Critical Appreciation of Jonathan Schaffer’s “The Contrast-Sensitivity of Knowledge Ascriptions”, Samuel Rickless

Author Information: Samuel Rickless, University of California, San Diego,

Rickless, Samuel. “Critical Appreciation of Jonathan Schaffer’s The Contrast-Sensitivity of Knowledge Ascriptions’.” Social Epistemology Review and Reply Collective 4, no. 4 (2015): 1-6.

The PDF of the article gives specific page numbers. Shortlink:

Please refer to:

Editor’s Note: With Samuel Rickless’ post, we initiate our “critical appreciation” series. In this series, we ask scholars to examine Social Epistemology’s most cited articles over the last 3 years (according to statistics sourced from CrossRef). We seek both a re-appraisal and re-imagining of the articles since their publication, and a sense of where the arguments and ideas might go in the future.

Image credit: Lars Ploug-Sørensen via flickr

Jonathan Schaffer’s 2008 article is part of a burgeoning trend, one that attempts to uncover previously unrecognized contrastive elements in a wide variety of different relations and properties (including knowledge, causation, freedom, belief, and confirmation of theory by evidence). My aim here is to provide a critical appraisal of the article, with a view to determining what it can teach us about how best to understand knowledge ascriptions, and how best to conduct research in epistemology and the philosophy of language more generally. 

Knowledge Ascriptions as Contrast-Sensitive

Schaffer argues that knowledge ascriptions are contrast-sensitive. This, by itself, is a remarkable and somewhat startling claim, even if it has some precursors with strong philosophical pedigree, notably Dretske (1970) and Craig (1990). Suppose I ask you, in the absence of contextual cues and completely out of the blue, whether you know that Barack Obama is President of the United States. I imagine that your answer will be immediate, unqualified, and in the affirmative. The same will be true if I ask you whether you know that 2+2=4, or that the Eiffel Tower is in Paris, or that torture is wrong. In considering how to answer my question, you will likely not be telling me that my question is ill-formed or that I need to give you more information in order to put you in a position to answer it.

Contrast this case with one in which I ask you whether you prefer chocolate. In this latter sort of case, you will scratch your head and give me a strange look. Then you will tell me that my question is ill-formed, or at least that you will need more information in order to answer it. You will say something like this: “It doesn’t make sense just to say that I prefer chocolate. Chocolate, if preferred, is always preferred to something else. So are you asking me whether I prefer chocolate to vanilla? Or whether I prefer chocolate to hazelnut? This is important, because my answer to the first question is yes, but my answer to the second question is no. So unless you fill in the relevant contrast-flavor, I won’t be able to answer your question.”

Despite the seemingly huge gap between the knowledge-question and the preference-question, Schaffer sees them as semantically similar. That is, just as we see the claim that you prefer chocolate as in some sense incomplete, Schaffer sees the claim that you know that Obama is the President as incomplete. On the contrastivist picture, to say that you know that Obama is the President is to say that you know that Obama rather than Romney is the President, or that Obama is the President rather than the Secretary of State (or maybe even that Obama is, rather than resembles, the President). In every knowledge ascription, no matter how seemingly straightforward and non-contrastive, Schaffer sees an implicit contrast proposition determined by what is in question. Thus, if what is in question is whether Obama or Romney is the President, then the ascription “Sally knows that Obama is the President” means the same as “Sally knows that Obama is the President rather than that Romney is the President”. And if what is in question is whether Obama is the President or the Secretary of State, then “Sally knows that Obama is the President” means the same as “Sally knows that Obama is the President rather than that Obama is the Secretary of State”.

If Schaffer is right, then classical epistemologists from Plato to, well, almost everyone, have been barking up the wrong tree for centuries. Descartes claims that he knows that he is thinking, on the basis of which he takes himself to know that he exists. And given knowledge of what he is thinking and knowledge of his own existence, he thinks he can prove (and hence acquire knowledge) that God exists. Whenever philosophers teach Descartes, this is what they tell their students (or so I hope). But, as Schaffer sees it, Descartes scholars should be qualifying what they say all over the place, making explicit all the contrast-propositions in question. “You see,” they should be saying, “Descartes claims that he knows that he is thinking rather than that he is walking, infers from this that he knows that he exists rather than that he goes bowling, and thence infers that he knows that God exists rather than that the Sultan of Brunei exists.” I am joking here, but really only half joking.

Contrastivists and Contextualists

I should emphasize that the urge to complicate the semantics of knowledge ascriptions sweeps in many more theorists than contrastivists. According to “contextualists”, knowledge ascriptions are sensitive, not to contrast propositions, but to other features of the context of utterance.

In the first of a famous pair of Bank Cases developed by Keith DeRose (1992; 2009), S and W wonder on Friday whether they can wait until the next day to drive to the bank to deposit a relatively unimportant check. S notes that he was at the same bank two Saturdays ago, and it was open then. W concludes (apparently, correctly) that S knows that the bank is open on Saturday. In the second case, S and W again wonder on Friday whether they can wait until Saturday to go to the bank, but this time it matters greatly to S and W that the check actually be deposited before Monday (e.g., they will lose their house to foreclosure). S again points out that the bank was open two Saturdays ago, but W wonders whether the bank might have changed its hours since then. If S can’t rule out the possibility that the bank has changed its hours, then it seems that S doesn’t know that the bank will be open on Saturday. From the apparent fact that the knowledge ascription in a low stakes Bank Case is true while the very same ascription in a high stakes Bank Case is false, DeRose (and others) conclude that knowledge ascriptions are context-sensitive, inasmuch as their truth-values are determined in part by the epistemic standards in the air at the time of utterance.

Why the urge to complicate matters in these ways? DeRose (and friends) are motivated at least in part to find an answer to the perennial problem of skepticism, the problem that if I can’t rule out the possibility that I am a brain in a vat on Alpha Centauri being stimulated to believe that I am typing on a laptop computer, then I really don’t know right now that I am typing on a laptop computer. In a low stakes case, says DeRose, it’s true to say that I know that I am typing, whereas in a high stakes case (as when I am pondering the skeptical problem in a philosophy class) it’s false to say that I know that I am typing. Schaffer’s motivations are similar. Only this time, whether the relevant knowledge claim is true or false is determined by the implicit contextually-provided contrast proposition: I can know that I am typing rather than that I am walking, but I might not know that I am typing rather than that I am a brain in vat to whom it appears that he is typing (see Schaffer 2004).

But both contextualists and contrastivists are focused, first and foremost, on the truth-conditions of knowledge ascriptions. They believe that our intuitive judgments about the truth and falsity of such ascriptions cannot be made to fit the old invariantist model, according to which they are a-contextual and a-contrastive. Schaffer’s article tries to hammer this point home, by pointing to utterances of the very same knowledge ascription that seemingly have different truth-values because different contrast propositions are made salient in the different contexts of utterance. For example, consider sentence (1):

(1) Debbie knows that Mary stole the bicycle.

If (a) what is in question is whether Mary or Peter stole the bicycle from the toy store, and detective Debbie finds Mary’s fingerprints at the crime scene, then (1) seems true; but if (b) what is in question is whether the bicycle or the wagon was stolen from the toy store, then, even if Debbie finds Mary’s fingerprints at the crime scene, (1) seems false.

In his article, Schaffer claims (rightly, as I see it) that there are three ways for an invariantist to respond to (a)-(b) pairs of linguistic intuitions: (i) reject the linguistic data, (ii) explain away the data; or (iii) offer an invariantist semantic explanation of the data. Schaffer then discusses and rejects three solutions of type (iii): the “Conditionals Strategy”, the “Conjunctions Strategy”, and the “Adjunctions Strategy”. Here is what these strategies tell us is being said by (1), depending on the situation, (a) or (b):

Conditionals Strategy
(a) Debbie knows that if Mary or Peter stole the bicycle, then Mary stole the bicycle.
(b) Debbie knows that if Mary stole the bicycle or the wagon, then Mary stole the bicycle.

Conjunctions Strategy
(a) Debbie knows that Mary but not Peter stole the bicycle.
(b) Debbie knows that Mary stole the bicycle but not the wagon.

Adjunctions Strategy
(a) Debbie knows that Mary rather than Peter stole the bicycle.
(b) Debbie knows that Mary stole the bicycle rather than the wagon.

One might object to one or another of Schaffer’s criticisms of these type (iii) strategies, but on the whole I agree with him that none of these three strategies is likely to work. That is, I do not believe that any of these strategies captures what is said by an utterance of (1) in different contexts. But it does not follow that there are no good type (i) or type (ii) invariantist responses. Indeed, I myself am drawn to option (ii).

Semantic or Pragmatic?

Language is a means of communication, but it is not the only such means. If you ask me what time it is while we are standing right in front of Big Ben, I might communicate the answer to your question (and much else besides, such as “Really, did you forget to put your contacts in this morning!”) simply by pointing directly at Ben’s face. Actions speak. But actions include more than moving one’s limbs, and human beings have discovered that they can communicate very complex thoughts by making complex utterances. Utterances are actions too. But utterances have meaning, and sometimes that meaning can be exploited for practical (i.e., pragmatic) purposes. This happens when, for example, language is used non-literally. Watching an obviously blown call on the field, I might tell a friend “Great call!” with exaggerated intonation. What I have said is that the call is great. But what I have conveyed is that the call is anything but.

The central question facing theorists at the intersection of epistemology and the philosophy of language right now is whether the phenomena that scholars such as Schaffer and DeRose identify as semantic are really pragmatic. If they are, then the linguistic evidence that drives contrastivism and contextualism (and potentially other contrastivisms and contextualisms too) is insufficiently probative.

At the end of his article, Schaffer sounds a refreshingly humble note. As he rightly points out, it is possible in principle to provide a pragmatic explanation of the linguistic data that support contrastivism. He adds, reasonably, that a successful pragmatic explanation “must invoke antecedently established pragmatic mechanisms”, and closes by inviting the invariantist “to provide the details”. As a parting shot, though, Schaffer claims that he has “no inkling as to how such an explanation might run” (242). In the space that remains, let me briefly see what I can do to adumbrate such an inkling.

Suppose we are in situation (a). Ann and Barbara both know that the bicycle was stolen, and Ann then notices that Debbie has found Mary’s fingerprints at the crime scene. Ann now tells Barbara: “Debbie knows that Mary stole the bicycle.” On my view, Ann’s utterance is pragmatically acceptable, even though it is actually false. Ann has adequate evidence that Debbie knows who committed the crime, but does not have adequate evidence that Debbie knows what was stolen. So Ann is not justified in believing that Debbie knows that Mary stole the bicycle. The most she is entitled to believe is that Debbie knows that Mary stole something. But, for the conversational purposes in which Ann and Barbara are engaged, it may simply not matter whether they come to agree that Debbie knows that Mary stole the bicycle or come to agree that Debbie knows that Mary stole something, something that they know to be the bicycle. If nothing in the conversation hangs on the difference between Debbie’s knowledge that Mary stole the bicycle and Debbie’s knowledge that Mary stole something (e.g., if Ann and Barbara are Mary’s parents, and what worries them, first and foremost, is that Debbie now knows enough to press criminal charges against their daughter), then Ann’s statement that Debbie knows that Mary stole (what both Ann and Barbara know to be) the bicycle, while strictly speaking false, satisfies the purposes of the conversational exchange without being misleading, and therefore counts as pragmatically acceptable. The thing about pragmatic acceptability is that it is easily confused with truth. So when we see nothing conversationally untoward about Ann’s utterance, we are inclined to think that Ann said something true, even if what she said is, strictly speaking, false.

Is the existence of pragmatic acceptability, as distinct from truth, “antecedently established”? Absolutely. In Rickless (2014), I discuss an example in which you ask me the time at 2:58pm and I reply: “It’s 3pm.” In a conversational context in which nothing hangs on the difference between its being 2:58pm and its being 3pm, my statement, while strictly speaking false, is pragmatically acceptable. After looking at your own watch, it would be bizarre and rude for you to reply: “Hey, you’re wrong, it’s not 3pm, it’s 2:58pm.” Similarly, in a situation in which everyone knows (and knows that everyone knows) that Elizabeth is the President, if Don says “Elizabeth is here”, it would be perfectly pragmatically acceptable to report Don as having said that the President is here, even if it is strictly speaking false that Don said that the President is here.

Much work in pragmatics in the past forty years has been concerned with articulating the mechanisms whereby propositions that are not semantically expressed by the words used in a conversational exchange are nevertheless (non-semantically) conveyed. The general idea is that speakers are engaged in an activity governed by conversational maxims (such as “be relevant!” and “avoid obscurity!” and “don’t say what you believe to be false!”—see Grice 1989). So if A and B are engaging in conversation and B violates one of the maxims, A can reasonably infer that B’s violation is in the service of conveying something different from what B actually said. (This is what happens in the “Great call!” case.) Less attention has been paid, I think, to the phenomenon of pragmatic acceptability. So one lesson I would take from the recent controversy over contrastivism and contextualism regarding knowledge ascriptions (and other statements too!) is that it would be good for philosophers of language to make sure that the data on which they build their semantic theories are intuitions of truth and falsity rather than intuitions of acceptability and unacceptability.

I suspect that when we take care to make this distinction, we will discover that the data that contrastivists and contextualists use to support their semantic theories of epistemological discourse are not semantic. If this is right, then we will find ourselves back in the position we were in before the “semantic” revolution in epistemology. This might look less exciting to some, but there are two silver linings. The first is that there is much exciting work in epistemology proper left to be done to solve the problems (such as the threat of skepticism) that have been dogging the field. And the second is that there is much exciting work in linguistics and the philosophy of language proper left to be done in the way of accommodating the phenomenon of pragmatic acceptability.


Craig, Edward. Knowledge and the State of Nature: An Essay in Conceptual Synthesis. Oxford: Oxford University Press, 1990.

DeRose, Keith. “Contextualism and Knowledge Attributions.” Philosophy and Phenomenological Research 52, no. 4 (1992): 913-929.

DeRose, Keith. The Case for Contextualism: Knowledge, Skepticism, and Context, vol. 1. New York: Oxford University Press, 2009.

Dretske, Fred. “Epistemic Operators.” Journal of Philosophy 67, no. 24 (1970): 1007-1023.

Grice, Paul. Studies in the Way of Words. Cambridge, MA: Harvard University Press, 1989.

Rickless, Samuel C. “The Contrast-Insensitivity of Knowledge Ascriptions.” Philosophy and Phenomenological Research 88, no. 3 (2014): 533-555.

Schaffer, Jonathan. “The Contrast-sensitivity of Knowledge Ascriptions.” Social Epistemology 22, no. 3 (2008): 235-245.

Schaffer, Jonathan. “From Contextualism to Contrastivism.” Philosophical Studies 119, no. 1-2 (2004): 73-103.

Categories: Critical Appreciation

Tags: , , , , ,

Leave a Reply