Definition
A44 §3A read context as evidence about a latent communicative situation z: P(y | c, u) = ∫ P(y | z, u) P(dz | c). A45 §10: “A frame is not necessarily a coordinate system. It can also be evidence that induces an effective interpretive model.” It gave three senses: “Representational: R_f(m); Inferential: P(z | c_f); Dynamical: K_f(s, ds′). The first changes the representation of meaning. The second changes beliefs about which communicative situation obtains. The third changes the interpreter’s state and subsequent dispositions. Our conversational path from relativistic frames to quantum cognition was, in part, a gradual discovery that these three operations should not be conflated.”
A48 §3.2 gave the same distinction by example: “Being addressed by a friend is a property of a situation. Being told that a speaker is a friend supplies evidence about the situation. Being instructed to interpret an utterance affectionately is an intervention on the interpretive process.”
C1 §2, the first full draft (U22), stated all three: a situation has a frame, interpreters infer one, and a framing intervention changes their inputs. “In an LLM experiment, the sentence ‘the speaker is a close friend’ is evidence about a supposed situation, not the social relationship itself.”
What it is not
Not the frame that obtains (TERM-030), nor an intervention (TERM-018). It adds the middle term that A39’s two-way split lacked. A45 §6: “no unique, identifiable latent variable z need exist that deserves to be called the pragmatic frame.” Kept in outline v4 (§2.3–2.4); the manuscript has none of the three-way distinctions.