Linguistics Lounge #3: Linguistics x Therapy

🛋️ Welcome to the Linguistics Lounge The Linguistics Lounge is where I talk to other writers on Substack about how language works in their fields and areas of expertise. From therapy to philosophy, billingualism to localisation—we’ll be learning how language and all of the linguistic frameworks I’ve written about apply to others. 🎥 🎧 A different format: video and audio posts that bring this work to life I sat down with Ellen Burns, PhD, a philosopher who writes AIs Without Minds on Substack, to talk about language, consciousness, and why the AI industry keeps getting this wrong. We landed on something she and I have talked about for a while: you can’t reduce language to statistics and then act surprised when the statistics reproduce society’s biases. We need more people in the humanities included in the AI discussion, something a lot of us call for.

This breakdown of sales language includes: * Why the most pressured sales pitch you’ll get this year won’t contain a single command * How the exit clause and the gift frame, two moves you already know from the meeting room, carry a sales pitch almost all the way to close * The modality shift: the moment a question about the product becomes a question about you * The split in the modals: why the seller’s claims come hedged and your obligations never do * Why a no costs more than a yes in these exchanges, and why that cost was designed in * Why “what’s the worst that can happen?” only counts one side of a cold email, and what a genuine ask leaves alone * Where the obligation register shows up in sales, and who ends up paying for it * Where the same register turns up next: the way AI gets sold

This post is about what’s actually happening underneath a phrase you’ve heard a hundred times — “let’s just chat” — and the specific ways a linguistic frame can shift under you before you’ve said a word. * Why “let’s just chat” can turn a career-affecting conversation into a casual one — and why that’s a specific linguistic move, not just a change of mood. * The three ways a frame can change hands: one disclosed to both people, one engineered by only one of them, one nobody’s doing at all — and how to tell, in the room, which of the three you’re standing inside. * Why “difficult,” “negative,” “not a team player” isn’t an improvised reaction to someone naming a frame out loud — it’s a role the frame had built and left empty before that person ever walked in. * What changes when the thing answering you is a language model: the frame you carry into a prompt shapes what comes back, and nothing on the other end had to read it to return it. * Where Goffman’s own account runs out — the question he never gets to about a frame, and what it costs to ask it yourself.

We’ve covered the framework this sits on in Part One, this post will cover where the chasm of meaning gets swallowed in one big gulp by implicature: * Why a sentence like can sit in a meeting doing real damage while remaining, on paper, a comment about something very normal. * Three specific constructions — not tone, not delivery, the grammar itself — that let a speaker mean something without the sentence ever committing to meaning it * Why the usual account of this (”deniability,” “reading between the lines”) undersells what’s actually happening: there is often no single hidden meaning to find, only a range of them, and that range is the whole mechanism * What it costs the person who has to turn that range into a single reading before they can respond to it — and why “you’re reading too much into that” is the exit the sentence was built with, not a comment on their judgement * Where the difference sits between a sentence built to hold two meanings open and a person simply being tactful

Today, we’re going to learn about a Grice’s Maxims of Cooperative Speech * Every message has two layers: what’s said, and what’s meant on top of it. The second does most of the work. * The maxims aren’t rules for speaking. They’re assumptions you run while listening — and meaning appears when someone breaks one on purpose. * Implicature isn’t “subtext.” It’s a specific claim you rebuild by inference, and rebuilding it is real work. * That work isn’t shared evenly. Cooperation assumes equals; most consequential talk isn’t between equals. * AI breaks the maxims differently — fluent surface, missing uncertainty markers, no way to answer only what was literally asked. You’re going to want to share this with your work bestie