Why it is a question
Footing and stance are what the line’s theses say survives or is lost. CLAIM-115’s defeat condition measures “pragmatic observables (stance, footing, alignment, expected uptake)”, CLAIM-061 says renderings can differ “in footing and illocutionary force”, and CLAIM-084 makes the primary outcome of a reconstruction study “pragmatic-frame retention judged blind to condition”. Each test of these claims is a judgement of footing or stance by somebody.
The record holds one footing judgement made twice of the same texts, and it disagreed with itself. CASE-012: the analyst’s four speaker labels for its own refrains “drifted” between A38 and A39, so that “‘The party always gets its cut’ went from cynical complicity to an informal rule without comment”. The judge knew what each text was meant to do, having written it, and still did not hold its labels between two passes. CLAIM-227, granted on 2026-10-10, adds that no design in the record gives judges a definition of footing, a response format beyond the analyst’s category list, a reliability target, or a validation against a contrast whose answer is fixed in advance.
So the question is not whether footing exists, nor whether lay readers can tell teasing from reproach. It is whether readers’ judgements in the categories the designs use (CASE-021’s “condemnation, solidarity, amusement and irony”, CASE-040’s “tone, stance, solidarity, and authority”) are reliable and measure what they are said to measure. Until that is known, a defeat test run on those categories tests the categories as much as the thesis.
The construct comes from Goffman’s “Footing”, in Forms of Talk (LIT-780), which decomposes the speaker into animator, author and principal. The record holds the book unread, so the definition judges would be given has no reading under it yet.
What would count as an answer
A study that reports, for a fixed definition and a fixed response format:
- Reliability between judges and between passes of one judge, each against a target fixed in advance. Which coefficient, and what target, needs a source on reliability coefficients that the record does not yet hold.
- Known-groups validity: judgements that separate renderings built to differ only in the attributed speaker, such as CASE-018’s reattributions of one sentence to a friend, a parent, a judge and a comedian, and that do not separate renderings built to differ only in wording with the speaker held fixed.
Stage 0 of the discriminating study (CASE-055) is that design. A negative answer would not show that footing is unreal. It would show that the designs’ categories cannot carry the line’s defeat tests, and the categories would have to be revised before the theses are tested.
Note of 2026-10-10: a documented case
The search of 2026-10-10 that the owner asked for found annotation studies that bear on this question, and CASE-058 files two of them.
What they answer. FactBank labels one event’s factuality separately for each source that reports it, so the proposition is held fixed while the source-relative label varies, and trained annotators agree at κ = 0.82. A stance label that is separable from the proposition, close to Goffman’s principal, can be annotated reliably when its categories are defined by discriminating tests and the task is decomposed. Kiesling et al. (2018) annotate Du Bois’s stance triangle (LIT-779), with an investment dimension defined as the split of animator from principal, and reach only α = 0.40 to 0.57, below 0.4 for investment and alignment in most of their material, with agreement that changes from one community to another. Interpersonal stance, the construct nearest the designs’ categories, is only moderately reliable even for trained judges. That is a negative-leaning answer, and it supports this question’s worry that a defeat test in those categories tests the categories.
What they leave. Neither reports agreement between two passes of one judge, which is the failure CASE-012 shows. Neither runs the known-groups test, renderings built to differ only in the attributed speaker. Neither uses the designs’ categories (CASE-021’s, CASE-040’s) or lay readers. No study found labels Goffman’s participation roles as such. The question stays open, and Stage 0 of CASE-055 is still the design that would answer it. FactBank suggests how to write that stage: decompose the judgement and give a test for each category. Kiesling et al. cite Craggs and Wood (2004) on Krippendorff’s α, a source on reliability coefficients this question says the record lacks.