The Contrast‐sensitivity of Knowledge Ascriptions

Abstract

Knowledge ascriptions are contrast‐sensitive. One natural explanation for this is that the knowledge relation is contrastive (s knows that p rather than q). But can the binary view of knowledge (s knows that p) explain contrast‐sensitivity? I review some of the linguistic data supporting contrast‐sensitivity, and critique the three main binary explanations for contrast‐sensitivity. I conclude that the contrast‐sensitivity of knowledge ascriptions shows that knowledge is a contrastive relation.

Keywords:

• Contrastivism
• Knowledge

Acknowledgements

Thanks to Martijn Blaauw, Walter Sinnott‐Armstrong, and the participants at the NAMICONA Epistemic Contrastivism Conference. The discussion there provided the inspiration for this paper.

Notes

[1] The contrastive theory emerges in the following passage from Fred Dretske: “To know that x is A is to know that x is A within a framework of relevant alternatives, B, C, and D. This set of contrasts … serve to define what it is that is known …” (1970, 1022). The theory has since been defended by Bredo Johnsen (2001), Adam Morton and Annti Karjalainen (2003), Walter Sinnott‐Armstrong (2004), Martijn Blaauw (2004), and Schaffer (2004, 2005, 2007). Thus Johnsen maintains that “what is known is always a contrastive proposition to the effect that P‐rather‐than‐any‐other‐member‐of‐category‐C is true …” (2001, 401). There are a variety of arguments for contrastivism, beyond the contrast‐sensitivity of our intuitions. These include the conceptual role of knowledge ascriptions, and the connections between knowledge, inquiry, and discrimination. I will not be discussing these other arguments further here. See Schaffer (2005) for further discussion, and see Ram Neta (2008) for a very helpful summary.

[2] Thus see, for instance, the partition semantics for questions presented in James Higginbotham (1996), the contrastive theory of focus developed in Mats Rooth (1992), and the “set of live options for the conversants” model of presupposition offered in Robert Stalnaker (1999a, 1999b).

[3] Dretske (1981) may have been the first to note the impact of contrast shifts on knowledge ascriptions: “Someone claiming to know that Clyde sold his typewriter to Alex is not (necessarily) claiming the same thing as one who claims to know that Clyde sold his typewriter to Alex … A person who knows that Clyde sold his typewriter to Alex must be able to rule out the possibility that he gave it to him, or that he loaned it to him … But he needs only a nominal justification, if he needs any justification at all, for thinking it was Alex to whom he sold it” (Dretske 1981, 373).

[4] I am indebted to Kent Bach, Michael Tooley, and René van Woudenberg for discussion on this point.

[5] Symbol key: I use “→” as a dummy connective for some conditional, and I reserve “⊃” for the material conditional, “⇒” for the indicative conditional, and “>” for the subjunctive conditional.

[6] See Sinnott‐Armstrong (2002) for a related argument against construing contrastive claims as material conditionals.

[7] Here I remain neutral on what exactly is the indicative conditional, only supposing that (i) it differs from the material and subjunctive conditionals, and (ii) it validates certain intuitively valid inference patterns. The reader who would deny (i) is welcome to just apply the criticism of using material or subjunctive conditionals here. The reader who would deny (ii) is invited to clarify what inferential role he/she sees the indicative conditional as playing. See Dorothy Edgington (2001) for a useful overview of accounts of the indicative conditional.

[8] A version of the conjunctions strategy has been suggested by David‐Hillel Ruben (1987) and Dennis Temple (1988), for reducing contrastive explanatory locutions. I am indebted to Jesper Kallestrup for further discussion here.

[9] That Mary stole only one item is presupposed by the contrastive locution of Rather (b). If Mary could have stolen both the bicycle and the wagon, then there would be no contrast between these options. This could all be made explicit by inserting “only”, to read “Mary stole only the bicycle”.

[10] An adjunct is a semantically optional element, as opposed to an argument, which is required to saturate a slot in the relation. For the contrastivist, the contrast is a required argument under K (it is the q slot in the Kspq relation). On the adjunctions strategy, the contrast is an optional element in p. Contrast clauses (like any other clauses) can serve as either adjuncts or arguments, depending on the relation. The preference relation, for instance, is a three‐place contrastive relation Pspq, so the contrast serves as an argument in:

Argument‐Prefer

I prefer that the Red Sox win the World Series, rather than that the Cardinals win.

The kissing relation, on the other hand, is a two‐place relation Kxy, so the contrast serves as an adjunct in:

Adjunct‐Kiss

Cindy kissed Billy rather than Tommy.

The adjunctions strategy then holds that the contrasts in knowledge ascriptions function as in Adjunct‐Kiss rather than in Argument‐Prefer.

[11] I am indebted to Thomas Blackson, Stephen Schiffer, and Daniel Howard‐Snyder for discussion here.

[12] All of the examples here in which the contrast clause is an argument (Prefer, Surprise, Regret) allow for the contrast to be left implicit:

Implicit‐Prefer

I prefer that the Red Sox win the World Series.

Implicit‐Surprise

Ann was surprised that Cindy kissed Billy.

Implicit‐Regret

I regret that Bush is president.

In the Implicit cases the contrast argument is to be recovered from context (from what is presupposed). The knowledge relation works the same way when the contrast is left implicit. This is the morale of Presupposition (a) and Presupposition (b) in The Contrastive Data section. The implicit contrasts for “the detective know that Mary stole the bicycle” are to be recovered from the question under discussion. (So is the contrastivist a contextualist? Yes and no: yes in that the truth‐value of binary knowledge that ascriptions comes out contextually sensitive, but no in that “knows” invariantly denotes the one and only K relation. What shifts with context is only the value of the contrast, and only when that is left implicit.)

[13] The reader may confirm that the contrast clauses in the Adjuncts cases do not impact truth‐values. For instance, if (i) Cindy kissed Billy, then (assuming only one kissing occurred) it follows that (ii) Cindy rather than Sally kissed Billy, (iii) Cindy kissed rather than chastely shook hands with Billy, and (iv) Cindy kissed Billy rather than Tommy. Indeed, assuming only one kissing occurred, (i)–(iv) are materially equivalent. To illustrate the point via focus: (v) Cindy kissed Billy, is materially equivalent to (vi) Cindy kissed Billy. Whereas (vii) Ann was surprised that Cindy kissed Billy, is not materially equivalent to (viii) Ann was surprised that Cindy kissed Billy.

[14] The problem of closure is clearest in the focus case: here the alleged closure inference runs from: (i) the detective knows that Mary stole the bicycle, to (ii) the detective knows that Mary stole the bicycle. How could these have any opportunity to differ in truth‐value (save by the grossest of logical blunders), if the contrast‐value was not playing the role of argument in the knowledge relation?