Essay add: 14-07-2017, 10:25   /   Views: 65

Contextualism, Invariantism and Instabilities in Meta-Epistemology.

1 Introduction

Who knows what? In identifying instances of knowledge, the competing Contextualist and Invariantist accounts of knowledge ascription both aim to account for our intuitions in cases where problematic and even paradoxical knowledge ascriptions arise. This paper explores the conflict between these competing views with a concern that the meta-epistemology of both might be complicated by instability in the characterisations of knowledge that lie at the heart of the discussions. Shifting characterisations of knowledge in the models affect their handling of our use of 'knowledge' terms in naïve practice. In short, the theories both suffer from the problems they are aiming to explain.

I look at how epistemologists use the term 'knowledge' in their discussions of knowledge ascription and how, while contextualism may have the upper hand as a working epistemology of knowledge ascription, it is not yet problem free.

2. The context of this discussion: Relativism, Contextualism and Invariantism

The Contextualist-Invariantist debate aims to explain whether and how rational thinkers can attribute different truth values to the same knowledge propositions.

Truth Relativism proposes that truth values for truth-apt propositions cannot be true or false in any absolute sense but, rather, true or false only in relation to particular human interests. Persons with different interests may then assign different truth value to the same proposition. This relativism can be distinguished from a merely expressive relativism in which there is reference (though perhaps implied, presumed, or otherwise elliptical) to a statement of relativisation, eg., "He is short,  for a professional basketball player".  Truth Relativism proper directly responds to the question of whether truth is absolute or relative. In Truth Relativism, 'what is said' is not considered to be relativised at all as it is in expressive relativism cases; the meaning of the proposition under consideration is held to be fixed and fully understood by all parties and the question is whether its truth value is absolute, or relative to some further factor. Propositions such as "divorce is wrong"/ "other people deserve consideration"/ "government is necessary"/ "democracy is effective" may be held by Truth Relativists to be true in some instances and not true in others without any adjustment to the meaning of the propositions. Their truth value, not their meaning, is understood to have a contextual component.

Truth relativism need not be a global phenomenon applicable to all truth everywhere. Global Truth relativism quickly gives rise to nonsense and provides a straw man relativism that is trivially overcome (since the truth claim of global truth relativism may itself  be relativised out of court and take up no more of our time). In particular domains Truth relativism can co-exist independent of and alongside Expressive relativism and the two should be differentiated in both attack and defence. But if truth, simpliciter, must be absolute, Truth Relativism is simply wrong.

If a proposition can be true in one environment (when suitably generalised or indexed to account for indexical and temporal aspects) then if it is untrue in another environment it appears we have a case of Truth Relativism.

This becomes relevant to semantic contextualism when we try to determine epistemic standards (those factors directly relating to what is rational to believe given the evidence ) for knowledge ascription. What varies with our epistemic standards is what it takes for a knowledge-ascription (or denial) to be true. We have an interaction between epistemic standards and truth-conditions. It then follows from the claim that there can be variation in the truth-conditions for uses of 'know' in different contexts that the meaning of 'know' shifts from one context to another. There are then as many senses of 'know' as there are epistemic standards.

Broadly speaking, Contextualist approaches to knowledge ascription have parallels with Truth Relativist approaches to Truth, whereas the Invariantist approaches have absolutist sympathies.

Contextualism holds that whether a belief a counts as knowledge can depend on contextual factors. Keith DeRose describes contextualism as a linguistic thesis holding that knowledge ascriptions are context-sensitive and that their truth values depend on the context of the ascriber, ie., the person making the knowledge evaluation. If one's standards for knowing (or ascribing knowledge to others) are higher than another's, then knowing might be differently attributed to instances of the same epistemic value on the basis of differing standards.

In contrast to Contextualism, Absolute Invariantism holds that there is no variation in epistemic standards across contexts and so the truth conditions of knowledge-ascriptions remain fixed. For invariantists, conversational contexts do not affect epistemic standards. The criteria for knowledge - how h3 one's epistemic position need to be to qualify a belief as knowledge - remain constant through variations in the context of utterance.

Absolute or Sceptical Invariantism represents the extreme end of a continuum of invariantist positions and is the claim that the semantic value of knowing is so high and rigid that all ordinary ascriptions of knowledge are false.

Moderate Invariantism allows that the semantic value of know permits nearly all  ordinary ascriptions of knowledge to be true. Classical Invariantists hold the truth conditions of knowledge ascriptions to be independent of context -whether the context of the ascriber or of the subject.

The debate takes place against a backdrop awareness of epistemic scepticism. Epistemic scepticism is the view that we can know nothing for certain- we cannot warrant absolute certainty in our knowledge of any matter whatsoever.

In circumventing radical scepticism and its nihilistic consequences for all epistemology we have to deal with the tension between developing too high a view of knowledge that permits little if any knowledge to exist, and too low a view that trivialises and devalues the same knowledge. The difficulty with the purest view of knowledge is that, given human fallibility, it actually does make knowledge vanish altogether, inasmuch as there appears to be literally nothing that we cannot be wrong about. Cartesian certainty seems not to be available.

In appealing to our intuitions on fictional (though mundane) scenarios and paradigm sample cases both the contextualist and the invariantist attempt to defend claims of knowledge against scepticism and yet both implicitly acknowledge an underlying sceptical principle. This is most noticeable in deliberations over the meaning and usage of "knowing". Dangerous for both views is the invocation of sceptical reasonings into their arguments and, because neither view can despatch the sceptic once and for all, the debate has to progress with the elephant of scepticism in the room, but never allowed to play. This creates a managed tension in the scope of meaning for knowledge terms in the arguments. I suspect that the meta-epistemic use of knowledge terms with these implicit  limitations is in itself a contextual factor that affects both positions, the contextualist and (more invidiously) the invariantist. 

Scepticism, like many other epistemological positions, exists in degrees. Few advocate an outright nihilistic scepticism that strips from us our motivation for epistemic enquiry. We need, rather, to proceed as best we can while retaining a measure of epistemic humility. The pertinent feature of recent literature appears to be an instability in how "knowing", should be applied in cases where "100% "or "perfect knowledge" claims are clearly inappropriate. It is in this domain that I hope to show the meta-language of epistemologists remains contaminated by the very problems the protagonists aim to address.

The modified invariantism it is most appropriate to consider in this regard is Sensitive Moderate Invariantism (more particularly Subject-Sensitive Moderate Invariantism) as set out by John Hawthorne. Hawthorne presents this view clearly and though critical of it, appears to favour it over contextualism. SMI (for short) is an invariantist position that holds the epistemic standards in an account of knowledge are set by the subject (the one "knowing").

DeRose observes that contextualism is inherently more flexible than invariantism as the speaker setting the standards allows the speaker to choose to adopt speaker, subject or third party theoretical standards as required. In so allowing, contextualism avails itself of permission to reproduce SMI's subject sensitive position when appropriate in a particular case. On the SMI view, it will always be the subject's context that dictates which standards are operative. First person cases can generate identical diagnoses from both views.

The contextualist-invariantist debate really covers two disputed issues simultaneously:

  1. whether "knows" has context sensitivity of content; and
  2. whether knowledge depends on non-epistemic factors, like the practical interests of -accessibility, reasonableness of effort required (practicability), consequences, probabilities, precedents or stakes in the truth of belief.

For consistency this paper adopts the following terminology in discussing knowledge ascription cases:

The speaker is the ascriber - ascribing a knowledge state to a person;

the subject is the person who's knowledge claim is being evaluated.

First person ascription (ie assertion) of one's own knowledge is legitimate and in such a case one may be both ascriber and subject. In Contextualism: the knowledge-ascriber sets the standards. In Subject-sensitive invariantism (SSI): the subject credited with knowledge (or its lack) sets the standard.

3. The Problem Arena

In this section three widely discussed paradigm examples are presented to explore the issues and observe the limitations in application of contextual and invariantist knowledge ascriptions,.

The first was originally presented by Jonathan Vogel  to raise issues with Hawthorne's SMI. I present the case as a set of propositions Pn with narrative as follows:

(P1) Jamie parks his car on Elm St. at 12:00 and leaves the area.

(P2) At 13:00, Jamie knows where his car is.

Vogel's example presumes that it is reasonable to conclude from Jamie remembering where he left his car, that he knows it is still there. At 14:00, say, a threat to Jamie's belief is made salient, such as Elm St.'s notoriety for car theft, and Jamie's response to this information is to consider that he does not any longer 'know' that his car is still parked on Elm St. (and has not become another crime statistic).  So,

(P3) At 14:00 Jamie does not know where his car is.

As presented, this example employs debatable constructions in regard to the content of Jamie's knowledge and the content of his rational responses. At P2, in particular, where rather than 'know where his car is', Jamie could at best be said to 'know where he left it' and, like the rest of us, 'be hoping it is still there, unharmed'. Jamie's knowledge of the current whereabouts of his car effectively ceases to be justified when unremarkable alternatives become feasible. In a simple case, as soon as he turns the corner from where he parked his car (and the thief waiting for him to do so deftly shims the door, hot wires the ignition and drives away in it) his knowledge of his car's location is no longer factive and, so, no longer knowledge.

Such inaccuracies in knowledge reporting, using terms informally as they might indeed be used in everyday speech, are problematic for the theorists' argument here, and lead to slippage in terminology and vagueness in reasoning.

The invariantist account continues: At P3, Jamie will still know where he parked his car, but now be legitimately more concerned about it still being there, since reasons for doubt have become salient. With increasing time there is more opportunity and hence probability that it may have been stolen.

Does it matter that this doesn't happen often when a car is parked? Car theft is not quite a sceptical hypothesis, and is better categorised as a lottery possibility. I'm observing only that, properly and quite reasonably speaking, for every day usage, we should only speak of knowing where we parked -and where our car was (not 'is') when we have left it. This practice is epistemically accurate,  maintains epistemic humility and means that our language can account for surprises without evoking any logical disparity such as might seem to occur with a looser use of 'knowing' and the occurrence of a subsequent surprise. Our very susceptibility to surprise underlines that we do not always know as we might casually (unreflectingly) think we do.

Contrary to Hawthorne, I would maintain that ordinary language has a vocabulary apt for expressing 'degree of knowledge' claims. We might not formulate the comparative and superlative from the root of "know", but analogously with "bad /worse/ worst" and "good/better/best" we can readily use "sure/surer/surest" and a number of other epistemic terms that can at least express a salient dimension of knowledge (eg confidence, understanding, certainty, appreciation)  for the communication in hand. Whether we use such degrees appropriately or not is a psychological issue, but we have the linguistic resources in place to be used. With their help the bare choice of knowing or not knowing, with its terse, bivalent, factive implicatures, should not be forced on us.

To explore Vogel's example further:

At 14:00, when Jamie "does not know" where his car is (whether it is still where he parked it), what would be his own ascription of his earlier state of knowledge? Hawthorne and Vogel ride this question to a seeming contradiction while discussing Hawthorne's Sensitive Moderate Invariantism (SMI). They proceed as follows:

Let (E) be the proposition: Jamie's car is on Elm Street and

(-T):  Jamie's car has not been stolen and driven away.

Since SMI holds the requirements for knowledge are subject-dependent, Jamie at 13:00 and 14:00 can be thought of as two subjects, Jamie13:00 and Jamie14:00 respectively. Jamie's situation at 14:00 does not affect what (and whether) Jamie13:00 knows at 13:00. But it does affect what Jamie14:00 knows at 14:00. Thus, SMI provides that at 13:00 Jamie knows both that (E) and that (-T), while at 14:00 Jamie does not know (-T), and hence does not know (E) either.

At 14:00, however, Jamie can consider his previous epistemic situation. According to SMI, Jamie14:00 would say, ''At 13:00, I knew that E and -T, but now I don't''. In the literature, SMI is reported as being in trouble here as this would allegedly be an odd thing to say. Hawthorne is aware of this difficulty, and offers as a reply:

"If at the later time I don't know the ordinary proposition, then I cannot assert that I used to know it, since knowledge is the norm of assertion. To properly assert that I used to know that p, one needs to know p now".

The thesis of this paper is that both the alleged issue, and responses such as Hawthorne's here, may be symptomatic of slippage in referents. Sensitive invariantists treat knowing as non-indexical, but knowledge of different objects, such as here, is conflated and reported with some confusion and does not invoke indexicality. Variations in the object of knowledge provide explanatory accounts of variations in knowledge ascription.

Taking knowledge as the norm of assertion, it is proper for Jamie14:00 to assert that he knew that E and -T at 13:00, only if (iff) he knows at 14:00 that he knew that E and -T at 13:00.

But knowledge being factive:

If Jamie13:00 knew that E and -T, then E and -T.

So, if Jamie14:00 could properly assert that he knew that E and -T at 13:00, he would have to know at 14:00 that E and -T.

However, at 14:00, Jamie doesn't know those propositions, because at 14:00 the possibility of car theft is salient to him and he acknowledges the hazard )to have been present, unrealised, at 13:00 -thereby rendering Jamie1300's knowledge erroneous.

According to Vogel, at 14:00 Jamie doesn't know that he knew E and -T at 13:00, and that's why Jamie can't properly say that he did. Vogel concedes that Hawthorne's SMI based explanation does yield the desired result -that Jamie doesn't know that he knew E and -T at 1300, but not in the clearest way.

Both Hawthorne's and Vogel's analyses appear to want a clear recognition from Jamie14:00 that he was in error to claim the knowledge he claimed at 13:00. Indeed, at 13:00 and at 14:00 all he knows is where he left his car and all else is merely reasonable supposition. All that has happened as a result of making car theft salient to Jamie is to revise his suppositions. At 14:00 he could accurately report that at 13:00 he had considered his car safely parked and at 14:00 he acknowledges the risk of having been stolen. By misrepresenting Jamie's relation to the location of his car when he is not checking it as knowledge, it is rendered possible for him to 'know' the potential results of different indirectly inferential hypotheses -his car being where he left it and his car not being there. This is hardly surprising since neither hypothesis is knowledge and nothing very mysterious is taking place epistemologically.

Hawthorne's SMI would allow that, at 13:00, Jamie knew whether E and whether -T. But since this 'knowledge' might be factually incorrect even by 13:00, I would object that this is not knowledge simpliciter. Nothing factual need change at 14:00 precisely (the car perhaps having already been stolen when Jamie becomes aware of the risk). But it is at 14:00 in the example that Jamie's additional information (the notoriety of Elm St) is obtained and he thereafter doesn't know whether his car is on Elm Street or whether his car has been stolen and driven away.

Under these circumstances it would be unacceptable for Jamie to report at 14:00 that at 13:00 he knew whether E or whether -T, but that at 14:00 he no longer does. Hawthorne's treatment of what is wrong with Jamie's saying ''At 13:00, I knew E and -T, but now I don't'' doesn't immediately explain why ''At 13:00, I knew whether E and whether -T'' is true, but not assertible by Jamie at 14:00.

Instead of considering what Jamie can properly assert, consider what Jamie ought to believe about his previous epistemic situation. At 14:00, the possibility of car theft has become salient to Jamie. Jamie reflects back on his conviction at 13:00 that E and -T. Ought Jamie14:00 to believe that at 13:00 he knew that E and -T, even though at 14:00 he doesn't know, and doesn't believe that he knows, that E (and, likewise, -T)? It seems not. But even if knowledge is the norm for assertion this doesn't help, since the issue is what Jamie ought to believe, not what Jamie may properly assert.

Further, if we extend the narrative: Imagine that at 15:00, Jamie returns to Elm Street, and his car is where he parked it. Jamie has been worrying about car theft, and he is now relieved to find his car safely where he left it. But, if Jamie now reflects back on his situation at 13:00, when the possibility of car theft hadn't yet been salient to him, he has to say that at 13:00 he did know that E and that -T. For, according to Hawthorne, the obstacle to Jamie's asserting that he knew at 13:00 would have to be his failing to know at 15:00 that E and that -T. But at 15:00 Jamie does know those propositions. SMI, then, has no explanation of why, at 15:00, Jamie would, or should, withhold the ascription of knowledge to himself, at 13:00, of E and -T.

Hawthorne's and Vogel's construction of the SMI diagnosis appears inadequate in tracking the knowledge content at each stage of Jamie' experience. By identifying a vulnerable inference as knowledge, then tracking the possible truth value of the inference as the salience of contributing premises are adjusted, Jamie's knowledge-so-called in this is example is shown to be unstable. In this first person case, contextualism would be subject to the same criticisms as the SMI model as the only point of view available is the subject's.

In a second example presented by Timothy Williamson, the following dialogue takes place at the zoo:

John: I know that this is a zebra.

Mary: How do you know it isn't a mule painted to look like a zebra?

John: For all I know it is a painted mule. So I was wrong. I don't know that it is a zebra after all."

The sensitive invariantist postulates that Mary's question causes John to cease knowing the animal is a zebra by making salient the serious epistemic possibility that it is not. This would not work if John's response had been 'I didn't know that it was a zebra after all', since then John is departing from the invariantist's diagnosis by negating the invariance of "know". Whatever tense John responds in, contextualists can account for his remarks by postulating that Mary's question shifts the reference of "know": knowing a zebra and knowing a zebra from a painted mule are two different levels of knowing. These are different contentual standards for knowing, as being able to dismiss the salient alternative requires additional knowledge of the animal in question (or of the appearance of dyed fur).

But Williamson observes that neither the invariantist's shift in John's perspective nor the contextualist's shift in the meaning of 'know' sits with John's admission of error ('So I was wrong') since both attempt to explain how John was not wrong in his first speech.  If John claims that his earlier claim was false, not all his claims are true.

Contextualists and sensitive invariantists might object that the inclusion of an error admission biases such an example. But in adjusting our epistemic claims to accommodate previously unnoticed sceptical scenarios as they are brought to our attention, we, as speakers, ordinarily acknowledge error and modify our apprehensions.  Difficulty in accommodating cases of imprecision and error within a theory of knowledge ascription bedevils examples like Jamie's car and John's Zebra.

Being just plain wrong needs to be isolated from the analyses as it swamps the distinctions being explored. Interpreters need to be able to recognise and filter out cases of linguistic incompetence (as in claiming knowledge of where a car is now when all that is available is knowledge of where it was parked). At the same time, endorsements of knowledge may often rest on a principle of charity, by which one should prefer to interpret subjects as speaking and thinking truly rather than falsely with faculties of perception, reason and memory functioning properly (ie no radical scepticism in play). Adjusting standards, as contextualists propose, seems to give us more flexibility to deploy our charity in respect of 'knowing' .

So far we have only indirectly alluded to a number of steps in a methodology for ascription of knowledge to a speaker. To make this more explicit -a knowledge ascriber must 1) asses eligibility for employing a principle of charity (to account for delinquent cases) 2) evaluate linguistic competence to detect subject errors, ambiguities and 'technical' misunderstandings 3) evaluate epistemic factors such as evidence, corroboration, precedent, 4) evaluate non-epistemic factors such as salience, interest, stakes, and consequences.  These areas are not equally amenable to evaluation, nor is the sequence of application fixed, but listing them helps show where our focus might be in any discussion. Delinquency is largely omitted from our example discussions, but error may (debatably) play a role in some of the propositional reasoning employed. I contend that Jamie's putative knowledge of his car's whereabouts is an example of error in formulation of the propositions evaluated as knowledge

Jason Stanley presents his view of interest-relative invariantism (IRI) as a model for resolution of a class of apparent semantic paradox where a subject may be said to know and not to know a proposition from different ascriber's positions simultaneously. This may result simply from a subject simultaneously meeting  a lower ascriber standard and failing to meet a higher ascriber standard. While Invariantism holds that the semantic terms have fixed, context-independent (ie invariant) meanings, Interest-relative Invariantism (IRI), does allow non-epistemic factors to influence our knowledge ascriptions. The meaning of  "knowing" is held not to change, but claims of knowledge are sensitive to an additional condition - eg beyond justification, truth and belief -that involves practical (non-epistemic) facts about the subject's  situation at the time of knowing. 

IRI allows non-epistemic factors such as strength of evidence  and relevance of alternatives to alter the criteria for knowledge in interest cases. Something can make a subject require h3er evidence to claim knowledge of a proposition at one time than at another. In "Relevant Alternatives IRI", different sets of alternate propositions must be ruled out at different times.  In both cases, the IR Invariantist claims that the meaning of "having knowledge" hasn't changed at all (contra the contextualist interpretation), but that the non-epistemic criteria for asserting "having knowledge" have.

Shifting the criteria for knowledge ascription in the way Stanley describes for IRI functioning is not immune to appearing contextualist. He develops an illustration in which an Ascriber and a Subject operate with different levels of interest in a piece of knowledge content. His specimen case is set up concerning knowledge of whether a bank is open on a Saturday. Bill has no pressing interest on whether the bank is open or not, but has been there previously on a Saturday. Hannah has a critical need to bank a cheque before Monday and has to make arrangements to go on Friday iff the bank isn't open on Saturday. The consequences for Hannah of her cheque not being banked before Monday are very serious.

IRI allows Bill to know the bank is open on his low grade evidence because of his low stakes on the truth of his knowledge. Hannah does not know the bank is open on Bill's testimony of his low grade evidence, because her high stake in the matter needs h3er evidence (eg to rule out salient possibilities, such as the possibility of bank hours recently changing, for any reason). The difficulty for IRI is how it should handle Hannah's ascription of Bill's knowledge. This is the high takes ascriber-low stakes subject (HSA-LSS) case.

Stanley explores our intuitions on the HSA-LSS case as to whether Hannah can conclude that Bill knows "by Bill's lights' while she does not know by the same. This simultaneous knowing and not knowing seems paradoxical, or at least in need of explanation.

Hawthorne makes two contributions to resolving this problem. First is an approach emphasising the factivity of knowledge. "To assert p, one must know p". So Hannah cannot assert "Bill knows the bank will be open" unless she knows the bank will be open.  However, Stanley notes that Hannah's "failure to assert Bill knows" is not equivalent to "asserting Bill does not know".

Secondly Hawthorne proposes Projectionism, which appeals to psychological features of the speakers, viz "overprojecting our (Hannah's) lack of knowledge onto others (Bill)". Keith deRose amends this to a "tendency to overproject our evidentiary standards onto other situations". This might be a plausible explanation in naïve practice, but it becomes a visible phenomenon when objectively considered, and so would be defeated in cool review.

Stanley offers an  IRI account developed from reflection on the High Stakes participant's purpose in enquiring whether someone else knows that p.

  1. High Stakes wants to know whether another person knows that p, eg because there are important consequences for High Stakes dependent on whether or not p.
  2. So, High Stakes' interest lies in establishing p; that is, in acquiring information that will allow her to know that p.
  3. What High Stakes is interested in finding out, then, is whether someone else's information state is sufficient for High Stakes to know that p.
  4. In short, the purposeHigh Stakes has in asking someone else whether or not p is true lies in finding out whether, if that person had the interests and concerns High Stakes does, that person would know that p.
  5. Since pis a serious practical question for High Stakes, she is not really worried about that person's own interests and concerns.

Stanley claims this is a perfectly intuitive explanation of our intuitions HSA-LSS cases. Hannah is worried about her impending bill, and so wants to know whether the bank will be open on Saturday. It is to resolve this question that she asks Bill. What she wants to know from Bill is whether he has evidence such that, were he in her practical situation, it would suffice as knowledge. She has to make a decision. Of course, were Bill to share Hannah's practical situation, he would be in a High Stakes situation, and so would not know, on the basis of the evidence that he actually has, that the bank will be open on Saturday. So Hannah is perfectly correct  to conclude that the answer to her actual concern- whether Bill would know that the bank will be open if he were in Hannah practical situation-is negative.

Stanley claims this explains our third-person intuitions about the case. Our intuitive recognition that what Hannah really cares about is whether Bill would know, were he in her practical situation. We recognise that the proposition Hannah really wants to know the truth value of-"Bill would know were he in our practical situation"-is false. So, we are h3ly inclined to go along with Hannah's judgment, since we recognise that she is correct about the information in which she is really interested.

But this reputedly invariantist interpretation contextualises the meaning of "knowing". The object of the verb "to know" has changed between the Ascriber and the subject. Bill's "know" is about the bank opening simpliciter; Hannah's "know" is about Bill's state of knowledge and whether it is of use to her. Stanley comments that, "we can recognise what it is Hannah really cares about", so there must be a semantic difference for us to independently recognise. The object of knowledge therefore differs between Bill and Hannah. This appears not to be recognised fully .

4. Invariantist Objection

To my objections on interpretation of the worked examples an invariantist might reply, as Stanley, that there are variations in truth-conditions for knowledge ascriptions that are due to factors other than the epistemic So the Invariantist could attempt to explain the subtle shifts in knowledge content I'm concerned about on non-epistemic bases.

My contention is that the shifts I see presented in the literature are, at bottom, cases of knowing different things and are a symptom of slippage in focus during the discussions. The shift occurs when a subject, in claiming propositional knowledge, represents that proposition in a particular linguistic form. When an ascriber assesses the knowledge claim, he does so in such a way that the proposition is altered (not just the language) the change may be inadvertently tailored to support the favoured meta-epistemic bias - invariantist or contextualist. This might occur through introduction of another factor or constraint conjoined on the initial proposition.  As in the case of John's zebra, two distinct criteria of knowledge are discussed as follows.

P1. This is a zebra. And P2. This is a zebra AND this is not a painted mule.

[cf: P1. This is a 500 Euro banknote. And P2. This is a real 500 Euro banknote and not a counterfeit.]

The invariantist defence that the shifts are non-epistemic seems not to apply in this case. The criteria for knowing, as  required in the first and second cases of these examples, shift epistemically. More detailed knowledge is required to knowP2 than to knowP1 as the counterfeit currency recognition example highlights. This assessment, as presented, is independent of any practicality or consequence of obtaining or possessing the necessary knowledge. That the example makes sense without any pragmatic colouring appears to make the point. Hawthorne observes that such questions as Mary's may add evidence for the alternative case ( a painted mule) and so change the context for the subject. For an SMI example, we would need a question from Mary that altered John's practical stake in the identity of the zebra eg:

John: I know that this is a zebra.

Mary: You know that if  it's a zebra it will be shot.

John: I don't know that it is a zebra after all.

The real difficulty in this area may arise from the first proposition, that is, on what it means to say  "I know [this is a zebra]".  If our ascription of knowledge of such a proposition is defeasible by a serious alternative epistemic possibility, we have to clarify what makes such a possibility serious. For example, whether we take the presence of a dyed mules seriously. The contextualist may answer "the context", which implies a range of possibilities that could be ranked in plausibility with an implied cut-off point. The acceptable range might then include " a near relative of a zebra that has a different name, eg a Zebroid" but not " a dyed mule". The invariantist highlights for us that the implied cut-off point may be due to a practical, non-epistemic feature of the sort he advocates as controlling our ascriptions. It seems implausible that a proposition's being a serious epistemic possibility should only be determined by facts about the subject and not also dependent on wider contextual considerations.

5 Conclusion

In seeking to increase the rigour of our methodology for ascribing knowledge there is a three way tension. It is, practically speaking, necessary to avoid the abyss of radical scepticism yet this limits the strength of the standard we can require for knowledge (by means which we might otherwise simplify knowledge ascriptions). But we require a breadth in our use of knowledge terms to allow their application in less than absolute cases. Advocating a contextual element in our use of  knowledge terms introduces a parameter which, while needing itself to be accounted for, appears to have explanatory success in accounting for variations and even oddities in our usage of knowledge terms. Determining the boundaries and limits for this contextualisation parameter proves problematic.

The current debate between contextualists and moderate invariantists, seems to be hindered by the drift that occurs in the informal ways we speak of knowledge, and in the absence of a settled terminology, or even conceptualisation, for the components of knowing. We are most hampered by our lack of an apparatus for expressing strength of knowing, and this paper has attempted to show that the lacuna is inadequately filled by substituting variations in content of knowledge (in its accuracy, precision and object) which blur our discussions. A number of instances of knowledge ascription (particularly first person) appear to allow both invariantist and contextualist accounts to be given, and the tension between the standpoints, premised on their being exclusive, can be questioned.

Since some degree of contextualisation is required to account for knowledge ascription and since absolute knowledge is precluded by our sceptical backdrop, a contextual element in knowledge ascription seems essential. Whether this should be primarily ascriber contextualist or subject invariantist or a hybrid is not resolvable at present because instability in terminology allows cases to be misconstrued.

Effort in meta-epistemology is required to tighten our grasp on the object of characterising the knowledge relation and identifying its parameters and character. Without greater precision it will remain difficult to tell whether a knowledge relation is changing across contexts as the Contextualist would have, or whether it remains constant as the invariantists say.


Word Count 5843


  • Black, Tim (2004) "Classic Invariantism, Relevance and Warranted Assertability Manoeuvres" The Philosophical Quarterly, 55(219):329-336
  • Brady, Michael & Pritchard Duncan (2004), " Epistemological Contextualism: Problems And Prospects" The Philosophical Quarterly, 55(219): 161-171
  • Brown Jessica (2004)"The Death of Invariantism" The Philosophical Quarterly, 55(219):264-285
  • ___________ (2004b) "Williamson On Luminosity And Contextualism" The Philosophical Quarterly, 55(219):320-327
  • Brueckner, Anthony (2004) "Contextualism, Hawthorne's Invariantism and Third-person Cases" The Philosophical Quarterly, 55(219): 315-318
  • Cohen, Stewart (2004) "Knowledge, Assertion, and Practical Reasoning", Philosophical Issues, 14, Epistemology 2004.
  • ____________ ("005) "Knowledge Speaker and Subject", The Philosophical Quarterly, 55(219): 200-212.
  • DeRose, Keith (1999) "Contextualism: An Explanation and Defense" in J. Greco and E. Sosa, ed., The Blackwell Guide to Epistemology, Blackwell .
  • ____________ (2004) "The Ordinary Language Basis for Contextualism, and the New Invariantism", The Philosophical Quarterly, 55(219):172-198
  • Harman, Gilbert & Sherman, Brett (2004) "Knowledge, Assumptions, Lotteries" Philosophical Issues, 14, Epistemology 2004.
  • Hawthorne, John (2004) Knowledge and Lotteries              Clarendon Press, OUP
  • ______________ (2004b) Replies Philosophical Issues, 14, Epistemology.
  • Lewis, David(1996) 'Elusive knowledge', Australasian Journal of Philosophy, 74: 4, 549 -567.
  • MacFarlane, John (2009) "Non-Indexical Contextualism" Synthese, Volume 166, Number 2 (January 2009), pages 231-250.
  • Moser, Paul K (ed.) (2002), The Oxford Handbook of Epistemology. Oxford University Press,
  • Stanley, Jason (2005)              Knowledge and Practical Interests, Clarendon Press, OUP
  • Vogel, Jonathan (2004) 'Speaking of Lotteries' Philosophical Issues, 14, Epistemology 2004.
  • Williamson, Timothy (2000) Knowledge and it Limits Clarendon Press, OUP
  • __________________(2005b), "Contextualism, Subject-Sensitive Invariantism and Knowledge of Knowledge", The Philosophical Quarterly, 55(219): 213-35.
  • Wright, Crispin. (2004)"Contextualism and Scepticism:Even-handedness, Factivity and Surreptitiously Raising Standards" The Philosophical Quarterly, 55(219):237-262
  • _____________, (2008)"Relativism about Truth Itself: Haphazard Thoughts about the Very Idea" , in Manuel Garcia-Carpintero and Max Kolbel, eds., Relativising Truth, OUP.


  1. DeRose, Keith. "Contextualism :An Explanation and Defense" pp 1-2, in J Greco and E Sosa, ed, "The Blackwell Guide to Epistemology", Blackwell ,1999
  2. Wright, Crispin. "Contextualism, Even Handedness and Factivity"  p239 in Phil. Quarterly Vol 55 No 219,
  3. Hawthorne, John ."Knowledge and Lotteries" p158  Clarendon Press, Oxford, 2004
  4. DeRose, Keith, "The Ordinary Language Case for Contextualism" p 189, in Phil Quart Vol  Apr 2005
  5. Vogel, Jonathan "Speaking of Knowledge" in Philosophical Issues, 14, Epistemology, 2004.
  6. Hawthorn John, op cit, passim, see especially p105 fn.120 and pp137-139
  7. Hawthorne does not present Jamie's Elm St car adventure, but Vogel expressly constructed it to parallel Hawthorne's Safari Ticket Lottery paradox example (Hawthorne 2004 op.cit pp160-162). I stick with Vogel as I find his example more accessible for the same arguments, but refer to Hawthorne mutatis mutandis for brevity
  8. Hawthorne (2004) p162
  9. Williamson, Timothy "Contextualism, Subject-sensitive Invariantism and Knowledge of Knowledge"pp 220 Phil.Quart. Vol 55 No 219
  10. Stanley acknowledges sensitivity to the temporal context of utterance, but categorises this as trivial.
  11. Stanley uses "Attributor" but "Ascriber" is used in this paper for consistency.
  12. Stanley, Jason "Knowledge and Practical Interests" pp96-98, Clarendon Press, Oxford, 2005. (minor liberties, such as omission of the character Sarah, have been taken for brevity.)
  13. Hawthorne (2004) pp162-166.
  14. Stanley (2005) p99.
  15. DeRose, Keith (Forthcoming) cited in Stanley (2005) p 100.
  16. Hawthorne John, "Knowledge and Lotteries" pp77-80. Clarendon 2004.

Article name: Meta-epistemology essay, research paper, dissertation