Contribution
A single model for what other traditions had split into separate kinds of stance: evaluation (assessment, appraisal), affective and epistemic positioning, and agreement or alignment. Du Bois argues that these are not types of stance but three aspects of every stance act, and gives the act a fixed structure of two subjects, one shared object and three directed relations among them. The structure is offered as both an analytic tool and a resource participants themselves use to infer what is left unsaid. He adds two methods for showing it in data: the diagraph from his dialogic syntax, and a “stance diagraph” that labels subjects, object and actions.
Key insight
No stance stands alone. “I’m glad”, “I know” and “I agree” are complete sentences and incomplete stances: each needs an object (glad about what, know what) and a prior stance it answers, and both are usually supplied by someone else’s earlier utterance. Because two people can orient to the same object, evaluating it also places each relative to the other, so alignment is not an extra act but a consequence of evaluating something someone else has evaluated. Hence the gloss: “I evaluate something, and thereby position myself, and thereby align with you.”
Assumptions
- Data: short excerpts of American English conversation, mostly from the Santa Barbara Corpus of Spoken American English and a few from the Longman Spoken American Corpus, both built under the author’s direction; paraphrases of the emergent stance are his.
- Stance subject ≈ speaking subject in nearly all examples; note 13 sets aside multivocality (Bakhtin; Goffman 1981, the book filed as LIT-780) for separate treatment.
- A shared stance object is assumed for the full triangle. Cases without one are acknowledged as a challenge; the reply (Section 6, note 29) is that, following Tomasello, all meaningful language use presupposes shared orientation, a claim deferred to future work.
- Alignment is scalar, convergent or divergent by degrees, not a binary aligned/disaligned.
- The too/either result depends on conditions only partly stated: a stance follow juxtaposed with a resonant stance lead, with subjective intentional predicates such as like, love, know, think and want (note 26), and it rests on a separate study (Du Bois 2004) not reported here.
Key results
- Three questions for interpreting any stance (Section 3): who is the stancetaker, what is the object of stance, and what stance is the stancetaker responding to. Attested utterances show each answer is routinely found outside the stance utterance: “it was really great” (a weekend) and “That stuff is great” (hibiscus cooler) are different stances; “I would love to go” is about Nicaragua, 40 lines earlier; Melissa’s “I totally agree” answers her mother’s directive 29 lines back.
- Subjectivity takes an object (Section 4). “I’m glad” and “I’m so glad” differ less by so than by what each is glad about (no significant problems to talk about; you guys having fun), which the prior turns supply. “I know” and “I don’t know” likewise take their objects from the partner’s prior utterance.
- Intersubjective marking (Sections 5–6). In “I don’t like those” / “I don’t either” and “I don’t know if she’d do it” / “I don’t know if she would either”, the second speaker’s either cannot be dropped without pragmatic anomaly, and the first speaker cannot use it. Stance lead and stance follow are named after Tomasello’s gaze following. Generalised: under the stated conditions a resonant stance follow must be marked, too for positive and either for negative utterances.
- The stance differential (Section 6). In Ken “I would love to go” → Lenore “Yeah.” → Joanne “Yeah?” → “I want to go too”, Joanne’s rising Yeah? and her want for love register a weaker alignment inside a formally parallel follow.
- The definition (example 49): “Stance is a public act by a social actor, achieved dialogically through overt communicative means, of simultaneously evaluating objects, positioning subjects (self and others), and aligning with other subjects, with respect to any salient dimension of the sociocultural field.”
- The triangle (Figure 1). Nodes: Subject 1, Subject 2, Object. Two sides are evaluative vectors from each subject to the shared object; the third side is alignment between the subjects, in either direction; positioning is a reflexive vector shown only by its arrowhead on each subject (“blowback” from evaluating, note 22). Six arrowheads in all: three acts, doubled for two subjects. Any unspecified part can be inferred from the rest: agreeing implies sharing the prior evaluation, and repeating the evaluation implies convergent alignment.
- Stance in one line (Section 8): “an act of evaluation owned by a social actor”, tying act, responsibility and value; proposed, if it binds the minimum structures of social action, as “perhaps, the smallest unit of social action”.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Evaluation, positioning and alignment are aspects of one stance act, not distinct stance types | moderate as an analysis; argued by showing co-occurrence in examples, no test of separability | Sections 2, 4–6 |
| C2 | A stance’s interpretation requires its subject, object and the prior stance it answers, often found only in prior turns | moderate: repeated corpus excerpts | Section 3 |
| C3 | Displays of subjectivity always make an object relevant (“Subjectivity takes an object”) | moderate: examples; “always” is not tested against counter-cases | Section 4 |
| C4 | A resonant stance follow must mark its intersubjective relation with too/either, and a stance lead cannot | moderate: examples here plus a cited corpus study (Du Bois 2004) whose figures are not given; conditions only partly specified | Sections 5–6, notes 26–27 |
| C5 | Participants use the triangle’s structure to infer unexpressed parts of a stance | weak: illustrated, no evidence about participants’ inferences beyond the analyst’s reading | Section 6 |
| C6 | Shared orientation to an object is general to all language use, so the triangle applies even without an evident shared object | weak: asserted with a citation, explicitly deferred | Section 6, note 29 |
| C7 | Stance may be the smallest unit of social action | speculative, offered conditionally | Section 8 |
Method
Two notations from dialogic syntax (Du Bois 2001). The diagraph sets two or more utterances in rows (intonation units, otherwise clauses) and aligns resonating elements in columns, with speaker labels and line numbers back to the transcript. The stance diagraph adds column headings for speaker, stance subject, the positioning/evaluating predicate, stance object, and alignment marker, with recovered material in braces ({to Nicaragua}). The analysis maps a stance diagraph onto the triangle.
Concepts
- stance: the definition at (49) above.
- evaluation: the stancetaker orients to an object of stance and characterizes it as having some quality or value.
- positioning: situating a social actor with respect to responsibility for stance and for invoking sociocultural value; affective (glad) or epistemic (know).
- alignment: calibrating the relationship between two stances, and so between two stancetakers; scalar, convergent or divergent.
- stance object / shared stance object: what is evaluated; when several speakers orient to it across turns, the cornerstone of intersubjectivity.
- stance lead / stance follow: the first stance and a resonant second one built on it.
- stance differential: the differences within a formally parallel follow (intonation, word choice) that calibrate alignment.
- other-positioning (note 16): proposing a candidate stance for the addressee, as ordinary questions do.
- intersubjectivity: the relation between one actor’s subjectivity and another’s, realized here through orientation to a shared object.
Connections
The chapter draws on dialogism (Bakhtin, Voloshinov, Linell), conversation analysis on assessment and agreement (Pomerantz, the Goodwins, Heritage and Raymond), linguistic anthropology on indexicality and responsibility (Silverstein, Hanks, Hill and Irvine), work on subjectivity in language (Benveniste, Langacker, Traugott), and Tomasello’s joint attention. It names Goffman’s Forms of Talk (LIT-780) once, in note 13, for the multivocality it sets aside: Goffman’s decomposition of the speaker is where the chapter’s simplification “stance subject ≈ speaking subject” would be tested. It closes on Wittgenstein: “There are no private stances.”
In this record, the shared object that binds two subjects is close in structure to the joint attention and shared-intention accounts filed among the social-ontology works (Tomasello, LIT-504; Gilbert’s “Walking Together”, LIT-477), both Deferred and unread; the connection is mine.
Bearing on the record
- No THEORY filed. The triangle is mainly a definition and an analytic frame, and its empirical claim (obligatory too/either on a stance follow) rests on a study not reported here. A candidate, not filed: “In English conversation a stance that resonates with a prior stance on the same object with a subjective intentional predicate must mark the relation with too or either, and the prior stance cannot.” Source this chapter; promote when Du Bois (2004) or another corpus count is read, with its conditions and exceptions.
- LIT-780 (Goffman). The chapter defers to Goffman on who the speaking subject is; the two filings bear on each other through the stance subject, and a reading of “Footing” would show whether the triangle survives a speaker split into animator, author and principal.
- The LIT as filed described the triangle correctly in outline. Two refinements: the triangle has two evaluative sides and one alignment side, with positioning reflexive at each subject, not one act per side; and alignment is scalar, covering divergence, not just alignment versus disalignment.
- No instruction for machine-learning practice.
Note of 2026-10-16: THEORY-693 now cites LIT-779 as a source.
Limitations
- Qualitative and illustrative. No counts, rates or systematic sampling; the excerpts are chosen, and the paraphrases of emergent stances are the author’s, which note 9 itself calls too explicit.
- One language and register: American English conversation.
- Speaker = stance subject is assumed; reported speech, animated voices and institutional speaking are left out (note 13).
- The shared-object requirement is not shown to hold generally; the defence is deferred.
- The too/either generalization is stated with conditions the chapter does not give in full, and too/either are not always intersubjective (note 27).
- A slip in the text: discussing example (48) the chapter says “Alice responds to Mary’s I don’t know if she’d do it”, although in (47) Alice speaks first and Mary second; the analysis that follows treats Alice as the lead, so the slip does not change the argument.
Open questions
- Can evaluation, positioning or alignment occur without the others? A case of alignment with no evaluated object, or evaluation that commits the speaker to no position, would break the “one act” claim; the chapter argues for co-occurrence without searching for such cases.
- How does the triangle treat a speaker who animates someone else’s stance? Goffman’s footing is the obvious test.
- How do stances invoke and reproduce systems of sociocultural value (note 28)? Named as the main open topic.
- Do the too/either conditions hold in other languages and in written or institutional discourse?