## The Tyranny of the Subjunctive

### David J. Chalmers

Department of Philosophy
University of Arizona
Tucson, AZ 85721

chalmers@arizona.edu

[[Unpublished. Presented at Princeton in October 1998, at ANU in January 1999, and at the Idaho conference on "Truth and Meaning" in March 2000.]]

(1) A CONTRAST BETWEEN INDICATIVE AND SUBJUNCTIVE CONDITIONALS

(1a) If Prince Albert Victor killed those people, he is Jack the Ripper (and Jack the Ripper killed those people).
(1b) If Prince Albert Victor had killed those people, Jack the Ripper wouldn't have (and Prince Albert wouldn't have been Jack the Ripper).

(2a) If XYZ is the liquid in the oceans and lakes, water is XYZ (etc).
(2b) If XYZ were the liquid in the oceans and lakes, water wouldn't be in the oceans and lakes (etc).

The above all seem to be intuitively correct, even though (1a) and (2a) have "metaphysically impossible" consequents.

What's the relevant difference? Indicatives take the antecedent as a hypothesis about the actual world, whereas subjunctives take it as a description of a way the world might have been but actually isn't (Fowler: "it belongs to utopia"). That is, indicatives consider the antecedent as actual; subjunctives consider it as counterfactual.

In terms of the 2-D semantic framework: evaluation of indicatives tracks the primary intensions of the expressions involved, whereas evaluation of subjunctives tracks their secondary intensions. "Metaphysical necessity" reflects secondary intensions, not primary intensions.

(2) THE ANALYSIS OF NECESSITY AND POSSIBILITY

S is necessary <-> S is true in all possible worlds.
S is possible <-> S is true in some possible world.

QUESTION: How to evaluate "S is true in W"?

SUBJUNCTIVE READING: "S is true in W" iff

If W had been the case, S would have been the case.

INDICATIVE READING: "S is true in W" iff

If W is the case, S is the case.

The first tracks familiar "subjunctive" (utopian) evaluation of possibilities. The second tracks "indicative" (epistemic) evaluation of possibilities. (Is the epistemic possibility that W is actual an instance of the epistemic possibility that P?)

There are two resulting sorts of "necessity": subjunctive necessity and indicative (epistemic) necessity. And two sorts of "possibility", etc.

Plausibly: S is subjunctively necessary if S's secondary intension is true in all possible worlds. S is indicatively necessary if S's primary intension is true in all (centered) possible worlds.

Primary intension: is S true of W considered as actual? (evaluated in the way in which we evaluate epistemic possibilities)

Secondary intension: is S true of W considered as counterfactual? (evaluated as we evaluate explicitly counterfactual possibilities)

e.g. S = "water is H2O". W = the Twin Earth world (with XYZ).

(3a) If W is the case, water is not H2O (it's XYZ).
(3b) If W had been the case, water would still have been H2O (and XYZ merely watery).
This mirrors the fact that the secondary intension of S is true in W, but the primary intension of S is false in W. So "water is H2O" is not indicatively (epistemically) necessary, although it may be subjunctive necessary.

(3) THE SYMMETRY THESIS

In contemporary philosophy (post-Kripke), "necessity" (simpliciter) is always read as subjunctive necessity, and "P is true in W" is always evaluated by the subjunctive reading.

In fact: subjunctive conditionals *ground* the contemporary evaluation of statements in worlds, and the consequent analysis of necessity. Kripke's arguments in Naming and Necessity are always ultimately grounded in intuitions about subjunctives.

QUESTIONS: Why is this? Arbitrary? Argued or stipulated? Is indicative necessity an equally good candidate for "necessity"?

SYMMETRY THESIS: Indicative necessity is as good a candidate for "necessity" as subjunctive necessity. And epistemic evaluation is just as reasonable a reading of "P is true in W" as subjunctive evaluation.

If the symmetry thesis holds, it should be just as reasonable to say that:

"Hesperus is visible in the evening (if it exists)" is necessary;
"water is H2O" is contingent;
"I am here now (if I exist)" is necessary;
"the meter stick in Paris is one meter long (if it exists)" is necessary;
the necessity of identity doesn't hold;

and

In the XYZ-world, water is XYZ;
It's not necessary that I am David Chalmers;
In some possible worlds, Cicero is not Tully;
etc.

(4) SIX POSSIBLE REASONS FOR FAVORING THE SUBJUNCTIVE

(a) Indicative necessity is "merely epistemic".

[Answer: So? Before 1970, almost everyone thought necessity was tied to the epistemic (cf. Pap's book). Kripke *argued* that necessity and epistemic notions came apart, by appeal to the subjunctive, but one can't simply presuppose it.]

(b) S is necessary <-> S "could not have been otherwise" -- subjunctive!

[Answer: (i) "Could not have been otherwise" is ambiguous between epistemic/subjunctive readings. (ii) This reading doesn't seem to have been in wide currency pre-1970.]

(c) Indicative necessity requires centered worlds.

[Answer: Possible worlds are a technical device, and there doesn't seem to be anything in the notion of necessity itself to bar an indexical element.]

(d) Subjunctive necessity supports quantified modal logic, whereas indicative necessity doesn't.

[Answer: True enough. But whether necessity allows quantifying in was a live issue -- it can't be a prior constraint on necessity that it does! (Though one can see why Kripke would be attracted to a notion of necessity that does.) At best a split verdict: Quine right about indicative necessity, Kripke about subjunctive necessity. cf. Burgess.]

(e) The relevant indicative conditionals aren't true/false, just assertible, due to epistemic relativity. (e.g. Sly Pete)

[Answer: This epistemic relativity doesn't affect epistemic evaluation of statements in worlds. At most it affects the move from incomplete antecedents to relevant worlds.

N.B. Possible-worlds analysis of indicative conditionals:

"If P, then Q" is correct/true/assertible <-> The epistemically closest P-world(s) is a Q-world
(where here a P-world is a world satisfying the primary intension of P).]

(f) Some indicative "necessities" will vary in truth-value between users, as primary intensions of some terms (e.g. names) can vary between users.

[Answer: True, but the same is the case for subjunctive necessities, e.g. "I am DC", as secondary intensions of some terms (e.g. indexicals) vary between users. At worst, bring in a sentence-meaning/utterance-content distinction.]

(5) THE TYRANNY OF THE SUBJUNCTIVE

Necessity plays a central role in almost every area of philosophy. So the analysis of necessity affects almost every area of philosophy. Kripke's analysis has been central in supporting essentialism in metaphysics, direct reference views in the philosophy of language, externalism in the philosophy of mind, and so on. Has led to a lot of interesting remaking of philosophical views, with "marvelous internal coherence". But.

Imagine an alternative universe in which Kripke* used indicative conditionals instead. His book Naming and Contingency might have

• defended a strong link between the a priori and the necessary (plausibly S is a priori iff it is indicatively necessary)
• have argued against the necessity of identity and de re necessities (identities aren't indicatively necessary, and indicatives don't allow substitution or quantifying in, so de re necessities don't make sense [except for numbers?]).
• have argued for a Fregean view of meaning and content (since primary intensions can plausibly play the role of sense -- they both determine reference and reflect cognitive significance).

Imagine that the ensuing philosophical analysis of necessity had been as biased toward indicative necessity as recent philosophy has been toward subjunctive necessity. The resulting dominant views in various areas of philosophy might have been very different:

• philosophy of language: Fregean views, not direct reference.
• philosophy of mind: internalism, not externalism
• metaphysics: nominalism, not essentialism; etc.

Neither our world nor W* is the right outcome. The best outcome, given the symmetry thesis, is one in which the notion of "necessity" is recognized as ambiguous. Kripke** (importantly) recognizes the modality of subjunctive necessity and its distinctive properties. This leads the pretheoretical notion of "necessity" to split into two varieties, along with two corresponding means of evaluation in possible worlds. As to specific philosophical areas: the outcomes here will depend on analyzing whether the modality in those areas is relevant qua subjunctive or qua epistemic.

• Metaphysics: A split verdict about essentialism, with Quine right about indicative necessity but essentialists right about subjunctive necessity. But "subjunctive essentialism" is a significant sort of essentialism (it's significant that there are ways an entity could not have been) -- so the door is reopened to essentialism in philosophy.
• Philosophy of mind, epistemology: Here it's plausible that the central role of modality by far (at least where belief content is concerned) is epistemic -- e.g. considerations of rational inference, cognitive significance, and so on. And that the most relevant application conditions of concepts to worlds are epistemic, not subjunctive. So the intensional analysis of mental content ends up internalist. (There are still issues about externalist belief attribution to handle, but they can be handled straightforwardly.) As a bonus, we have a notion of belief content that is much more useful in epistemology.
• Philosophy of language: A split verdict? Expressions have complex semantic value, with something like a primary intension (more or less Fregean) and a secondary intension (more or less referential, for some terms). The latter are needed to analyze subjunctive conditionals and maybe belief attributions (though: scope analysis?). The former are needed to analyze indicative conditionals, belief attributions, etc; and to analyze the rational connections among statements, and to analyze the connection between language and thought. The modality inherent in Frege's notion of sense is arguably epistemic (sense determines the epistemic application-conditions of a concept), so Fregean sense is vindicated as above (modulo the introduction of an indexical element and the occasional variation of sense between tokens).