Definition
A191 §11 distinguishes “three senses of structure”, each with its question:
| Type | Question | Example |
|---|---|---|
| Descriptive | What relationships are observed? | Two speakers are judged to be aligned. |
| Modal or causal | What would happen under changes in conditions? | Removing shared background knowledge changes whether irony is recognized. |
| Constitutive | Which relationships make this the kind of act it is? | A promise establishes a commitment under relevant conventional and normative conditions. |
A relation of a communicative act is descriptive when it is observed in the data, causal when it holds under interventions on the conditions, and constitutive when its alteration changes which act was performed. A191 §4 pairs these with three kinds of identification: categorical identification (“In a specified category, the full representable profile determines an object up to isomorphism”), operational identification (“A selected family of measurements can distinguish some communicative realizations from others”) and ontological constitution (“Certain relations partly make a communicative act the kind of act it is”).
A203 §16 restates the table in other words (“Constitutive | Which relationships partly determine the act’s identity?”) and adds: “Causal intervention results can support a constitutive account, but they do not by themselves prove metaphysical constitution.” A218 §2 keeps the distinction as an open classification: “some observed relations are merely diagnostic, some causally influential, and others constitutive”.
What it is not
- Not a chain of entailments. A191 §11: “These form a useful progression in the research, but not an automatic chain of logical entailments. Descriptive structure alone does not determine causal structure. Causal stability alone does not establish constitution.” A191 §4: “Neither of the first two logically entails the third.”
- Not a ranking of evidence. Causal evidence is not a stronger grade of descriptive evidence for constitution. A constitutive claim needs a theory of the act. A191 §6: it “must be grounded in a substantive theory of the act and its normative conditions, not only its measured classification.”
- Not the distinction between situation and intervention. That one is about what an experiment manipulates (CLAIM-130). This one is about what kind of relation a result could establish.
Note of 2026-10-27: asked without a domain
The distinction is drawn for a communicative act. The exchange of 2026-10-27 asked it of any entity (A10 §8: “which distinctions are merely descriptive, which are causally consequential, which are constitutive”), after the owner set the record’s scope as unrestricted (U6), and it is filed so in QUESTION-284. Whether to widen this term, a new version with the domain dropped, is left to the owner; nothing filed uses it outside its domain.