Contribution
Before this paper, nearly every theory of dispositions tied their modality to a conditional, usually a counterfactual from stimulus to manifestation. The partial exceptions were Lowe (2011), which Vetter noticed only late (fn. 3), and Aristotle’s dynamis. Vetter states a worked-out alternative: dispositions are individuated by manifestation alone, and their modality is possibility. For threshold terms like ‘fragile’ this is a proportion-based graded possibility over maximally varied cases. For terms like ‘breakable’ it is plain possibility. She also argues that philosophers’ intuitions about the technical locution “disposed to … if …” cannot count as neutral data against the view.
Key insight
Dispositional adjectives (‘fragile’, ‘soluble’, ‘irascible’) name only a manifestation and paraphrase as “can (easily) M”, not as a conditional. So a disposition can be modelled as a restricted possibility, with the bearer’s intrinsic make-up held fixed. That is the same device Lewis (1976) used for ‘can’ ascribing abilities. Degrees of the disposition are then how many varied cases manifest it, not how close the nearest manifesting world is. The stimulus drops out of the disposition’s identity: where it seems to matter, it is either part of a causally complex manifestation (claustrophobia: being caused by enclosed rooms to feel anxiety) or a further fact about the typical causes of that kind of process.
Assumptions
- Dispositions are modal properties (p. 1). The paper addresses the which modality question, not the how question: reductive analysis, non-reductive model, or grounding of the modality in dispositions (p. 6). The possibility conception is “entirely neutral on the how question” (p. 7).
- A possible-worlds framework for restricted possibility, with contextually variable restrictions (Lewis 1976, Kratzer 1981/1991) (pp. 7–8).
- Intrinsicality idealisation. The relevant worlds hold fixed the bearer’s intrinsic constitution, or its physical base, and let external circumstances vary. This is admitted to be “strictly false” because of extrinsic dispositions (McKitrick 2003). It is kept as a first idealisation, and extrinsic dispositions are accommodated by also holding fixed external facts “selected by relations to that object” (fn. 6, p. 8).
- Ordinary-language evidence about dispositional adjectives counts as evidence about the properties (p. 3). Intuitions about the philosophers’ term ‘disposed to’ do not (§3.1).
- Scope: the claims concern ordinary dispositions “as we know them from ordinary life and language”, not the natural properties of fundamental science (p. 25).
Key results
- Thesis (§2.1, pp. 6–7). Standard conception: (1) a disposition is individuated by stimulus–manifestation pair(s), “a disposition to M when C”; (2) its modality is a counterfactual, “If x were S, x would M”. Possibility conception: (1) individuated by manifestation alone, “a disposition to M, fullstop”; (2) modality is possibility, “x can M”, to a first approximation.
- Against closeness-grading (§2.2, pp. 9–11). If “easily possible that p” means p holds in at least one relevant world sufficiently close to an ideal world, then: (a) the ideal cannot be actuality, or anything that actually breaks would be maximally fragile; (b) for any ideal there is a counterexample recipe. Stipulate a block with an Achilles’ heel that breaks only in a maximally normal world. It then counts as fragile, and as more fragile than a champagne glass. “One world is not enough.”
- Proportion-grading (§§2.2–2.3, pp. 11–14). x is more fragile than y iff x breaks in more of the relevant cases. x is fragile simpliciter iff it lies above a contextually fixed point in that ordering. The needed proportion is small (“a few”), near the “at least one” end of the quantificational spectrum. Two refinements: (1) quantify over cases (world, time, object), not worlds, because the irascible get angry many times per world; (2) to avoid cardinality problems with non-denumerable sets, use a finite subset of cases with the greatest variation in (causally relevant) external circumstances. Vagueness means no canonical subset is needed (p. 14). The Ann-and-Betty example: breadth of circumstances, not raw frequency, drives the degree.
- Plain vs graded possibility (§2.4, pp. 14–16). For ‘breakable’, ‘transmissible’, ‘soluble’, ‘walkable’ and ‘readable’, one actual manifestation suffices, so the modality is possibility simpliciter and any positive degree is enough; even the Achilles’-heel block is breakable. Variation across terms spans only “at least one” to “a few” and never approaches necessity (p. 15). Kjellmer (1986) observed this split empirically: more frequent ‘-ble’ adjectives tend to take the “easy possibility” reading (fn. 16).
- Stimulus-bearing ascriptions (§3.1, pp. 16–20). Both parsings of “disposed to M if C” fail on the possibility view: the wide-scope conditional, and the conditional inside the possibility. Vetter’s response is to discredit the data. A COCA survey (425M words) found 341 occurrences of “disposed to”, 115 discarded as irrelevant, leaving 226. Of these, 221 had agents as subjects and the other 5 abstracta. 190 were followed by action verbs and 35 by verbs of sensation or sentience. Only 5 interacted with ‘if’ and 1 with ‘when’, each scope-ambiguous. Not one ascribed a permanent intrinsic tendency to a concrete inanimate subject (pp. 18–19). So “disposed to” is a theoretical term, and intuitions about it encode the conditional theory rather than testing it.
- Stimuli and causation (§3.2, pp. 20–22). Water-solubility, hay fever, claustrophobia and acrophobia (same manifestation, anxiety, but different dispositions) are handled by complex manifestations (being caused by X to M) or by facts about the typical causes of a process type. The OED gloss on ‘-able’ (“capacity … of being subjected to … the action”) supports the passive or causal reading. The trade-off is stated as open: the conditional conception gives more structure and ties dispositions to causation, while the possibility conception gives more flexibility, for example for spontaneous decay (p. 22).
- Stronger modalities (§3.3, pp. 22–25). (i) Stability, robustness and sturdiness are the negatives of susceptibilities: ‘unbreakable’ is restricted impossibility, and ‘robust’ is “graded necessity”, the negation of ‘fragile’. (ii) Solubility’s lawfulness belongs to the manifestation (dissolving is law-governed), not to dispositionality: a substance S that dissolved only unsystematically would still count as ‘soluble’ per the OED. (iii) Dispositional essentialism (Bird 2007): natural properties that ground laws need nomological necessity, so they are not “just like” ordinary dispositions. The clash is “merely nominal” and hits only the usual way of introducing the thesis via fragility. “Perhaps there is more than one variety of modal properties” (p. 25).
- Conclusion (§4, pp. 25–26). The view is defended, not shown superior; that argument is left to Vetter (forthcoming, “Multi-Track Dispositions”). The conditional conception is not obviously the default starting point.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Ordinary dispositional adjectives express possibility (‘can’), not conditionals, and name only a manifestation | moderate | OED entries (pp. 2–3); Kjellmer (1986) survey of ‘-ble’ dictionary entries, “the conditional plays no role whatsoever” (fn. 1); Kratzer (1981) on German -lich/-bar; cross-linguistic assertion (fn. 2) |
| C2 | A disposition is individuated by its manifestation alone | weak–moderate | informal argument: multi-track and stimulus-less dispositions (Fara 2005; Manley and Wasserman 2008) plus the §3.2 recasting of stimuli as parts of complex manifestations; comparative superiority explicitly not argued (p. 6) |
| C3 | Closeness-based easy possibility cannot model fragility for any choice of ideal world | moderate | counterexample recipe (Achilles’-heel block, pp. 10–11), general across ideals; rests on intuition that the block is not fragile |
| C4 | Degree of a disposition = proportion of manifesting cases in a finite, maximally varied set of cases (world, time, object) | weak | informal proposal (pp. 12–14); the finite-subset choice is admitted to be non-unique and defended only by tolerance of vagueness; no formal measure given |
| C5 | Some disposition terms (‘breakable’, ‘soluble’) take possibility simpliciter; threshold terms (‘fragile’, ‘irascible’) take graded possibility; none approach necessity | moderate | inferential test from one actual manifestation (pp. 14–15); Kjellmer’s frequency observation (fn. 16) |
| C6 | ”disposed to” is a philosophers’ theoretical term, so intuitions about “disposed to M if C” are not neutral data | moderate | COCA counts: 226 relevant of 341, 221/226 agentive subjects, 0 inanimate intrinsic-tendency uses (pp. 18–19); the inference from corpus absence to “theoretical term” is argument, not measurement |
| C7 | The possibility conception does not solve finks, masks or antidotes and does not need to | assertion + informal argument | p. 5: these show dispositions concern the object, not its surroundings; p. 7: masks and mimics arise on both conceptions |
| C8 | Robustness is “graded necessity” as the negation of fragility | informal argument | duality with ‘unbreakable’/‘breakable’ (p. 23) |
| C9 | The clash with dispositional essentialism is “merely nominal” | informal argument | pp. 24–25; depends on accepting that natural properties are a different variety of modal property, which is asserted (“Perhaps …”) |
| C10 | The possibility conception is a serious alternative, not the default loser | moderate (as scoped) | the paper’s whole defence; the stronger claim of superiority is explicitly deferred (p. 26) |
Method
Conceptual analysis in a possible-worlds framework, driven by lexicographic and linguistic evidence: OED entries, Kjellmer 1986, Kratzer’s ordering sources. It also uses one small corpus study, a manual classification of all 341 COCA hits for “disposed to”. Objections are taken from Manley and Wasserman (2008, 2011), a referee, and Alastair Wilson.
Concepts
- Possibility conception — the conjunction of manifestation-only individuation and possibility as the disposition’s modality (p. 7).
- Standard / conditional conception — stimulus–manifestation individuation plus counterfactual modality (p. 6).
- Which vs how question — which modality dispositions involve, as against how they involve it (reduction, model or grounding) (p. 6).
- Easy possibility — possibility “of a sufficient degree”, with the threshold set by context (p. 9).
- Closeness conception — p is more possible than q iff some p-world is closer to an ideal world than any q-world (p. 9). Rejected.
- Proportion conception — p is more possible than q iff p holds in more of the relevant worlds or cases (p. 10). Adopted.
- Case — a centred world: a triple of world, time and object (p. 12).
- Graded necessity — Vetter’s term for ‘robust’ or ‘sturdy’: a little less than a restricted necessity of remaining intact (p. 23).
- Physical base — the part of the intrinsic constitution held fixed across relevant worlds (p. 8).
Connections
The metaphysical thesis is that dispositions are intrinsic modal properties whose modality is possibility. The paper presupposes realism about dispositions as properties of objects, and it stays neutral on whether modality is reducible to them. It sits in the Martin–Lewis–Bird–Manley–Wasserman debate over dispositions and conditionals. It extends Lewis’s (1976) restricted-possibility treatment of ‘can’ from abilities to dispositions, reversing Vihvelin’s and Fara’s strategy of treating abilities as dispositions. It borrows the proportional idea from Manley and Wasserman (2008) but keeps the proportions small, near possibility, where theirs are often majorities akin to necessity (fn. 12). It is the seed of Vetter’s Potentiality (2015) and of the “Multi-Track Dispositions” paper it cites as forthcoming.
Among held works, the nearest is LIT-144 (Pernu, causal exclusion and downward counterfactuals). Pernu relies on a counterfactual semantics that this paper argues is the wrong modality for ordinary dispositions, though it says nothing about counterfactual theories of causation. LIT-097 (the extended mind) treats dispositional beliefs, but it uses “dispositional” in the functionalist sense, not this one. No other held work engages the metaphysics of powers or dispositions (grep of literature.d for dispositions, powers, Vetter, Manley, Kratzer).
metaphysics tag: justified. The paper is squarely about the modal nature and individuation of a category of property, and it takes a position that bears on dispositional essentialism and the grounding of laws (§3.3). That is “what it is to exist … laws” in the tag’s blurb. metaphysics is correctly the primary and only topic. linguistics would be a defensible second tag, since the positive case is built on ‘-able’ semantics and a corpus study.
Bearing on the record
It carries nothing for ML practice. There is one conceptual point of contact that the anthology might want by name, not as a practice. Talk of model “capabilities” and “propensities” is dispositional talk. Vetter’s distinction bears on how capability claims are evaluated: ‘breakable’-type, where one exhibition suffices, versus ‘fragile’-type, which needs a sufficient proportion of varied circumstances. The distinction and her insistence on variation across external circumstances over raw frequency map onto the difference between “the model can do X” (one elicitation) and “the model reliably does X” (robust across varied prompts). This is an analogy, and the paper does not draw it. No THEORY document is supported or contradicted.
Note of 2026-10-16: THEORY-670 now cites LIT-109 as a source.
Limitations
- By its own statement it does not argue that the possibility conception beats the conditional one (pp. 6, 26), so it cannot be cited as having refuted the conditional analysis.
- The proportion measure is not defined. “Figuratively” counting cases in a non-unique finite subset is a gesture toward a measure, and the cardinality problem is sidestepped rather than solved (pp. 13–14).
- It does not engage finks and masks beyond placing them on both views. The reductionist sketch inherits the intrinsicality idealisation, which the paper concedes is false (fn. 6).
- The linguistic evidence is English-centred with cross-linguistic assertion (fn. 2). The corpus study is small (226 tokens) and manually coded, with no inter-rater check reported.
- It sets aside dispositions of fundamental science (charge) by positing “more than one variety of modal properties”, which restricts the view’s scope more than it answers the essentialist.
- Read from the penultimate draft; wording and pagination may differ in Mind.
Open questions
- Can a principled measure over maximally varied cases be given, making degrees of fragility determinate enough to compare objects that break in disjoint circumstance types?
- Does the possibility conception handle multi-track dispositions better than conditional accounts in a head-to-head comparison? (Deferred to Vetter, “Multi-Track Dispositions”.)
- How does the possibility conception relate to causation, if stimuli are only parts of complex manifestations (§3.2, left “an open question”)?
- If natural properties require a stronger modality, what unifies them with ordinary dispositions beyond both being “modal”?
Corrections to the seeded skim
- The dossier and LIT summary state the thesis as a finding (“a fragile vase is one that can break easily, not one that would break if struck”). The paper says it will “not … address the question which conception is to be preferred” (p. 6) and aims only to show the view “merits serious consideration” (p. 6; p. 26, “I leave the argument for another day”). The summary should be framed as a proposal.
- Dossier: “§§2.2–2.4 develop graded possibility (‘easily’, ‘more fragile than’) and go beyond it.” §2.4 (“Beyond graded possibility”) goes weaker, not stronger: ‘breakable’, ‘soluble’, ‘transmissible’ take possibility simpliciter (one case suffices). Going beyond possibility (toward necessity) is §3.3 (“Beyond possibility?”), which rejects it for ordinary dispositions.
- Dossier’s check “whether it avoids the finkish and masking counterexamples”: the paper explicitly says the proposal “is not going to solve” finks, masks and antidotes (p. 5) and admits masked and mimicked dispositions just as the conditional view does (p. 7). Its only move is a reductionist sketch that holds the bearer’s intrinsic make-up fixed and varies external circumstances (pp. 8, 12–13).
- The graded-possibility semantics is proportion-based, not closeness-based: closeness to any ideal world is rejected via Manley and Wasserman’s “Achilles’ heel” block (pp. 10–11), and proportions are taken over cases (triples of world, time, object) within a finite, maximally varied subset of them (pp. 12–14). The dossier left “how ‘can easily M’ is made precise” open; it is this.
- The dossier says the view clashes with “one way of formulating” dispositional essentialism. More exactly, Vetter says the clash is “merely nominal”: natural properties like charge need a modality stronger than possibility, so they are not “just like” ordinary dispositions but share only modal character (pp. 24–25).
- Additional topic, optional:
linguistics, since the argument rests on the semantics of ‘-able’/‘-ible’ adjectives (Kjellmer 1986, Kratzer 1981) and a COCA corpus study (§3.1). The test is met only weakly, because the linguistics is used as evidence, not studied, so this is a note rather than a firm correction. - primary topic: metaphysics is right; no change.