Forced-March Sorites Arguments and Linguistic Competence

Forced-March Sorites Arguments and Linguistic CompetenceJonas Åkerman†

Abstract

Agent relativists about vagueness (henceforth ‘agent relativists’) hold that whether or not anobject x falls in the extension of a vague predicate ‘P’ at a time t depends on the judgementaldispositions of a particular competent agent at t. My aim in this paper is to critically examinearguments that purport to support agent relativism by appealing to data from forced-marchSorites experiments. The most simple and direct versions of such forced-march Sorites argu-ments rest on the following (implicit) premise: If competent speakers’ judgements vary in acertain way, then the extensions of ‘P’ as used by these speakers must vary in the same way. Thispremise is in need of independent support, since otherwise opponents of agent relativism cansimply reject it. In this paper, I focus on the idea that one cannot plausibly reject this premise, asthat would commit one to implausible claims about linguistic competence. Against this, I arguethat one can accommodate the data from forced-march Sorites experiments in a way that iscompatible with a plausible picture of linguistic competence, without going agent relativist.Thus, there is reason to be sceptical of the idea that such data paired with considerations aboutlinguistic competence can be invoked in order to lend any solid support to agent relativism.Forced-march Sorites arguments of this kind can, and should be, resisted.

1. Introduction and overview

The aim of this paper is to critically examine a kind of argument that has been putforward in different forms by various proponents of agent relativism about vague-ness (henceforth ‘agent relativism’). According to agent relativism, the extensionsof vague predicates are relative to particular agents and times. More precisely,agent relativists hold that whether or not an object x falls in the extension of avague predicate ‘P’ at a time t depends on the judgemental dispositions of aparticular competent agent at t. This is a view that has become increasinglypopular in recent years, although it remains controversial. It is also a view thatcomes in various forms, some of which will be briefly introduced in the moredetailed presentation given in section 2.1

The kind of argument for agent relativism to be examined in what follows willbe referred to as forced-march Sorites arguments. A forced-march Sorites argu-ment is an argument that purports to support agent relativism by appealing to datafrom a forced-march Sorites experiment, i.e. an experiment in which a competentspeaker is led through a Sorites series and is forced to make a judgement of the

† Department of Philosophy, Stockholm University, SE-106 91 Stockholm, Sweden;Email: [email protected]

1 References to the relevant literature will be given in the course of this paper.

bs_bs_banner dialecticadialectica Vol. 67, N° 4 (2013), pp. 403–426

DOI: 10.1111/1746-8361.12038

© 2014 The Author dialectica © 2014 Editorial Board of dialectica

mailto:[email protected]

form ‘x is P’ or ‘x is not P’, for each x in the series.2 These experimentsand arguments are further described in section 3, where I also argue that themost simple and direct versions of forced-march Sorites arguments fail. Thereason for this failure is that they rely on an assumption that anti-agent relativistscan happily reject, namely the assumption that if competent speakers’ judge-ments vary in certain ways, then the extensions of ‘P’ as used by these speakersmust vary in the same way (a more precise formulation of this assumption isgiven below).

In section 4, I turn to the idea that considerations about linguistic compe-tence can lend additional, independent support to this assumption. The crucialclaim on the part of the proponent of the forced-march Sorites argument can bestated as follows: If we were to deny the assumption on which the argumentrelies, we would be committed to a view according to which ordinary speakersare incompetent in their use of vague language, and since such a view wouldbe absurd, we had better accept the assumption. This argumentative strategyis closely related to the more general idea behind forced-march Sorites argu-ments, namely the idea that only agent relativism can make sense of the datafrom forced-march Sorites experiments. I challenge this idea, and argue that theconstraint on linguistic competence invoked by the proponent of the forced-march Sorites argument is too strict; a speaker may remain competent evenwhen she makes semantically incorrect judgements. Thus, the argument can beresisted.

In section 5, I address two outstanding questions. The first concerns therelationship between competent use and extensions. I argue that the general andplausible claim that the extensions of vague predicates are somehow determinedby how competent speakers apply them is compatible with various non-agent-relativist theories of vagueness. This means that agent relativists are not in a betterposition than their rivals when it comes to accommodating this claim. The secondquestion concerns how one may handle the apparent contradictions that competentspeakers end up with in a standard forced-march Sorites experiment, withoutgoing agent relativist. I present a couple of different options and discuss thembriefly.

2. Agent relativism

Agent relativists hold that for any vague predicate ‘P’ and any object x, whether ornot x falls in the extension of ‘P’ at a time t depends on the judgemental disposi-tions of a particular competent agent at t. This means that from the agent-relativistpoint of view, a proper understanding of vague predicates requires that we

2 The term ‘forced-march Sorites’ is due to Horgan (1994).

Jonas Åkerman404


relativize their extensions to particular agents and times. Laurence Goldstein, arecent proponent of agent relativism, puts the idea as follows:3

[T]o ask ‘What is the snapping point, simpliciter, for unrelenting noise at 80 deci-bels?’ would be nonsense. And similarly, it is nonsense – a category mistake – to askwhere the red/non-red borderline is, not for a particular agent but objectively. Thesame can, of course, be said about the heap/non-heap borderline, the infant/non-infant borderline, the wealthy/non-wealthy borderline etc.: no such objective border-lines exist. (Goldstein 2012, 63)

What is being claimed in this passage is not that we cannot make sense of vaguepredicates having extensions at all, but merely that we cannot make sense ofvague predicates having extensions that are not relativized to particular agentsand times. Just as it does make sense to ask for the point at which a particularagent at a particular time will snap if exposed to unrelenting noise at 80 deci-bels, it does make sense to ask for the borderline of the extension of any vaguepredicate ‘P’ relative to a particular agent at a particular time. Consider a Soritesseries for ‘P’, i.e., a series ranging from items that are clearly P to items that areclearly non-P, and such that adjacent members of the series are very similar(perhaps even indiscriminable) with respect to the dimension of comparison rel-evant to ‘P’ (e.g. colour if P = red, or number and arrangement of hair on thehead if P = bald). For any time t and agent A there is a point in the series atwhich A would (be disposed to) switch from a ‘P’-verdict to a non-‘P’-verdict,and according to agent relativism, that point marks a (non-objective) borderlinefor ‘P’ relative to A and t (cf. Goldstein 2012, 61–63). Since the switching pointfor a specific agent may shift over time, and since the switching point at aspecific time may be different for different agents, the extension of any vaguepredicate ‘P’ may vary across agent and times, and should thus be relativized toparticular agents and times; the extensions of ‘P’ are ‘subjective’ rather than‘objective’.

So, on the agent relativist picture, there is a very intimate link between theextensions of vague predicates and the way that competent speakers use them.Let us take a closer look at how one may conceive of this link. According toDiana Raffman (1994, 69–70; 1996, 182), the extensions of vague predicates aredetermined by certain aspects of competent agents’ psychology, namely theirjudgemental dispositions. Such dispositions constitute what she calls “psycho-logical” or “internal” contexts, and the relation between such dispositions and

3 As an anonymous referee for dialectica pointed out, it may be questioned whether‘snapping points’ and boundaries of extensions of vague predicates really are similar in therespects that would be required in order to make a comparison of this kind pertinent to adiscussion of the problems of vagueness. For discussion, see Fine (1975, 286) and Egré(forthcoming).

Forced-March Sorites Arguments and Linguistic Competence 405


the extensions of a vague predicate ‘P’ is simply that, relative to an agent A, anobject o falls in the extension of ‘P’ at t, if and only if A is disposed to apply‘P’ to o at t.4

On Raffman’s view, then, the psychological context co-determines the agent’sactual judgement and the semantic status of the item judged. In contrast, on theagent-relativist views defended by Scott Soames (1999) and Stewart Shapiro(2006), the judgemental dispositions determine the semantic status of the itemjudged indirectly, via the actual (public) judgement.5 This is but one of manyinteresting differences among the agent relativist theories in the offing, but forpresent purposes we need not go further into this matter. The crucial point is thatall of these theories take there to be an intimate link between the extensions ofvague predicates and the way that competent speakers use them; they all comprisethe thesis that vague predicates are judgement dependent in a way that guaranteesthat ordinary competent speakers’ actual applications of vague predicates do notdiverge from their extensions.6

That ‘P’ is judgement dependent is not supposed to entail that competentagents have complete access to, and control over the extensions with which theyuse ‘P’. First, the conditions in which the judgements are made may be less thanoptimal. For instance, the lighting conditions may be abnormal, the agent’s clas-sificatory abilities may be temporarily substandard, or the agent may be misin-formed about relevant facts about the subject matter or about what comparisonclass is operative. Second, even when the conditions are ideal, extensions are likelyto shift quite often, as the relevant dispositions may shift whenever agents shifttheir attention from one item to another. The agents will typically not be aware ofsuch subtle shifts, and much less be able to track them. Third, Raffman suggests,plausibly enough, that competent speakers are never disposed to judge two adja-cent items in a Sorites series differently at the moment when they are consideringthem simultaneously as a pair. This makes the switching point (i.e., the non-objective borderline) elusive in the sense that if an agent was previously disposedto switch between two items, the very considering of them as a pair will trigger ashift to the effect that the switching point is not there anymore (Raffman 1996,

4 In fact, Raffman’s view as developed in the 1996 paper is a bit more complicated, butas argued in Åkerman (2009, 81–90), it seems that the simpler version of the view is the moreplausible one.

5 It is interesting to note that although Shapiro (2006, 26–27) focuses on more commu-nal aspects of language use, he takes Raffman’s account to be “more basic,” as the collectivejudgements in Shapiro’s conversational version of the forced-march Sorites experiment aredetermined by the individual judgements of the conversationalists.

6 One could also use the term ‘response dependence’ to refer to this idea (cf. Goldstein2012, 63). See Åkerman (2009, 36–38) for a more detailed discussion of different versions ofjudgement dependence.

Jonas Åkerman406


178).7 Fourth, since competent agents cannot predict the dispositional shifts, theycan never know the switching point in advance. At any time t, they are ignorant ofthe (non-objective) boundary’s location at t.

The first of these points shows that although there is a sense in which judge-ment dependence entails that competent agents are infallible, the infallibility is, asRaffman puts it, “highly circumscribed” (1994, 70–71). After all, normal compe-tent speakers make all sorts of mistakes in their daily application of the predicatesof their own language, and sometimes they even (genuinely) contradict them-selves. On a charitable understanding of agent relativism, there should be room formistakes on the part of competent speakers due to unfavourable external circum-stances, misinformation, or (temporary) dips in the agent’s level of performance.Let us refer to conditions in which such sources of error have been eliminated as‘optimal conditions’. For present purposes, then, the judgement-dependence thesismay be formulated as follows:

(JD) If ‘P’ is a vague predicate, then, provided that the conditions are optimal, thejudgemental dispositions of an ordinary competent speaker S determine theextensions of ‘P’ in a way that guarantees that if S applies ‘P’ to x at t, then xfalls in the extension of ‘P’ (as used by S at t).

According to (JD), an ordinary competent speaker’s actual application of ‘P’ at atime t cannot diverge from the extension with which S uses ‘P’ at t, as long as theconditions are optimal. Of course, on the agent-relativist view, the extension hereis ‘subjective’ rather than ‘objective’; hence the relativization to S.

The idea that vague predicates are agent relative has become increasinglypopular during the last couple of decades, most notably through its central role invarious contextualist theories of vagueness.8 But although this idea is part andparcel of (standard) contextualism about vagueness, one need not endorse acontextualist view in order to accept agent relativism. For instance, Goldsteinagrees with (standard) contextualists that vague predicates are agent relative, buthe is bound to disagree with their claim that vagueness consists in a form ofcontext sensitivity since he takes vagueness to remain even when all contextualfactors are held fixed. He illustrates this point with a science fiction scenario inwhich a subject is presented, at time t1, with the first item in the Sorites series and,after the verdict has been passed “is transported back to t1 when the subject’s brainreconfigures to the exact state, b1, it was in at that time” (Goldstein 2012, 61). Thisprocedure is then repeated for each of the items in the series. Goldstein maintainsthat although all contextual variation is eliminated in this scenario, vagueness ispreserved. Thus he takes sides with those critics of contextualism about vagueness

7 It has been argued, for instance by Delia Graff Fara (2000, 70–71) that this feature putsagent relativists in a particularly good position to explain the appeal of soritical reasoning. Forcriticism and further discussion, see Keefe (2007) and Åkerman (2009, 100–105; 2011; 2012).

8 For a critical survey, see Åkerman (2009; 2012).



who argue that even if vague predicates are typically sensitive to various contex-tual factors, vagueness is distinct from context sensitivity, since vagueness remainswhen the context is held fixed or the effects of the contextual variations areblocked.9

The fact that Goldstein’s approach is (irreconcilably) different from standardcontextualism about vagueness illustrates that different versions of the generalagent-relativist idea can be invoked in various ways, as part of different approachesto vagueness.10 For present purposes, however, we need not go into further detailabout the differences between the various contextualist theories on the market,11 orabout the extent to which Goldstein’s theory differs from these.

3. Forced-march Sorites arguments

In presenting and arguing for their views, agent relativists typically appeal to anexperiment of a certain kind, which has become known as the ‘forced-marchSorites’. It is often conducted as a thought experiment set up in something like thefollowing way.12 We are asked to consider what would happen if an agent – acompetent speaker of English – were to be led through a Sorites series of the kinddescribed above, and forced to make a judgement of the form ‘x is P’ or ‘x is notP’, for each x in the series.13 It is then assumed that the agent would, at some pointduring the march, switch from ‘x is P’ to ‘x is not P’. Moreover, it is assumed thatif we were to let her do several runs through the series, starting in different placesand going in different directions, the switch would occur at different points(although it would always occur outside of the range of clearly P items and clearlynon-P items). In these thought experiments, the outputs consist in what seem to bethe most plausible assumptions about how the agent would behave in the scenariodescribed. But of course, the forced-march Sorites may also be conducted as anempirical experiment, and to the extent that this has been done, the results appearto support the outputs of the thought experiments.14

9 For objections of this kind, see Williamson (1994, 215), Keefe (2000, 10), Stanley(2003), and Heck (2003, 120). For detailed discussion of these objections see Åkerman &Greenough (2010a), and Åkerman (2009, 105–121; 2012, 476–477).

10 Thanks to an anonymous referee for dialectica for pointing out the need to clarify this.11 I have done this elsewhere. See Åkerman (2009; 2012).12 For instance, see Goldstein’s sci-fi version above, or the scenarios described in

Raffman (1994; 1996).13 Of course, one could allow a wider range of responses, but in order not to complicate

matters more than necessary, let us follow Raffman (1994, 45–46; 1996, 177) in restrictingourselves to experiments in which only two forms of judgements are allowed. The judgement ‘xis not red’ may also be taken to cover all forms of ‘non-red’ verdicts, including gap- andglut-judgements (cf. Goldstein 2012, 61).

14 For instance, see Raffman (2014), and Egré et al. (forthcoming).

Jonas Åkerman408


The purpose here is not to question any of these assumptions, so presently weneed not worry about how solid these results are, or about whether or not theassumptions of the thought experiments are correct. We can safely assume thatthey are adequately grounded, empirically or otherwise, and thus treat them asconfirmed facts. Accordingly, they will henceforth be referred to as results or datarather than as assumptions.

The results of the forced-march Sorites experiment show that competent speak-ers’ actual applications of vague predicates vary across speakers as well as overtime, even when familiar contextual factors such as comparison class are heldfixed.15 Let us now turn to the question of how such results can be invoked in orderto argue for agent relativism.

Consider the following simple forced-march Sorites argument. Given thatdifferent competent speakers in a linguistic community judge the items in themiddle of the Sorites series for ‘P’ differently (even under optimal conditions), andthat one and the same speaker judges the same items differently on differentoccasions, there cannot be any shared or stable extension for ‘P’ in this community.This is what might be called a “direct” forced-march Sorites argument. It startsfrom the forced-march Sorites data – which show that there are certain variationsacross competent speakers and over time when it comes to how vague predicatesare actually applied – and then proceeds directly to the claim that the extensions ofthese predicates must vary in the same way, which in turn yields the conclusionthat there cannot be any objective, shared extension for ‘P’.16

In order to see clearly what the problem with this argument is, it will be helpfulto observe a very basic distinction between the use of a predicate and the extensionwith which the predicate is used. In the simple cases that we are concerned withhere, to use a predicate ‘P’ is to make judgements of the form ‘x is P’ and ‘x is notP’ with respect to a certain range of objects, while the extension with which ‘P’ isused by a speaker S at a time t is the set of objects of which ‘P’ is true (relative toS and t). Now, the crucial move in direct forced-march Sorites arguments like theone above can be seen to rely on the following claim concerning the connectionbetween use and extension:

15 It will be assumed throughout that the variations cannot be explained solely in termsof uncontroversial forms of context sensitivity like relativity to comparison class, which arearguably distinct from vagueness. (For instance, even if many paradigmatically vague terms, like‘tall’, are relative to comparison class, these two phenomena are analytically distinct. On the onehand, we could fix the comparison class – as in ‘is tall relative to the class of professionalAmerican basketball players in 2013’ – without eliminating vagueness. On the other hand, wecould eliminate vagueness without fixing the comparison class by stipulating that one falls in theextension of ‘tall’ if and only if one is above the (precise) average height of the members of thecontextually relevant comparison class (cf. Åkerman 2012, 471). This is something on whichboth proponents and critics of contextualism and agent relativism agree.

16 This is, in effect, how Goldstein (2012, 62–64) argues for agent relativism.



(UE) (i) If two competent speakers S and S′ use ‘P’ differently with respect to oneand the same range of objects R (under optimal conditions), then S and S′ use‘P’ with different extensions (at the respective times of judgement), and(ii) If a competent speaker S uses ‘P’ differently with respect to R at t and t′(under optimal conditions), then S uses ‘P’ with different extensions at t andt′.

Unless we accept (UE), we have no reason to accept that competent speakers use‘P’ with different extensions merely on the basis of the observation that they judgecertain items in the Sorites series differently (i.e., apply ‘P’ differently to one andthe same range of objects).

Now, if one is already a convinced agent relativist, one is not likely to findanything problematic about accepting (UE). Indeed, given (JD), competent agents’actual application of ‘P’ (under optimal conditions) cannot diverge from theextensions with which they use ‘P’, and thus any difference in actual applicationentails a corresponding difference in extension; if S judges x to be P at t, then theextension with which S uses ‘P’ at t must include x, and if S′ judges that x is notP at t, then the extension with which S′ uses ‘P’ at t cannot include x. It follows thatif two competent speakers S and S′ (or one and the same competent speaker at twodifferent times) apply ‘P’ differently to one and the same (range of) object(s)(under optimal conditions), then they thereby use ‘P’ with different extensions (atthe respective times of judgement).

However, anyone who rejects agent-relativism will conceive of these mattersquite differently. On an objectivist17 view, the extensions with which competentspeakers use ‘P’ need not be taken to be determined in accordance with (JD). Thus,an objectivist can allow that, even under optimal conditions, two competentspeakers S and S′ can use ‘P’ with the same extension even if they apply itdifferently, and that one and the same competent speaker S can use ‘P’ with thesame extension at t and t′ even if S applies ‘P’ differently at t and t′. In other words,an objectivist can, and will typically deny (UE).18

Note that in denying (UE), one need not commit oneself to any particular formof objectivism. The extensions of ‘P’ may be taken to be either sharp or blurred,determinate or indeterminate, fixed or shifting, etc., as long as they are not takento be relative to agents and times in the way that agent relativists take them to be.19

17 Here, and henceforth, the term ‘objectivist’ is used as shorthand for ‘anti-agent-relativist’.

18 An objectivist might even take the all of the judgements in the forced-march Soritesexperiment to be correct but deny that the extensions shift with the dispositions of the agent if sheis willing to make more radical deviations from classical logic and semantics. For instance, adialethist may take all of the competent judgements to be true simultaneously, despite some ofthem being in genuine contradiction with each other. (Thanks to an anonymous referee fordialectica for drawing my attention to this possibility.)

19 Contrary to what was assumed by an anonymous referee for dialectica, the plausi-bility of denying (UE) does not depend on the plausibility of epistemicism. It merely depends on

Jonas Åkerman410


Denying (UE) simply amounts to leaving room for the possibility that competentspeakers’ application of ‘P’ at t may deviate from the extension with which theyuse ‘P’ at t. Since all objectivists deny that the extension with which a competentspeaker use ‘P’ at t depends on her judgemental dispositions at t, they shouldindeed be expected to leave room for this possibility, and deny (UE). To simplyassume that (UE) is correct would thus amount to begging the question against theobjectivist. In order to turn a direct forced-march Sorites argument into a dialec-tically efficient argument, (UE) must be given additional, independent support. Atthis point, it is natural to turn to what appears to be the general driving idea behindforced-march Sorites arguments, namely the idea that making sense of the forced-march Sorites data requires that we adopt principles like (JD) and (UE).20

Here is how such an argument would go. If we were to follow the objectivistin denying these principles, then we would have to ascribe certain (semantic)mistakes to the agent in the forced-march Sorites, even when the conditions areoptimal. Not only would we have to take some of the judgements to be false, wewould also have to take some of them to be in genuine contradiction. This, theargument goes, does not make sense. Instead, we should follow the agent relativistin accepting (JD) and (UE). Firstly, given the close link between actual applicationand extension guaranteed by (JD) and (UE), semantic misclassifications would beruled out (under optimal conditions). Second, from the agent relativist’s point ofview, the contradictions that may seem to arise during the course of a series offorced-march experiments would be merely apparent, as there would be semanti-cally relevant differences between any apparently contradictory judgements. Forinstance, on Raffman’s view, such judgements are made with respect to differentpsychological contexts. Just like there is no contradiction between my saying ‘I’mhungry’ and your saying ‘I’m not hungry’, there is, on Raffman’s view, no genuinecontradiction between utterances of ‘x is P’ and ‘x is not P’, where ‘P’ is a vaguepredicate and x is a non-clear case of application for ‘P’, as long as these utterancesare made with respect to different psychological contexts.21,22

the prima facie plausibility of the claim that competent speakers’ application of ‘P’ at t maydeviate from the extension with which they use ‘P’ at t. Embracing epistemicism is just one ofmany ways of being an objectivist (as defined in footnote 17).

20 For instance, see Soames (1999, 213–214); Shapiro (2006, 26).21 This holds mutatis mutandis for Soames’s and Shapiro’s views as well. It should be

pointed out though, that when it comes to so-called non-indexical contextualism, this comparisonis not entirely adequate. However, the general point would hold anyway. For further discussionsee Åkerman (2009, 45–58; 2012, 471–472) and Åkerman & Greenough (2010b).

22 As long as we take the utterances to be relative to the relevant agent/time-specificfactors, this would be the obvious way to explain away the apparent contradictions. However, ifone were to take the utterances to be objective, i.e. as saying something like ‘Item #545 isobjectively P’ and ‘Item #545 is objectively not P’, respectively, the agent relativist may followGoldstein (2012, 64) in saying that the utterances are neither true nor false but nonsensical, sinceit does not make sense to take ‘P’ to have objective extensions. In this case, agent relativists will



A (direct) forced-march argument complemented in this way would not bequestion begging in the obvious way observed earlier. But of course, its force nowdepends on whether or not there are any good independent reasons for acceptingthe claim that the only way to make sense of the forced-march Sorites data is toaccept principles like (JD) and (UE). In the following discussion, I focus on theidea that considerations about linguistic competence can provide such reasons, andI argue that the prospects for this strategy are dim. As already mentioned, theforced-march Sorites data themselves will not be questioned, nor will the claimthat objectivists are committed to ascribing classificatory mistakes to the agents inforced-march Sorites scenarios. I do, however, have some reservations concerningthe claim that objectivists must ascribe genuine contradictions to the agents. But asthis issue is not my main focus here, I shall leave it to one side for now, and insteadaddress it (albeit briefly) in section 5.

4. Linguistic competence

The idea that considerations about linguistic competence could be used to arguefor agent relativism is far from new. The following passage is from one ofRaffman’s earliest papers on vagueness, published nearly 20 years ago:23

[S]ince our actual applications of vague predicates vary with psychological context,it follows that if the (true) extensions of these predicates do not thus vary, then we arelinguistically incompetent in their use. But of course we are not incompetent in theuse of these words. (Raffman 1994, 66)

For present purposes, we can grant that actual application varies with psychologi-cal contexts. The relevant question is what this tells us about (UE), i.e., whether ornot such variation entails variation in extension. The claim made in this passage isin effect that denying (UE) would lead to the absurd conclusion that ordinaryspeakers are incompetent in their use of vague language. This claim in turndepends on the (implicit) assumption that in order for a speaker S to count aslinguistically competent with a vague predicate ‘P’, S’s actual applications of ‘P’at a time t must not diverge from the extension with which ‘P’ is used by S at t (i.e.,the (true) extension of ‘P’ at t). So what we have now is a forced-march Soritesargument that depends on a certain constraint on linguistic competence.

not be in a position to say that the judgements are correct (true), but given that nonsensicalutterances do not genuinely contradict anything, they can still avoid ascribing any genuinecontradictions to the agents in the forced-march Sorites experiment. Thanks to an anonymousreferee for dialectica for drawing my attention to this.

23 The argument is repeated in Raffman (1996, 190).

Jonas Åkerman412


4.1. Different aspects of linguistic competenceBefore we turn to the discussion of this argument, a few preliminary remarks arein order. First, the constraint on linguistic competence needs to be qualified in thesame way as (JD):

(LC) Linguistic competence with a vague predicate ‘P’ requires that one’s actualapplication of ‘P’ at t (under optimal conditions) does not diverge from theextension with which one uses ‘P’ at t.

The rationale for the qualification is the same as in the case of (JD); we do not wantthe constraint to exclude the possibility of mistakes due to unfavourable externalconditions, misinformation, or temporary dips in performance level.24

Second, being a competent speaker of a language containing vague predicatesinvolves a lot of things. So, let us put aside some of those that are not at issue here.One aspect of competence with vague predicates that we can safely put aside islinguistic competence in the Chomskian sense, as defined by the set of rules (thegrammar) manifested in the speakers’ understanding of syntactically acceptableusage (cf. Chomsky 1965, 3–9). Clearly, there is nothing about the forced-marchdata that commits an objectivist to deny that ordinary speakers are competent inthis sense. Rather, what is at issue is what might be called classificatory compe-tence, in virtue of which ordinary speakers can correctly classify objects as falling,or not falling in the extension of a predicate.

However, there is also a certain aspect of classificatory competence that we cansafely put aside for present purposes, namely classificatory competence withrespect to the clear cases of application and non-application of ‘P’ at the beginningand at the end of the Sorites series for ‘P’. It is characteristic of such clear casesthat, under optimal conditions, competent speakers do not disagree about theirsemantic status, and that their judgements on them are stable over time. Even onan agent-relativist view, there will be significant overlaps between different agent/time-relativized extensions of vague predicates, and when it comes to clear cases,the overlap will be complete, at least under optimal conditions (cf. Raffman 1994,67; Goldstein 2012, 64). Consequently, there will be no variation in application of‘P’ to clear cases across different runs of the forced-march Sorites experiment(under optimal conditions), and thus there will be nothing about the forced-march

24 Of course, in order for the argument to work as intended given this qualification of theconstraint, we need to assume that the pattern of application would continue to vary acrossdifferent runs of the forced-march Sorites experiment, even under optimal conditions, sinceotherwise there would be no pressure to adopt the agent relativism in order to satisfy the (LC).For present purposes, this assumption will be accepted, even though it may be hard to confirmempirically. It should also be noted that the forced-march Sorites argument might work even ifwe weaken the conditions in (LC), but as that would yield a stronger constraint, I prefer to keepit as it is for present purposes. As my aim here is to question the constraint, I want to consider itin its weakest and most plausible form.



data that commits an objectivist to deny that ordinary speakers are perfectlyclassificatorily competent with respect to clear cases.

The crucial question, then, is if linguistic competence with vague languageshould be taken to entail perfect classificatory competence (under optimal condi-tions) with respect to non-clear cases.25 If we were to accept (LC) and take it toapply in full generality, we would be committed to a positive answer to thisquestion. However, as shall be argued in what follows, there is scope for denying(LC), at least when it comes to non-clear cases, and, contrary to the line ofargument laid out above, the objectivist can make sense of the ascriptions ofclassificatory mistakes to the agents in the forced-march Sorites experiment.

4.2. Unclarity and classificatory mistakesLet us start with a few remarks regarding ordinary speakers’ epistemic access tothe extensions of vague predicates. Everyone can agree that vagueness gives riseto a certain kind of unclarity.26 Whether this unclarity is due to ignorance aboutexisting facts of the matter or whether it is due to genuine non-epistemic indeter-minacy is of course a matter of controversy, but everyone can agree that when itcomes to certain items in the middle of the Sorites series for a vague predicate ‘P’,we do not know what their semantic status is. That is to say, for some x, we neitherknow that Px nor that not-Px (cf. Greenough 2003, 257). If there is a fact of thematter here, we are ignorant of it, and if there is no fact of the matter, there isnothing to know, and thus we cannot know it. Moreover, even if our lack ofknowledge were a matter of ignorance, it seems that no amount of (non-semantic)information about the subject matter could relieve us from this ignorance. So, nomatter how epistemically well placed we are, we will still lack knowledge of thesemantic status of some items.27 This means that partial lack of semantic knowl-edge is to be expected even from a competent speaker in optimal conditions.

Given this, it seems that it should not be too difficult for objectivists to makesense of the idea that ordinary competent speakers could make classificatorymistakes in the forced-march Sorites. First, they can happily accept the forced-march data, and agree that on their view, the agents make certain classificatorymistakes. Second, they can point out that as the experiment is set up, the agents are

25 Since, on the agent-relativist view, the extensions are determined subjectively ratherthan objectively, there will be no objective extension, and no objective fact of the matter to beperfectly correct about. Still, as we have seen, forced-march Sorites arguments rest on the claimthat competent agents’ judgements are perfectly correct relative to the extensions with which theyuse the predicate on the occasion of the judgement (under optimal conditions).

26 Several authors have taken this to be a suitable neutral starting point for theorizingabout vagueness. See Williamson (1994, 2); Sainsbury (1995, 64); Keefe (2000, 6).

27 It would not help to introduce a ‘borderline’ semantic status, as there would still besome items whose status remained unclear. For instance, an object may be neither clearly P norclearly borderline P.

Jonas Åkerman414


forced to make judgements about things about which they lacks sufficient knowl-edge (or about which there is no fact of the matter). Thus, it makes sense to expectsome mistakes on the part of the agents, even if they are a normal competentspeakers and the conditions are optimal. Indeed, objectivists could say that themistakes are likely to occur because the agents are forced to make judgementsabout things of which they lack knowledge.

Agent relativists will have to tell a rather different story, but there is no reasonto think that they will deny that there are items x such that we ordinary competentspeakers neither know that Px nor know that not-Px. For instance, recall that onRaffman’s view, there is no time t such that the agent knows the semantic status ofevery item in the series at t, and this would arguably remain the case even underoptimal conditions. That is to say, even normal competent speakers with access toall relevant non-semantic facts concerning the objects in the Sorites series shouldbe expected to lack some semantic knowledge. That the extensions are ‘subjec-tively’ determined does not mean that the relevant subjects have full epistemicaccess to them.

Moreover, from an agent relativist point of view, it seems that this idea – thatfor some x, the agent knows neither that Px nor that not-Px – is perfectly consistentwith the idea that we should expect all the agent’s judgements – which are eitherof the form Px or of the form not-Px – to come out as semantically correct. First,in order to be reliable in one’s classifications in the forced-march, one need not ateach point know the semantic status of each item in the series. It suffices that oneknows the semantic status of each object at the moment when one’s judgement onit is passed. Of course, in view of the above considerations about unclarity, it maybe doubted that such knowledge can appropriately be ascribed to the agent. But,second, classificatory reliability does not even require that one has propositionalknowledge of the content of one’s current judgements, or even a confident beliefin these contents. Neither lack of knowledge of, nor lack of confidence (due tounclarity) in propositions of the form ‘x is P’ and ‘x is not P’ excludes that the agentcan reliably get all of her judgements right. Such lack of knowledge or confidenceis perfectly compatible with her having an ability to reliably make semanticallycorrect classifications. Agent relativists could make sense of such an ability interms of (JD), and in general, there is nothing mysterious about the combination ofsuch abilities and ignorance of the relevant kind. To take an example that shouldbe familiar to most of us, one may have the ability to find and maintain theequilibrium required in order not to fall off one’s bicycle without ever being ableto specify where one’s point of gravity must be located (relative to the vehicle andthe ground) in order to reach and maintain this equilibrium. This is an ability thatone has in virtue of being a competent cyclist. In analogy, the agent relativist mightargue that the ability to classify the non-clear items correctly in the forced-marchSorites experiment is an ability that the agents have in virtue of being competent



speakers.28 Even when it is unclear to them which side of the boundary they are on,as it were, they are able to get the judgement (semantically) right.

Where, then, does this leave us as regards the dialectical situation betweenobjectivism and agent relativism? Well, even if both of the stories are coherent,neither of them, taken on its own, constitutes an argument against the opposingview. In particular, the agent relativist story does not in itself give us any inde-pendent reason for accepting (LC) or (JD), it merely shows how different aspectsof the view cohere. Thus, the question remains whether or not there is any(independent) reason for thinking that an ability of the kind appealed to in theagent-relativist story is essential to competence with vague predicates of the kindthat we should expect from ordinary speakers. Absent such a reason, it is hard tosee how (LC) could be invoked in order to clinch the above forced-march Soritesargument.

4.3. Learning a vague languageIt might be thought that the above analogy between riding a bicycle and classifyingnon-clear cases can be of help in providing independent reasons for accepting(LC). In the bicycle case, it seems clear that one would not count as competentunless one had the ability to find and maintain the required equilibrium underfavourable conditions. After all, learning to ride a bicycle includes learning how tobalance on it, so this is something we should expect from any competent bicyclist.

However, an analogous argument concerning linguistic competence and theability to correctly classify non-clear items in the forced-march experiment doesnot seem very compelling. Learning a vague language does not (typically) involvelearning how to make semantically correct classifications of non-clear cases. Thetypical way in which one comes to understand a vague predicate is quite different,as Soames points out in the following passage:

Often we are shown clear and unproblematic examples of its application andnonapplication. [. . .] From these instructions, we learn both how to characterize theparticular examples shown to us and how to generalize the use of the predicate tocharacterize many items outside the original sample. For example, in coming tounderstand the predicate bald, we learn that whether or not people are bald dependson how much hair they have on their heads. People with little or no hair – no morethan the bald individuals in our original sample – are bald. People with a lot of hair– at least as much as the “not bald” individuals in the original sample – are not bald.(Soames 1999, 209–210)

28 At a first glance, it may seem odd to take this kind of ability to be an aspect ofcompetence. However, we can make sense of this if we leave room for a ‘minimalist’ notion ofcompetence according to which the question of competence is not distinct from the question ofhow the relevant facts determine meaning and extension (cf. Wikforss 2010, 3).

Jonas Åkerman416


By being shown certain clear cases (under suitable circumstances), we learn whatproperties are relevant for the application of the predicate, and knowing this isclearly something that we should expect from any competent speaker. Moreover,in learning how to generalize from the clear cases of application and non-application, we acquire (implicit) knowledge of so called penumbral connections,such as ‘Any man who is balder than a bald man is also bald’.29 Now, these areaspects of learning and competence that objectivists and agent relativists can agreeabout.30 But what about the ability to make semantically correct classifications ofnon-clear cases in the forced-march Sorites experiment (under optimal condi-tions)? Although there may not be anything incoherent about taking competentspeakers to have an ability of this kind, we still have not seen any independentmotivation for doing so. Indeed, insofar as the above account is on the right track,it is far from clear that there is any aspect of the way we typically learn vaguelanguage that could plausibly be regarded as the source of such an ability.

Perhaps it could be claimed that in learning vague predicates, we learn tofollow our judgemental dispositions, and this would, on the agent-relativistpicture, be tantamount to learning how to make semantically correct classificationsof non-clear cases. But, just like the above story, this merely shows how agentrelativists might make sense of how we acquire this kind of ability in the processof learning vague language. Such an account would be of no help when it comesto arguing for agent relativism, since it presupposes a close link between judge-ments and extensions of the very kind that is under dispute. In other words, anaccount of this kind would fail to provide the very thing that would be needed inorder to make the forced-march Sorites argument work, namely, an independentreason for accepting (LC).

4.4. Handling non-clear casesIt may fairly be pointed out that the simple picture sketched above is incompletein that it leaves out important aspects of how we learn to use vague language invarious everyday situations in order to achieve certain practical goals. It seemsplausible to assume that ordinary speakers will typically acquire abilities of thiskind in the process of becoming competent language users, and it seems ratherclear that we expect ordinary competent speakers to have such abilities. One thingthat is of particular interest for present purposes is the fact that competent speakersof vague language often manage to apply vague predicates to non-clear cases in

29 The term ‘penumbral connection’ is due to Fine (1975).30 It is a further question as to whether these aspects of learning and competence can be

invoked to support any particular understanding of vagueness. An anonymous referee fordialectica suggested that they might favour an inferentialist view, but I will not pursue this issuehere. For relevant discussion, see Zardini (2008) and Cobreros et al. (2012).



ways that seem perfectly appropriate given the situation at hand.31 Consider thefollowing example, given by Alice Kyburg and Michael Morreau (2000, 577):

There are two pigs in a pen: the first a runt, the second quite round but not really fat.Now the farmer says, ‘The fat pig won a prize.’ His neighbor could protest thatneither pig is really fat. But he does not protest. He understands that the rounder ismeant, and both of them know it.

The farmer’s (indirect) classification of the rounder pig as fat seems perfectlyappropriate given the situation and his practical goals. Moreover, this is a com-pletely ordinary everyday situation, which we should expect any competent user ofvague language to be able to handle. Nothing beyond the competence that weexpect from normal speakers of English seems to be required, either from thefarmer or from his neighbour.

Let us consider another example, adapted from Mark Sainsbury (1997, 259–260). In a paint shop, there are two shelves, marked ‘Red’ and ‘Orange’, respec-tively, and a set of cans, ranging from (clearly) red to (clearly) orange. The shopassistant has been given the task of arranging these cans on the two shelves in anappropriate way. The task turns out to be more difficult than expected, as the cansform a Sorites series from red to orange. Fortunately, they are numbered accordingto the relative redness/orangeness of the paint they contain. This makes it easy toavoid putting a can on the ‘Red’ shelf that is less red than some can on the ‘Orange’shelf. In other words, the assistant can quite easily avoid violating any penumbralconnections in classifying the cans. But in order to fulfil his task, the assistant willhave to classify some non-clear cases as falling in one of the two availablecategories, and, in this sense, a boundary (of sorts) will have to be drawn in theseries. Thus, the assistant ends up in a situation that resembles a forced-marchSorites experiment in several respects.32 Nevertheless, we would typically expectordinary competent speakers to handle situations like this without too muchtrouble.

If this is right, then, it seems that ordinary speakers’ competence with vaguepredicates comprises an ability to handle various situations in which one needs tomake classifications of non-clear cases, including situations which are quitesimilar to forced-march Sorites experiments. Now, it may be thought that in thelight of this, we should not expect ordinary speakers to have any particular troublein handling the forced-march Sorites either.

In one sense, I think that this is precisely right, but the sense in which it is rightis not the one that would be needed in order to provide us with a good reason foraccepting (LC). What we have good reasons to ascribe to normal competentspeakers is an ability to apply vague predicates to non-clear cases in order to fulfil

31 For an empirical study and discussion of this phenomenon, see Syrett et al. (2009).32 See Shapiro (2006, 22) and Åkerman (2009, 67–68) for further discussion of these

similarities.

Jonas Åkerman418


certain practical purposes, and to understand other speakers when they do the samething.33 In particular, we should expect normal competent speakers to be able tocommunicate efficiently with vague language, and this will sometimes requireclassification of non-clear cases. This does not mean that we should expect theseclassifications to be semantically correct, as we can communicate efficiently bysaying things that are not really true (and by using descriptions that do not reallyfit the objects about which we want to communicate something). Thus, althoughwe do have reasons to expect ordinary competent speakers to be able to applyvague predicates to non-clear cases in a way that is appropriate in order to achievecertain practical (e.g., communicative) purposes, we do not thereby have anyreason to expect them to be able to make semantically correct applications ofnon-clear cases in the forced-march Sorites experiment.

4.5. Doing right while being wrongAgent relativists sometimes emphasize that there does not seem to be anythingwrong about the way in which the agents apply vague predicates in the forced-march Sorites experiment.34 This point is related to the point about handlingnon-clear cases discussed in the previous section, and again, I think that it is rightin one sense, but not in a sense that would help with respect to the argument we areconcerned with here. The sense in which it is right is that the agents’ judgementsare perfectly appropriate given the situation they find themselves in. As explainedabove, the way that competent speakers handle the forced-march Sorites experi-ment and similar situations will typically be perfectly in order from a pragmaticpoint of view, and there will typically not be anything inappropriate about the waythey apply vague predicates in these situations. But the classifications need not besemantically correct in order to be appropriate. The impression that it seems out ofplace to accuse the agents in the forced-march Sorites experiment of having doneanything wrong may be explained in terms of appropriateness rather than truth,and thus this observation does not provide any independent reason for acceptingthe assumptions on which the forced-march Sorites argument rests.

Indeed, the impression that the agents’ judgements are appropriate appears toremain even under the assumption that some of the applications are semanticallyincorrect. Suppose that it was somehow discovered that the agents had made somefalse (or untrue) judgements concerning items in the middle of the series (despite

33 Following David Lewis (1979), we can account for this in terms of (pragmatic)accommodation: sometimes we are willing to accept (without qualification) sentences that arenot clearly (or determinately) true given that they are true enough in the situation at hand. OnLewis’s view, accommodation does not mean that the extension with which the predicate is usedchanges as a function of what is true enough or what the speaker and her interlocutors accept indifferent situations. The appropriate application of the predicate to non-clear cases in a certainsituation is up for negotiation among competent language users, but the extension is not.

34 For instance, see Shapiro (2006, 26).



the conditions being optimal). It would still seem just as odd to accuse them ofhaving done something inappropriate. After all, they have just done what they hadto do given the rules of the game, as it were, so in that sense, they did not doanything wrong (cf. Shapiro 2006, 26). But not doing anything wrong (in onesense) is of course compatible with being wrong (in some other sense).

An analogy borrowed from Rosanna Keefe (2003, 79) may be helpful here.The forced-march Sorites experiment resembles somewhat a situation in whichone is trying to guess the weight of something, e.g., a giant pumpkin at a farmersmarket, in order to win a prize, e.g., the pumpkin in question. In guessing, one willtry to get as close as possible, but there will be nothing inappropriate about saying‘It weighs 453 pounds’ although it neither does weigh exactly that, nor does onereally believe that it does. In order to enter the contest, one has to say something,but it should not be taken as a genuine assertion expressing a belief about the exactweight of the pumpkin. Thus, just like in the scenario described above, it wouldseem completely out of place to accuse someone of having done somethinginappropriate in guessing the wrong weight. But of course, that does not changethe fact that the guess was wrong (cf. Keefe 2003, 80).

Let us take stock. We have seen that we should expect ordinary speakers to beable to classify clear cases in a semantically correct way (at least under optimalconditions), to have (implicit) knowledge of penumbral constraints, and to have apragmatic competence in virtue of which they can make use of vague language inorder to achieve their everyday (communicative) purposes. These all seem to becentral aspects of ordinary competent use of vague language, although I do notclaim that this list is exhaustive. For instance, perhaps we should also expectcompetent speakers to take vague predicates to be (in some sense) tolerant tosufficiently small changes along the relevant dimension of comparison, or toconsider certain disagreements concerning non-clear cases to be legitimate (cf.Eklund 2005, 41; Goldstein 2012, 63)).35 However, from a neutral point of view,there seems to be no reason to take ordinary competence with vague predicates torequire an ability to classify non-clear cases in a semantically correct way, not evenunder optimal conditions. The upshot of this is that we should at the very leastleave open the possibility that linguistic competence with vague predicates becompatible with error concerning non-clear cases, even under optimal condi-tions.36 If this is right, then, objectivists can quite appropriately deny principles

35 The notion of tolerance was introduced by Crispin Wright (1975, 333–334). SeeÅkerman (2009, 38–45) and Åkerman & Greenough (2010a, 276) for examples and discussionof different versions of tolerance.

36 I am not the first to endorse a claim of this kind (cf. Keefe 2003, 79), but so far, neithercritics nor defenders of agent relativism have gone into sufficient detail about how such claimsshould be understood and evaluated in relation to the forced-march Sorites argument.

Jonas Åkerman420


like (LC), and thus resist forced-march Sorites arguments that rely on suchprinciples.

5. Competent use, extensions and contradictions

Before concluding, I would like to address briefly two outstanding issues concern-ing (i) the relationship between competent use and extensions, and (ii) howobjectivists can handle the apparent contradictions in the forced-march Soritesexperiment.

Let us start with (i). It seems plausible to take the extensions of vague predi-cates to be determined, in some way, by how they are used by competent speakers.Given this, the fact that agent relativism posits a very close and direct relationbetween (individual) judgements and extensions may seem to count strongly in itsfavour. However, adopting principles like (JD) and (UE) is not the only way toaccommodate the idea that competent use determines the extensions of vaguepredicates. To be sure, the agent relativist account is neat and simple, but that is notenough to give us good reasons to accept it. First, we should ask ourselves whetherwe have any reason to think that the actual relationship really is as simple as thisaccount suggests. If not, it may be regarded as simplistic rather than simple (cf.Keefe 2003, 78–79). Second, even if we want to acknowledge the importance ofcompetent speakers’ judgements, agent-relativist accounts like Raffman’s may betaken to be too individualistic. Why should we relativize extensions to individualspeakers (or even small groups of conversationalists) rather than, say, letting theextensions be determined by the totality of the judgemental dispositions of therelevant linguistic community? One might, for instance, prefer a collectivist view,according to which for any vague predicate ‘P’, an object’s being P depends on itspropensity to be judged to be P by all (or at least a high preponderance of)competent agents (under optimal conditions).37

One may object that since ordinary competent speakers’ judgemental disposi-tions are in conflict and in flux (even under optimal conditions) it is hard to seehow the determination of extensions is supposed to work on any such collectivistview.38 Given the assumption that the extensions must be fixed and sharplybounded, this point is well taken, but the objection appears to lose much of itsforce if this assumption is dropped. And this assumption can indeed be dropped. Adenial of (individualistic) agent relativism in favour of a collectivist view iscompatible with the claim that the extensions are neither sharply bounded norfixed, since the function from collective use to extension may either be determinateor indeterminate, and certain variations in the overall use or dispositions maycause the extension to change over time.

37 Thanks to an anonymous referee for dialectica for helpful comments on this issue.38 This objection was raised by an anonymous referee for dialectica.



In sum, the general and plausible claim that the extensions of vague predicatesare determined (in some way) by how they are used by competent speakers is, atleast prima facie, compatible with theories of vagueness of very different kinds,and it is far from clear that (individualistic) agent-relativist views are better suitedto accommodate it than their rivals.39

Let us now turn to (ii).40 One option for the objectivist would be to draw onKeefe’s analogy with guessing, and deny that the judgements about non-clearcases in the forced-march Sorites experiment are genuine assertions. The semanticcontents of these judgements may still be in contradiction, but since neither ofthem is uttered with genuine assertoric force, there is a sense in which the agentshave not genuinely contradicted themselves; none of them have asserted twocontradictory propositions.41

Another option would be to emphasize (again) that due to the unclarity thatcomes with vagueness, the semantic status of non-clear cases is not something thatordinary speakers can be expected to be very confident about. In view of this, weshould not expect judgements on such cases to be very stable, and thus, it is notvery surprising that the agent in the forced-march Sorites experiment tends tojudge some non-clear cases differently across different runs through the Soritesseries.42 Now, if we were to take the extension to remain fixed across thesedifferent runs it would be hard to deny that these judgements genuinely contradicteach other.43 However, we would not thereby have to take the agents to becommitted to any genuine contradictions. Instead, it could be claimed that thereason that the agents in the forced-march Sorites experiments tend to makedifferent judgements on different runs is simply that they tend to change theirminds concerning the non-clear cases across the different runs.44 As an agent

39 For similar considerations, see Williamson (1994, 205–212) and Keefe (2000, 79–83;2003, 78).

40 In the present context, I confine myself to suggesting a couple of different strategiesavailable to the objectivist. I do not intend to suggest that they are all straightforward orunproblematic, but since this falls outside the main focus of this paper, I postpone furtherdiscussion to another occasion.

41 Of course, since this solution does not appeal to semantic differences, the sense inwhich the apparent contradiction is not genuine is not the same as on the agent-relativist accountdescribed above.

42 As suggested by an anonymous referee for dialectica, one may take the oscillation inthe competent agent’s judgements to reflect a symmetry between conflicting judgemental dis-positions. Rather than taking the extensions to shift depending on which of the dispositions ispresently dominating, one may, if one is prepared to deviate from classical semantics, take thereto be a gap or an overlap between the extension and the anti-extension.

43 After all, the judgements are about one and the same object x, they have the form ‘xis P’ and ‘x is not P’, respectively, and ‘P’ expresses the same property and has the same extensionin the two situations in which the judgement is made.

44 Objectivists could even invoke something like Raffman’s account of the psychologi-cal mechanisms behind the judgements in order to explain why the pattern of application shifts

Jonas Åkerman422


changes her mind about a non-clear case, and passes a new judgement on it, shethereby implicitly retracts her previous judgement.45 This means that a theory mayentail that the forced-march Sorites data include judgements whose semanticcontents contradict each other, without thereby entailing that the agents’behaviour in the forced-march Sorites experiment commit them to any genuinecontradictions.

Yet another option would be to deviate more radically from classical logic andsemantics, for instance by adopting a dialethist view according to which contra-dictions involving non-clear cases are true. On such a view, the appearance ofcontradiction between some of the judgements in the forced-march experiment canbe taken at face-value, and their contents may even be considered as jointlyassertible. To many, accepting a view on which contradictions can be true wouldseem to be an unacceptably high cost for accommodating the forced-march Soritesdata. Nevertheless, it deserves to be mentioned in this context, since insofar itis taken to be a viable contender, it should be included among the objectivistalternatives.46

As a final remark, let me point out that since the purpose here has been todiscuss forced-march arguments, the scope has been limited to forced-marchSorites data, which concern simple judgements of the form ‘x is (not) P’. Recently,there has been some discussion concerning data of other kinds, including resultsfrom empirical studies indicating that ordinary competent speakers are quitewilling to accept statements of the form ‘x is P and x is not P’ (i.e., outrightcontradictions) when ‘P’ is a vague predicate and x is a non-clear case.47 This raisesmany interesting issues, including questions concerning how these data should beinterpreted, and how they could be brought to bear on issues like the ones dis-cussed above. But these are topics for another day.

in the way that it does, as long as they deny that these mechanisms also determine the extensions.See Åkerman (2009, 101–105; 2011) for further discussion.

45 In order to model how this works, one may invoke something similar to the notion ofa conversational scoreboard that Shapiro (2006, 12) borrows from Lewis (1979). Among otherthings, such a scoreboard contains, at any time t, all the propositions that the speaker – and herinterlocutors, if there are any – are committed to at t. When a proposition is asserted, it goes onthe score (provided that none of the interlocutors protests). What happens to the score if the agentin the forced-march Sorites experiment denies something about a non-clear case that has previ-ously been asserted is simply that the previous judgements is eliminated from the conversationalscoreboard as the new judgement is added. In fact, this is just what happens on Shapiro’s view.Now, in accordance with his agent-relativist view, he also takes such shifts to entail shifts in theextension of the vague predicate, and this is something that an objectivist must deny. But thisdoes not mean that Shapiro’s notion of a conversational score cannot be invoked on the pragmaticlevel, as it were, in order to model changes in the agent’s commitments.

46 Indeed, recent years have seen an increased interest in theories of this kind. For arecent example, see Ripley (2013). See also references given in footnote 47.

47 See Ripley (2011), Alxatib and Pelletier (2011), Cobreros et al. (2012), and Alxatibet al. (2013).



6. Conclusion

Let me conclude by summing up the main points of the above discussion. First,simple ‘direct’ versions of forced-march Sorites arguments rely on (UE), which inturn stands in need of independent support. Second, the strategy of appealing to(LC) in order to support (UE) fails, as there is scope for denying (LC) as appliedto non-clear cases. Indeed, given the lack of semantic knowledge concerning thesecases, objectivists can make good sense of competent speakers’ making classifi-catory mistakes in the forced-march Sorites experiment. Third, even if linguisticcompetence with vague predicates includes being able to apply them to non-clearcases in a way that is appropriate in order to achieve certain practical goals, weneed not take it to include an ability to make semantically correct applications tonon-clear cases in the forced-march Sorites experiment. The impression that thereis nothing wrong with competent agents’ behaviour in the forced-march Soritesexperiment can be explained in terms of appropriateness rather than truth, and thuswe can do justice to it without taking the extensions of vague predicates to co-varywith competent speakers’ actual application of them.48 Fourth, we need notembrace agent relativism in order to retain a plausible picture of the relationshipbetween use and extension. One can accommodate the general and plausible claimthat extensions are somehow determined by competent use without thereby com-mitting oneself to principles like (JD) and (UE). Fifth, there are various ways tohandle the apparent contradictions between judgements in forced-march Soritesexperiments without going agent relativist.

In sum, objectivists of various kinds can accommodate the data from forced-march Sorites experiments, without committing themselves to implausible claimsabout ordinary speakers’ competence with vague language or the relation betweenuse and extension. The upshot of this is that we should be sceptical of the idea thatsuch data paired with considerations about linguistic competence can be invokedin order to lend any solid support to agent relativism. In the absence of other waysof backing up (UE) and kindred principles, the prospects for constructing aconvincing forced-march Sorites argument look dim.*

References

Åkerman, J. 2009, Extensions in Flux: An Essay on Vagueness and Context Sensitivity, Stockholm:Doctoral dissertation. Online: http://urn.kb.se/resolve?urn=urn:nbn:se:su:diva-30080.

48 As pointed out in footnote 18, if we are prepared to go dialethist, we could even takeall of the judgements in the forced-march Sorites experiment to be semantically correct, withouttaking the extensions to shift with the agent’s judgemental dispositions.

* An earlier version of this paper was presented at the Logic, Language and MindSeminar at the department of Philosophy at Stockholm University. Thanks to the participants forvaluable discussion and feedback. Special thanks to Nat Hansen, Sören Häggqvist, Peter Pagin,

Jonas Åkerman424


http://urn.kb.se/resolve?urn=urn:nbn:se:su:diva-30080

Åkerman, J. 2011, “Vagueness, Semantics and Psychology”, Philosophical Quarterly, 61, pp. 1–6.Åkerman, J. 2012, “Contextualist Theories of Vagueness”, Philosophy Compass, 7, pp. 470–480.Åkerman, J. and Greenough, P. 2010a, “Hold the Context Fixed – Vagueness Still Remains”, in: R.

Dietz and S. Moruzzi, eds, Cuts and Clouds – Vagueness, its Nature, and its Logic, Oxford: OxfordUniversity Press, pp. 275–288.

Åkerman, J. and Greenough, P. 2010b, “Vagueness and Non-Indexical Contextualism”, in: S.Sawyer, ed., New Waves in Philosophy of Language, Basingstoke: Palgrave Macmillan, pp. 8–23.

Alxatib, S. and Pelletier, F. J. 2011, “The Psychology of Vagueness: Borderline Cases and Con-tradictions”, Mind and Language, 26, pp. 287–326.

Alxatib, S., Pagin, P. and Sauerland, U. 2013, “Acceptable contradictions: Pragmatics or Seman-tics? A reply to Cobreros et al.”, Journal of Philosophical Logic, 42, pp. 619–634.

Chomsky, N. 1965, Aspects of the Theory of Syntax, Cambridge, MA: MIT Press.Cobreros, P., Egré, P., Ripley, D. and van Rooij, R. 2012, “Tolerant, Classical, Strict”, Journal of

Philosophical Logic, 41, pp. 347–385.Egré, P. forthcoming, “Vagueness: Why Believe in Tolerance?”, Journal of Philosophical Logic.Egré, P., de Gardelle, V. and Ripley, D. forthcoming, “Vagueness and Order Effects: Evidence

for Enhanced Contrast in a Task of Color Categorization”, Journal of Logic, Language andCommunication.

Eklund, M. 2005, “What Vagueness Consists In”, Philosophical Studies, 125, pp. 27–60.Fara, D. G. 2000, “Shifting Sands: An Interest-Relative Theory of Vagueness”, Philosophical Topics,

28, pp. 45–81. (Originally published under the name ‘Delia Graff’.)Fine, K. 1975, “Vagueness, Truth, and Logic”, Synthese, 30, pp. 265–300.Goldstein, L. 2012, “The Sorites is Nonsense Disguised by a Fallacy”, Analysis, 72, pp. 61–65.Greenough, P. 2003, “Vagueness: A Minimal Theory”, Mind, 112, pp. 235–281.Heck, R. 2003, “Semantic Accounts of Vagueness”, in: J. C. Beall, ed., Liars and Heaps, New York:

Oxford University Press, pp. 106–127.Horgan, T. 1994, “Robust Vagueness and the Forced-March Sorites Paradox”, Philosophical Per-

spectives, 8, pp. 159–188.Keefe, R. 2000, Theories of Vagueness, Cambridge: Cambridge University Press.Keefe, R. 2003, “Context, Vagueness, and the Sorites: Comments on Shapiro”, in J. C. Beall, ed., Liars

and Heaps, New York: Oxford University Press, pp. 73–83.Keefe, R. 2007, “Vagueness Without Context Change”, Mind, 116, pp. 275–292.Kyburg, A. and Morreau, M. 2000, “Fitting Words: Vague Language in Context”, Linguistics and

Philosophy, 23, pp. 577–597.Lewis, D. 1979, “Scorekeeping in a Language Game”, Journal of Philosophical Logic, 8, pp. 339–359.Raffman, D. 1994, “Vagueness Without Paradox”, Philosophical Review, 103, pp. 41–74.Raffman, D. 1996, “Vagueness and Context Relativity”, Philosophical Studies, 81, pp. 175–192.Raffman, D. 2014, Unruly Words: A Study of Vague Language, Oxford: Oxford University Press.Ripley, D. 2011, “Contradiction at the Borders”, in: R. Nouwen, R. van Rooij, U. Sauerland and H-C.

Schmitz, eds, Vagueness in Communication, Berlin: Springer, pp. 169–188.Ripley, D. 2013, “Sorting out the Sorites”, in: K. Tanaka, F. Berto, E. Mares and F. Paoli, eds,

Paraconsistency: Logic and Applications, Dordrecht: Springer, pp. 329–348.Sainsbury, M. 1995, “Why the World Cannot be Vague”, Southern Journal of Philosophy, 33, pp.

63–81.Sainsbury, M. 1997, “Concepts Without Boundaries”, in R. Keefe and P. Smith, eds, Vagueness: A

Reader, Cambridge: MIT Press, pp. 251–264.Shapiro, S. 2006, Vagueness in Context, Oxford: Oxford University Press.

Mikael Pettersson, Daniel Rönnedal, the editors of dialectica, and several anonymous refereesfor reading and providing very helpful comments on earlier drafts. The research leading to theseresults has received funding from the European Research Council under the European Commu-nity’s Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement no 229 441-CCC,and from the Swedish Research Council (Dnr 435-2010-454 and 421-2010-1542).



Soames, S. 1999, Understanding Truth, Oxford: Oxford University Press.Stanley, J. 2003, “Context, Interest Relativity and the Sorites”, Analysis, 63, pp. 269–280.Syrett, K., Kennedy, C. and Lidz, J. 2009, “Meaning and Context in Childrens’ Understanding of

Gradable Adjectives”, Journal of Semantics, 27, pp. 1–35.Wikforss, Å. 2010, “Are there Understanding-Assent Links?”, The Baltic International Yearbook of

Cognition, Logic and Communication, 5, pp. 1–17.Williamson, T. 1994, Vagueness, London and New York: Routledge.Wright, C. 1975, “On the Coherence of Vague Predicates”, Synthese, 30, pp. 325–365.Zardini, E. 2008, “A Model of Tolerance”, Studia Logica, 90, pp. 337–368.

Jonas Åkerman426


Documents

Forced-March Sorites Arguments and Linguistic Competence