Relativism about Predicates of Personal Taste and Perspectival Plurality

Abstract

In this paper we discuss a phenomenon we call perspectival plurality, which has gone
largely unnoticed in the current debate between relativism and contextualism about
predicates of personal taste (PPTs). According to perspectival plurality, the truth value
of a sentence containing more than one PPT may depend on more than one
perspective (subjects, experiencers or judges). Prima facie, the phenomenon
engenders a problem for relativism and can be shaped into an argument in favor of
contextualism. We explore the consequences of perspectival plurality in depth and
assess several possible responses on behalf of advocates of relativism.

Full text

Relativism about Predicates of Personal Taste and Perspectival
Plurality
Abstract
In this paper we discuss a phenomenon we call perspectival plurality, which has gone
largely unnoticed in the current debate between relativism and contextualism about
predicates of personal taste (PPTs). According to perspectival plurality, the truth value
of a sentence containing more than one PPT may depend on more than one perspective
(subjects, experiencers or judges). Prima facie, the phenomenon engenders a problem
for relativism and can be shaped into an argument in favor of contextualism. We
explore the consequences of perspectival plurality in depth and assess several possible
responses on behalf of advocates of relativism.
Keywords: Predicates of personal taste. Relativism. Contextualism. Perpsectival
plurality.
Acknowledgments: Markus Kneer would like to thank Peter Pagin, Max Kölbel, Isidora
Stojanovic, François Recanati, John Perry, Edouard Machery and Robyn Carston, and
the audience of the 2014 PLM workshop in San Sebastián for helpful feedback on a
dissertation chapter on which parts of this article are based. Agustin Vicente and Dan
Zeman’s research for this work has been funded by Projects IT769-13, from the Basque
Government, and FFI2014-52196-P, from the Spanish Ministry of Economy and
Competitiveness (MINECO). Dan Zeman also acknowledges the financial help of
Project FFI2012-37658, as well as that of a Juan de la Cierva grant from the same
source. Both authors would like to thank Elena Castroviejo and Javier Ormazabal, the
other members of the HiTT group at the University of the Basque Country and the
audiences at the 2015 PLM workshop in Oslo, the 1st Context, Cognition and
Communication conference in Warsaw and the Pervasive Context conference in
Reading. All three authors thank the two anonymous referees for this journal for their
suggestions and comments and suggestions.

1. Introduction
Predicates of personal taste (henceforth PPTs) – expressions like ‘tasty’, ‘fun’,
‘disgusting’, ‘boring’, ‘cool’ etc. – have been at the center of ongoing debates in
contemporary semantics. Although there is no definitive characterization of PPTs in the
literature1, one feature that most authors agree upon is their semantic dependence on
subjects, experiencers or judges. To circumvent the debate regarding which of the latter
three notions is most appropriate, we will take PPTs to be semantically dependent on
perspectives broadly conceived. This dependence is perhaps best brought to the fore by
attending to the fact that utterances of the same sentence containing a PPT can vary in
truth value across contexts in which the relevant perspectives are different. To have a
working example, imagine John and Marie, two eight-year olds who have just tried
licorice for the first time. John found it tasty, but Marie did not. Assume that each of
them separately utters
(1) Licorice is tasty.
Intuitively, the truth values of the two utterances of (1) differ: John’s utterance of (1) is
true, while Marie’s is false. What explains this variation in truth value is the fact that
John liked the taste of licorice, while Marie was repelled by it – that is, a difference in
John and Marie’s perspectives.
The two mainstream views that have shaped the recent debate surrounding PPTs
are contextualism and relativism. Although we will focus on the latter in this paper, they
are best defined in tandem. We thus assume a broadly Kaplanian framework in which
sentence types or characters and contexts jointly determine the content of an utterance.
Contents, in turn, are evaluated at a series of parameters drawn from the context of
utterance, which form the circumstances of evaluation. Circumstances standardly
comprise a time and a world, but might well include locations, epistemic standards or –
our focus here – perspectives. Although contextualism and relativism capture the
perspective-dependence of PPTs equally well, they do so differently. In the framework
sketched above, the difference comes down to attributing distinct roles to perspectives
in the semantic apparatus. According to contextualism, the perspectives on which the
interpretation of PPTs depends are specified in the content of the utterances of sentences
1 See for instance Lasersohn (2005), Stephenson (2007), Sæbø (2009), Collins (2013) and Snyder (2013)
for interesting discussion.

containing them. According to relativism, by contrast, the perspectives on which the
interpretation of PPTs depends are values of a specific parameter in the circumstances
of evaluation with respect to which the relevant utterances are evaluated. This
difference can be expressed formally as follows, where (2) captures the contextualist
truth conditions of (1) and (3) its relativist truth conditions:
(2) [[Licorice is tasty]] c, w = 1 iff licorice is tasty-according-to-p in w.
(3) [[Licorice is tasty]] c, w, p = 1 iff licorice is tasty in w according to p,
where c signifies the context of utterance, w the parameter for possible worlds and p that
for perspectives (other possible parameters are ignored for simplicity). As can be seen,
in (3) there is a parameter for perspectives in the circumstances, while in (2) there is not,
the perspective entering directly into the content, and as a result making it vary across
contexts (the predicate ‘tasty-according-to-p’ is meant to capture precisely that).2
Though contextualism and relativism differ in their approaches to formalizing
the perspective-dependence of PPTs, the view that PPTs do give rise to such
dependence is uncontroversial. In this article, we explore a related phenomenon that has
received very little attention so far, namely the possibility of perspectival plurality of
sentences invoking more than one predicate of personal taste. In section 2 we briefly
introduce perspectival plurality. In sections 3 and 4, we show how plurality gives rise to
two problems for relativism, one regarding the operator argument, the other regarding
syntactic binding. In section 5 we canvass several responses the relativist might give. In
section 6 we explore what might be the relativist’s best shot to cope with perspectival
plurality, the ‘paraphrasing strategy’. Though this approach has considerable merits, it
2 Both contextualism and relativism come in many forms. Some contextualist proposals treat PPTs as
containing variables for perspectives in their logical form (e.g., Stojanovic (2007), Schaffer (2011)).
Others conceive of PPTs as akin to gradable adjectives, and therefore as manifesting a more complex
structure than a simple variable for perspectives (Glanzberg (2007)). Finally, PPTs could be treated as
indexicals, perhaps following Rothschild and Segal’s (2009) treatment of color terms. Relativist positions
differ with respect to the type of context that provides the value of the perspective parameter. So-called
‘moderate’ versions of relativism argue that the perspective value is drawn from the context of utterance
(e.g., Kölbel (2004), Recanati (2007), Brogaard (2008)). More ‘radical’ versions take it to be provided by
the context of assessment (e.g., Lasersohn (2005, 2013), MacFarlane (2014)). An orthogonal issue arises
as to whose perspective is relevant in interpreting an utterance containing a PPT. Usually, the relevant
perspective is taken to be that of the speaker, but this need not be so: it could be that of the salient person
in a context, that of a group (as Recanati (2007) and Huvenes (2012) have suggested), a generic
perspective (e.g., Stojanovic (2007), Moltmann (2010), Snyder (2013)), or any of the above, depending
on the context (Cappelen and Hawthorne (2009)). Since our discussion below is not affected by any
choices made in connection to these issues, we ignore questions pertaining to them in what follows.

is ultimately found incapable of resolving the problems for relativism raised by
plurality.
2. Perspectival Plurality
Sentences like (1) are mono-perspectival. They contain a single PPT and thus are
dependent on a single perspective in a particular context – even if said perspective is
shared by a group of subjects. But consider the following examples3:
(4) We took the kids to a resort in Italy this summer. The wine was delicious
and the water slide was great fun.
(5) On Halloween, the kids did the rounds. They either got a delicious treat,
or else played a silly prank on us neighbors.
(6) Even your vegetarian husband would love the new restaurant. The steaks
are of course delicious, but the broccoli burgers are very tasty, too.
(7) At the world chemistry conference, professionals were encouraged to
bring their partners. The talks were very interesting and the DIY
experiments a lot of fun.
(8) At the educational book fair, every parent bought their child an
informative yet fun new book.
Examples (4)-(8) contain multiple perspective-dependent predicates.4 Though all
examples allow for various interpretations, the default reading of each relativizes the
different predicates to different contextually salient perspectives. In example (4) and
(5), for instance, it makes sense to assume that the adults considered the wine delicious
or the Halloween pranks silly, whereas the treats were delicious for the children, just as
the water-slide was fun from a point of view which is exclusively theirs, or at least
includes it. Example (6) invokes different culinary tastes: a vegetarian pallet versus a
non-vegetarian one. The steak will only please the latter, whereas the broccoli burger
will find fancy with vegetarians, or vegetarians and non-vegetarians alike. This shows
that sentences featuring predicates of taste (and others) can – and in perfectly standard
3 The examples and parts of the following sections are taken from or based on Kneer (2015).
4 ‘Informative’ is not a PPT, but for all that matters here it works in the same way. What is informative
for a pre-school child is not necessarily informative for an adult, and vice-versa. More generally, what
this shows is that perspectival plurality is not a phenomenon restricted to predicates of personal taste only.

cases do – give rise to multi-perspectival readings, i.e. readings which draw on multiple
perspectives provided by a single context of utterance.5
3. Perspectival Plurality and the Operator Argument
Perspectival plurality is an interesting phenomenon in its own right. It might also
constitute an important, to date largely unexplored dimension in the ongoing debate
regarding which type of semantic framework – relativist or contextualist – best
accommodates perspectival predicates. We will examine perspectival plurality with
regards to two central arguments that have had a significant impact in the literature: the
operator argument and the argument from binding. We will begin with the former.
According to a standard Kaplanian picture, we said, the truth-value of a sentence
depends on the context of utterance and the circumstance of evaluation: The semantic
value of a character, or sentence type, is a function from contexts and circumstances to
truth-values, and the semantic values of a content, or sentence-in-context, is a function
from circumstances to truth-values. But what determines which kinds of parameters the
circumstance is to comprise of? ‘The amount of information we require from a
circumstance’ Kaplan writes, ‘is linked to the degree of specificity of contents, and thus
to the kinds of operators in the language.’ Hence, ‘[w]hat sorts of intensional operators
to admit seems […] largely a matter of language engineering. It is a question of which
features of what we intuitively think of as possible circumstances can be sufficiently
well defined and isolated.’ (1989: 502)
Kaplan, who conceives of operators as modifying contents of sentences6, argues,
for instance, that time must be amongst the parameters of the circumstances. His reasons
are twofold: (i) tenses in English, he holds, are best conceived as sentential operators;7
and (ii) given that the purpose of temporal operators consists in shifting the time of
evaluation, they must operate on entities that are variable in truth across times. Kaplan
explains:
If we built the time of evaluation into the contents (thus removing time from the
circumstances leaving only, say, a possible world history, and making contents
5 In two studies in experimental linguistics (Kneer 2015, Kneer ms.) it is shown that a large majority of
ordinary language speakers do interpret examples like (4)-(8) as depending on multiple perspectives. The
phenomenon is also shown to extend to claims invoking multiple epistemic modal expressions.
6 This premise is questioned by various scholars, cf. for instance Lewis (1980) and Richard (1981).
7 King (2003), amongst others, famously disagrees.

specific as to time), it would make no sense to have temporal operators. To put
the point another way, if what is said is thought of as incorporating reference to
a specific time [...], it is otiose to ask whether what is said would have been true
at another time [...]. Temporal operators applied to eternal sentences (those
whose contents incorporate a specific time of evaluation) are redundant. (1989:
503)
Philosophers who consider claims regarding taste, aesthetics and morality to be truth-
relative advocate a ‘proliferation’ of parameters that accommodates the relevant types
of perspective. Although not necessary for relativism (cf. e.g., MacFarlane (2014)),
operator arguments can help motivate such parameter proliferation. Kölbel (2009), for
instance, argues along these lines that the English language contains (something like) an
operator that shifts the standard of taste, ⌜FOR t⌝. If FOR is prefixed to a singular term
designating an individual, the taste claim p in a construction ⌜FOR t, p⌝ must be
evaluated with regards to t’s taste. ‘For Anna, whale meat is tasty’, for instance, is true
if whale meat agrees with Anna’s pallet, i.e. if ‘Whale meat is tasty’ is true for Anna, or
from her perspective. The taste operator is characterized by the following rules:
(S1) For all sentences φ and all singular terms α, ⌜FOR α, φ⌝ is a sentence.
(S2) For all φ, α, w, s and a: if φ is a sentence and α is a personal name
referring to a, w is a possible world, and s is a [perspective]: ⌜FOR α, φ⌝
is true in a circumstance <w, s> iff φ is true in <w, s(a)> (where s(a) is
a’s [perspective]) (2009: 384)
Does our ordinary language term ‘for’ work in ways equivalent to the taste operator
FOR? Kölbel argues it does. Just as one can make a case for the existence of modal
operators by aid of examples such as (9), one can adapt them to make a similar case for
the existence of taste operators in ordinary English. (10), it seems, is no less true in
certain contexts than its modal pendant, (9):
(9) In possible world W, whales are extinct, but whales are not extinct.
(10) For Anna, whale meat is tasty, but whale meat is not tasty. (2009: 10)

What is more, in combination with the quantifying expressions ‘everyone’ and
‘someone’ the FOR operator mirrors the standard modal operators ‘necessarily’ and
‘possibly’. ‘FOR everyone, p’ and ‘FOR someone, p’ shift the taste parameter in just the
same ways ‘necessarily’ and ‘possibly’ shift the world parameter.8
Though prima facie not implausible, this argument in support of a taste or
perspective parameter, and the taste-neutral contents it entails, runs into trouble when
confronted with multi-perspectival readings. As Cappelen & Hawthorne (2009: 75,
fn.10) point out, sentences like (11), at least on some interpretations, invoke more than a
single perspective:
(11) Maria ate something that was tasty for Anna in a dignified way.
On the most salient reading, ‘tasty’ is relativized to Anna’s perspective, yet ‘dignified’
is not – it is, much rather, dependent on what the speaker deems dignified. But if not all
perspective-dependent predicates fall under the scope of the perspective operator FOR,
then the latter cannot be a sentential operator in the vein proposed by Kaplan.
‘Dignified’, one might object, is not a predicate of personal taste, so it is no
surprise that it does not fall under the ‘standard of taste’ operator which manipulates
‘tasty’. The circumstance might include various perspective parameters, one related to
personal taste, another one to style. Naturally, even if one restricts oneself to predicates
pertaining to taste proper, it doesn’t take much to fix the example, as Kölbel – who
provided example (13) – is ready to admit:
(12) Frank showed John how to cook something tasty for his wife in a fun
way.
(13) She ate something tasty for Anna on a pretty plate. (2011: 144)
In (12) the perspective relevant for the interpretation of ‘tasty’ is that of John’s wife (as
specified by the phrase ‘for his wife’), while the one relevant for the interpretation of
‘fun’ is (presumably) John’s. Similarly, in (13) the perspective relevant for the
interpretation of ‘tasty’ is Anna’s (as specified by the phrase ‘for Anna’), while the one
relevant for the interpretation of ‘pretty’ is (presumably) the speaker’s. Assuming that
8 The possibility of FOR combining with variables is taken by Kölbel to follow straightforwardly from
the definition of the operator. See Kölbel (2011: 145) for discussion.

the two for-phrases are modeled after Kölbel’s FOR operator, since FOR is a sentential
operator we get the result that the perspectives relevant for the interpretation of the
whole sentences (and thus of both PPTs) in (12) and (13) are those of John’s wife and
Anna’s, respectively. Since this follows from the semantics of FOR, the multi-
perspectival readings of (12) and (13) are excluded – a bad result.
We should be clear that we discuss this objection to the operator argument with
the sole purpose of illustrating how perspectival plurality has been appealed to in the
debate we are concerned with, and that we don’t necessarily commit to its dialectical
effectiveness. Thus, it might be possible for Kölbel to avoid the objection engendered
by plurality. In fact, in responding to Cappelen & Hawthorne’s counterexample, he
suggests to conceive of ‘for’ as an intensional predicate operator rather than a sentential
operator – a view that is shared by Lasersohn (2008) and MacFarlane (2012). According
to such a view, ‘for Anna’ operates on a single PPT (‘tasty’) and not on all PPTs that
occur in the sentence. While such an approach is reasonable in itself, it is no longer
clear whether it has much argumentative bite, since restricting the scope of such
operators from sentences to predicates means that we have little in the way of an
operator argument along Kaplanian lines.9 Another way out for Kölbel would be to
claim that the relevant for-phrases only sometimes work as sentential operators, which
would be enough for the introduction of a parameter for perspectives in the
circumstances.10 Finally, a last line of defense would be to cling to the initial
assumption that they are fully sentential operators and recast the problematic examples
in ways that preserve it (see, e.g., Lihoreau (2012)). Discussing such answers in detail
would take us too far afield; suffice it to say that perspectival plurality blocks sentential
operator arguments of the type discussed under standard interpretations of its premises.
4. Perspectival Plurality and the Binding Argument
The second argument in the context of which perspectival plurality occurs is the
discussion of the Binding Argument. Lasersohn (2008) provides a response to this
argument, which has been recently used against relativism about PPTs (e.g., Schaffer
(2011)).11 In a nutshell, the argument starts from the premise that there are bound
9 See, however, Zeman (2013) for a reconstruction of the operator argument without taking the relevant
phrases to be sentential operators.
10 We thank an anonymous reviewer for suggesting this alternative.
11 The argument from binding has been employed for a variety of expressions by Stanley in a number of
papers now collected in his (2007). Stanley, however, was concerned with unarticulated constituents, that

readings of the target expression, and via certain bridging principles between syntax and
semantics concludes that the target expression has a variable of a certain kind in its
syntactic configuration. Since claiming that an expression has a variable of a certain
kind in its syntactic configuration (whose value enters into the content of utterances of
sentences containing the expressions in question) amounts to a contextualist view about
that expression, the argument supports such a view over the rival relativist view.
Schaffer constructs the argument against relativism about PPTs on the basis of the
bound reading of
(14) Everyone got something tasty,
where the thing that each person under the range of ‘everyone’ got is tasty for that
person – that is, the perspective to which ‘tasty’ pertains to is bound. Schaffer
concludes that ‘tasty’ has a variable for perspectives in its syntactic configuration.
Lasersohn avoids this conclusion by providing an alternative account of binding
that allows one to ‘quantify directly on the [perspective parameter], setting and resetting
its value in tandem with the variable introduced by a quantifier’ (2008: 324). This
alternative account – ‘index-binding’ – replaces quantification over variables for
perspectives in the object language with quantification over such variables in the meta-
language – in the present case, over the parameter for perspectives in the circumstances
of evaluation (‘index’, in Lasersohn’s jargon). Formally, this idea is implemented by
introducing a second sentence-abstract-forming operator, μ, in addition to the more
common λ-operator, and to treat quantifier phrases as such sentence-abstract forming
operators. This treatment allows quantifier phrases to bind both variables in the object
language and variables in the meta-language (parameters in the index) when the variable
in the meta-language is of a different kind than the one in the object language. Thus,
when they bind variables in the object language, their effect could be described as in
(15); when they bind variables in the meta-language – and here we focus specifically on
the parameter for perspectives – their effect could be described as in (16):
(15) [[λnφ]]M, c, w, u, g= {x ∈U | [[φ]]M, c, w, u, ,g[x/n] = 1} (2008: 313)
(16) [[μnφ]]M, c, w, u, g = {x ∈U | [[φ]]M, c, w, x, ,g[x/n]= 1} (2008: 324),
is, propositional constituents not articulated at the surface level of the sentence. These, he suggested,
should not be conceived as pragmatically supplied variables, but as implicit arguments, which are part of
the syntax of the expression.

where φ is a sentence, M is a model, c a context, w a possible world, u an individual12, g
an assignment function, g[x/n] is that sequence in which x is the n-th element and which
agrees with g in all other positions and U the set of individuals. As (15) shows, λ
manipulates the assignment function (the familiar lambda-abstraction), whereas, as (16)
shows, μ manipulates the parameter for the perspective in the index, so that the truth of
φ is dependent on the values taken by x. To illustrate how this complicated account
works, let us consider the following example:
(17) Every man rode some ride that is fun
(17) is represented in Lasersohn’s system as
(18) [[every man]µ1 [[some [ride that λ2 [pro2 is-fun]]]λ3 [pro1 [rode pro3]]]],
where the quantifier phrase ‘every man’ binds both a variable for individuals in the
object language (pro1) and the parameter for perspectives in the index. Thus, Lasersohn
claims, ‘[(17)] receives a reading in which each man rode a ride which is fun for him,
even though there is no variable corresponding to for him in the syntactic representation.
The index itself is ‘bound’, but this is part of our meta-language, not the object
language.’ (2008: 325).
All its technical ingenuity notwithstanding, index binding has considerably less
expressive power than variable binding, as Lasersohn is ready to admit. Once there are
multiple predicates of personal taste in a phrase, the variable approach allows for
multiple binding, since each PPT might come equipped with its own syntactic variable.
In index binding, by contrast, the quantifier phrase binds the single perspective
parameter. Lasersohn’s example, (19), could in principle be interpreted in (at least) four
different ways (17a-d):
(19) Every man gave some woman a fun ride and a tasty dish. (2008: 325)
(19a) Every man gave some woman a ride and a dish, which were tasty and fun
according to the speaker.
12 We interpret Lasersohn’s parameter for individuals u as a parameter for perspectives. Since
perspectives are held by (at least) one individual, the modification is innocuous.

(19b) Every man gave some woman a ride and a dish, which were tasty and fun
according to each man.
(19c) Each woman received a ride and a dish, which were tasty and fun
according to her standards.
(19d) Every man gave some woman a ride fun by his standards, as well as a
dish tasty according to her standards.
Index binding can represent readings (19a)-(19c) without problems, because each of
these interpretations invokes only a single perspective. This, Lasersohn writes, makes
perfect sense of one basic tenet of relativism:
[In index binding] we are employing a single, systematic parameter, relative to
which all denotations are assigned; and if an operator manipulates this
parameter, it will do so for all expressions in its scope. (…)
The intuition behind this pattern can perhaps be expressed this way: In a
relativist theory, in order to assess a sentence for truth or falsity, one must adopt
a stance – that is, truth assessment is always done from a particular perspective.
Operators in the sentence may shift the perspective from which truth assessment
is to be done, or quantify over such perspectives; and when they do so, the
relevant perspective must be adopted for the entire scope of the operator.
Because such operators shift the perspective from which truth is assessed, rather
than shifting the denotation of some particular expression like a pronoun, they
cannot selectively shift only certain items in their scope. (2008: 326)
So far, so good. However, in order to make sense of (19d) – the interpretation in which
the two PPTs are relativized to two different perspectives – we need flexible variable
binding. The hidden argument of ‘fun’, on a contextualist rendition of this
interpretation, is bound by ‘every man’, whereas the variable of ‘tasty’ is bound by
‘some woman’. According to Lasersohn, this is not a problem, because ‘the sentence
[i.e. (19)] cannot mean that each man gave some woman a ride that was fun for him,
and a dish that was tasty for her’ (2008: 325, italics added). This, in turn, he takes to
‘show that [PPTs] cannot have arguments freely chosen from a set of pronouns similar
to pro1, pro2, pro3, …’ (2008: 326). Now, we agree with Lasersohn that the
interpretation (19d) of his example sentence (19) may sound rather awkward, and that it

is not easy to get a multi-perspectival reading of that sentence.13 However, this does not
mean that claims involving multiple PPTs never allow for perspectival plurality. In fact,
in all of the examples (4)-(8) provided above, the multi-perspectival reading is not only
perfectly felicitous, but arguably constitutes the default interpretation. Naturally, similar
examples can be devised to include quantifiers, and thereby make a case in favor of
multiple perspective variables in a single claim:
(20) On Halloween, every child would play a silly trick on some adult or else
get a delicious treat.
(21) Every Argentinean took some vegetarian friend to the Sunday barbecue
for some tasty steaks and delicious corncobs.
(22) On father’s day, the fair was in town. Every dad took some child to the
fairground to taste the delicious local brew and try out the fun new rides.
(23) Every historian took an engineering friend along to the university fair and
there were great new robots and highly interesting history talks.
One perfectly felicitous interpretation of (20), for instance, would have it that the
victims of the Halloween pranks, i.e. the adults, consider them silly, whereas the
recipients of the treats, i.e. the kids, consider said treats delicious. This is not only a
perfectly plausible reading, but the default reading: the kids won’t find their own tricks
silly, whereas the adults quite possibly think little of the cheap chocolate handed out to
appease the youngsters. To take just one more example, on the most natural
interpretation of (22), the perspective variable of the ‘delicious local brew’ is bound by
‘every dad’, whereas the variable of the ‘fun ride’ is bound by ‘some child’. Lasersohn’s
discussion of (19) shows that, by means of index binding, the relativist can
accommodate a considerable number of interpretations of claims involving multiple
predicates of personal taste – namely, all the ones in which the various PPTs are
relativized to a single perspective. However, it is clear that the limited expressive power
of index binding is insufficient to account for all the readings sentences similar to (19) –
in particular, for the multi-perspectival readings illustrated by our examples (20)-(23).
13 The reading might be acceptable given enough contextual support. However, instead of trying to devise
a context where the multi-perspectival reading of (19) sounds fine, we will make use of more natural (but
structurally equivalent) examples.

Though our principal focus lies with relativism, a brief note on contextualism.
Prima facie, plurality can be easily accommodated by contextualism, given that on such
an approach each PPT comes part and parcel with its own free perspective variable.
There is no restriction regarding what type of perspective can saturate the variable, even
in cases of coordination as the above. Variables can be filled by an exocentric
perspective, by a generic perspective or the subjective perspective of the speaker. In
cases where a sentence contains multiple PPTs, there is no reason to suppose that the
two different variables need to be saturated by the same perspective or even the same
type of perspective. Hence, contextualism does not face the same obstacles generated by
plurality as relativism does.
5. Relativist Responses
5.1. Available Options
We have suggested that perspectival plurality engenders two important problems for
relativism: It questions the possibility of justifying the postulation of perspective
parameters by aid of operator arguments, and it raises severe difficulties to account for
binding. We see two types of responses that can be made on behalf of the relativist: A
negative strategy would attempt to challenge the central premises of our arguments. A
positive strategy would amend the relativist framework so as to accommodate plurality.
As regards negative options: To deal with both arguments wholesale, what must
be undermined is the empirical premise common to both, that is, the premise that
ordinary language speakers do in fact interpret the above example sentences in multi-
perspectival fashion. However, in a series of five studies in experimental semantics
involving over 600 participants, Kneer (ms.) finds conclusive proof for perspectival
plurality both with respect to taste claims and epistemic modals. Given that a wholesale
attack on the empirical plausibility of plurality is thus unlikely to succeed, relativists
will have to question the conceptual assumptions of each argument in a piecemeal
fashion.
Though we here assume that parameters must be justified, and that one of the
most plausible ways to do so consists in operator arguments à la Kaplan, alternative
strategies can be envisioned.14 Furthermore, the requirement to account for syntactic
binding could be called into question, for instance on the basis that embedding is not
14 MacFarlane (2009, 245), for instance, casts doubt on Stanley’s (2005, 69-72) contention that parameter
proliferation can only be motivated through operator arguments.

semantically innocent15 or that the binding argument over-generates.16 Pursuing such
questions in depth would carry us too far afield. We would merely like to note that this
line of defense is only successful if both the binding argument and the requirement for
parameter justification are brought to fall, most likely on independent grounds since
they make different assumptions. Note that the view that parameter proliferation must
be justified by an operator-type argument and consistent with syntactic binding is a
common one and shared by many relativists. Undermining the premises of both
arguments would have considerable revisionary consequences, which might be seen as
both an ad hoc and excessive response to the phenomenon of perspectival plurality.
Alternatively, a positive proposal could be envisaged, which amends the
relativist framework in ways suitable to accommodate perspectival plurality. We can
think of two ways this could, at least in principle, be achieved: further parameter
proliferation or ‘multi-indexing’ on the one hand, or else a general strategy according to
which complex sentences invoking multiple PPTs are broken down into mono-
perspectival units whose truth is evaluated individually. In the following we will briefly
voice some reservations as regards the first option (section 5.2). The second, novel,
alternative – the ‘paraphrasing strategy’ – will be developed in a bit more depth in
section 6, as we consider it the relativist’s best shot. We ultimately find fault with this
type of response, too, and must conclude that the problems plurality raises for relativism
are not easily overcome.
5.2. The Uniqueness of Perspective Constraint
In sections 3 and 4 we have seen that perspectival plurality poses a challenge for
relativism both with respect to operator arguments and with respect to syntactic binding.
As regards the latter, one might think that the problem lies in a certain particularity of
Lasersohn’s account of binding. But, as the passage we quoted from Lasersohn makes
clear, the problem runs deeper: according to Lasersohn himself, it is a core commitment
of relativism that is in tension with the multi-perspectival readings we have put forward.
Lasersohn’s reason for rejecting such readings of (19) had to do with the adoption of a
stance, which amounts to the claim that each sentence has to be evaluated at a unique
15 Cf. Recanati (2002), cf. also Sennet (2007) for helpful discussion.
16 Cf. for instance Capellen & Lepore (2002: 274) and Recanati (2002); for a response cf. Martí (2006).

perspective.17 Let us flag this ‘one sentence – one stance – one perspective’ claim for
further reference, by dubbing it the
Uniqueness of Perspective Constraint (UPC)
A sentence has to be evaluated for truth relative to one and only one perspective
of a certain kind.
We take it that the UPC follows from a natural view of what relativist semantics is.18
Relativists claim that perspectives belong to circumstances of evaluation in the
Kaplanian divide between content and circumstances of evaluation, along with world
and time parameters. These parameters are functions from contents to truth-values. This
explains why there can only be one of each of these parameters per content evaluated:
the work of each parameter is to help determine whether a certain content is true or false
(or true or false relative to the other parameters). What doesn’t seem feasible is to have
two perspective parameters (or two world or time parameters) in the circumstances of
evaluation. It would be feasible to have two different perspective parameters if these
worked at the local level, i.e., at the level of the predicate, before the content is built by
a process of composition. However, if perspectives work at the global level, i.e. the
level of sentential content, there seems to be no way to make two different parameters
work in the way the relativist would want them to work. In other words, the UPC is in
tension with perspectival plurality, and the problem is general.19
17 A more detailed discussion about taking a stance in the relativist framework can be found in Lasersohn
(2009, section 3).
18 Kissine (2012) makes the same claim about what he calls ‘circumstance-of-evaluation-relativism’ –
which is what we simply call ‘relativism’. Although his focus is a relativist view about gradable
adjectives, the objection he raises involves multi-perspectival readings in which one of the perspectives is
made explicit. His objection, however, is more complex in the sense that he investigates short exchanges
in which one of the interlocutors use the locution ‘I agree’ to signal the fact that there is agreement
concerning a conjunctive sentence whose conjuncts each contains a gradable adjective that is to be
evaluated with respect to a different comparison class. One premise in his argument against relativism
about gradable adjectives is the assumption that each conjunctive sentence has to be evaluated with
respect to a unique comparison class – that is, he assumes the UPC.
19 The problem posed by perspectival plurality can be circumvented if one thinks that the semantics of
PPTs doesn’t need to be uniform. Thus, MacFarlane (2014) thinks that PPTs that are used differently get
different treatments: if a PPT is used exocentrically (that is, from another’s perspective) or is bound, the
predicate has a variable for perspectives in its logical configuration, while if a PPT is used egocentrically
(that is, from one’s own perspective) the variable is lacking. (A similar idea appears in Stephenson
(2007), where both uses of PPTs have a variable for perspectives in their logical configuration, but the
variables are of different types (pro and PROJ).) Since perspectival plurality consists in various
combinations of egocentric, exocentric and bound uses of PPTs, this approach avoids the problem
because there is always one perspective the sentence will be evaluated against – that relevant for the
interpretation of the predicate(s) used egocentrically – thus rescuing the UPC. However, MacFarlane

Before we move on, a clarificatory remark concerning the precise formulation of
the UPC above. According to the constraint, a sentence has to be evaluated for truth
relative to one and only one perspective of a certain kind. As various philosophers have
pointed out, relativists face a choice as regards the determinate set-up of the
circumstances. Each type of perspective-dependent expression (aesthetic predicates,
taste predicates, epistemic modals etc.) might require its own parameter. Or else, one
might combine them all into a ‘catch-all standard of evaluation’ (Kölbel, 2009: 384), i.e.
a general perspective parameter which all (or most) types of such expressions depend
on. The latter approach is just as problematic as the first, if not even more so, in
particular since it means that sentences invoking multiple perspectives are even more
ready at hand. Consider:
(24) The cat food might be tasty. (Stephenson, 2007, p. 499, credited to
Danny Fox)
Stephenson employs (24) to argue that predicates of personal taste can, whereas
epistemic modals cannot, be understood relative to an exocentric perspective. This
alleged difference is, in our view, rather spurious, but off-topic. What the example can
be used to show, however, is that a catch-all parameter for perspectival features is
problematic: There simply seems no plausible alternative to a reading according to
which the tastiness of the cat food must depend on the cat’s perspective, and the
epistemic modal ‘might’ on the speaker’s perspective.20 What this suggests is that
advocates of parameter proliferation best help themselves to different parameters for
different sorts of perspectives, so as to avoid that the problems here raised arise in even
more drastic form with regards to a single catch-all parameter. However, some of the
examples above show that it is simply not possible to save relativism by distinguishing
doesn’t provide any justification for the different treatment of the egocentric, exocentric and bound uses
of PPTs, and a solution of this sort thus sounds rather ad hoc. Besides being ad-hoc, this approach makes
PPTs ambiguous between a one-place predicate (when used egocentrically) and a two-place predicate
(when used exocentrically and bound), thus leading to a hybrid (that is, relativist and contextualist) view.
Ambiguity, we take it, is better avoided as a semantic solution and used only as a last resort.
20 Kölbel (2009) mentions another problem in passing. Having introduced a FOR operator for taste
predicates (see above) and a POSSIBLY operator for epistemic modality he acknowledges that ‘the two
operators might cause some complication to do with scope. Consider, for example, the sentence ‘Whale
meat might have been tasty for Anna’. Should this be construed as ‘POSSIBLY (FOR Anna, whale meat
is tasty)’ or as ‘FOR Anna, POSSIBLY whale meat is tasty’?’ Given that we already run into problems
regarding scope when different parameters depend on the outlook of a single individual, things are bound
to get considerably worse if different parameters invoke different perspectives.

different kinds of perspectives. It can be said, concerning our example (5), for instance,
that whereas ‘tasty’ is clearly a predicate of personal taste, ‘silly’ is a different kind of
predicate, which relates to a different kind of perspective. This move is not feasible in
cases like (6), where we find two predicates of personal taste proper, ‘delicious’ and
‘tasty’. We doubt that we can individuate kinds of perspectives in a way so fine-grained
that we could get relativism off the hook in this way.21
5.3. Multiple Perspective Parameters
The train of thought proposed in the previous section, perhaps, moves a little too
quickly. Relativism, one might hold, is either committed to the UPC, or must provide a
plausible picture according to which there are multiple perspective parameters with
regards to a particular domain. But contrary to what was alleged above, there is good
precedent regarding multiple parameters of the same sort, at least as concerns time. In
fact, multi-indexing was subject to extensive debate in the 1970s, and if the strategy
holds promise regarding time, it might carry over to various sorts of perspective
parameter. Kamp (1971) and Vlach (1973) argued that expressions such as ‘now’ and
‘then’ call for double temporal indexing in frameworks that treat tense as index-shifting
sentence operators. Consider the following examples:
(25) One day, all persons alive now will be dead.
(26) Once all persons alive then would be dead.
The sentence in (25) makes reference to two different times, the time of utterance, as
well as a certain day in the future when everyone alive at the time of utterance is dead.
The truth conditions of (25) cannot appropriately be expressed by invoking a standard
Priorean future tense operator only. The time the indexical expressions ‘now’ or ‘at the
present moment’ refer to would also be shifted into the future by the future tense
operator. What is required (it seemed) is double temporal indexing, that is, one index
shiftable by tense operators plus another rigid one which gets picked up by the indexical
21 We thank an anonymous reviewer for pressing this point. Let us add that the problem related to
perspectival plurality does not arise only with respect to PPTs. It is also possible to devise examples
where two different epistemic modals (e.g. two different occurrences of ‘might’) in a sentence relate to
two different perspectives (Kneer 2015, Kneer ms.). In these cases, it makes no sense to argue that the
two occurrences of ‘might’ relate to different kinds of perspectives.

‘now’. The first index corresponds to the time of evaluation, the second to the time of
reference.
Though we will not investigate the possibility of imitating this strategy with
respect to perspective parameters in detail, two brief considerations why it holds little
promise. Firstly, whether multi-temporal indexing is a helpful idea is highly
contentious. King (2003: 222-223) for instance, demonstrates that there are good
syntactic reasons to treat tenses as quantifiers over times rather than as index shifting
sentence operators. But if the temporal index shifting operators are best avoided then,
King concludes, we don’t need temporal coordinates either. So the hope to mimic multi-
index approaches in tense logic as regards other parameters might prove frustrating:
syntactic concerns pertaining to multiple time coordinates bring out the attractiveness of
time-specific (or eternalist) propositions. Multiplying other parameters to account for
perspectival plurality might raise similar worries which, ultimately, make a case for
taste-specific propositions and the like. Naturally, the attempt to spell out such an
argument makes little sense in the absence of a detailed story – to be provided by the
relativist – regarding how such a multiple perspective parameter picture is to work in
the first place.
Secondly and more importantly, even in the absence of a detailed proposal, it
seems that the analogy between multiple time indexing and the multiple taste indexing
we would require to circumvent the problems posed by our examples is a poor one. In
the tense-logic case, multiple indexing is motivated by the fact that certain indexicals
like ‘now’ (as well as expressions which incorporate some indexical element like ‘the
current president of the US’) leap out of the scope of the tense operator. The problem
posed by ‘now’ is resolved (if it is) by postulating two very different time coordinates,
each of which has a specific function. The time of evaluation interacts with tense and is
shiftable, the time of reference interacts with temporal indexicals and is unshiftable.
In our examples above, however, such a clear-cut division of labor is not
available: different occurrences of predicates of personal taste must be evaluated
relative to the same type of thing – a personal standard of taste – and hence the same
type of parameter or coordinate. So whereas the different temporal parameters discussed
are devised in such a way that they can only interact with the relevant parts of the
sentence, multiple taste parameters in our examples are, as it were, ‘unoriented’. The
problem is that none of the PPTs in our examples so far behaves like a reference-fixing
device (we will discuss examples in which PPTs behave like this below). These

examples involve PPTs that are of equal status: either both of them contribute with a
perspective to the content of the sentence, or both of them relate to perspectives in the
circumstances of evaluation. The relativist is committed to this latter view, but such a
view runs is in conflict with the UPC.
6. The Paraphrasing Strategy
The relativist, we said, faces a dilemma: holding fast to the UPC means plurality is left
unaccounted for, accommodating the latter means giving up a central tenet of relativism.
A final attempt to disarm the dilemma consists in modifying the UPC in ways that
reduce the scope of Lasersohn’s ‘one stance’ commitment from entire sentences to
meaningful units of truth-evaluation. Differently put, Perspectival uniformity could be
satisfied at the level of individual sub-sentences, while the truth of the content as a
whole might still be relative to multiple perspectives. Consider this revised version of
the UPC:
Uniqueness of Perspective Constraint* (UPC*)
A unit of evaluation has to be assessed for truth relative to one and only one
perspective of a certain kind.
Restricting the UPC in scope makes room for the following proposal: Suppose that
multi-perspectival sentences can be paraphrased as concatenations of simpler
propositional units, each of which contains but one single PPT. The resultant mono-
perspectival units can be evaluated individually in conformity with the UPC* (i.e.
invoking not more than one perspective each). The truth-value of the sentence as a
whole is calculated on the basis of the values of the individual units. Although the truth-
value of each individual unit is relative to a single perspective at most, the truth-value of
the content of the sentence as a whole can be sensitive to multiple perspectives.
6.1. Applications and fundamental premises
Very roughly, the paraphrasing strategy, as we will call it, assumes that multi-
perspectival claims can be paraphrased into a series of mono-perspectival units joined
by logical connectives. Furthermore, the strategy requires that the process of truth-
assessment can be conceived (or at least reconstructed) as proceeding in a piecemeal
fashion in which the truth-value of the sentences as a whole is determined on the basis

of the values of its individual units. To see how the strategy might deal with some of the
problematic cases, let’s begin with a simple example. A complex sentence such as (27)
is factored out into a conjunction of two simpler individual units, (28):
(27) Mary has a biscuit and a coffee.
(28) Mary has a biscuit & Mary has a coffee.
Assessing (27), the suggestion goes, involves assessing ‘Mary has a biscuit’ and ‘Mary
has a coffee’ individually. Once a verdict for each of the two individual units has been
reached, the latter are used to compute the truth-value of the complete content in
conformity with the basic rules of propositional calculus. (27) is true if and only if both
of the conjuncts of (28) are true. Things are no different when it comes to sentences
containing predicates of personal taste. (29), for instance, can be paraphrased as (30):
(29) Mary had a fun ride and ate a tasty dish.
(30) Mary had a ride & Mary ate a dish & the ride was fun & the dish was
tasty.
Just as in the previous example, (29) is true if and only if every single one of the
conjuncts of (30) is true. In line with the UPC*, the individual units ‘the ride was fun’
and ‘the dish was tasty’ must be evaluated with respect to exactly one perspective each.
Since they are assessed individually, however, nothing prevents us from using different
perspectives for each of the units. Whether this is advisable depends, of course, on
which perspectives are salient in the particular context at hand. For instance, in a
context in which the speaker considers himself an authority in culinary matters, but
defers to experiencers as regards rollercoasters, the ride might be evaluated as fun
relative to Mary’s perspective, yet the dish as tasty relative to the speaker’s perspective.
The UPC* is satisfied with respect to each of the conjuncts containing a PPT, while the
truth-value of the sentence as a whole can still be sensitive to more than one
perspective.
Consider further (31), which does not contain a conjunction but a conjunctive
DP. (31) can be paraphrased as (32):
(31) Mary had a fun ride and a tasty dish.

(32) Mary had a ride & Mary had a dish & the ride was fun & the dish was
tasty.
Again, the perspective parameters of the two individual units containing PPTs can be
set to different values, and again (31) is true iff all of the conjuncts of (32) are true –
potentially with respect to distinct perspectives.
So far we have outlined a very general strategy relativists might pursue in order
to counter the worries posed by perspectival plurality. Fleshing out this proposal in a
satisfying manner would constitute a paper in itself. Here we limit ourselves to briefly
sketching some of the core assumptions on which the paraphrasing strategy relies. There
are at least four such assumptions:
A1: Sentences that contain n PPTs can always be decomposed into at least n
units of evaluation.
A2: Each unit of evaluation will contain no more than one PPT in a
predicative position.22
A3: The truth of each constitutive unit can be assessed independently.
A4: The truth of the entire sentence can be derived from the truth values of its
individual units.
Each of the assumptions raises problems and would saddle the relativist with
commitments that go well beyond the problematic here raised. We will briefly discuss a
few of these.
6.2 Assumption A1: The possibility of paraphrasing
Though A1 merely states that multi-perspectival claims can be factored out into
multiple units which involve at most one perspective, a satisfying account must state a
general recipe for the paraphrasing process. One possibility to do so focuses on atomic
sources of error. Suppose we want to assign a truth-value to the sentence ‘This is a
green tomato’. We would check whether the designated object is a tomato, and
thereafter whether the object is green. Given that there are two potential sources of
error, i.e. two ways in which the sentence can misrepresent the world, one needs to have
22 The reason we talk here about predicative position will become clear in section 6.3.

an eye on each. Since there are two sources of error that must be assessed
independently, the sentence must be broken down into two distinct units of evaluation.
Each unit is assessed individually, and the value ‘true’ can be assigned to the sentence
as a whole if both sources of error can be ruled out, that is, if each unit is evaluated as
true. This strategy is rather general in nature, because it does not engender further
commitments with respect to whether ‘green’ and ‘tomato’ denote properties, sets, or
functions, although it is in principle plausible that the way we verify whether a predicate
applies to a subject is by thinking about the predicate as denoting a property.
Alternative theories of truth-value assignment can be envisioned. Here is one:
when evaluating ‘This is a green tomato’ we focus on the intersection of the set of green
things and the set of tomatoes (or the conjunctive property green tomato) and check
whether the demonstrated object is present in the set (or whether the object instantiates
the conjunctive property). On this proposal, there would be just one possible source of
error and so just one unit of evaluation. When evaluating sentences with multiple PPTs
in predicative position, we have to construe the intersection of several sets or properties,
each of which may relate to a different perspective. Suppose we say
(33) Marie had a funny-looking, tasty dish,
where the dish was funny-looking to us and tasty for Marie. If we use this type of
verification procedure, then before we assign a truth-value to (33) we have to
contextualize the denotations of ‘funny-looking’, and of ‘tasty’, so that the truth or
falsity of (33) would not be relative to a perspective. Naturally, this type of verification
procedure is at odds with the paraphrasing strategy and the trouble plurality spells for
relativism is back full force. What this means is that adopting the paraphrasing strategy
engenders a commitment to the first type of verification procedure, which tracks atomic
sources of error. The best support for this view would be evidence that demonstrates
that ordinary language speakers in fact proceed along these lines in truth-value
assessment. Unfortunately, very little is known about the psychology of truth-
assignment to sentences in natural language23, so a claim of this would constitute a
23 See, however, Pietroski, Lidz, Hunter & Halberda (2009) for some partial results regarding the
expression ‘most’. The authors compare different ways (‘verification procedures’) of assigning truth-
values to ‘most’ sentences. Such verification procedures yield truth-conditions that are extensionally
equivalent, but they are very different in computational terms. They focus in particular on the ‘one-to-one
correspondence strategy’ and the ‘cardinality comparison strategy’ as applied to a simple sentence such as

somewhat daring empirical conjecture. Alternatively, one might consider it sufficient to
provide a plausible, rational reconstruction of said underlying psychological process,
which would be subject to revision once empirical linguistics makes progress. Proposals
of this sort are not uncommon. For example, on some interpretations Grice’s theory of
implicature should be understood as a plausible rational reconstruction of the
psychological process that underpins the processing of implicatures.24 Advocates of the
paraphrasing strategy would hence need to provide some general, empirically plausible
rationale regarding why truth-value computation might proceed in piecemeal fashion in
the first place and, on the basis of the former, a recipe of how the complex sentence is to
broken down into smaller, meaningful units.
6.3. Assumptions A2 and A3: Mono-perspectival units and independent truth-
assessment
Let’s turn to A2, according to which each unit of evaluation must contain no more than
one PPT in a predicative position. As an anonymous reviewer suggested, certain types
of adverbs might give rise to tricky cases. Consider:
(34) The dog food is astonishingly tasty.
Since the dog food itself cannot reasonably be deemed astonishing, it is not obvious
how the two PPTs25 in predicative position can be pulled apart so as to form two distinct
units of evaluation. Perhaps (34) could be factored out into (35):
(35) The dog food is tasty & this is astonishing.
Some support that a paraphrase along these lines is plausible comes from authors like
Katz (2005) and Nouwen (2005) who have argued that ‘astonishingly’, in sentences
such as the one under consideration here, behaves like a special intensifier or degree
‘most xs are F’ (e.g., ‘most dots are blue’). The one-to-one correspondence strategy in a simple setting
where only Fs and Gs are relevant, consists in checking (i) whether for each thing that is an F there is a
corresponding thing that is a G, and then (ii) whether there is an excess of Fs, thus verifying that most xs
are F. The cardinality comparison strategy compares the cardinalities of the Fs and the Gs and assigns
truth if the cardinality of the Fs is bigger than the cardinality of the Gs. After running some experiments,
their conclusion is that we assign a truth-value to a ‘most’ sentence by comparing cardinalities.
24 Cf. for instance Geurts & Rubio-Fernández (2015).
25 Strictly speaking, ‘astonishing’ is not a PPT. However, it is similar enough as it is also a perspective-
dependent expression in the relevant sense.

modifier such as ‘very’, not as a sentential modifier as is sometimes assumed. Hence,
(34) entails ‘the degree to which the dog food is tasty is astonishing’.26 Although this
seems a viable rendering of (35), the relativist is saddled with certain assumptions about
‘astonishing’ that might not be completely innocuous.
Note, on the other hand, that paraphrasing (34) as (35) raises a worry regarding
the independence of the conjuncts (A3 above): the use of the demonstrative ‘this’ in the
second conjunct of (35) explicitly refer back to the entire first conjunct. Naturally, it
seems unreasonable to ban any connection between conjuncts whatsoever – in ‘John
played a silly prank and he had a lot of tasty licorice’, ‘he’ refers back to John yet
paraphrasing it as two conjuncts doesn’t look problematic. However, the more
interdependence is deemed acceptable, the more assumption A3 comes under pressure,
that is, the assumption that the truth of the distinct units can be assessed independently
from one another. While we will not pursue this matter further here, suffice it to say that
the relativist needs to clearly specify the sense in which the two units are taken to be
independent from each other, and ensure that A3 holds good.
Assumption A2 seems to come under pressure from other expressions as well.
Consider:
(36) The book was fun yet informative.27
Paraphrasing (36) into mono-perspectival units such as ‘The book was fun’ & ‘the book
was informative’ loses the contrastive force of ‘yet’ and it is not clear what should be
added in order to account for it. Perhaps one could follow Potts (2005), according to
whom the contrastive force of ‘yet’ can be captured in a separate unit such as ‘For the
most part, interesting books are not fun’. But the problem merely resurfaces, since an
extra unit of this sort contains (37), which itself can be multi-perspectival:
26 This way to go might amount to a satisfactory solution due to the fact that, after all, it is possible to
claim that we can distinguish two truth-evaluable units in (35). A first unit is expressed by the claim: ‘the
dog food is tasty’. This unit has to be evaluated according to a perspective (the dog’s perspective) and to
the standard fixed by the context. Once it is decided whether the dog food is tasty for the dog according to
the standard we have fixed (a standard related to the price we paid, for instance), we have to decide
whether it is astonishing to find the dog’s food in the point of the scale of tastiness where we find it. In
order to do this, we do not have to relativize again to the dog’s tastes. The dog’s tastes were relevant to
fixing what counts as tasty, and thus to evaluating the first atom. Now, in the second step, we are only
concerned with whether finding the food there, where we find it, is astonishing.
27 If one worries about ‘informative’ not being a PPT, here is an example with two PPTs in the strict
sense: ‘I took the kids to Pomp, Duck & Circumstance. The dinner was fun yet delicious.’

(37) Interesting books are not fun
Here is a possible solution. As mentioned above (assumption A2), the defender of the
paraphrasing strategy is committed to the idea that we will not find more than one PPT
in a predicative position in each of the units of evaluation corresponding to any given
sentence. As we hope it is clear, (37) has a multi-perspectival reading according to
which the books are interesting for the parents but not fun for the kids. Nonetheless, it
seems that in (37) there is a single unit of evaluation and two PPTs. Now, it could be
said that in order to know whether (37) is true we have to first find the interesting books
(i.e. books that are interesting for us, parents), and then check whether those books are
fun, i.e., whether the kids find them fun. That is, the PPT in the subject position,
‘interesting’, helps us determine the extension of the NP. Once we have that extension,
we only have to check whether it is true that the individuals in the set have the property
of being fun. To evaluate this we have to relativize to the kids’ perspective. In sum,
once we have fixed the denotation of the NP, the unit of evaluation contains just one
PPT: ‘the books in the set (the ones parents find interesting) are not fun’. One of the
PPTs would contribute to the content of the sentence, while the perspective related to
the other one, the one in the VP, which has a predicative use, is located in the
circumstances of evaluation. This move would draw on the multi-indexing strategy
explained above. The reason why multi-indexing would be allowed in this case (if it is)
is that the different PPTs play different roles in the sentence: one has a referential role,
while the other is purely predicative.
In the previous section we contended that it was not possible to account for our
examples (4)-(8) and (20)-(23) by distinguishing two roles perspectives might play:
supplying information to the content, and being part of the circumstances of evaluation.
In a sentence like ‘the current president of the US is handsome’, we can distinguish two
elements sensitive to time. Each plays a different role: the first (‘current’) has a role to
play in building the content by fixing a reference, while the second (the verbal tense)
relates to an index in the circumstances of evaluation. (At least, that would be the
relativist’s story.) Now, as all our examples until then concerned two PPTs that were in
straightforward predicative positions, both of them had to play the same role: either
both contributed to the content (as the contextualist would have it) or both related to
different indexes in the circumstances of evaluation (which could be a relativist move,
but we said we do not see how it can be implemented). However, example (37) is

different. Example (37) is structurally analogous to the example involving ‘the current
president…’, so the relativist seems to be allowed to deal with it in a similar way, i.e. by
treating the first PPT as contributing to fixing the reference of the object of predication.
Now, even if responses such as the ones tried here were to be seen as
satisfactory, we are convinced that there are many more examples that will pose
problems for a paraphrasing strategy (particularly ones involving quantification, as no
food is both tasty and amusing). While we take it that the strategy is probably the best
option for the relativist, we leave the question whether it is indeed a promising strategy
open.
6.4. Assumption A4: Truth-value computation
The final assumption, A4, requires that it be possible to derive the truth-value of the
complex sentence from the truth-values of the individual units. This premise, too, can
be called into doubt. One problem regards the question whether the constituent units
into which the complex sentence gets split up provide sufficient guidance in deciding
relative to which perspectives they must be evaluated. It is standard practice to speak
about context ‘providing’ or ‘supplying’ the relevant parameters of the circumstance of
evaluation (Kaplan, 1989: 591). This manner of speaking, however, considerably
oversimplifies what is actually going on. In evaluating a sentence in context, it is
frequently not the case that context puts a ready-formed world-time-perspective tuple at
the interlocutor’s fingertips. Instead of the passive reception of the ‘provided’
parameters, at times an active investigation of the context is required, in particular when
the options are plentiful (as, for example, in the wine and water-slide example, which
allows for at least twelve reasonable different readings as regards perspective).
Lewis, for instance, highlights that some features of context are more accessible
than others. Whereas certain features such as time, place, world and the speaker might
be relatively obvious, ‘[t]he audience, the standards of precision, the salience relations,
the presuppositions (...) of the context are given less directly. They are determined, so
far as they are determined at all, by such things as the previous course of the
conversation that is still going on at the context, the states of mind of the participants,
and the conspicuous aspects of their surroundings.’ (1980: 86). The question thus arises
whether paraphrasing, giving rise to a (slight) change in co-text, and thus context, will
leave the salience of perspectives unchanged. Differently put, relative to which
perspective the first constitutive sentence must be evaluated, and relative to which the

second, might either be less obvious, or once torn apart, the PPTs in the individual
constituent sentences might call for different perspectives than the ones that prove
salient when the sentence is evaluated as a whole.
Another, not unrelated worry: The precise mechanics of truth-evaluation
according to the paraphrasing strategy might be deemed somewhat mysterious. The
constitutive sentences of a multi-perspectival claim get evaluated separately, the
proposal goes, each with regards to a single, appropriate perspective. Since parameters
are ‘provided’, ‘supplied’ or ‘given’ by context, and since the perspective parameters
for each of the simple sentences differ, do we not require two different contexts, too? If
we understand contexts, with Lewis, as concrete situations of utterance, it is important
to recall that ‘features of context do not vary independently. No two contexts differ by
only one feature. Shift one feature only, and the result of the shift is not a context at all.’
(1980: 86; cf. also Kaplan’s ban on improper contexts, though recall that for Kaplan
‘contexts’ are technical constructs.) Now, if there are several units of evaluation instead
of one complex one, there may also be several contexts (differing in terms of more than
just one feature), and thus several circumstances of evaluation or indices. These indices
might in principle differ considerably, since they are determined by different contexts.
So assumption 4, too, is far from innocent. All this is not to say that the paraphrasing
strategy cannot be eventually made to work, or that the relativist is out of options in
general. What it does show, however, is that perspectival plurality constitutes a rather
serious challenge to relativism, and that overcoming it is not a simple affair.
7. Conclusion
In this paper we have introduced a phenomenon we called ‘perspectival plurality’,
according to which the truth value of a sentence with multiple predicates of personal
taste can depend on more than one perspective. Plurality, we suggested, might raise
difficulties for those who attempt to justify parameter proliferation in the circumstance
of evaluation by means of Kaplanian operator arguments. What is more, the
phenomenon blocks what might be considered the most promising relativist account of
binding.
We assessed two broad relativist strategies to respond to plurality. We rejected
the first – denying the data – by pointing to robust experimental results obtained by
Kneer (ms.). The second broad strategy we considered accommodates plurality at the
conceptual level. Two proposals of this sort were explored: the postulation of multiple

perspective parameters, and what we called the ‘paraphrasing strategy’. Whereas we
raised doubts that the former is viable at all, the latter, we argued, holds at least some
initial promise. However, certain tricky sentences might resists paraphrasing, and a
number of more general worries may afflict this approach, too. Further research will
prove whether the strategy can be successfully developed, or whether perspective
plurality turns out to be an insuperable obstacle to relativism.
References:
Brogaard, B. (2008). Moral Contextualism and Moral Relativism. The Philosophical
Quarterly, 58, 385-409
Cappelen, H. & Lepore, E. (2002). Indexicality, Binding and A Priori Truth. Analysis,
62, 271– 81
Cappelen, H. & Hawthorne, J. (2009). Relativism and Monadic Truth. (Oxford
University Press)
Chung, S. & Ladusaw, W.A. (2006). Chamorro evidence for compositional asymmetry.
Natural Language Semantics, 14, 325-357
Collins, J. (2013). The Syntax of Taste. Philosophical Perspectives, 27, 51-103
Geurts, B. & Rubio-Fernández, P. (2015) Pragmatics and Processing. Ratio 28, 446-46
Glanzberg, M. (2007). Context, Content, and Relativism. Philosophical Studies, 136, 1-
29
Huvenes, T. T. (2012). Varieties of Disagreement and Predicates of Taste. Australasian
Journal of Philosophy, 90, 167-181
Kaplan, D. (1989). Demonstratives. (In J. Almog, J. Perry & H. Wettstein (Eds.),
Themes from Kaplan (pp. 481-563). Oxford University Press.)
Katz, G. (2005). Attitudes toward Degrees. (In E. Maier, C. Bary & J. Huitink (Eds.),
Proceedings of SuB 9. Radboud Universiteit Nijmegen.)
King, J. (2003). Tense, Modality and Semantic Values, Philosophical Perspectives 17,
195-245
Kissine, M. (2012). From Contexts to Circumstances of Evaluation: Is the Trade-off
Always Innocuous? Synthese, 184, 199-216
Kneer, M. (2015). Perspective in Language, dissertation, Institut Jean Nicod / ENS
Paris.
Kneer, M. (ms.). Perspectival Plurality in Natural Language.

Kölbel, M. (2004). Indexical Relativism vs Genuine Relativism. International Journal of
Philosophical Studies, 12, 297-313
Kölbel, M. (2009). The Evidence for Relativism. Synthese 166, 375-394
Kölbel, M. (2011). Objectivity, Relativism and Context Dependence. (Fern Universität
Hagen)
Lasersohn, P. (2005). Context Dependence, Disagreement, and Predicates of Personal
Taste. Linguistics and Philosophy, 28, 643-686
Lasersohn, P. (2008). Quantification and Perspective in Relativist Semantics.
Philosophical Perspectives, 22, 305-337
Lasersohn, P. (2009). Relative Truth, Speaker Commitment, and Control of Implicit
Arguments. Synthese, 166, 359-374
Lasersohn, P. (2013). Non-World Indices and Assessment-Sensitivity, Inquiry, 56, 122-
148
Lewis, D. (1980). Index, Context, and Content. (In S. Kanger & S. Ohman (Eds.),
Philosophy and Grammar (pp. 79-100). Dordrecht: Reidel.)
Lihoreau, F. (2012). Personal Taste Ascriptions and the Sententiality Assumption. The
Reasoner, 6, 143-144
MacFarlane, J. (2009). Nonindexical contextualism. Synthese, 166, 231-250
MacFarlane, J. (2012). Relativism. (In D. Graff Fara & G. Russell (eds.), The Routledge
Companion to the Philosophy of Language (pp. 132-142). New York:
Routledge.)
MacFarlane, J. (2014). Assessment Sensitivity: Relative Truth and Its Applications.
(Oxford University Press)
Marti, L. (2006). Unarticulated Constituents Revisited. Linguistics and Philosophy, 29,
135-166
Moltmann, F. (2010). Relative Truth and the First Person. Philosophical Studies, 150,
187-220
Nouwen, R. (2005). Monotone Amazement. (In P. Dekker & M. Franke (Eds.),
Proceedings of the Fifteenth Amsterdam Colloquium (pp. 167-172). ILLC.)
Potts, C. (2005). The Logic of Conventional Implicatures. (Oxford: Oxford University
Press)
Pietroski, P., Lidz, J., Hunter, T. & Halberda, J. (2009). The Meaning of ‘Most’:
Semantics, Numerosity and Psychology. Mind and Language, 24, 554–585

Recanati, F. (2002). Unarticulated Constituents. Linguistics and Philosophy, 25, 299-
345
Recanati, F. (2007). Perspectival Thought: A Plea for Moderate Relativism. (Oxford
University Press)
Richard, M. (1981). Temporalism and Eternalism. Philosophical Studies, 39, 1-13
Rothschild, D. & Segal, G. (2009). Indexical Predicates. Mind and Language, 24, 467-
493
Sæbø, K. J. (2009). Judgment ascriptions. Linguistics and Philosophy, 34, 327-352
Sennett, A. (2008). The Binding Argument and Pragmatic Enrichment, or, Why
Philosophers Care Even More than Weathermen about ‘Raining’. Philosophy
Compass 3, 135–57
Schaffer, J. (2011). Perspective in Taste Predicates and Epistemic Modals. (In A. Egan
& B. Weatherson (Eds.), Epistemic Modality (pp. 179-226). Oxford University
Press.)
Snyder, E. (2013). Binding, Genericity and Predicates of Personal Taste. Inquiry, 56,
278-306
Stanley, J. (2005). Knowledge and Practical Interests. (Oxford University Press)
Stanley, J. (2007). Language in Context. Selected Essays. (Oxford University Press)
Stephenson, T. (2007). Judge Dependence, Epistemic Modals, and Predicates of
Personal Taste, Linguistics and Philosophy, 30, 487-525
Stojanovic, I. (2007). Talking about Taste: Disagreement, Implicit Arguments, and
Relative Truth. Linguistics and Philosophy, 30, 691-706
Zeman, D. (2013). Experiencer Phrases, Predicates of Personal Taste and Relativism:
On Cappelen and Hawthorne's Critique of the Operator Argument. Croatian
Journal of Philosophy XIII (39), 375-398