Abstract
There is a long-standing view in epistemology that perception is a way of knowing (WOK). There is a less long-standing but increasingly popular view that knowledge attributions have a relativist semantics (RKA). I discuss three things here. First, I show that it is a consequence of the logic of RKA that WOK and RKA are incompatible. Second, I argue that, even if WOK is incompatible with the main rivals to RKA, this is not a consequence of the logics of these views. RKA comes with a hitherto unobserved philosophical consequence that its main rivals do not come with. Third, I consider some responses to the effect that it does not matter that RKA comes with this consequence, and argue that they are all unsatisfactory as they stand. I conclude that, at the very least, the onus is placed on relativists to engage with more epistemology than they currently do to show that we do not need to worry about the incompatibility of WOK and RKA. I conclude with some suggestions about the broader relevance and implications of this paper.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
I use ‘knowledge attributions’ to refer to both knowledge attributions proper and knowledge denials.
I am grateful to an anonymous reviewer for pressing me to spell out this aspect of the dialectic.
For a given circumstance of evaluation, read down from the worlds columns (‘\(\hbox {w}_{\mathrm{c1}}\)’, ‘\(\hbox {w}_{\mathrm{c2}}\)’) and across from the epistemic standards rows (‘\(\hbox {e}_{\mathrm{a1}}\)’, ‘\(\hbox {e}_{\mathrm{a2}}\)’); the circumstances at which \(\hbox {ALICE}_{1}\) and \(\hbox {ALICE}_{2 }\)are true are marked with ‘\(\hbox {ALICE}_{1}\)’ and ‘\(\hbox {ALICE}_{2}\)’ respectively.
I am grateful to an anonymous reviewer for pointing out that there could be an epistemic standard that is so high that if one knows that p relative to that epistemic standard, then one knows that p relative to any epistemic standard, i.e. one could be in the strongest possible epistemic position with respect to p. For instance, Alice could be in the strongest possible epistemic position with respect to the proposition that Humpty Dumpty had a fall. Here it would not be the case that there must be a circumstance of evaluation at which \(\hbox {ALICE}_{1}\) is true but \(\hbox {ALICE}_{2}\) is false. Taking our toy model from the main text for demonstration, \(\hbox {ALICE}_{2}\) could be true at both \(<\hbox {w}_{\mathrm{c1}},\hbox {e}_{\mathrm{a1}}>\) and \(<\hbox {w}_{\mathrm{c1}},\hbox {e}_{\mathrm{a2}}>\), since Alice’s epistemic position could meet both standards \(\hbox {e}_{\mathrm{a1}}\) and \(\hbox {e}_{\mathrm{a2}}\). However,recall that what is at stake here is the compatibility of WOK and RKA, so in order to make this response work for RKA one has to suppose that for any pair of propositions expressed by a perception sentence and a knowledge attribution, like the propositions \(\hbox {ALICE}_{1}\) and \(\hbox {ALICE}_{2}\), the subject of these sentences is in the strongest possible epistemic position with respect to the object of these sentences. In other words, one has to commit to what we can call perception infallibilism (PI): for any subject x and any proposition p, that x perceives that p puts x in the strongest possible epistemic position with respect to p. PI is controversial and obviously not part of WOK or RKA. Of course, there may well be good reasons to believe that PI is true, but the onus would be on relativists to give these reasons. Moreover, we will see in Sect. 5 that the main rivals to RKA do not have the same issues with WOK, so it follows that they do not need to appeal to PI. Therefore, even if relativists could give good reasons for accepting PI, RKA would still come with some controversial philosophical consequences that its main rivals do not.
We have defined WOK in terms of logical implication, but one might wonder if WOK could be defined in terms of an alternative notion and whether the present argument would still go through in that case. For example, WOK*: for any subject x and sentence \(\Phi \), \(\ulcorner \) x knows that \(\Phi \)\(\urcorner \) is a logical consequence of \(\ulcorner \) x perceives that \(\Phi \)\(\urcorner \), where logical consequence is defined as preservation of truth at a context of utterance and a context of assessment. Firstly, note that WOK* works with sentences rather than propositions. We might wonder whether this gets to the heart of the view that perceiving entails knowing. Specifically, we might want to say that, unlike WOK*, this view is about more than just a relation between perception sentences and knowledge attributions. Secondly, our argument still goes through with WOK*. Recall that circumstances of evaluation are defined as pairs of the world of the context of utterance and an epistemic standard set by a context of assessment. Suppose for sentences (1) and (2) the context of utterance is \(\hbox {c}_{1}\) and the context of assessment is \(\hbox {a}_{2}\). By analogy with our toy semantic model above (see Table 1), it is possible for sentence (1) to be true at \(\hbox {c}_{1}\) and \(\hbox {a}_{2}\) and for sentence (2) to be false at \(\hbox {c}_{1}\) and \(\hbox {a}_{2}\), in which case truth is not preserved at a context of utterance and a context of assessment and therefore (2) is not a logical consequence of (1). Thanks to an anonymous referee for pressing this point and for defining WOK*.
We might think that there are facts about perception, for example low-level facts about a subject’s subpersonal perceptual processing, that determine the strength of a subject’s epistemic position, and these facts can be expressed by assessment-insensitive sentences (I am grateful to a reviewer for this suggestion and wording). However, notice that if we follow this suggestion through, then that one is in a certain perceptual process should logically imply that one perceives something. For instance, that Alice is in a certain perceptual process - call this proposition \(\hbox {ALICE}_{3}\) - should logically imply that Alice is in a certain perceptual process (call this proposition \(\hbox {`ALICE}_{3}\)’) should logically imply that Alice perceives something; should logically imply \(\hbox {ALICE}_{1}\). Now, just as \(\hbox {ALICE}_{1}\) is expressed by sentence (1), so presumably \(\hbox {ALICE}_{3}\) is expressed by (3): ‘Alice is in a certain perceptual process’. But if we hold that (3) is assessment-insensitive and (1) is assessment-sensitive, then we get the exact analogue of the problem discussed in Sect. 3, viz. \(\hbox {ALICE}_{3}\) does not logically imply \(\hbox {ALICE}_{1}\) after all. More generally, it follows that that one is in a certain perceptual process does not logically imply that one perceives something (cf. Sect. 7 and Tarasov 2014: pp. 177–178).
I am grateful to a reviewer for querying whether we cannot be convinced the truth-values of these propositions might co-vary in the right way after all. They have in mind Dretske (1970: pp. 1015–1016) and Goldman (1976: pp. 772–73) style scenarios in which a subject x appears to perceive zebras or barns, but they turn out to be cleverly painted mules or barn facades. The suggestion is that in such cases we would say that x only appears to perceive zebras or barns, but does not really, just as we would say that x only appears to know there are zebras or barns in front of x, but does not really. This seems to show that we are willing to retract an utterance of a perception sentence just when we are willing to retract an utterance of a knowledge attribution, which in turn might show that the truth-values of the propositions expressed by these sentences co-vary in the right way, and in particular that the truth-values of the propositions expressed by knowledge attributions and perception sentences are both sensitive to changes in epistemic standards. In fact, I think these examples work in our favour. Firstly, they lend indirect support to WOK; we would not bother retracting an utterance of a knowledge attribution that was based on an utterance of a perception sentence if we did not think that perception was sufficient for knowledge in the first place. Secondly, the examples only work if we assume that the only basis for the claim that x knows that there are zebras or barns is that x perceives that there are. But often the basis for knowing something is non-perceptual, e.g. deductive or testimonial, or a combination of perceptual and non-perceptual. Thus, we can argue that whatever constitutes x’s epistemic position and, as a corollary, an epistemic standard, is richer than whatever might constitute a standard for perception (if there is such a thing). As a result, the truth-values of the propositions expressed by perception sentences and knowledge attributions are not both sensitive to an epistemic standard. Lastly, recall that these examples are cases where x appears to perceive, but does not perceive. Although they might be suggestive of what happens in cases where x actually perceives, they do not say anything directly about these cases.
For a detailed comparison of the semantics of the views discussed here, see Tarasov (2014).
I am very grateful to an anonymous referee for pressing me to spell out the wider implications of this paper, and especially for suggesting an argument along the lines of the second point below.
I am grateful to a reviewer for reminding me that, if we follow Tim Williamson (2000: p. 34), then WOK can be understood as an instance of a more general schema, which we can call Ways of Factiving (WOF): for any subject x and any proposition p, that x \(\upvarphi \)s that p logically implies that x knows that p, where \(\upvarphi \) is any factive verb (e.g. ‘realise’, ‘recognise’ etc.). The reviewer suggests that if it is a consequence of my findings that RKA is incompatible not only with WOK, but also with WOF, then this immediately makes my findings less plausible. I disagree; if my findings generalise, as I suggest above they do, then we should expect consequences just like the incompatibility of WOF and RKA, and these consequences should be seen as an additional issue for RKA.
References
Brown, J. (2006). Contextualism and warranted assertibility manoeuvres. Philosophical Studies, 130(3), 407–435.
Chisholm, R. (1989). Theory of knowledge. Foundations of philosophy (3rd ed.). Upper Saddle River: Prentice-Hall, Inc.
Daly, C., & Liggins, D. (2011). Deferentialism. Philosophical Studies, 156(3), 321–337.
DeRose, K. (2009). The case for contextualism: Knowledge, skepticism, and context (Vol. 1). Oxford: Oxford University Press.
Dretske, F. (1970). Epistemic operators. The Journal of Philosophy, 67(24), 1007–1023.
Edgington, D. (1995). On conditionals. Mind, 104(414), 235–329.
Goldman, A. (1976). Discrimination and perceptual knowledge. The Journal of Philosophy, 73(20), 771–791.
Hawthorne, J. (2004). Knowledge and lotteries. Oxford: Oxford University Press.
Kaplan, D. (1989). Demonstratives: An essay on the semantics, logic, metaphysics, and epistemology of demonstratives and other indexicals. In J. Almog, J. Perry, & H. Wettstein (Eds.), Themes from Kaplan (pp. 481–563). Oxford: Oxford University Press.
Kölbel, M. (2008). Motivations for relativism. In M. García-Carpintero & M. Kölbel (Eds.), Relative truth (pp. 1–38). Oxford: Oxford University Press.
Kolodny, N., & MacFarlane, J. (2010). Ifs and oughts. The Journal of Philosophy, 107(3), 115–143.
Lewis, D. (1980). Index, context, and content. In S. Kanger & S. Öhman (Eds.), Philosophy and grammar (pp. 79–100). Dordrecht: Springer.
MacFarlane, J. (2005). The assessment sensitivity of knowledge attributions. In T. Gendler & J. Hawthorne (Eds.), Oxford studies in epistemology: Volume 1 (pp. 197–233). Oxford: Oxford University Press.
MacFarlane, J. (2014). Assessment sensitivity: Relative truth and its applications. Oxford: Oxford University Press.
Moore, G. E. (1953). Some main problems of philosophy. London: George Allen & Unwin Ltd.
Russell, B. (1948). Human knowledge: Its scope and limits. New York: Simon and Schuster.
Sosa, E. (1980). The raft and the pyramid: Coherence versus foundations in the theory of knowledge. Midwest Studies In Philosophy, 5(1), 3–26.
Stephenson, T. (2007). Indicative conditionals have relative truth conditions. Proceedings from the Annual Meeting of the Chicago Linguistic Society, 43(1), 231–242.
Tarasov, L. (2013). Contextualism and weird knowledge. The Philosophical Quarterly, 63(252), 565–575.
Tarasov, L. (2014). The Semantics of Knowledge Attributions: A Defence of Moderate Invariantism. Ph.D. thesis, The University of Manchester, Manchester.
Turri, J. (2010). Does perceiving entail knowing? Theoria, 76(3), 197–206.
Williams, M. (1996). Unnatural doubts: Epistemological realism and the basis of scepticism. Princeton: Princeton University Press.
Williamson, T. (2000). Knowledge and its limits. Oxford: Oxford University Press.
Wright, C. (2008). Relativism about truth itself: Haphazard thoughts about the very idea. In M. García-Carpintero & M. Kölbel (Eds.), Relative truth (pp. 157–185). Oxford: Oxford University Press.
Acknowledgements
I wrote the first draft of this paper in 2014. Since then it has been through a long process of re-write and refinement. I am grateful to everyone who has contributed to this process, especially several anonymous reviewers, Jessica Brown, David Liggins, Joel Smith, Graham Stevens and Thomas Uebel, and also to John MacFarlane and Crispin Wright for some early discussions that got me thinking about these issues. I am especially grateful to Rob Knowles, Thomas Uebel and Jack Winter for discussions (I am sorry for bothering them for three years). Separate thanks to Rob, who also read a near-final draft and provided extensive feedback and encouragement.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix I: Some simplifications and assumptions used throughout this paper
The framework that we have worked with throughout this paper simplifies MacFarlane’s view in several respects. In order to pre-empt any claims that we have worked with an inaccurate characterisation of relativism, below I explain which simplifications were made and why. I also note that the argument should generalise beyond MacFarlane’s brand of relativism to other relativist frameworks.
Firstly, throughout this paper we have appealed to the notion of propositions. Relativists accept the standard assumptions about propositions. For example, John MacFarlane accepts that ‘sentences are true or false because they express propositions that are true or false’ and that ‘[propositions] are the contents of assertions and beliefs, and the things we call “true” or “false” in ordinary discourse’ (MacFarlane 2014: p. 71). However, relativists are also careful to distance themselves from any commitments to the nature of propositions that are not essential to their theory (see e.g. MacFarlane 2014: p. 72). We can follow suit; although we have worked with propositions, nothing we say should be taken to commit us to claims about their nature.
Secondly, I sketched the relativist framework in Sect. 2 on the basis of David Kaplan (1989) well-known semantic framework for the analysis of indexical expressions and MacFarlane’s (2014: Ch. 4) modifications to this framework. Kaplan’s framework makes use of four key notions: (i) contexts, understood to represent possible occasions of utterance of a sentence and defined as sequences of at least individuals, times, locations and possible worlds (Kaplan 1989: p. 543); (ii) characters, defined as functions from contexts to propositions (Kaplan 1989: pp. 505–506); (iii) propositions (contents in Kaplan’s terminology), defined as functions from circumstances of evaluation to truth-values (Kaplan 1989: pp. 500–501); and (iv) circumstances of evaluation, defined as sequences of the worlds and times parameters of the contexts and a parameter for assignments of values to variables, where each parameter can vary independently of another (Kaplan 1989: pp. 502–504).
We need to note three things in relation to this framework. Firstly, MacFarlane (2014: Ch. 3) defines several key notions—most relevantly for our purposes, the notion of logical implication—using David Lewis (1980) rather than Kaplan’s framework. Lewis’ framework works only with sentences and open formulas, not propositions. However, we have already seen that MacFarlane accepts that sentences have truth-values because they express propositions and propositions have truth-values. Moreover, his discussion of propositions indicates that what he has to say about sentences extends mutatis mutandis to propositions (see MacFarlane 2014: Ch. 4, especially pp. 90–92). In light of this, in Sect. 3 I extended the relevant claims MacFarlane makes about sentences—in particular, his definition of logical implication—to propositions. Secondly, MacFarlane (2014: pp. 58–60) distinguishes semantics proper from what he calls post-semantics. Roughly put, semantics proper deals with the compositional aspect of semantics, i.e. the contribution the meanings of sentential constituents make to the meanings of the sentences in which they occur. Post-semantics deals with the aspect of semantics which, as MacFarlane (2014: p. 57) puts it, has ‘direct pragmatic relevance’, i.e. the aspect which tells us whether a sentence is true or false as used on a particular occasion. The distinction affects the way one should define truth for sentences and propositions (see e.g. MacFarlane 2014: pp. 67, 91), but it is not important for our purposes. In light of this, I sketched the relativist framework in Sect. 2 in a way that is consistent with the distinction, but that ignores it. Thirdly, we have already noted that Kaplan treats contexts as sequences of at least individuals, times, locations and possible worlds, and circumstances of evaluation as sequences of the worlds and times parameters of the contexts and a parameter for assignments of values to variables. We did not need such a rich picture of contexts and circumstances for our purposes; in particular, I did not countenance individuals, times, locations or assignments in my sketch of the relativist framework in Sect. 2.
Next, note that there is a distinction between what Max Kölbel (2008: pp. 18–20) calls moderate and radical relativism, or what MacFarlane (2014: p. 88) calls non-indexical contextualism and relativism. The difference is roughly this: whereas moderate relativists add unorthodox parameters (i.e. parameters besides possible worlds and times) to circumstances of evaluation and say that the values of these parameters are determined by contexts of utterance, radical relativists say that the values of both orthodox and unorthodox parameters of circumstances of evaluation may be determined by so-called contexts of assessment. Radical relativism is the most popular relativist account of the semantics of knowledge attributions (MacFarlane 2005, 2014: Ch. 8), so it is the view I have worked with throughout this paper. However, as we saw in Sects. 3 and 4, my claim that WOK an RKA are incompatible depends mainly on the idea that the proposition expressed by a knowledge attribution and the propositions expressed by some other sentences change truth-values with respect to different parameters of the circumstances of evaluation, where the values of these parameters vary independently of each other. This idea is common to both moderate and radical relativists, so what I had to say about radical relativism about knowledge attributions should also apply to moderate relativism about knowledge attributions.
Finally, MacFarlane (e.g. 2014: pp. 49–52) points out that the key aspect of radical relativism is not the addition of unorthodox parameters to circumstances of evaluation, but the idea that the values of the parameters of circumstances of evaluation may be determined by contexts of assessment. However, we just noted that radical relativists think that the proposition expressed by a knowledge attribution changes truth-values with respect to an unorthodox parameter of circumstances of evaluation the value of which is determined by contexts of assessment, so, given that we are focusing on knowledge attributions here, it does not matter if we ignore the thought that the values of orthodox parameters of circumstances of evaluation may be determined by contexts of assessment.
Appendix II: Demonstration of a point in Sect. 4
In Sect. 4 we argued that there are good reasons to expect the truth-values of the propositions expressed by a knowledge attribution and a perception sentence to vary with respect to different parameters determined by contexts of assessment, where the values of these parameters vary independently of each other. On this basis, we inferred that we should expect the truth-values of these propositions not to co-vary in the right way, and therefore for non-local relativism to be incompatible with WOK. Here I demonstrate that this inference is sound.
Make all the assumptions we made in Sect. 3, except now suppose that both sentences (1) and (2) are assessment-sensitive, the truth-value of \(\hbox {ALICE}_{1}\) varies with what we will call visual standards \(\hbox {v}_{\mathrm{a}}\) determined by contexts of assessment, but not epistemic standards, and the truth-value of \(\hbox {ALICE}_{2}\) varies with epistemic standards \(\hbox {e}_{\mathrm{a}}\) determined by contexts of assessment, but not visual standards. (This is consistent with the argument in Sect. 4 that the truth-values of \(\hbox {ALICE}_{1}\) and \(\hbox {ALICE}_{2}\) vary with respect to different parameters of circumstances of evaluation; there we also briefly touched on the notion of visual standards). Our semantic framework in Sect. 2 can be straightforwardly extended to accommodate this: circumstances of evaluation can be understood as triples of worlds of the context of utterance and visual and epistemic standards determined by contexts of assessment: \(< \hbox {w}_{\mathrm{c}}, \hbox {v}_{\mathrm{a}}, \hbox {e}_{\mathrm{a}},>\). However, for the sake of the argument, let us assume that if \(\hbox {ALICE}_{1}\) and \(\hbox {ALICE}_{2}\) are true, they are true at all possible worlds, and therefore drop worlds from our model. We know that it is the interaction between the values of \(\hbox {v}_{\mathrm{a}}\) and \(\hbox {e}_{\mathrm{a}}\) that needs to do the work in order for the argument in Sect. 4 to go through, and dropping worlds from our model will help to keep things as simple as possible.
Now imagine we are working in a toy semantic model that has four, and only four, circumstances of evaluation: \(< \hbox {v}_\mathrm{a1}, \hbox {e}_\mathrm{a1}>,< \hbox {v}_\mathrm{a1}, \hbox {e}_\mathrm{a2}>,< \hbox {v}_\mathrm{a2}, \hbox {e}_\mathrm{a1} >, < \hbox {v}_\mathrm{a2}, \hbox {e}_\mathrm{a2}>\). Assume \(\hbox {ALICE}_{1}\) is true at \(< \hbox {v}_\mathrm{a1}, \hbox {e}_\mathrm{a1}>\). Then \(\hbox {ALICE}_{1}\) is also true at \(< \hbox {v}_\mathrm{a1}, \hbox {e}_\mathrm{a2}>\) and false at \(< \hbox {v}_\mathrm{a2}, \hbox {e}_\mathrm{a1}>\) and \(< \hbox {v}_\mathrm{a2}, \hbox {e}_\mathrm{a2}>\). Assume that \(\hbox {ALICE}_{2}\) is likewise true at \(< \hbox {v}_\mathrm{a1}, \hbox {e}_\mathrm{a1}>\). Then \(\hbox {ALICE}_{2}\) is also true at \(< \hbox {v}_\mathrm{a2}, \hbox {e}_\mathrm{a1}>\) and false at \(< \hbox {v}_\mathrm{a1}, \hbox {e}_\mathrm{a2}>\) and \(< \hbox {v}_\mathrm{a2}, \hbox {e}_\mathrm{a2}>\). This is represented in Table 2.
There is a circumstance of evaluation, viz. \(< \hbox {v}_\mathrm{a1}, \hbox {e}_\mathrm{a2}>\), at which \(\hbox {ALICE}_{1}\) is true but \(\hbox {ALICE}_{2}\) is false, so \(\hbox {ALICE}_{1}\) does not logically imply \(\hbox {ALICE}_{2}\).
Rights and permissions
About this article
Cite this article
Tarasov, L. Semantic relativism and ways of knowing. Synthese 196, 2089–2109 (2019). https://doi.org/10.1007/s11229-017-1572-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11229-017-1572-z