Quantifier Domain Restriction in the Minimalism/Contextualism Debate
by
Carleine van Schie
A thesis submitted in partial fulfilment of the requirements for the degree of
MASTER OF ARTS IN PHILOSOPHY at the UNIVERSITY OF AMSTERDAM supervisor Prof. Dr. M. J. S. Stokhof second reader Dr. P. J. E. Dekker September, 2014
Abstract
This thesis investigates, by means of conceptual analysis, in which manner the phenomenon of quantifier domain restriction can contribute to the debate between Semantic Minimalism, Moderate Contextualism and Radical Contextualism. By means of exploring the different contextual mechanisms employed by each position, a review is given on whether these mechanisms can give an adequate account of the phenomenon of quantifier domain restriction. Furthermore, since the debate between the three positions depends on the extent and manner one is willing to admit the influence of context on meaning and, conversely, on the extent and manner in which meaning is restricted by linguistic features, the debate boils down to whether a systematic, recursive theory of meaning is possible. The constraints on the formulation of the interpretation of a natural language given by the Principle of Compositionality serve as a guideline to determine if this is the case for treatment of quantifier domain restriction in the three positions. By the assumption of compositionality, the phenomenon of quantifier domain restriction turns out to be most adequately captured by the Moderate Contextualist.
Acknowledgements
I would like to express my sincere gratitude to my supervisor, dr. prof. Martin Stokhof, for his continuous guidance, insightful remarks, reassurances, patience, and vast knowledge. Without his engagement, I would not have been able to write this thesis.
Furthermore, I would like to thank prof. Paul Dekker, for being my second reader, and for introducing me to Natural Language Semantics in his memorable and enjoyable lectures.
Also, I would like to express my gratitude to prof. Elsbeth Brouwer, who, in previous years and throughout various courses, has introduced me to Philosophy of Language in such a manner that I immediately fell for it.
Last but not least, I would like to thank my friends and family, especially my parents, Jack and Sabineke, who have always had me in their thoughts and supported my every step.
Thank you, Carleine
TABLE OF CONTENTS ABSTRACT...I ACKNOWLEDGEMENTS...II INTRODUCTION...1 1. CONTEXT-SENSITIVITY IN SEMANTICS...3 1.1 INTRODUCTION...3
1.1.1 The Principle of Compositionality...4
1.2 ASSIGNING SEMANTIC VALUES...5
1.2.1 Literal Content...5
1.2.2 Semantic Content...6
1.2.3 Speech Act Content...8
1.3 INTUITIONSABOUT CONTEXT-SENSITIVITY...9
1.3.1 Incompleteness...10 1.3.2 Context Shifting...11 1.4 CONTEXTUAL MECHANISMS...12 1.4.1 Saturation...12 1.4.1.1 Indexicals...13 1.4.1.2 Semantic Parameters...14 1.4.1.3 Syntactic Ellipsis...16 1.4.2 Modulation...17 1.3.2.1 Free Enrichment...17
1.4.2.2 Loosening and Semantic Transfer...19
1.4.3 Implicatures...19
2. SEMANTIC MINIMALISM, MODERATE CONTEXTUALISM, AND RADICAL CONTEXTUALISM...21
2.1 SEMANTIC MINIMALISM...21
2.1.1 Arguments against Semantic Minimalism...23
2.1.1.3 Underdeterminacy...24
2.1.1.1 Availability...24
2.2.1 Arguments against Moderate Contextualism...26
2.2.1.1 A Slippery Slope to Radical Contextualism...26
2.3 RADICAL CONTEXTUALISM...28
2.3.1 Arguments against Radical Contextualism...28
2.3.1.1 The Miracle of Communication...28
2.4 THE COREOFTHE DEBATE...29
2.4.1 The Principle of Compositionality Revised...30
3. QUANTIFIER DOMAIN RESTRICTION...32
3.1 QUANTIFIERSIN NATURAL LANGUAGE...32
3.1.1 Intuitions about the Context-Sensitivity of Quantifier Domains...33
3.1.1.1 Incompleteness...33
3.1.1.2 Context Shifting...34
3.2 QUANTIFIER DOMAIN RESTRICTIONIN SEMANTIC MINIMALISM...34
3.2.1 Quantified Expressions as Indexicals...35
3.2.1 Minimal Quantified Expressions and Availability...37
3.2.3 Quantifier Domain Restriction, Semantic Minimalism, and Compositionality...38
3.2.3.1 The Contextual Principle of Compositionality...38
3.3 QUANTIFIER DOMAIN RESTRICTIONIN RADICAL CONTEXTUALISM...40
3.3.1 Quantifier Domains as Unarticulated Constituents...40
3.3.2 Quantifier Domain Restriction, Radical Contextualism, and Compositionality...42
3.3.2.1 The Modulated Principle of Compositionality...43
3.4 QUANTIFIER DOMAIN RESTRICTIONIN MODERATE CONTEXTUALISM...47
3.4.1 Quantifier Domains and Syntactic Ellipsis...47
3.4.2 Quantifier Domains and Semantic Parameters...48
3.4.2.1 The Binding Argument...49
3.4.2.2 Characterization of a Semantic Parameter for Quantifier Domains...53
3.4.3 Quantifier Domain Restriction, Moderate Contextualism, and Compositionality...54
3.4.3.1 The Live Principle of Compositionality...55
4. CONCLUSION...58
Quantifiers, Context, and Compositionality:
Quantifier Domain Restriction in the Minimalist/Contextualist Debate
Quantification is a topic which has interested linguists, philosophers, and logicians over many decades. However, whereas formally, logical quantifiers generally exhibit certain a priori behaviour by ranging over all the elements in a domain of discourse, this is not the case for quantifiers in natural language. In ordinary linguistic communication, it is rarely the case that speakers are talking about all the elements in a universal set of objects over which a quantifier ranges. Rather, they are speaking about a specific set of objects in the context in which they are communicating.
The behaviour of logical quantifiers has been mostly investigated by ideal language philosophers, who studied formal languages and, via them, language in general. Ideal language philosophers were concerned with syntax and semantics, that is, determining meaning by means of investigating the formal relation between expressions, and the relation between expressions and the objects to which the expressions are applicable. Quantifiers in use were a topic of investigation of ordinary language philosophers, who supported a more practical approach and considered the abstract nature of logic to obscure, rather than disclose, the properties of natural language. Ordinary philosophers were concerned with pragmatics, that is, determining meaning by means of investigating the relation between expressions and interpreters[CITATION Rec041 \t \l 1043 ] [ CITATION Sza05 \l 1043 ].
However, the sharp distinction between semantics and pragmatics have vanished over the years. Since linguistic communication is an actual, real phenomenon, the formal approach to language was insufficient to determine the application of expressions to objects by natural language speakers. Furthermore, the pragmatic approach could not sufficiently determine what the relation is between expressions and interpretation without mediation of certain invariant features of language [CITATION Rec041 \t \l 1043 ].
The disappearance of the sharp distinction between semantics and pragmatics has influence philosophers of language to give an account of meaning that accounts for both the contextually
invariant properties of natural language, as the influence that context has on the meaning of those properties. The extent and manner in which context influences the meaning of expressions has given rise to several new positions: Semantic Minimalism, Moderate Contextualism and Radical Contextualism.
The character of the debate between Semantic Minimalism, Moderate Contextualism and Radical Contextualism is threefold: i) philosophically, the debate is about the manner in which a theory of meaning may ascribe meaning to an expression, and to the sentences and utterances in which they occur, ii) conceptually, the debate involves the relation between contextual features and linguistic expressions, iii) empirically, the debate entails how hearers of utterances actually interpret natural language expressions.
The aim of this thesis is to investigate, by means of conceptual analysis, in which manner the phenomenon of quantifier domain restriction can contribute to the debate between Semantic Minimalism, Moderate Contextualism and Radical Contextualism. By means of exploring the different contextual mechanisms employed by each position, a review will be given on whether these mechanisms can give an adequate account of the phenomenon of quantifier domain restriction. Furthermore, since the debate between the three positions depends on the extent and manner one is willing to admit the influence of context on meaning and, conversely, the extent and manner in which meaning is restricted by linguistic features, the debate boils down to whether a systematic, recursive theory of meaning is possible. The constraints on the formulation of the interpretation of a natural language given by the Principle of Compositionality will serve as a guideline to determine if this is the case for treatment of quantifier domain restriction in the three positions.
1. Context-Sensitivity in Semantics
1.1 Introduction
In order to formulate an adequate semantic theory, it has to be clear what a semantic theory is supposed to do. A common reply is that the theory has to explain what a sentence means1. A dominant way to describe what a sentence means is to specify under which circumstances the sentence is true. This implies that each sentence represents the world in a certain way, and that the world must satisfy certain conditions in order for the sentence to be true. However, speakers might not always talk about the way things actually are. They may speak about how the world has been yesterday, or how the world would have been if something did or didn’t happen. [CITATION Soang \t \l 1043 ]. In order to account for these types of discourse, an intensional semantics has been formulated2.
An intensional semantics has a set of metaphysical possible worlds as its foundation. This means that the truth-conditions of a sentence do not just take into account the actual world, but also other possible worlds (such as the actual world at a different time). In a possible world semantics, the extension, or referent, of a singular term in a possible world is an individual. The intension, or ‘individual concept’, of a singular term is a function from possible worlds to individuals. The extension of a predicate in a possible world is a set of individuals, the intension of a predicate is a function from possible worlds to sets of individuals. The extension of a sentence in a possible world is a truth-value, the intension of a sentence is a function from possible worlds to truth-values. The latter, a function from possible worlds to truth-values, is called a proposition [ CITATION Mon70 \l 1043 ].
On this view, the meaning, or semantic value, of a singular term, is identified with a function from possible worlds to individuals. The meaning, or semantic value, of a predicate, is identified with a function from possible worlds to sets of individuals, and the meaning, or semantic value, of a
1 Considering the scope of this thesis, only declarative sentences will be of concern.
2 Although there are other semantic theories that are significantly different from intensional semantics, an intensional semantics serves present purposes adequately.
sentence, is identified with a function from possible worlds to truth-values, hence with the proposition that the sentence expresses [ CITATION Mon70 \l 1043 ].
1.1.1 The Principle of Compositionality.
An important methodological assumption is that the semantic value of a sentence, hence the proposition expressed by a sentence, is determined by the semantic values of its constituents and the way these constituents are arranged. In order for language uses to be able to make sense of an infinite set of sentences that is generated from a finite set of words by a finite set of recursive, syntactic rules, one has to postulate the cognitive reality of some sort of a system, as the human brain is finite as well. To achieve that, a principle has been postulated that accounts for linguistic systematicity. Since the referents of the constituents in all possible worlds ultimately determine the intension of a sentence, and hence the proposition expressed by that sentence, the assignment of a proposition to a sentence is a bottom-up process. Since the semantic value of a sentence ultimately depends on the semantic values of its constituents and the way these constituents are arranged, the proposition expressed by the sentence is composedcompositionally. This notion can be captured by the following principle [CITATION Pag07 \p 47
\t \l 1043 ]:
(PoC) The meaning of a complex expression is a function of the meanings of its constituents and their mode of combination.
(PoC) μ(σ(t1,…tn)) = ρ(σ)(μ(t1),…, μ(tn))
In this rendering, μ is the semantic function that maps syntactic terms on semantic values. t1, …tn are the syntactic terms that are the constituents of the complex expression σ(t1,…tn), that is built up from those constituents and the syntactic operator symbol σ. μ(t1),…, μ(tn) are the semantic values of the constituents. ρ is a composition function that maps a syntactic operator, or, mode of combination, σ, and the semantic values of the constituents onto the semantic value of the complex expression [CITATION Pag07 \t \l 1043 ][CITATION Pag071 \t \l 1043 ].
The Principle of Compositionality is of a methodological character, and does not limit natural languages themselves, but the formulation of their interpretation [CITATION Dekng \l 1043 ]. By means of postulating this systematic and recursive principle of interpretation, several features of natural language communication can be guaranteed3: (i) a compositional account of meaning assures the learnability of a natural language. By means of a finite set of lexical expressions and a finite set of syntactic rules, an infinite amount of meaningful sentences can be learned as having distinct meanings, (ii) a compositional language guarantees the understanding of novelty of meaningful sentences. It entails that speakers are able to interpret and understand the meaning of infinitely many sentences, even though they have not heard the sentence before, (iii) compositionality secures the possibility that speakers have the ability to generate infinitely many meaningful sentence. This argument is called the argument of productivity, (iv) a compositional account of natural language ensures intersubjectivity and communication. It explains that speakers and hearers of the same language interpret novel sentences similarly, and that they are able to agree, disagree, discuss and hence participate in successful communication[CITATION Pag10 \t \l 1043 ].4
1.2 Assigning Semantic Values
A semantic theory should be able to assign a semantic value to sentences, which is determined by the intensions of the constituents of those sentences, and hence ultimately by their referents in all the possible worlds. However, the question then arises how expressions are assigned their semantic value, and, more specifically, whether the proposition that a sentence expresses can be determined purely compositionally. In order to investigate this, the aspects which contribute to the proposition expressed by a sentence will be outlined.
1.2.1 Literal Content.
Firstly, the semantic value of a sentence is determined by the semantic value of various linguistic features of a sentence. The lexical meaning of the constituents in3 Although several additional arguments are formulated in favour of compositionality, the arguments stated serve present purposes.
4 Although the principle of compositionality and the presented arguments in favour of it are not uncontested, they will be assumed here, as the accompanying constraints placed on the formulation of natural language interpretation are of serious consideration in the debate between Semantic Minimalism, Moderate Contextualism and Radical Contextualism. This will be elaborated in section 2.4.1.
a sentence are taken as a starting point, as without knowing the lexical meaning of words, little can be said about the meaning of sentences containing them.
Consider the sentences:
(1) The book is on the table (2) The book is on the shelf
Sentence (1) will express a different proposition than the sentence (2), due to the different semantic values of ‘table’ and ‘shelf’. Whereas the other lexical constituents and the way these constituents are arranged remains the same, due the principle of compositionality, the semantic value and hence meaning of the sentence as a whole changes when the expression ‘table’ is substituted for the expression ‘shelf’.
Secondly, the way constituents are arranged, i.e. the syntactic structure, determine partially what a sentence means.
(3) The boy looked at the girl with glasses
Which proposition is expressed by (3) depends on whether the expression ‘with glasses’ specifies an attribute of the girl or of the way the boy was looking. Again, the principle of compositionality is at play here. Although the semantic values of the lexical expressions in () do not vary, their mode of composition, that is, their syntactic structure, does. This results in a different composition of the semantic values of the expressions, resulting in a different expressed proposition. These linguistic features (i.e. the lexical meaning of the constituents and the syntax of the sentence) together constitute what will be called the literal meaning of a sentence.
When talking about the literal meaning of a sentence, one is talking about the lexical meanings of the expression it contains, and the way these expressions are syntactically structured. This entails that the literal meaning of a sentence is context-independent. [CITATION Rec041 \t \l 1043 ]. The phrase literal content will be used for the conventional and context-independent linguistic meaning.
1.2.2 Semantic Content.
However, some expressions express something different if they are placed in a context. Suppose that a speaker wants to make a claim about pandas, either in a context about the panda populations, or about the pandas in a zoo. Consider the sentences:(4) The panda is an endangered species (5) The panda was set free
In both (4) and (5), the same expression ‘the panda’ is used. However, in (4), the expression refers to the panda species, whereas in (5), the expression refers to a specific instantiation of the panda species, i.e. a particular panda that was held in captivity. Although the same expression is used in (4) and (5), the reference of the expression depends on the context in which the sentences are uttered. Furthermore, an expression containing such instantiations of a general concept such as (5), may itself vary with context, as the truth-conditions of (5) depend on which panda is referred to. This means that the context in which a sentence is uttered may influence the semantic value of a sentence.
Since contextual features hence can influence the semantic value of linguistic features of expressions, the linguistic features of an expression may not be enough for a sentence to be
propositional, but a context is required to determine the semantic value of an expression or sentence. The context, that is, a range of extra-linguistic features, in which the sentence is used, then plays a role in assigning meaning to expressions. Interpreting these features requires pragmatic skills, that is, the ability of speakers and hearers to incorporate extra-linguistic contextual factors into linguistic communication. This entails that truth-conditions (sometimes) can only be specified by means of incorporating contextual effects on semantic content. For example, in order to interpret the meaning of (5), the hearer will have to resort to his knowledge about the context in order to determine which panda the expression ‘the panda’ refers to and assign semantic values and truth-conditions to an utterance of (5). However, in order to give an adequate theory of meaning, it has to be determined to
which extent these pragmatic effects intrude semantic content and thereby truth-conditions.
As shown in section 1, the meaning of a sentence is identified with the proposition it
expresses. In order to determine whether context has a role in determining the meaning of a sentence, we have to examine whether the sentence expresses a proposition at all without placing it into
context, and/or whether the sentence expresses a different proposition if uttered in different contexts. The literal meaning of a sentence may not be sufficient to assign a semantic value and hence referent to an expression, but it may nevertheless be needed in order to do so. This is the case when the sentence contains an expression that is context-sensitive in one way or another. The literal content of the context-sensitive expressions then supplies a function which, given a context, allows the hearer to determine the semantic content of the expression in that context [CITATION Rec041 \t \l 1043 ]. This entails that the literal content is insufficient by itself to determine what proposition the sentence expresses and what its truth-conditions are, but nevertheless is necessary for doing to. The context in which a sentence is uttered then also plays a role in evaluating it. Consider:
(6) I am a philosopher
What proposition is expressed by (6) depends on which person is uttering the sentence. The linguistic meaning of the pronoun ‘I’ stays the same, namely ‘the one who is speaking’. In this case, this linguistic meaning or literal content of the pronoun provides the rule that determines the referent of the expression given a certain context. Hence, contextual information is necessary in order to make the sentence propositional and truth-evaluable. Therefore, a distinction is made between the literal
meaning of a sentence, and the proposition that is expressed by an utterance of the sentence,
i.e., what is said [CITATION Ken05 \t \l 1043 ]. The phrase semantic content will be used for the linguistic meaning of an expression plus the necessary contextual information to make the utterance propositional, that is, what is said in uttering the sentence. This will be explained in more detail in section 1.4.1.
1.2.3 Speech Act Content. However, this is not the only manner in which the meaning of an utterance of a sentence is subject to context. Consider a person who has gotten himself a minor cut and is whining about it. Someone could utter (7) in response (example due to Bach, referred to in Recanati, 2004, p. 8):
(7) You are not going to die
Obviously, in this context, the speaker meant to say that, as the injury won’t be lethal, the injury is too minor to whine about. However, although one might not die from a minor cut, it is inevitable that
the injured person is going to die, some day. Although what is said by the utterance is false, what is implied in the utterance is that he is not going to die from that cut, which is true. Therefore, another distinction is made between the proposition that is expressed, i.e. what is said, and the proposition that is conveyed by the utterance, i.e. what is implied [CITATION Rec041 \t \l 1043 ]. The phrase speech act content will be used for the proposition that is conveyed by a speaker in a certain context, taking in account extra-linguistic features, that is, what is implied by the utterance5 [CITATION Cap \t \l 1043 ][CITATION Rec041 \t \l 1043 ]. This will be explained in more detail in section 1.4.2.
1.3 Intuitions about Context-Sensitivity
As shown above, a trichotomy of approaches to meaning is made, dividing it in the literal content, semantic content, and speech act content. In order to give an adequate theory of meaning, the mere evaluation of lexical meanings and syntactic structures has shown to be insufficient. In order to specify whether a proposition is expressed and if so, which proposition, one first has to know the lexical meaning of its expression in order to determine its referents, but in order to determine its referents, one (at least sometimes) has to place the sentence in context.
There are different intuitions about the extent and manner in which context can influence the proposition expressed. Firstly, according to some philosophers, certain sentences do not seem to express a full proposition, but are semantically incomplete. Their semantic content is to be completed by the addition of contextual features. This entails that the semantic content, that is, ‘what is said’ results from adding contextual material to the semantic ‘skeleton’ in order to make an utterance of the sentence propositional. In these cases, the literal content of a sentence is not fully propositional, but serves as input to a contextual mechanism, which adds semantic values to constituents in sentence as it is uttered[CITATION Rec041 \t \l 1043 ]. These arguments will be called Incompleteness Arguments, as they appeal to intuitions about the incompleteness of semantic content [ CITATION Bor07 \l 1043 ] [CITATION Cap \t \l 1043 ].
5 This distinction is not endorsed by everyone. Recanati, (2004) gives examples of several authors that argue for more levels that just the three distinguished here.
Secondly, some philosophers argue that certain sentences are propositional, but that the proposition they express in uttering them changes from context to context [CITATION Rec041 \t \l 1043 ]. These arguments will be called Context Shifting arguments, as they appeal to intuitions about shifting propositions in shifting contexts [ CITATION Bor07 \l 1043 ] [CITATION Cap \t \l 1043 ].
Incompleteness arguments overlap with Context Shifting arguments. Whenever a sentence is semantically incomplete, and hence needs contextual input in order to express a full proposition when uttered, different contexts result in different propositions. Hence, when a sentence is uttered in different contexts, the truth-conditions of the utterance will shift as the context shifts. This will be elaborated in section 1.4.1. However, an appeal to Context Shifting arguments does not entail that an expression is incomplete as well, but may be used in order to show that only in some contexts, an expression is incomplete. This will be explained further in section 1.4.2.
1.3.1 Incompleteness. Some sentences seem to lack the possibility of being assigned truth-conditions. This entails that their propositional content is, either in all contexts or in some contexts, not yet fixed and that contextual contribution is necessary. When one has the intuition that a sentence is semantically incomplete, the question is whether it expresses a complete proposition and whether it has truth-conditions, and hence take truth value, without placing it into context. If this is not the case, contextually provided input is necessary to make an utterance of the sentence propositional.
Consider the sentences: (8) I am a student (9) It’s raining
In (8), the first personal pronoun is again present. Without the sentence being uttered, the expression ‘I’ does not have a referent. Only when the sentence is uttered by someone, its referent will be fixed and the expression will be assigned a semantic value. This hence means that, without the sentence being uttered, it does not express a complete proposition. This means that the expression is semantically incomplete [CITATION Rec041 \t \l 1043 ]. It has to be noted that, in order to know what the semantic value of ‘I’ is, the hearer has to know who uttered the sentence. In case the hearer
merely overhears the utterance but does not see who the utterer is, the hearer may know that the semantic value of ‘I’ is fixed, but does not know which value the expression has .
In (9), things are a bit more complex. According to some philosophers6, although (9) does not contain any syntactical expression of which the semantic value has to be fixed, the specification of truth-conditions are problematic. What the sentence literally expresses is the proposition that it rains in one place or the other. However, in order to determine truth-conditions of the sentence, it has to be clear in which location it is raining: “The semantic incompleteness is manifest to us as a felt inability to evaluate the truth value of an utterance of (9) in the absence of a contextually provided location (or range of locations)” [CITATION Cap \p 34 \t \l 1043 ]. Context hence needs to provide a location in order for an utterance of the sentence to be propositional7. This entails that the sentence, without contextual input, does not seem to express a full proposition and therefore should be regarded as incomplete in some contexts. The aim of arguments about Incompleteness is to show that the sentence, which according to some8 might be propositional, is in fact (in certain contexts) semantically incomplete and hence context-sensitive [CITATION Cap \t \l 1043 ][CITATION Rec041 \t \l 1043 ].
1.3.2 Context Shifting. In order to determine whether context plays a role in evaluating an utterance, one can imagine uttering the same sentence in different contexts and evaluate whether the resulting truth-conditions vary or not. Even if the utterance is (made) propositional, i.e., does not suffer from incompleteness, the expressed proposition might differ from context to context. If this is the case, the speech act content varies from context to context. By imagining different contexts of the same utterance, one might be able to say whether ‘what is said’, i.e., the proposition expressed by, or the truth-conditions of the utterance changes systematically with changes in contexts[CITATION Cap \t \l 1043 ]. These intuitions about different propositions in different contexts are called Context Shifting arguments. Context Shifting claims are used by those who argue for the context-sensitivity of sentences. Consider (Example due to Travis, referred to in Cappelen & Lepore, 2005, p. 47): 6 E.g. Recanati (2004).
7 Some (e.g. Perry, 1986) argue that the verb ‘raining’ not only supplies a relation to a location, but also a relation between times and locations. The time value is then supplied by the present tense. However, as noted earlier, tense is fixed throughout the thesis, hence only the location value will be discussed here.
(10) Justine is a philosopher.
In a context in which friends of Justine are talking about her character, some may consider Justine a rather bland individual that goes through life without much contemplation. In this context, and utterance of (10) may be regarded as false, while Justine might, in fact, be a professor in philosophy. In another context, the group of friends is at a philosophy conference, talking about the professions of their acquaintances. An utterance of (10) in this context expresses the proposition that Justine is a professional philosopher, which may be regarded as true, while her personality might lack all characteristics that a philosopher has[CITATION Cap \t \l 1043 ].
1.4 Contextual Mechanisms
The intuitions about the context-sensitivity of expressions or sentences have led to different mechanisms are used to incorporate the contextually provided input. Some of these mechanisms are linguistically mandatory in order to make a sentence, as it is uttered, propositional, which is the case for Incompleteness Arguments. These mechanisms are instances of a process called Saturation. On the other hand, other mechanisms are not linguistically mandated, because the sentence itself does not contain any expressions that are semantically incomplete, but explain how, e.g. in Context-Shifting claims, the proposition expressed by uttering the sentence changes as the context in which the sentence is uttered changes. These mechanisms instances of a process called Modulation [CITATION Rec041 \t \l 1043 ].
1.4.1 Saturation. As mentioned, some constituents require contextual information in order to be assigned a semantic value. This process is called Saturation. By means of this process, semantic values are contextually supplemented to semantic ‘slots’ associated with certain context-sensitive constituents. Since these semantic slots are associated with the literal content of constituents in a sentence, this process is linguistically mandated in in order to arrive at a full proposition [CITATION Pag07 \t \l 1043 ]. Besides the contextual provision of semantic values to context-sensitive expressions in a sentence, context may also provide a semantic value for a phonologically unmanifested constituent, if such a constituent is required in order to make the sentence
propositional. However, as saturation is linguistically mandated, this constituent has to be justified on independent grounds, usually of a syntactic nature [CITATION Rec041 \t \l 1043 ].
Saturation is characterized as a bottom-up process, as the contextual fixing of semantic values of expressions in a sentence are triggered by elements of the sentence in question. This entails that the process of saturation applies to Incompleteness arguments, and that these Incompleteness arguments should be valid in every context and results in variations of semantic contents with variations of contexts. The characteristic feature of saturation hence is that the contextual influence on semantic content is syntactically tractable and required in order to make an utterance of a sentence propositional[CITATION Rec041 \t \l 1043 ].
Depending on how much contextual influence on semantic content one admits of, different types of semantic slots are be identified and, correspondingly, different mechanisms are formulated in which context can complete the semantic value of an utterance of a sentence to make it propositional. A distinction can be made between three different types of ‘semantic slots’: Indexicals, Semantic Parameters and Syntactic Ellipsis.
1.4.1.1 Indexicals. One way saturation can take place is when the sentence contains an
indexical expression. The linguistic meaning of these expressions contains a semantic slot that can be filled with contextual information in order to assign a semantic value to the expression. There is a list of obviously indexical expressions postulated, called the Basic Set. This list includes expressions such as personal pronouns such as ‘I’, ‘you’, demonstratives such as ‘this’, ‘that’, place indications like ‘here’, ‘there’, and time indications like ‘now’, ‘yesterday’. In order to explain the context-sensitivity of indexicals, a distinction between the content and character of an indexical expression can be made. Whereas the content of an indexical is represented by a function from possible worlds to individuals, the character of an indexical is represented by the function (or rule) which assigns to each context that content which is represented by the function from possible worlds to individuals. Recall the literal meaning of the personal pronoun ‘I’. With the distinction between character and content, the character of the indexical ‘I’ is represented as a constant function from contexts to the individual concept of ‘the speaker’ in that context; whereas its content is represented as the function
from possible worlds to the individual that is speaking [CITATION Kap79 \t \l 1043 ]. Hence, in the above terminology: the literal content of indexicals is represented as a function from contexts to semantic contents, the semantic content of an indexical is represented as a function from possible worlds to individuals.
Consider the sentences:
(11) I am a philosopher
(12) Yesterday, you went to the theatre.
Although the character, or literal content, of the indexical ‘I’ provides the rule or function from context to contents, nevertheless, without (11) being uttered the expression 'I' does not have a determinate semantic content and hence (11) as such is not propositional. In order to make the sentence propositional, it needs to be uttered in order for a semantic value to be assigned to the expression. Similarly, the semantic value of ‘you’ in (12) needs to be assigned a semantic value that is given by context as the person who the utterance is directed to, in order to make the sentence propositional. Furthermore, without reference to a time of utterance, the expression ‘yesterday’ does not have a fixed semantic value, as the day that is referred to depends on the time of utterance and hence on the context [CITATION Pag07 \t \l 1043 ]. The linguistic meaning of indexicals hence determines their semantic value as a function of particular contextual variables[CITATION Ken05 \t \l 1043 ]. Consequently, for semantic interpretation, contextual saturation of indexicals is a mandatory process to make an utterance propositional.
Besides the obvious context-sensitive expressions there is little agreement about which expressions should be included in the list of indexicals. According to some, there may be surprise indexicals associated with expressions that do not immediately seem to be context-sensitive, but on further inspection, can be claimed to be, since their semantic value is fixed in the same manner as the established indexicals. According to some philosophers9, this is the case for epistemological expressions such as the verb ‘to know’. Whereas the literal content of ‘know’ does not change, it is argued that the semantic value of the expression systematically varies as the context of utterance varies with epistemic standards. This indicates that its literal content provides a function from 9 According to Cappelen & Lepore (2005), Lewis, DeRose, and Cohen are amongst these philosphers.
contexts to semantic contents, and that an utterance of ‘know’ determines the semantic value of the expression. According to this view, the expression may not be as obviously indexical as the expressions in the Basic Set, it behaves in the same manner and hence should be considered as indexical, albeit a surprising one[CITATION Cap \t \l 1043 ].
1.4.1.2 Semantic Parameters. Besides the surprise indexicals that are treated in the same
way as indexicals such as ‘I’ and ‘yesterday’, there might be otherwise syntactically triggered semantic slots to ‘fill in’. According to some philosophers10, there are semantic parameters11 associated with expressions that do not immediately seem to be context-sensitive, but on further inspection, are claimed to be. Whereas indexicals are represented as a constant function from contexts to semantic content, a semantic parameter is characterized differently. Although both indexicals and semantic parameters require a contextually assigned semantic value, the function or rule which contextually determines the semantic content of an indexical is constant. However, this is not the case for semantic parameters. Their linguistic meaning does not provide a strict rule that directly determines their content, given a context. Consider the sentence (example due to Stanley & Szabó, 2000, p. 233):
(13) John is tall
Whereas the literal content of e.g. the indexical ‘I’ is characterized as a function from contexts to the speaker in a context (that is, to the individual concept of ‘the speaker in a context’, not to the individual that is speaking), there is no such determined function for the comparative adjective ‘tall’ in (13). Rather, to determine which proposition is expressed by an utterance of (13), the relevant comparison class with regard to which John is said to be tall has to be contextually provided. An utterance of (13) may be true if the relevant comparison class only comprises John’s colleagues, while an utterance of (13) may be false if it also includes a range of professional basketball players 10 For example, Stanley & Szabó (2000).
11 C&L call this view the ‘Hidden Indexical Strategy’. However, whereas indexicals inherently are characterized as a constant function from contexts to semantic contents, this is not the case for what C&L call ‘hidden indexicals’. Although there might be a semantic slot that needs to be filled in, the semantic values that can be assigned in these cases are not given by a constant function that is provided by the literal content of the expression. C&L acknowledge these phenomena and admit that cannot be explained by means of the characterization of indexicals, but claim that therefore the expressions in question are not context-sensitive, instead of making a distinction between strict indexicals and more complex syntactically triggered variables. This view is endorsed by e.g. Borg, (2007) and Maitra, (2007).
[CITATION Sta001 \t \l 1043 ]. Nevertheless, there is no reason to assume that there is a grammatically overt expression referring to this class present in (13). Rather, the relevant comparison class is the value of a contextual parameter. An utterance of (13) then may expresses a proposition relative to the value of the contextual parameter [CITATION Sta001 \t \l 1043 ].
Instead of claiming that a phonologically expressed constituent of the sentence actually should be regarded as an indexical (as is the case in the surprise indexical strategy), the semantic parameter approach justifies alleged context-sensitivity by revealing or postulating a hidden semantic slot or, semantic parameter, of which the semantic value is contextually assigned. Such a phonologically unarticulated, ‘hidden’ syntactic element can be disclosed by means of examining the
‘Logical Form’ of a sentence. This logical form can be described as an underlying structure of the sentence that might not be recognized in its superficial grammatical form [CITATION Sta002 \t \l 1043 ].
The syntactic representation of the basic functional and structural properties associated with lexical expressions is, by means of transformational operations, mapped onto the superficial grammatical form of a sentence [CITATION May \l 1043 ]. The operations responsible for this mapping are represented by the logical form, which hence functions as a mediation between the grammar and meaning of a sentence. Only elements that are grammatically contributed or syntactically expressed are represented at this level. The assumption in that rules of semantic interpretation are applied to representations at the level of logical form, rather than the superficial grammatical form. [ CITATION Hua94 \l 1043 ]. By examining the structure of the logical form, the syntactic structure may turn out rather different than its surface grammatical structure. Investigating a sentence’s logical form entails assigning overt complexity to natural language expressions. However, postulating such additional complexity needs to be argued for empirically [CITATION Sta002 \t \l 1043 ]. This can be done by revealing that the semantic value of an expression is determined by the semantic value of another expression, as this demonstrates that there is a free variable present in the dependent expression. An example of such empirical evidence can be found in section 3.4.2.
1.4.1.3 Syntactic Ellipsis. Another way to account for semantic incompleteness is a
mechanism called Syntactic Ellipsis. By means of this mechanism, the articulated sentence may be semantically incomplete, but the semantic value required to make an utterance of the sentence propositional is provided by context. Although this constituent is not part of what is articulated, it is nevertheless present in the utterance as a syntactic component which has no phonologically manifestation. The semantic value of this unarticulated syntactic element is then supplied by the context in which the sentence is uttered. Consider the sentences (example due to Stanley & Szabó, 2000, p. 232):
(14) Sam plays chess on Sundays. Max does too.
(14*) Sam plays chess on Sundays. Max plays chess on Sundays too.
According to the Syntactic Ellipsis approach, context operates to guarantee that (14) and (14*) express the same proposition in the given context. By examining the logical form of the sentences in question, the unarticulated phrase ‘plays chess on Sundays’ is covertly present in the form of a syntactic, hidden element that is not phonologically manifested. In this manner, context provides the syntactic constituent that is required to make an utterance of a sentence propositional [CITATION Sta001 \t \l 1043 ].
1.4.2 Modulation. As mentioned, the process of saturation is, depending on how much context-sensitivity one admits of, a linguistically mandated process in order to reach a full proposition. Modulation, on the contrary, is not linguistically mandated. It is characterized as a top-down process, as the contextual influence on semantic content is not syntactically tractable but pragmatically required in order to determine the truth-conditional content of an utterance. Incompleteness intuitions may be resolved by means of modulation. However, as the provision of a semantic value to sentences susceptible to modulation is not linguistically mandated, but rather pragmatically required in certain contexts, it has to be shown that not in every context this contextual addition is necessary, but only when the pragmatic need is felt. For, if the contextual provision of a semantic value is required in every context, it would be an instance of saturation. Hence, since
modulations are optional, they apply in some rather than every context[CITATION Rec041 \t \l 1043 ].
1.3.2.1 Free Enrichment.
The literal content of a sentence may not only be bottom-upcontextually saturated by means of saturation, but may also be top-down augmented by the addition of semantic material. This can be done by means of Free Enrichment, where the added semantic material is not linguistically mandated due to indexicals or other semantic slots that need to be assigned semantic values, but optionally added to the uttered sentence in order to reach an appropriate interpretation [CITATION Rec041 \t \l 1043 ]. Recall the sentence:
(9) It’s raining
Whereas those who argue for indexical or semantic parameter strategies claim that there is a syntactic element that requires contextual input to provide a location in order to make (9) propositional, some philosophers argue that there is no such syntactic element traceable, neither in syntax nor logical form. This entails that, if necessary, the location is pragmatically supplied, without being syntactically triggered. This location then is called an Unarticulated Constituent, as the constituent is not phonologically realised but nevertheless may be pragmatically required for an utterance of (9) to be propositional in certain contexts. If an unarticulated constituent would be linguistically mandated and hence an instance of saturation, every utterance in which context does not provide such a constituent would be semantically incomplete and hence not truth-evaluable. However, for examples such as (9), contexts can be imagined in which no contextual provision of a location value is necessary. For example, in a context in which rain has become an exceptional but indispensable phenomenon, and rain sensors have been distributed over the entire surface of the terrain (perhaps over the entire surface of the earth), the signalling of one such rain sensor, no matter which one on which location, would result the attendee to utter (9). In this context, the attendee expresses the complete proposition that is it raining in one location or another, without the need of a location value. This suggests that a location value is not linguistically mandatory to make an utterance of (9) propositional, and hence no instance of saturation, but pragmatically required in order to determine the proposition that is expressed in those contexts in which the need for the provision of such a
location value is indeed felt. This entails that the contextual delivery of a location value is an instance of optional free enrichment by means of adding an unarticulated constituent. A similar claim can be made of an utterances of (example due to Recanati, 2004, p. 8):
(15) The table is covered with books
(15*) The table of our living room is covered with books
(15) does not lack any mandatory semantic component to be propositional, but, with the use of the determiner 'the', 'what is said' entails that there is only one table in the universe and that it is covered with books. Although the sentence does express a complete proposition, it is absurd to conventional communication standards. Therefore, context provides information on which table it is that is covered with books. An utterance of (15) would then have a content like that of (15*)[CITATION Rec041 \t \l 1043 ].
1.4.2.2 Loosening and Semantic Transfer. A second type of modulation is loosening. This
involves the pragmatic extension of the application conditions of an expression in order to modulate the semantic value of that expression. An example of loosening is (example due to Recanati, 2004, p. 42):
(16) The ATM swallowed my credit card
Here, the application conditions of the verb ‘to swallow’ are not merely those conditions in which someone consumes something by mouth, but also to the process of the ATM not returning the credit card [CITATION Rec041 \t \l 1043 ].
A third type of modulation is semantic transfer, in which the semantic value of another, pragmatically provided expression is transferred to the expression used in an utterance. An example of semantic transfer is (example due to Recanati, 2004, p. 42):
(17) The ham sandwich left without paying
where the semantic value of the expression ‘the ham sandwich’ is used as referring to the person who ordered the ham sandwich. Hence, the semantic value of ‘the orderer of the ham sandwich’ is semantically transferred to ‘the ham sandwich’ [CITATION Rec041 \t \l 1043 ].
For the present purposes, only modulation in the form of free enrichment will be of concern. Although the type of figurative speech in cases of loosening and semantic transfer may play a role in quantified noun phrases ('every ham sandwich left without paying'), it is not the specific phenomenon that is investigated here, as the quantifier has to restrict the set of objects, no matter if the meaning of the objects is taken literally or figuratively.
1.4.3 Implicatures.
Whereas saturation and modulation are primary contextual processes, as they incorporate contextual features in order to reach propositionality, secondary contextual processes are characterized as taking the result of these primary processes as input. These secondary pragmatic processes are different types of implicatures. Whereas ‘what is said’ by an utterance is controlled, either linguistically or pragmatically, by the literal content of the uttered sentence, these constraints do not apply to implicatures. Consider the sentence (example due to Recanati, 2004, p. 5):(18) I am French
‘What is said’ in an utterance of (18) is that the speaker has the French nationality. However, an utterance of (18) in response to the question ‘are you a good cook?’ implies that the utterer of (18) is indeed a good cook. What is implicated here does not merely depend on the proposition in ‘what is said’, but depends on various background assumptions of the French cuisine. The hearer interprets the utterance by means of inference and the knowledge he has about these background assumptions. Given enough background, implicatures may express any proposition, as they are not controlled by the literal content of an utterance but by an indefinite inference chain [CITATION Rec041 \t \l 1043 ].
For similar reasons as those for modulations as loosening and semantic transfer, implicatures will not be of concern here, as the question is how the domain of a quantified expression of utterances of identical sentences may vary across different contexts, not which unrestricted inferences these utterances may generate.
2. Semantic Minimalism, Moderate Contextualism, and Radical Contextualism
The manner and extent to which pragmatic effects are acknowledged in the assignment of propositional, truth-evaluable content to sentences/speech acts divides philosophers of language into three main positions: Semantic Minimalism, Moderate Contextualism, and Radical Contextualism.
Semantic Minimalism identifies only a few expressions as context-sensitive and hence accepts little contextual effect of an utterance on the semantic content of that utterance. 'What is said' in an utterance of a sentence is according to Semantic Minimalists propositional and hence truth-evaluable, and although speech act content of an utterance may differ from semantic content, they maintain that in all cases, semantic content is the minimal proposition that is common to all speech acts concerning the utterance of the sentence.
Moderate Contextualism admits more context-sensitive expressions. As a result, the semantic content of a sentence may be incomplete and ‘what is said’ in an utterance of such an incomplete sentence needs to be contextually enriched in order to express a full proposition.
Radical Contextualism claims that all propositional content is essentially dependent on context. There is nothing that can be said about the truth-conditions or ‘what is said’ in a sentence, as there are no limits to the contextual input when a sentence is uttered.
2.1 Semantic Minimalism
Semantic Minimalism is characterized as recognizing only very few context-sensitive expressions, namely, only the expressions that are in the Basic Set. Because of this finite list of context-sensitive expressions, the Semantic Minimalist admits of very limited effects of the context of an utterance on the semantic content of an utterance[CITATION Cap \t \l 1043 ]. The literal meaning of a sentence is the primary source for semantic content; only when necessary will contextual mechanisms provide input for semantic content and this is will be the case only when the sentence contains one of the expressions of the Basic Set. This entails that the distance between the literal meaning of a sentence and what is said by uttering the sentence is minimal, hence the name Semantic Minimalism [CITATION Rec041 \t \l 1043 ]. According to the Semantic Minimalist, semantic content is a minimal proposition that is closely connected to literal content and in the large majority of cases, these will indeed be the same. If the sentence requires contextual input in order to be propositional, this context-sensitivity should be syntactically triggered. This entails that the semantic content of lexical constituents has to be fixed by means of contextually provided information. Apart from fixing the semantic value of the expressions that are obviously context-sensitive, the context of the utterance of a sentence has no influence on the proposition that is expressed semantically. I.e., according to the Semantic Minimalist, context-sensitivity is restricted to a limited number of context-sensitive expressions which was already referred to as Basic Set, which includes personal pronouns such as 'I', demonstratives such as 'that', place indications such as ‘here’ and time indications such as ‘now’. Only when syntactically triggered does context contribute to semantic content. This means that the Semantic Minimalist will only admit of saturation applied to indexicals, which require a contextually fixed semantic value in order to make an utterance of a sentence propositional. The proposition that results after this mandatory saturation is the minimal proposition that is expressed in 'what is said'. Of course, the Semantic Minimalist does acknowledge that the minimal proposition might differ from the intuitive proposition that a speaker would attribute to an utterance, but it is nevertheless the minimal proposition that is common to all speech acts concerning the uttered sentence. On this view, the semantic content of a sentence is that proposition
that all utterances of the sentence express (provided that the semantic value of indexicals are fixed) [CITATION Cap \t \l 1043 ].
Since the distance between literal content and semantic content is minimal, the semantic content, that is, 'what is said', is predominantly determined by linguistic rules of sentences (i.e. lexical meaning, syntax, logical form), rather than by pragmatic inferences of utterances. According to the Semantic Minimalist, this is the only way in which it can be explained that the same content can be expressed, asserted, etc. in extremely different contexts. The semantic content of a minimal proposition assures that hearers in different contexts can interpret each other’s utterances and hence, communicate. This is possible precisely because the semantic content of minimal propositions is mainly insensitive to contextual features [CITATION Cap \t \l 1043 ].
However, speech act content, that is, 'what is implied', may vary from semantic content considerably since it can be subject to a variety of contextual factors. This view is called Speech Act Pluralism, and states that, instead of an utterance expressing a single proposition, indefinitely many propositions are expressed. ‘What is implied’ depends on various extra-linguistic features other than the features that semantic content depends on, i.e. contextual features that determine the semantic value of indexicals. Speech Act Pluralism entails that an utterance can convey a proposition that may not even be a logical implication of the proposition that is expressed semantically. And speech act content may even be incompatible with semantic content, as semantic content is strongly related to literal content while speech act content may be affected by a range of contextual factors. [CITATION Cap \t \l 1043 ]. However, the minimal proposition of sentence is the proposition that is common to all utterances concerning that sentence.
A consequence of Speech Act Pluralism is the claim that there cannot be a systematic account of speech act content. This claim entails a motivation to separate theories of semantic content from theories of speech act content. This is the only manner in which semantic theory can be kept systematic itself [CITATION Pag07 \t \l 1043 ]. The Semantic Minimalist has to deny that semantic content is speech act content, as it is clear from the intuitions about context-sensitivity that, if there are minimal propositions, these are (in many cases) not the propositions that are asserted in linguistic
communication. However, minimal propositions will always be available as a fall-back interpretation when communication fails. Whatever proposition speech act content might have been, the hearer will always be able to correctly report the minimal proposition as the proposition that is expressed in the utterance, as this is the proposition common to all speech acts concerning the uttered sentence. Therefore, semantic content will always be part of a possibly larger speech act content [ CITATION Bor07 \l 1043 ].
2.1.1 Arguments against Semantic Minimalism. Some philosophers12 have argued that the Semantic Minimalist's sharp distinction between semantic content and speech act content is based on the assumption that the interpretation of sentences, as opposed to utterances, can result in fully propositional content. [CITATION Rec041 \t \l 1043 ]. However, the resulting minimal propositions are subject to claims about semantic underdeterminacy and availability. Arguments about underdeterminacy may be due to a lack of assigning semantic values by means of saturation (top-down), or modulation (bottom-up), but in either way, they amount to the claim that no truth-evaluable content can be ascribed to the alleged minimal proposition. Arguments about availability amount to the claim that the minimal proposition as construed by the Semantic Minimalist is not cognitively available to speakers and hearers and hence does not explain which proposition is expressed in an utterance.
2.1.1.3 Underdeterminacy. The interpretation of utterances, even with fixed semantic values
for indexicals, may not result in complete propositions, rather, it only delivers semantic fragments that need contextual enrichment. The contextual fixing of the semantic values of indexicals only appeals to a limited part of context, but in addition there is a range of other contextual factors that co-determine the proposition that is expressed. i.e., there are more context-sensitive expressions that are semantically underdeterminate. For example, an utterance containing the expression ‘John’s car’ depends on more wide-ranging contextual features, as it has to be determined if, say, the car is actually owned by John, the car is borrowed by John or that it is the car that John last drove in. The function that relates the expressions ‘John’ and ‘car’ needs to be contextually determined. If this relation is not specified, the expression ‘John’s car’ does not have a semantic value and, according to 12 E.g. Recanati (2004) and Travis (2006).
compositionality, the sentence containing the expression by itself is not propositional. This is the case for all expressions that are semantically underdeterminate: whereas the semantic value of an indexical merely depends on a function of one out of a limited range of simple contextual factors, other expressions depend on a more complex function. This entails that interpretation as construed by the Semantic Minimalists cannot determine ‘what is said’ in the utterance of an semantically underdeterminate sentence[CITATION Rec96 \t \l 1043 ].
2.1.1.1 Availability. Another problem with minimal propositions as construed by the
Semantic Minimalist is that they cannot account for the fact that the propositions expressed by the sentence are intuitively very different from the propositions that speakers themselves would assign to their utterances, or that the minimal propositions may not even be cognitively available to speakers or hearers. Consider the sentence (example due to Recanati, 2004):
(19) John has three children
The minimal proposition of an utterance of (19) would, from a Semantic Minimalist point of view, express that John has at least three children. Although there might be a context imaginable in which the speaker intends to communicate that John at least three children, for example in a conversation about a price reduction for families with three or more children, in most other contexts, this is not the proposition that the speaker wants to express. If someone is asked how many children John has, and the speaker replies by uttering (19), the proposition that is meant to be communicated is that John has exactly three children. Besides the fact that the minimal proposition is not the intended proposition, the speaker is only conscious of his intended proposition. This entails that speakers may not even be aware of the minimal proposition that they arguably express by Semantic Minimalist’s standards. On the contrary, the speaker may even deny the truth-conditions of the minimal proposition because the minimal proposition is not consciously available. The conception of minimal propositions hence lacks generality as it does not account for intuitive propositions and truth-conditions[CITATION Rec041 \t \l 1043 ].
2.2 Moderate Contextualism
The Moderate Contextualist argues that context plays a more significant role in determining what proposition is expressed. According to the Moderate Contextualist, there are more expressions than those in the Basic Set that admit of context-sensitivity. This entails that many of the sentences that the Semantic Minimalist assigns semantic values to, are in fact semantically incomplete and not truth-evaluable as such. The Moderate Contextualist hence appeals to Incompleteness arguments. The context-sensitive expressions in these cases are regarded as ‘semantic skeletons’ that need ‘fleshing out’. This entails that they are not fully propositional and that contextual input is required in order to assign the context-sensitive expressions a semantic value. In these cases, only utterances of sentences in context containing context-sensitive expressions are truth-evaluable and propositional. [CITATION Cap \t \l 1043 ]. Whereas the Semantic Minimalist only admits of the Basic Set assumption, the Moderate Contextualist does not endorse this restriction. The Moderate Contextualist argues not only that there are more context-sensitive expressions than those in the Basic Set, but also that there are more contextual mechanisms than just the saturation of indexicals. However, these mechanisms have to be syntactically triggered.13 For example, the Moderate Contextualist may argue for the existence of surprise indexicals, as explained in section 1.4.1.1. On this view, the Moderate Contextualist appeals to the same saturation mechanism by which indexicals are defined, as the Moderate Contextualist argues that the additional context-sensitive expressions need their semantic value fixed as well. However, as not all context-sensitive expressions the Moderate Contextualist argues for behave like indexicals, they may resort to other, more complex saturation processes such as syntactic ellipsis and semantic parameters, as these mechanisms too are linguistically mandated in order to arrive are a full proposition by contextually determining semantic values. This entails that, if Incompleteness arguments are appealed to, it has to be shown that in every context, the sentence is 13Cappelen & Lepore (2005) argue that the Moderate Contextualis expands the Basic Set of context-sensitive expressions, whereas the Radical Contextualist holds that all expressions are context-sensitive. However, one of the main features of the expressions in the Basic Set is that their context-sensitivity is syntactically triggered. But this is not what Radical Contextualists aim for. Rather, they argue that some contextual effects are not triggered syntactically, but imposed by top-down effects. The extent to which the Basic Set is expanded hence does not provide an adequate differentiation between Semantic Minimalism, Moderate Contextualism and Radical Contextualism. As the Semantic Minimalist only admits of the Basic Set, and the Radical Contextualist argues for top-down pragmatic effects, the Moderate Contextualist should be defined as a middle ground, not by merely expanding the Basic Set but by means of more complex syntactic triggers which assign semantic values to context-sensitive expressions.
incomplete. Furthermore, this involves that the Moderate Contextualist claims that all context-sensitivity should be triggered syntactically [ CITATION Bor07 \l 1043 ].
2.2.1 Arguments against Moderate Contextualism. According to some philosophers, Moderate Contextualism collapses into Radical Contextualism. The intuitions about Incompleteness and Context Shifting that supposedly demonstrate the context-sensitivity of certain contested expressions supports, on this view, Radical Contextualism [CITATION Cap \t \l 1043 ].
2.2.1.1 A Slippery Slope to Radical Contextualism. As mentioned in section 1.3.1, one of the
intuitions in Incompleteness arguments is that the sentence in question fails to answer questions that need to be answered in order to be able to specify truth-conditions. In order to answer these questions, contextual features should be added to the content of the utterance in order to make the utterance propositional. This entails that 'what is said' is not propositional and that a syntactically triggered element (for 'It's raining' a location value) should be added in order to arrive at the proposition that is communicated. However, these elements only answer one (or a few) of the questions that, according to Moderate Contextualist, should be answered. There is a range of questions possible that needs to be answered in order to determine the truth-conditions of the utterance, including the syntactically triggered element. A question that needs answering may be: ‘How many raindrops should fall in a certain time range and on a certain surface for it to rain?’. This entails that the supposedly propositional utterance, including the syntactically triggered constituent, remains underdeterminate, as it doesn't answer every possible question which should be answered in order to determine its truth-conditions. Incompleteness arguments hence cannot demonstrate that semantic content is incomplete but can be completed, but can only illustrate that semantic content is not even near propositionality and never will be. On this view, Incompleteness claims can hence only contribute to Radical Contextualism, as there are many more contextually required features to be implemented in the uttered sentence in order to determine its truth-conditions [CITATION Cap \t \l 1043 ].
As mentioned in section 1.3.2, Context Shifting claims are about imagining a range of contexts in which a sentence, which is supposed to be propositional, changes in truth-value as the
