP 1980b: 632). Thus, mastering the permissible moves in a language game by mastering the concepts that are at play in it is an essential part of coming to understand what counts as a reason for what within that language game (OC 1969: 18). And what counts as a reason for what is in turn causally determined by – though it need not logically reduce to – what we treat as a reason for what. It thus rests on our ways of acting, which are themselves based on our primitive (instinctive, unreflective, prelinguistic) reactions: “instinct comes first, reasoning second” (RPP 1980b: 689), and “it is our acting, which lies at the bottom of the language-game” (OC 1969: 204). For these reasons, language games form what Wittgenstein at one point calls “the base” (OC 1969: 609) from which we give reasons.

One straightforward consequence of making reasons dependent on language games in this manner is that the language games not only enable, but also limit the giving of reasons. This aspect is most emphatically brought out by Wittgenstein in On Certainty, where it is treated in terms of the importance of the “background” or “world-picture” in determining what counts as a reason for what: my “world-picture”, says Wittgenstein, “is the substratum of all my enquiring and asserting” (OC 1969: 162); it is “the inherited background against which I distinguish between true and false” (OC 1969: 94), a “system” which “belongs to the essence of what we call an argument” (OC 1969: 105). Even what looks like it must be uncontroversially and universally accepted as a reason depends on there being a certain kind of practice in place for it to be a reason. In On Certainty, Wittgenstein wonders whether one could possibly deny that the “propositions Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 112]

of physics” (OC 1969: 608) constituted a “good ground” (OC 1969: 608) or a “telling reason” (OC 1969: 609): Supposing we met people who did not regard that as a telling reason. Now, how do we imagine this? Instead of the physicist, they consult an oracle. (And for that we consider them primitive.) Is it wrong for them to consult an oracle and be guided by it? – If we call this “wrong” aren’t we using our language-game as a base from which to combat theirs? (OC 1969: 609) I said I would ‘combat’ the other man, – but wouldn’t I give him reasons? Certainly; but how far do they go? At the end of reasons comes persuasion. (Think what happens when missionaries convert natives.) (OC 1969: 612) In these passages, Wittgenstein can be seen to draw some of the consequences that his conception of reasons has for his conception of reason and rationality. If an utterance constitutes giving a reason in virtue of the ways in which it functions in certain contexts of use, then a particular kind of constraint will be provided on what someone can discover that he or she has reason to do. If giving and asking for reasons are activities which can only exist within a language game, then any reason one may give depends, in order to be accessible to the addressee as a reason, on there being some common ground, a shared complex of activities, purposes and language-use in which there is minimal agreement on what counts as a reason for what. Reasons go, as Wittgenstein intimates, only so far. Absent this requirement, it is not the reasons that must be altered, but the conditions under which they are recognised as such – something that is done not by providing more reasons, but through persuasion or conversion.

## 3. The Chain-Method

In what sense do reasons link up with each other to form chains? The image of the chain of reasons makes its appearance in Wittgenstein’s writings in the early 1930s, and occurs at least fourteen times in his middle period. Talk of a chain of reasons highlights the fact that a reason may both provide justification for a reason and itself stand in need of justification by a reason. A reason both supports and is supported by further reasons with which it interlocks, forming a chain of reasons which can be followed up, reason after reason, with iterated “Why?”questions. While Wittgenstein resorts to the chain image in earlier writings already (TLP 1961: 2.03; TLP 1961: 4.22), it is only in the 1930s that he connects it to reasons. Its importance lies in suggesting distinct entities made to hang together not by means of some additional, connective medium, but by being directly locked into each other.

Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 113]

Wittgenstein’s extensive employment of the chain image in the 1930s does not come from nowhere, and it is worth stepping back from its application to reasons for a moment to consider the ways in which it informs Wittgenstein’s methodology more generally. As a means of rendering perspicuous the organisation of a range of clearly individuated but nonetheless interrelated elements, the chain image was highly influential in Western thought – as was its sister image of the ladder, which Wittgenstein employs in the Tractatus (TLP 1961: 6.54). For centuries, the scala naturae and the Great Chain of Being were the traditional representations of the continuity, gradation and hierarchical ordering of all things (Lovejoy 1970: 185). This imagery was still very much in use in the days of Goethe, whose thought it underpins both in his literary works (Wyder 1998: 63) and in his so-called “morphological method”, which revolves around the search for Urphänomene or primordial phenomena, the last links in serial arrangements of phenomena which render their interconnections and their overall organisation perspicuous (Nisbet 1972: 39; Glock 1996). The following passage from Goethe’s Theory of Colours brings out the connection between primordial phenomena and the chain of interlocking considerations which can be followed up to their last link to shed light on the whole: […] everything is gradually arranged under higher rules and laws, which, however, are not to be made intelligible by words and hypotheses to the understanding merely, but, at the same time, by real phenomena to the senses. We call these primordial phenomena, because nothing appreciable by the senses lies beyond them, on the contrary, they are perfectly fit to be considered as a fixed point to which we first ascended, step by step, and from which we may, in like manner, descend to the commonest case of every-day experience. (Goethe 1840: 175) Not only was Wittgenstein familiar with Goethe’s writings from an early age, but around 1930, at a time just preceding the occurrence of this figurative expression in Wittgenstein’s writings, he is also known to have been a predominantly enthusiastic reader of Oswald Spengler, who explicitly acknowledges his profound methodological debt to Goethe (Schulte 1990: 33 – 34). In Wittgenstein’s hands, these ideas become what Joachim Schulte calls the “chain-method”, the arrangement of things “in such a way that they can be seen to hang together and as it were explain their own characteristics by means of their positions in this chainlike arrangement” (Schulte 2003: 69). Where this method is successful, it produces what Wittgenstein calls “the understanding that consists in ‘seeing connections’” (PI 2009: 122). In conversations recorded by Friedrich Waismann, Wittgenstein said: “Here seeing matters essentially: as long as you do not see the new system, you have not got it” (WWK 1979: 123). To connect two things, we do not need a tertium quid which Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 114]

makes the connection. “Things must connect directly, without a rope, i. e. they must already stand in a connection with one another, like the links of a chain” (WWK 1979: 155). At some level, there must be such an immediate connection, as Wittgenstein argues by way of a regress argument reminiscent of Plato’s third man argument in the Parmenides (1996: 132a – b): “If a connection between two things always consists in a mediation by a third thing that is connected to each of the two, then two things are never connected to each other.” But that doesn’t mean: A connection is never achieved; it’s just that it makes no sense to say “A connection is achieved” (and therefore neither does its opposite). That is, it makes no sense to talk about “connection”; the concept of “connection” has in no way been explained. (BT 2005: 399) Connections – including the kind of rational connection we are interested in – must be immediate at some level, because assuming the contrary, namely that any connection must be made by some third term, makes it impossible to see how any two things could ever be connected: if two connected things A and B could only be connected by way of some third term C connected to each of the two, the connection between A and C would in turn have to be explained by way of a mediating term D, itself connected to each of the two in a way that would in turn require mediation, and so on. The resulting regress robs the very notion of a connection of its intelligibility. As Wittgenstein hastens to point out, this does not demonstrate the impossibility of ever bridging the gap between two things, but it shows the underlying notion of a connection to be misconceived. If talk of connections makes any sense, it must do so under the assumption that where there are connections, some of them are immediate, like the connections between the links of a chain. The connections between the links of chains of reasons are made in our reasoning practices: it is through the endorsement of certain inferential patterns in the “everyday practice” (PI 2009: 197) of reasoning that the connections between the links in the chain are effected. To make these connections apparent, no theory is required. Things only need to be arranged in a certain way. This is the chain-method.

## 4. Finite and Infinite Chains

Returning from the chain image as a methodological idea to its more substantive use in articulating conceptions of reasons, we find that Wittgenstein first deploys the chain image not to articulate his own view, but to describe a conception of reasons he opposes – this is the conception of reasons as forming an infinite chain: Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 115]

Now there is the idea that if an order is understood and obeyed there must be a reason for our obeying it as we do; and, in fact, a chain of reasons reaching back to infinity. (BBB 1958: 15) This picture of an infinite chain of reasons is what underlies the intuition that every “Why?”-question could, in principle, be answered by giving a reason. “We would like to give reason after reason after reason. Because we feel: so long as there is a reason, everything is all right” (BT 2005: 187). But when one fails to live up to the demand for reasons, the same intuition engenders the impression of groundlessness and arbitrariness. In order to avoid this impression, and in the conviction that it must be possible to give a reason, one strives to answer “Why?”-questions even where the chain of reasons has come to an end. And when one does give an answer, ones sometimes fails to see that what one has given is not a reason, but a cause: “When the chain of reasons has come to an end and still the question ‘why’ is asked, one is inclined to give a cause instead of a reason” (BBB 1958: 15). Such conflations of reasons with causes leave one in the grip of the expectation that there always has to be a justificatory answer to “Why?”-questions. Once our inability to give reasons for our practices becomes apparent, these practices appear groundless and arbitrary. If, however, we hold on to the realisation that actual chains of reasons are not infinite, there is nothing mysterious about the idea that the enterprise of working our way along these chains by reiterating “Why?”-questions will end where the chain begins. “If […] you realize that the chain of actual reasons has a beginning, you will no longer be revolted by the idea of a case in which there is no reason for the way you obey the order” (BBB 1958: 15). “The difficult thing here is not to dig down to the ground; no, it is to recognize the ground that lies before us as the ground” (RFM 1978: 333). Indeed, for Wittgenstein, it is a condition on the meaningfulness of the idea of a chain of reasons that it must come to an end somewhere. The notion of justification would lose its point if the chain of reasons were infinite, because the enterprise of justifying a statement or an action could never come to rest, and the justification would never be complete.

The image which leads us astray here – the image of the infinite chain of reasons – itself has its roots in the topological metaphor of deliberative routes (BBB 1958: 15), which suggests that the possibility of giving a reason is similar to the possibility of retracing one’s footsteps. Since the possibility of doing the latter nearly always seems intelligible, one is coaxed into thinking that one should likewise indefinitely be able to give a reason for what one says or does. But Wittgenstein rejects the idea that chains of reasons are infinite in Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 116]

this way. The misleading impression to the contrary stems from the following confusion: If […] you had said, “wherever you are, you could have got there from another place ten yards away; and to that other place from a third, ten yards further away, and so on ad infinitum”, if you had said this you would have stressed the infinite possibility of making a step. Thus the idea of an infinite chain of reasons arises out of a confusion similar to this: that a line of a certain length consists of an infinite number of parts because it is indefinitely divisible; i. e., because there is no end to the possibility of dividing it. (BBB 1958: 15) Just as it does not follow from the possibility of dividing up a distance to be covered into indefinitely many steps that it is actually covered in indefinitely many steps, it does not follow from there being an endless variety of conceivable chains of reasons that any actual chain of reasons is infinite. The mistake, encouraged perhaps by the topological idiom of deliberative routes, is to model chains of reasons, which cannot be followed up indefinitely, on causal chains of events, which usually can.

A similar idea is invoked at the beginning of the Philosophical Investigations, where Wittgenstein urges his interlocutor to refrain from saying: “There isn’t a ‘last’ explanation” (PI 2009: 26). Drawing an analogy between a chain of reasons and a row of houses, Wittgenstein argues that saying that “there isn’t a ‘last’ explanation” is like saying: “There isn’t a last house in this road; one can always build an additional one” (PI 2009: 26). Here also, the idea is that it does not follow from the possibility of building indefinitely many houses that the row actually consists of indefinitely many houses. There is, at any one moment, a last house in the road, even though which house counts as the last one might well change over time. And the same is true of the “ancient city of language” (PI 2009: 18). There is, at any one moment, something which we accept as the last explanation or justification in a given context, even though we might always come to develop a further one.

On Wittgenstein’s view, then, chains of reasons are bound to peter out; the only question is where they do so, which depends on the language game they are given in. He illustrates this point by comparing two language games: in the first, one determines the colours of objects using a sample placed “beside objects to test whether the colours match” (PG 1974: 96); in the second, one specifies the colours of objects “without a sample”, but “in accordance with the words of a word-language” (PG 1974: 96), that is, using nothing but colour-words. In both variants, there comes a point at which the demand for reasons becomes senseless. The only question is where that point is: Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 117]

Suppose I am now asked “why do you choose this colour when given this order; how do you justify the choice?” In the one case I can answer “because this colour is opposite the word ‘red’ in my chart.” In the other case there is no answer to the question and the question makes no sense. But in the first game there is no sense in this question: “why do you call ‘red’ the colour in the chart opposite the word ‘red’”? A reason can only be given within a game. The links of the chain of reasons come to an end, at the boundary of the game. (Reason and cause.) (PG 1974: 96 – 97) While the choice of colour in the first game (played with a sample) can be justified by pointing to the sample, the demand for justification in the second game (played only with words) already fails to make sense – there is no reason why we call a certain colour “red”; we might give a causal explanation of this fact by referring to contingent facts of linguistic history or human physiology, but this would only be to name causes of our use of language, not to adduce reasons for it.j If one got the impression that the colour sample in the first game secures a form of objectivity and prevents the chain of reasons from ending, however, this impression dissipates as soon as the “Why?”-question is reiterated: the sample only delays the question’s loss of meaning by one step, because as soon as we ask why we call the sample “red”, we again reach the limits of the language game of colours. Our calling the sample “red” may justify other statements and actions within the game, but it cannot itself be justified within that game. We widely agree to call such a sample red, and our doing so makes it possible to develop an elaborate language game based on that fact. In this sense, it is true that we use the word without justification. But, as Wittgenstein reminds us in the Philosophical Investigations, to “use a word without a justification does not mean to use it wrongfully” (PI 2009: 289). The chain of reasons ends where the language game begins.

Precisely because the chain of reasons available within the language game only holds in virtue of the language game’s being played, however, it cannot serve to justify the language game. “The primitive language game we originally learned needs no justification” (RPP 1980b: 453), and the very idea of providing reasons for our language games is misguided, because “not until there is a language game are there reasons” (RPP 1980b: 689): � To be precise: investigating the causal history of our colour vocabularies might reveal them to be contingent in two respects: one form of contingency concerns the question whether we could have used a different colour term; the other concerns the question whether we could have carved up the colour space differently. The first is rather trivial, while the second has given rise to extensive debates; see Glock ����.

Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 118]

Giving grounds, however, justifying the evidence, comes to an end; – but the end is not certain “propositions” striking us immediately as true, i. e. it is not a kind of seeing on our part; it is our acting, which lies at the bottom of the language-game. (OC 1969: 204) Insofar as our ways of reasoning can be said to have a foundation, that foundation lies not in some indubitable or self-evident propositions, but in our shared communal practices. In virtue of their matter-of-factual status, the language games embedded in our practices are able to play their foundational role as “last court of appeal” (PI 2009: 230) when it comes to giving reasons. We have no choice but to “accept the familiar language game” (RPP 1980b: 453). A language game is “not based on grounds. It is not reasonable (or unreasonable). / It is there – like our life” (OC 1969: 559). This is an insight which runs fairly constantly through Wittgenstein’s writings from the 1930s onwards. Already in the remarks published as the Philosophical Grammar, he confronts an outraged interlocutor who exclaims: “Surely the rules of grammar by which we act and operate are not arbitrary!” (PG 1974: 110). Wittgenstein responds: “Very well; why then does a man think in the way he does, why does he go through these activities of thought? (This question of course asks for reasons, not for causes.)” (PG 1974: 110). It is in the very attempt to answer it that the question’s fruitlessness emerges: “Well, reasons can be given within the calculus, and at the very end one is tempted to say ‘it just is very probable, that things will behave in this case as they always have’ – or something similar. A turn of phrase which masks the beginning of the chain of reasons” (PG 1974: 110 – 11). Wittgenstein is adamant that any attempt to answer the question by giving reasons for the way we think and count certain things as reasons from some Archimedean point outside our language games (or outside the calculus, as he still thought of it at the time) is bound to fail, since it must either presuppose what it wants to justify or fail to be intelligible. It must either “lead us from one such game to another” without managing to move beyond the rule it wants to justify, or, where it does step “outside the province of these games” (PG 1974: 111; BT 2005: 229 – 31), it will no longer be recognisable as a justification. If we conflate reasons and causes, the fact that we can nearly always answer “Why?”-questions by forming conjectures about the causal chains that led up to our actions can generate the expectation that we should always be able to produce reasons for them. When the expectation fails to be met, our understanding of our own actions appears defective. If, on the other hand, we distinguish between the “causal” and the “logical dependence” (PI 2009: 220) of an action, between its causes and the reasons for it, it becomes clear that “Why?”-questions demanding reasons come to an end not when they are answered, but when they become senseless. Demands for reasons become senseless at the point at which Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 119]

we reach the limits of the language game, a point marked by the rules that constitute the language game and enable demands for reasons to be meaningfully raised within it. Sometimes, a reason that forms a terminus in one context may be supportable by further reasons in another context. But as long as one does not switch into another language game in which the question of why the rules are as they are can meaningfully be raised, the demand for reasons will gain no traction. And even when one does switch into another language game, “reasons will soon give out” (PI 2009: 211). One will eventually reach a case where it is simply not possible to switch into yet another language game: just as there is bound to be a last house in the road at any one point in time, there are bound to be reasons which cannot be given support by further reasons, neither in the language game one started out with, nor in any other. This is most perspicuous in the case of “Why?”-questions demanding reasons for the norms of reasoning themselves: we cannot give reasons for the way we are guided by reasons, since any such justification must presuppose what it is meant to justify if it is to be accessible to the addressee.

In light of this, the right attitude to take when the chain of reasons comes to an end is to give what Robert Fogelin has called a “defactoist” (2009: 27) response: deferring to what we, in practice, accept as a reason. An example is Wittgenstein’s concern with the question whether we legitimately generalise from past experience. Suggestions to the effect that we may only do so when we understand the mechanisms underlying those experiences are brushed aside in the spirit of Hume with the remark that such an understanding would itself have to be based on past experience, which raises concerns of circularity (BT 2005: 399). Instead, Wittgenstein stresses the fact that we “simply call” a certain kind of statement about past experience a reason for certain kinds of predictions: Now if one asks: But how can a previous experience be a reason for assuming that later on this or that will happen? – the answer is: What general concept of a reason for such an assumption do we really have? Well, this kind of a statement about the past is simply what we call a reason for assuming that this will happen in the future. – And if you are surprised � The expression “switching into another language game” should be taken cum grano salis. Language games do not come neatly packaged and delineated from each other. One cannot make a list – even an open one – of all the language games our language consists of. One can describe a language game that is clearly distinct from other language games by describing a model, but since it is only a model, it does not compete with other language games. In fact, one language game may cross or shadow into another, forming a “complicated network of similarities overlapping and criss-crossing” (PI ����: ��). Talk of language games thus primarily serves explanatory purposes.

Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 120]

that we are playing such a language-game, then I refer you to the effect of a past experience (to the fact that a child who has been burned fears fire). (BT 2005: 396) Wittgenstein’s defactoist response is not the summoning of a further reason, but a reference to our communal practices. The fact that we attribute such authority to past experience is something which, on pain of circularity, we cannot justify. We can ask why we reason as we do, but, taken as a demand for reasons, the question is misguided. Wittgenstein writes: “What counts as a reason for an assumption can be given a priori and determines a calculus, a system of transitions” (PG 1974: 110). That is to say, the criteria for what counts as a reason for what are fixed from the very beginning of the formation of a calculus and settle its mode of operation. He continues: “But if we are asked now for a reason for the calculus itself, we see that there is none” (PG 1974: 110). Only if the “Why?”-question is taken as a demand for causes can we perhaps answer – adopting an aetiological perspective on our reason-giving practices – that what, causally, makes something a reason is the fact that we treat it as a reason; and, still from an aetiological perspective, the fact that we treat it as a reason will hardly seem surprising in many cases. The consequences of a language game may indeed explain its emergence: given our needs and interests, and given the way the world is, the child that burnt itself was no doubt better off if it feared fire as a result. But this does not justify the language game. This seems to me to be the force of Wittgenstein’s remark in On Certainty: “This game proves its worth. That may be the cause of its being played, but it is not the ground” (OC 1969: 474). While it provides no justification, appealing to the consequences of a language game does go some way towards answering the worry that “the calculus” is “something we adopt arbitrarily” (PG 1974: 110): we adopt it no more arbitrarily “than the fear of fire, or the fear of a raging man coming at us” (PG 1974: 110). Just as conventional practices in general are elaborations of rudimental patterns of natural behaviour, our linguistic practices are rooted in primitive reactions � This suggestion was elaborated in a Wittgensteinian spirit by Jane Heal (����: ��� – ��) and by Robert Brandom under the title of a “Wittgensteinian conception of the normative” (Brandom ����: ��). According to these approaches, it is the endorsement of certain inferential patterns in our implicitly normative practices which ultimately determines the inferential relations we make explicit using the vocabulary of reasons, i. e. in terms of what is a reason for what. Crucially, however, this is not to say that norms are reduced to mere patterns in behaviour or mere regularities. Following Wittgenstein (PI ����: ��� – ��), Brandom emphasises that what counts as a “regular” performance is itself a normative question, and thus the appeal to regularities (individual or communal) either only postpones the normative question of how to understand the distinction between what is done and what is to be done, or it loses the distinction by assimilating the normative to a dispositional understanding of what is done (see Brandom ����: �� – ��).

Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 121]

which, far from being “the result of thought”, are “prototypes” of a way of thinking (RPP 1980a: 916).

## 5. A priori Reasons and Empirical Reasons

Wittgenstein’s claim that what “counts as a reason for an assumption can be given a priori and determines a calculus, a system of transitions” (PG 1974: 110; cf. BT 2005: 181) would seem to require qualification, however. In many cases, what counts as a reason for what can indeed be given a priori, namely whenever the inferential relations in question are given in virtue of the conceptual contents of the propositions involved. That one sees lightning now, for instance, is a reason to think that one will hear thunder soon (at least on a conception of lightning for which it is part and parcel of the concept that it creates sound). One knows in virtue of mastering the concepts involved in this inference that one is a reason for the other. Likewise, one can say a priori – in virtue of having mastered the language – that past experience provides reasons for certain beliefs about the future. Hence Wittgenstein’s remark in the Investigations: “How do I recognize that this colour is red? – One answer would be: ‘I have learnt English’” (PI 2009: 381).

In other cases, what does or does not count as a reason cannot be given a priori. Whether something is a reason for something else also depends on what other things are true, on laws of nature and contingent facts, that is, on what claims can serve as collateral premises in inferences. Whether tight monetary policy is a reason to think that there will be an economic crisis not only depends on a great number of other factors, but is something which must be determined empirically. Finally, and crucially in this context, whether the uniformity of nature will continue to hold is also an empirical question, as is the question whether we shall continue to be able to successfully generalise from past experience. In a discussion which prefigures Peter Strawson’s writings on inductive reasoning, Wittgenstein introduces the distinction between, on the one hand, the idea that the fact of our counting past experience as a reason can be given a priori and cannot be justified, and, on the other hand, the fact that whether a particular past experience counts as a reason depends on the context, on other experiences one has made, and on the sense one can make of a particular experience as part of an overall picture. The latter are what we might call em- � Strawson argues in his influential Introduction to Logical Theory that it is a non-contingent, a Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 122]

pirical reasons, which are reasons not in virtue of their role in a calculus or a language game, but in virtue of their role within an overall empirical picture. Wittgenstein discusses the example of one who is asked: “Why do you believe that this movement of your arm will be accompanied by pain?”, and who answers: “Because sometimes it produces it and sometimes it doesn’t” (BT 2005: 395). Wittgenstein waves off: “But that isn’t any reason for your assumption. […] You seem to be giving the cause (the psychological cause) for your assumption, but not the reason” (BT 2005: 395). Only within a certain context does the observation form a reason for the belief: “Why do you believe that that will happen?” – “Because I’ve observed it several times; and evidently this is how it happens: …” (and now an extensive hypothesis is laid out). But this hypothesis, this total picture, must make sense to you. Here the chain of reasons does not continue. – (It would be more correct to say that it comes to an end.) (BT 2005: 396) In this passage, Wittgenstein points to the fact that a particular experience one had in the past forms a reason only as part of an “extensive hypothesis” or a “total picture”. One might, for instance, have hurt one’s arm in an accident and been diagnosed with an elbow sprain. In this context, that is, given our understanding of human anatomy, injury, bodily movements, pain and their relations to each other, the observation that my arm hurt in the past is indeed a reason to believe that it will do so in the future; but not, as it were, absolutely, in any imaginable situation and in isolation from any other belief. It makes sense to us as part of an overall picture.

As with language games, once we have accepted an observation as providing a reason for a belief within an overall picture, we cannot go on to provide reasons for the overall picture itself. This overall picture “must make sense” (the German original reads: “muss einleuchten”). As Wittgenstein writes, “[h]ere the chain of reasons does not continue”, and it would be “more correct to say that it comes to an end”. The “more correct” is somewhat puzzling in the English translation, as it implies that, strictly speaking, it is still false to say that the chain of reasons comes to an end, even though Wittgenstein repeatedly uses priori matter that induction is rational or reasonable, while it is a contingent, factual matter that it is sometimes possible to form rational opinions concerning what specifically happened or will happen when the universe is such that induction will continue to be successful. The uniformity of nature is a contingent fact, and so is our ability to form rational opinions through inductive reasoning. But, as Strawson emphasises, “neither this fact about the world, nor any other, is a condition of the necessary truth that, if it is possible to form rational opinions of this kind, these will be inductively supported opinions” (Strawson ����: ���).

Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 123]

this formulation without qualifications. But a look at the German original (“eher könnte man sagen, daß sie sich schließt”) reveals the rather different idea that the chain closes, or comes full circle, which might be incorrect only insofar as it takes the image of a chain of reasons somewhat too seriously. Yet the picture of the closed chain of reasons is revealing in suggesting not a chain which abruptly comes to an end, but a self-contained system of beliefs which, though not tied to anything outside themselves, mutually support each other, and thus, as a whole, manage to convince us: The picture and the data convince us, but they don’t lead us further – towards other reasons. We say: “These reasons are convincing”; and here it isn’t a matter of premises, from which what we were convinced of follows. (BT 2005: 396) In the words of On Certainty, one might say that it is “a system in which consequences and premises give one another mutual support” (OC 1969: 142) that convinces one. Observations can convince us by entering into an overall picture without that picture needing to be mediated by or derived from something else. They provide non-inferentially acquired beliefs we treat as reasons for other beliefs without feeling the need to justify them further. The chain of reasons ends with such non-inferentially acquired beliefs. And it is precisely the fact that justification can come to an end in experience in this fashion which enables it to fulfil its justificatory function: “Justification by experience comes to an end. If it did not, it would not be justification” (PI 2009: 485).

## 6. First-Order and Second-Order Reasons

It is helpful at this point to introduce a distinction between two forms of reasongiving. When a reason is given, the demand for further reasons can take two forms. One is to demand reasons for the reason (what we might call first-order reasons); the other is to demand reasons for taking it to be a reason (what we might call second-order reasons). Suppose we take R1 to be a reason for thinking that p is the case. There will then be, on the one hand, the first-order reason R2 we have for thinking that R1 is the case, and, on the other hand, the second-order reason R1’ for taking R1 to be a reason for thinking that p is the case. The first form of the demand, which asks for a reason R2 for thinking that R1 is the case, is the one Wittgenstein addresses with the remark that demands for reasons become senseless at the limit of the language game, a limit marked by the rules that constitute the language game and enable demands for reasons to be raised within it. The second form, however, asks why we take R1 to be a Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 124]

reason in the first place. It does not demand reasons for thinking that p, but reasons for thinking tout court. As Wittgenstein writes: “It is one thing to justify a thought on the basis of other thoughts – something else to justify thinking” (BT 2005: 229).

It is against this second form of the demand, which asks for a justification of the way we reason, that Wittgenstein’s defactoist response that this is “simply what we call a reason” (BT 2005: 396) is directed. But here our earlier distinction comes in, the one between the sense in which what counts as a reason can be given a priori, and the sense in which it can be taught by experience. In many cases, second-order reasons (reasons for taking something to be a reason) can be given: “Why do you assume that he’ll be in a better mood because I told you that he’s just eaten? Is that any kind of reason?” – “That’s a good reason because, based on past experience, eating influences his mood.” And that could also be put this way: “Eating really does make it more probable that he will be in a good mood”. (BT 2005: 398) Here, past experience provides a second-order reason (R1’) to think that one’s having eaten is a reason (R1) to believe that one will be in a better mood (p). But if one asks why past experience should be regarded as a reason in this respect, the chain of reasons ends once more: But if one wanted to ask: “And is everything you put forth about past experience a good reason to assume that this time too this is the way it will be?” then I can’t say: “Yes, because that makes the occurrence of the assumption probable.” In what I said earlier I justified my reason using a standard for a good reason; but now I can’t justify the standard. (BT 2005: 398) While experience can in principle give us both first- and second-order reasons, Wittgenstein’s reflections aim to show that the fact that we take experience as a standard for what counts as a reason for what cannot itself be justified by experience: “Here there can be no further talk of justification” (BT 2005: 398). This is an insight Wittgenstein considers worth reiterating in On Certainty: But isn’t it experience that teaches us to judge like this, that is to say, that it is correct to judge like this? But how does experience teach us, then? We may derive it from experience, but experience does not direct us to derive anything from experience. If it is the ground of our judging like this, and not just the cause, still we do not have a ground for seeing this in turn as a ground. (OC 1969: 130) Justifying our thinking and judging in this sense would require more than giving a cause of our thinking and judging as we do. The demand for such a justifica- Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 125]

tion is thus revealed to be the self-contradictory demand for a reason “outside the game of reasoning” (AWL 1979: 4).

In one regard, the moral is the same as in the case of demands for first-order reasons for reasons: if the giving of reasons is to perform any function, there has to be a point at which our reason for believing there to be a rationalising relation between a reason and what it is a reason does not derive from experience, for otherwise a regress threatens, and with it the pointlessness of engaging in justifications in the first place: If the relationship of a reason to what it is a reason for were taught by experience, one would have to ask the next question: “And what is your justification for taking that as a reason for this belief?” And it would go on in this way; and belief would never be justified. (BT 2005: 399) One needs some standard of measurement in order to measure the weight of any considerations on the balance of reasons. Failing such a standard, the question whether something can weigh in as a reason cannot be meaningfully raised: “If these aren’t reasons, what are? […] ‘Not reasons’ – as opposed to what?” (BT 2005: 397). This passage adumbrates Strawson’s point that it is a non-contingent, a priori matter that induction is rational or reasonable (1952: 262). The fact that experience provides a standard for justification cannot, in the same sense of “justification”, be independently justified, but is a constitutive norm of the game of reasoning as we play it – the “defining equation simply has no solution. We haven’t established any method for its solution” (BT 2005: 399). Thus, while chains of causes can usually be prolonged by hypothesising further causes (BBB 1958: 88), chains of reasons come to an end in the “defining equations” that constitute the reasoning calculus and enable reason-giving to get off the ground in the first place.

## 7. Explanation and Justification

The pursuit of chains of reasons and causes to their ends is of course an old theme, epitomised in the Kantian “quest for the unconditioned” (1929: A 307/B 364).✏ Exploiting this fact, we can throw Wittgenstein’s construal of the dichotomy into relief by holding it up against contrasting construals proposed by Kant � According to Kant, the “principle peculiar to reason […] in its logical employment is: to find for the conditioned knowledge obtained through the understanding the unconditioned whereby its unity is brought to completion” (����: A ���/B ���).

Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 126]

commentators. One such is A. C. Ewing, who draws a strict parallel between chains of reasons and chains of causes: If the cause is viewed as explaining or giving the reason of the effect, then this suggests that there must be something which is its own cause, otherwise causation gives no ultimate explanation or reason at all, just as it would be futile to give a chain of the reasons for accepting a proposition if none of them could, any more than the original proposition, be seen to be true in their own right. (Ewing 1938: 218) For Ewing, it is a condition on successful causal explanation that the cause given can be traced to “something which is its own cause”, a causa sui. Likewise, chains of reasons must lead to self-evident propositions if they are to provide support for the justificandum. On this view, reasons and causes are on a par – they systematically lead on to other reasons and causes, respectively, and both fulfil their purpose only insofar as the chains thus followed up come to an end in some unconditioned condition. This conception would seem to differ from Wittgenstein’s in its denial that there is an asymmetry between chains of reasons and chains of causes.

By contrast, another Kant commentator, Jonathan Bennett, agrees with Wittgenstein that there is an asymmetry, but conceives of it in precisely the opposite way. According to Bennett, it is not chains of causes that always lead on to further causes, but chains of reasons that lead on to further reasons until some secure basis is reached: It is true that my deriving Q from P gives you no reason to believe Q unless you already believe P, but it is not analogously true that my showing you how e1 led to e2 gives you no explanation of e2 unless you can explain the occurrence of e1. A justification needs a justified basis, but an explanation does not need an explained basis. (Bennett 1974: 185) On Bennett’s account, a reason R1 to believe in a proposition p provides support only to the extent that there is a reason R2 to believe that R1 is the case, and the support provided by R2 is in turn conditional on there being a further reason R3 to believe R2, etc., until one reaches a proposition which one already accepts, the “justified basis” of justification. Explanations by appeal to causes are different insofar as determining the cause e1 of an event e2 is by itself sufficient to explain e2, no matter whether e1 is in turn explained or not.

We seem to have here three competing construals of the dichotomy between chains of reasons and chains of causes: either they both lead on to some unconditioned condition, or only chains of reasons do, or only chains of causes do. Can Ewing’s and Bennett’s accounts be reconciled with the Wittgensteinian model? By way of conclusion, I shall briefly set out how and to what extent they can Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 127]

be brought together, and in the process, we shall answer our last question: how does the distinction between chains of reasons and chains of causes relate to the distinction between explanation and justification?

Ewing, in his claim that chains of reasons and causes alike must end in some unconditioned condition, fails to distinguish between causation and causal explanation: e1 causes e2, but it is the fact that e1 happened which causally explains the fact that e2 happened (cf. Strawson 1992: 112 – 13). Further, to cite the fact that e1 causes e2 and the fact that e1 happened is to give reasons to believe that e2 happened, and therefore to place them in the chain of reasons. Ewing’s claim that chains of causes must lead to a causa sui must therefore be disentangled from the claim that causal explanations must come to an end if they are to be explanations at all. The first claim is one which Wittgenstein neither contradicts nor affirms, since he never asserts that chains of causes are literally infinite, but only that it is usually “easy enough” (BBB 1958: 88) to hypothesise further causes. Ewing’s second claim, meanwhile, is in line with Wittgenstein’s insistence that chains of reasons come to an end, since the reasons in question may be either explanatory reasons (answering the question “Why did X happen?”) or reasons to believe that X happened. In this sense, chains of reasons are chains of causal explanation, and Wittgenstein concurs with Ewing in the opening paragraph of the Investigations when he writes: “Explanations come to an end somewhere” (PI 2009: 1).

Where Bennett’s account is concerned, which at first appears to be the exact mirror image of Wittgenstein’s, a similarly conciliatory reading can be offered. Bennett claims, first, that one’s deriving q from p gives one no reason to believe that q is the case unless one already believes that p is the case; second, he claims that one’s showing how e1 causally led to e2 explains e2 independently of whether one can explain the occurrence of e1. Wittgenstein could assent to both propositions. Bennett’s first claim points at the regress of justification, the fact, as he puts it, that every justification needs a “justified basis”. Bennett’s point is not, however, the sceptical one encapsulated in Agrippa’s Trilemma, the claim that attempts at justification must either lead to an infinite regress, circularity or arbitrary dogmatism. It is merely that in order for p to justify q, it is not enough to show that q can be derived from p – p must itself be known to be the case if it is to serve as a basis for the justification of q. Yet the principle that for one proposition to justify another, the former must be more certain than the latter, is one which Wittgenstein recognises straight at the outset of On Certainty: When one says that such and such a proposition can’t be proved, of course that does not mean that it can’t be derived from other propositions; any proposition can be derived from other ones. But they may be no more certain than it is itself. (OC 1969: 1) Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 128]

Wittgenstein can therefore accept that every justification needs a justified basis, because what Bennett brings out is not an asymmetry between chains of reasons and chains of causes, but an asymmetry between justification and explanation. The fact that the distinction between chains of reasons and chains of causes does not line up with the distinction between justification and explanation becomes particularly perspicuous when we consider the matter in terms of the contrast between intensionality and extensionality: reason relations, and therefore chains of reasons, differ from causal relations, and therefore from chains of causes, in that they are intensional: “…is a reason for…” is a relation that holds not between singular terms or events, but “between facts or truths” (Strawson 1992: 109). Causal relations, meanwhile are extensional: “…is a cause of…” is a relation that holds between particular events. By contrast, explanation and justification are both intensional. It may be that a certain event e1 caused a certain event e2, but it is the fact that e1 occurred which explains the fact that e2 occurred, and it is the fact that p is the case which justifies the belief that q is the case. Now, Bennett’s point is that justification requires a justified basis, while explanation does not require an explained basis. But this is something which Wittgenstein is in no way committed to denying, and it does not affect his claim that chains of reasons come to an end. Quite the contrary, since explanation also requires a justified basis: e1 may explain e2 independently of whether e1 is itself explained, but it does not explain e2 independently of whether one is justified in believing, first, that e1 causes e2, and second, that e1 occurred. Therefore, it is as crucial to explanations as it is to justifications that chains of reasons come to an end and thereby provide the required support for the claim that acts as explanans. Where Bennett and Wittgenstein may well part ways is over the question of what counts as a “justified basis”. Both Bennett and Wittgenstein agree that one’s deriving q from p gives one no reason to believe that q is the case unless one already believes that p is the case. But, as we saw with regard to the claim that experience justifies something’s being a reason without itself being justified in its justificatory status, Wittgenstein is committed to denying that the justificans p must always be justified over and beyond its being believed: “At the foundation of well-founded belief lies belief that is not founded” (OC 1969: 253). Certain beliefs are more certain than any others and can serve as a justificatory basis for other beliefs, but not because they are indubitable or self-evident, and without being justified in turn by even more certain beliefs. “It is not single axioms that strike [one] as obvious”, but “a system in which consequences and premises give one another mutual support” (OC 1969: 142): Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 129]

The child learns to believe a host of things. I.e. it learns to act according to these beliefs. Bit by bit there forms a system of what is believed, and in that system some things stand unshakeably fast and some are more or less liable to shift. What stands fast does so, not because it is intrinsically obvious or convincing; it is rather held fast by what lies around it. (OC 1969: 144) We come to master linguistic and reasoning practices through demonstration, imitation, and repetition in training situations that are themselves rooted in our natural capacities and primitive reactions. It is such training situations which provide the “justified basis” or “bedrock” (OC 1969: 498), the justifier which, because it is inaccessible to questions of justification, is able to play a foundational role as the highest court of appeal in questions of justification. “Sure evidence is what we accept as sure, it is evidence that we go by in acting surely, acting without any doubt” (OC 1969: 196). Insofar as justifications have a foundation, they have a foundation in our practices, and it is in virtue of the role certain beliefs play in our language games that they possess their air of indubitability, held in place by the communal practices they are part of – just as the “foundation-walls are carried by the whole house” (OC 1969: 248). When viewed through the lens of the concept of reasons, Wittgenstein’s writings can thus be seen to offer us a rich and nuanced conception of reason and rationality. Reasons, whether a priori or empirical, first-order or second-order, link up immediately to form finite chains of reasons anchored in and held in place by our reason-giving practices. The resulting conception of reason and rationality could be said to be both liberating and limiting. It is liberating in the sense that in bringing out the complexities and particularities of reason-giving practices, Wittgenstein dissuades us from modelling chains of reasons on chains of causes; the must of rational necessity is altogether different from the must of causal necessity, and realising this allows us to view the fact that chains of reasons come to an end not as limitation on reason, but as an enabling condition – it is precisely because of their finitude that chains of reasons can guide, justify and explain our thoughts and deeds. Yet this conception of reasons is also limiting, in the sense that it firmly chains reason to the particular ways we have of going on, to the shared complex of activities, purposes and language-use in which there is minimal agreement on what counts as a reason for what. Reasoning-practices constrain what one can discover that one has reason to do, and reasoning-practices may differ from one form of life to the next. Both these aspects might be summed up by saying that though we may be everywhere in chains, these chains are not everywhere the same.

Brought to you by | Universitaetsbibliothek Basel Authenticated

[p. 130]

Bibliography Alvarez, Maria: Kinds of Reasons. An Essay in the Philosophy of Action, Oxford 2010. Bennett, Jonathan: Kant’s Dialectic, Cambridge 1974.

Blackburn, Simon: The Majesty of Reason, in: Blackburn, Simon: Practical Tortoise Raising and Other Philosophical Essays, Oxford and New York 2010.

Brandom, Robert: Making It Explicit. Reasoning, Representing, and Discursive Commitment, Cambridge, Mass., and London 1994.

Ewing, Alfred Cyril: A Short Commentary on Kant’s Critique of Pure Reason, London 1938. Fogelin, Robert John: Taking Wittgenstein at his Word. A Textual Study, Princeton and Oxford 2009.

Glock, Hans-Johann: Overview, in: Glock, Hans-Johann: A Wittgenstein Dictionary, Oxford 1996. –: Relativism, Commensurability and Translatability, in: Ratio 20 (4) 2007, 377 – 402. Goethe, Johann Wolfgang: Goethe’s Theory of Colours, trans. by Charles Lock Eastlake, London 1840.

Heal, Jane: ‘Back to the Rough Ground!’ Wittgensteinian Reflections on Rationality and Reason, in: Ratio 20 (4) 2007, 403 – 421.

Kant, Immanuel: Critique of Pure Reason, trans. by Norman Kemp Smith, London 1929. Lovejoy, Arthur Oncken: The Great Chain of Being, Cambridge, Mass. 1970. Nisbet, Hugh Barr: Goethe and the Scientific Tradition, London 1972. Plato: The Republic, trans. by Paul Shorey, Loeb Classical Library, Cambridge, Mass. 1930. –: Parmenides, trans. by Mary Louise Gill and Paul Ryan, Indianapolis 1996. Ryle, Gilbert: The Concept of Mind, ed. by Julia Tanney, London 2009. Schulte, Joachim: Experience and Expression. Wittgenstein’s Philosophy of Psychology, Oxford 1993. –: Goethe and Wittgenstein on Morphology, in: Breithaupt, Fritz, Raatzsch, Richard & Kremberg, Bettina: Goethe and Wittgenstein, Frankfurt a. M. 2003. Strawson, Peter Frederick: Introduction to Logical Theory, London 1952. –: Causation and Explanation, in: Strawson, Peter Frederick: Analysis and Metaphysics. An Introduction to Philosophy, Oxford and New York 1992.

Wyder, Margrit: Goethes Naturmodell. Die Scala Naturae und ihre Transformationen, Köln, Weimar and Wien 1998.

Brought to you by | Universitaetsbibliothek Basel Authenticated