Notes / Inferentialism
Scorekeeping
Working note — a draft kept in the open. Tell me where it’s wrong.
Start with the picture most of us absorb without choosing it. A word stands for a thing; a sentence pictures a fact; to understand language is to know which words go with which bits of the world, and inference is what we do afterwards, with meanings already in hand. Call this representationalism: reference first, reasoning second. Brandom’s wager — inherited from Sellars, and behind him from Kant and Hegel — is to turn the order around. Reasoning first. To grasp a concept just is to know your way around the inferences it stands in: what follows from it, what it follows from, what it rules out.
To say “that is copper” is to have committed yourself to its conducting electricity, to its being a metal, to its melting at a certain heat; it is to have ruled out its being an insulator. Someone who could pair the word with the substance but did not know any of this would not have the concept copper; they would have a label. Meaning lives in the web of material inferences a term licenses, not in a relation of standing-for that could be grasped in a single act.
The game of giving and asking for reasons
Brandom’s name for the practice is deontic scorekeeping. When you assert something you do two things at once: you commit yourself (you may now be asked to defend it, and you have authorised others to use it) and, if you are entitled, you license others to inherit the commitment on your say-so. The rest of us keep score — tracking what you are committed to, what you are entitled to, where your commitments collide. Meaning is the shape of this scorekeeping. It is irreducibly normative and irreducibly social: there is no last private fact about what I meant, only my standing in a practice of mutual accountability.
Saying “I” is taking up a position in the space of reasons — making oneself liable to the question why? after Sellars and Brandom
I find this the most honest description of what we are actually doing when we talk. It also explains, without strain, the things the representational picture has to bolt on: why disagreement is possible (our scores conflict), why testimony works (entitlement can be inherited), why understanding comes in degrees (you can know more or less of a term’s inferential surroundings).
The bill at the border with the world
Now the cost, which I do not think Brandom fully pays. If meaning is inferential role — relations among sentences — what keeps the whole game from spinning frictionless, a closed web of words answerable only to itself? Concepts are supposed to be about copper, not merely about other concepts. Brandom’s answer is that the practice has language-entry moves (perception: the world prompts non-inferential reports) and language-exit moves (action: avowals issue in deeds). The web is pinned to the world at its edges, by observation and by action, not by a representing relation in the middle.
That is the right shape of answer, and I am not sure the edges can bear the weight placed on them. A reliable disposition to say “red” in the presence of red is, by itself, the sort of thing a thermostat has; what makes it a report, a move in the space of reasons rather than a mere reaction, is that the reporter grasps its inferential consequences. So the entry move already presupposes the inferential interior it was meant to ground. The friction with the world may be real and still be thinner than representationalists want — enough to make the game about something, not enough to recover the comfortable idea that meaning is, at bottom, a name pinned to a thing.
Why this matters for the rest of the notebook
Two consequences I keep drawing. First, on vagueness: if meaning is inferential role in a social practice, the absence of a sharp boundary is no embarrassment — the practice simply does not include a rule that decides every case, and competence is knowing how to extend it answerably. Inferentialism and the pragmatist reading of the sorites are the same thought. Second, on law: a legal system is the game of giving and asking for reasons with sanctions attached and officials appointed to keep score. “Holding” a person to a rule is, quite literally, tracking their commitments and entitlements and acting on the score. That the model of meaning and the model of legal order turn out to be one model is, to me, the most promising thing in this whole region — and the thing I am least sure how to make precise.
Touching: Sellars (1956), Brandom Making It Explicit (1994) and Articulating Reasons (2000). See the canon §C.