The plan of the paper is as follows. In section 1, I present Geach's text and analyse its scope and connections with [OI]. In Section 2, I defend the neutrality of [OI] with respect to traditional metaphysical and semantic controversies. In section 3, I argue that the two instances of the argument pattern mentioned in the Geach's text, one involving the predicate ‘is true’ and another one involving the predicate ‘is bad’, present notable differences concerning their logico-semantic properties. I contend that, within the class of normative notions, functions of propositions, e.g. ‘is true’/ ‘it is true that’, ‘ x knows that’, and ‘x believes that’, should be distinguished from functions of concepts (e.g. ‘is bad’, ‘is good’), and also from functions of objects, e.g. ‘is beautiful’ and ‘is tasty’.Footnote 1 The ‘is-true’ argument involves a function of propositions, while the ‘is-bad’ argument involves a function of concepts. Section 3 is devoted to showing that FGA differently affects arguments involving functions of these two categories. The ‘is-true’ argument is analyzed in sub-section 3.1. I argue that its validity provides no reason to accept or reject a realist approach to truth. A non-representationalist, ‘deflationary’ explanation of truth can accommodate the validity of the ‘is-true’ argument as smoothly as a substantive, representationalist approach that treats truth as a ‘genuine’ property. Subsection 3.2 focuses on the ‘is-bad’ argument. The validity of the ‘is-bad’ argument, by contrast, confronts descriptivism/representationalism with its least defensible consequences, i.e. the interpretation of normative notions as ordinary properties. The semantic assimilation of ‘is bad’ to descriptive predicates such as ‘is triangular’ prompts well-known metaphysical difficulties and goes against any naturalist approach to abstract entities compatible with the scientific outlook.Footnote 2 The good news is that nothing in Geach's argument, nor in its reception by his followers and critics relies on a representationalist interpretation of ‘x is bad’ or on a realist approach to values. The realist, representationalist assumption is an independent addition that is not prescribed by FGA, even if it might well be what Geach himself had in mind. On the contrary, the general acknowledgement of the validity of the ‘is-bad’ argument that supports FGA rests only on the organic insight, [OI], which is a neutral claim with respect to semantic and metaphysical debates. In section 4 I will address some reasonable criticisms that have been made about the non-descriptivist, non-compositionalist approach to propositions.
1. The Organic Intuition
To be a proposition is simply to bear propositional properties. I call this claim the ‘organic intuition’ [OI]Footnote 3 about propositional recognition:
[OI] To be a proposition is to possess propositional properties (where this includes standing in inferential relations).
Making this intuition more acceptable is the main aim of this paper. In my defense of [OI], I will focus on the so-called ‘Frege-Geach Argument’ (hereafter, FGA), which develops some claims made by Peter GeachFootnote 4 and has become the most far-reaching argument ever directed against semantic non-descriptivism.Footnote 5 My contention is that Geach assumed, immediately and non-inferentially, something like [OI], which has also been accepted without resistance by all parties in the debate between descriptivism and non-descriptivism. Thus, Geach's argument will be used to build a case for a claim, [OI], that many would see as a direct target of Geach's criticism.
According to Geach,
there is a theory that to say ‘what the policeman said is true’ is not to describe or characterize what the policeman said but to corroborate it; and a theory that to say ‘it is bad to get drunk’ is not to describe or characterize drunkenness but to condemn it. […] There is a radical flaw in this whole pattern of philosophizing […] for that would mean that arguments of the pattern ‘if x is true (if w is bad), then p; but x is true (w is bad); ergo p’ contained a fallacy of equivocation, whereas they are in fact clearly valid.Footnote 6
This text has two distinguishable parts. In the first paragraph, Geach suggests his own semantico-pragmatic position, in which he assumes that only acts of describing or characterizing, as opposed to acts of corroborating or condemning, produce items apt to bear inferential properties or, alternatively, apt to fill the gaps of truth functions.
The second paragraph states the validity of the following argument patterns:
1. If x is true, then p; x is true; therefore p.
2. If w is bad, then p; w is bad; therefore p.
Most non-descriptivist views have felt the need to reject, discuss, amend, etc. the semantic part of Geach's argument, i.e. his descriptivism. Some amendments along the lines of non-descriptivism consist in offering innovative strategies to combine conative attitudes with descriptive contents,Footnote 7 or in attempting to explain how states of mind turn into propositional contents.Footnote 8 Expressivists try to control the damage by recognizing descriptive/representational layers in normative ascriptions.Footnote 9 Anti-expressivists welcome FGA as exposing expressivism’ s fatal misconceptions.Footnote 10 But very few people, if any, have questioned the second part, i.e. that the items that occur as the premisses and conclusions of arguments are propositions. All consumers of FGA, whether expressivist or anti-expressivist, descriptivist or anti-descriptivist, acknowledge the validity of the pattern.Footnote 11 By giving credence to FGA, all parties implicitly accept that Geach's examples are valid and that normative claims can safely fill propositional slots. I will rely on this broad agreement for my argument.
The FGA is a development of Geach's point, in which a particular semantic view (ascriptivism, in this case) and the generalized intuition about propositionality are mixed together. The argument has different versions, some of them independently known as the ‘embedding problem’, and also as the ‘negation problem’.Footnote 12 These versions can be formulated as follows: The a rguments of truth functions are by definition truth-bearers. If normative claims do not produce any dysfunction when they occur as the antecedents or consequents of conditionals, or when they fall under the scope of negation, then they are truth-apt and possess propositional content.
But the [FGA] includes something else. In Geach's argument patterns, normative claims also occur in free-standing clauses. In the second premise of the patterns the normative sentences are not embedded, but nevertheless trigger the m odus p onens. To do that, and avoid the risk of fallacy, they need to possess the same status, and so the same content, as the antecedent of the first premise. The fallacy of equivocation that Geach mentions in his text would occur whenever (i) the free-standing instance of the normative claim, i.e. Premise 2 in Geach's patterns, is said not to express a proposition, while (ii) its corresponding embedded instance in the antecedent of Premise 1 is fully propositional. SearleFootnote 13 uses a similar argument in support of his claim that non-descriptivism is fallacious.
The gist of Geach's argument is that these patterns are nevertheless valid. What I want to stress is that their validity is immediate, and does not require any further explanation. It agrees with our basic semantic intuitions as competent speakers and qualified reasoners. [OI] is a generalization of this intuition to all other propositional properties and relations. Notice that [OI] is an identifying, recognition criterion for propositions, but neither a particular method of individuating them,Footnote 14 i.e. a method that establishes when we face the same proposition twice, nor a principle or axiom of any particular semantic view on the meaning of natural language sentences.
There is no circularity in [OI] inasmuch as the essential propositional properties and relations are independently listed. According to the logico-semantic paradigm we find ourselves in, and which this paper does not challenge, propositions are truth-bearers, the arguments of truth-functions (‘not’, ‘or’, ‘and’, ‘if’), the arguments of propositional-attitude verbs (‘know’, ‘believe’, ‘doubt’, ‘assume’, ‘reject’) and the kind of entity capable of entering into inferential relations (basically, implication, and incompatibility). The core of [OI] is that there is nothing more to possessing propositional status than being able to fill propositional argument-slots. [OI] sharply distinguishes two sets of entities, the set of propositions and its complement set, and that is all that is required from a precise definition.
Adapting Gibbard's felicitous explanation of the status of his slogan ‘means implies ought’,Footnote 15 I will say that [OI] belongs to the meta-theory of a theory of propositions, whereas the different descriptivist and non-descriptivist alternatives developed to answer Geach's argument constitute substantive theoretical proposals about the particular nature of propositions, their properties, and the semantics of the linguistic items by means of which speakers express them.Footnote 16
Belonging to the meta-theory, [OI] is prior to and strictly neutral about how propositions are individuated and the particular semantics of declarative sentences. What FGA conclusively refutes is what might be called the ‘interjection’ view of normative claims – that is, the view that assimilates normative claims to pure expressions of feeling without propositional content, e.g. some reductionist readings of classical emotivism.Footnote 17 The interjection view irremediablyFootnote 18 falls victim to the ‘embedding problem’, since it is unable to explain how non-propositional products of linguistic actions possess properties and stand in relations reserved for truth-bearers. Nevertheless, most contemporary non-descriptivist approaches are as far from the interjection view as can be, and are not affected by FGA in the way that has sometimes too quickly been assumed.
2. The neutrality of the organic intuition in the debate descritivism/anti-descriptivism
The validity of Geach's examples is immediately accepted by all parties, although the theoretical reasons provided to explain this vary widely. Certainly Geach himself, and each of his followers and critics endorses one or another of the semantic and metaphysical theories available on the philosophical market. But neither FGA, nor its versions, the embedding problem and the negation problem, make use of any of them. [OI], and not Geach's descriptivism, is the basic assumption of FGA, the assumption that triggers criticism of the interjection view, that lends FGA its strength, and that explains its extraordinary success. In the debate between descriptivism and non-descriptivism, it should be kept in mind that something like [OI] has been assumed by admirers and detractors of FGA, without any extra independent argumentation, and thus that [OI] should not be counted among the merits or demerits of any one of the contending views.
Some of FGA's success and the prevalence of descriptivism is explicable in terms of some kind of justificatory metonymy, in which the plausibility of [OI] has been improperly inherited by a particular semantic theory, in this case descriptivism.
[OI] offers a criterion for identifying or recognizing propositions that proceeds ‘from above’, relying on their functional properties, and disengaged from their particular ‘inner nature’. This, I believe, is the main lesson to draw from the entire debate: that competent speakers naturally accept the propositional character of normative claims. Thus, the simplest criterion for propositional identification that FGA supports, and the criterion that best exposes the flaws of the interjection view is [OI], i.e. that to be a proposition is to be able to bear propositional properties and to stand in inferential relations.
[OI] entitles us to maintain that normative claims express propositions because they play the roles of premises and conclusions in inferences, they are believed and doubted, are subject to rational debate, and can be disagreed with.Footnote 19 From [OI] it follows that the criterion of propositionality, on the one hand, and the semantics for declarative sentences, on the other, are distinguishable aspects of the general task of explaining normative discourse.
Once propositions are pinpointed via [OI], the door is open to tackle other semantic and metaphysical issues. At some stage, propositions can be characterized as essentially structured entities made from the concepts, and sometimes also objects, which are their building blocks.Footnote 20 Alternatively, one might favour a non-structured view of propositions as sets of possible worlds,Footnote 21 or else as nodes in inferential networks.Footnote 22 Propositions and meanings, on the other hand, stand at different levels. This is the classical difference between intensions and meaningsFootnote 23 or between propositional contents and the semantic values of sentences.Footnote 24 Lewis considers the principle of compositionality to be essential to any grammar of English,Footnote 25 whereas Hintikka and SanduFootnote 26 argue that the principle cannot explain the functioning of quantifiers. Besides, there are externalistFootnote 27 and also internalistFootnote 28 theories of meaning. At a different level, true propositions can be explained as somehow mirroring states of affairs.Footnote 29 Alternatively, we might characterize true propositions as those contents that we endorse and promote as premises for future inferential acts.Footnote 30 Still a different debate is whether mental states ground and/or individuate contents or else whether contents are instead individuated by objective/intersubjective relations. Classical expressivists took the former path, modern inferentialists take the latter.
And so on and so forth. But the enquiries into the inner nature of propositions, into the semantics of linguistic items, or into the metaphysical statuses of the entities that we know, believe, or doubt stand beyond a basic acknowledgement that whatever behaves like a proposition is a proposition. Because it belongs to the meta-theory, [OI] remains neutral about particular semantic and metaphysical implementations. This neutrality is a welcome consequence of [OI] which keeps normative claims within the realm of the inferentially structured discourse, of rational evaluation and rational criticism, but out of the partisan debates over the nature of propositions or the semantics of natural language sentences. A related desirable effect is that it explains in the simplest way how descriptive and normative claims co-exist in inferences.
3. The logico-semantic structure of Geach's valid patterns
In sections 1 and 2 I have argued for the meta-theoretical nature of [OI], which places this intuition beyond partisan debates between the theories that would provide its semantic and metaphysical implementation. In this section I will show that, although [OI] is compatible with both descriptivist and anti-descriptivist positions, different theories applied to different arguments with different logico-semantic structures incur different commitments. Not all theories produce the same results or have to answer to the same kinds of objection, and the outcome also depends on the particular structure of the arguments at issue. The two arguments that Geach mentions are instances of modus ponens, but the propositions that constitute them admit distinct analyses from a logico-semantic perspective. In particular, the two patterns that Geach mentions show some peculiarities that are worth considering.
There are two argument-patterns in Geach's text, the ‘is-true’ pattern and the ‘is-bad’ pattern:
The way in which Geach includes these two arguments in his text makes it difficult to notice that they possess different logico-semantic structures and properties. To begin with, ‘is true’ and ‘is bad’ apply to entities of different logico-semantic categories, even if from a grammatical viewpoint they are both predicables. Truth is a higher-level notion that primarily applies to propositions (0-adic concepts), whereas the arguments of ‘is bad’ are either individuals or n-adic concepts (n > 0). The semantic options for explaining the validity of both arguments are explained in the following sub-sections. The ‘is-true' argument will be the topic of subsection 3.1; the ‘is-bad’ argument will be analysed in subsection 3.2.
3.1 The ‘is true’ argument
Let x = what she said and p = Theresa May resigned as PM. The following argument, [‘is-true' argument 1], is an instance of Geach's ‘is-true’ argument:
[‘is-true’ argument 1]
Premise 1: If what she said is true, then Theresa May resigned as PM
Premise 2: What she said is true
Conclusion: Theresa May resigned as PM.
To make this argument completely explicit, let us assume that what she said is that Theresa May resigned as PM, i.e. that p in Geach's pattern discloses the content of x.
The validity of the ‘is-true’ argument is immediately seen by anyone who understands it. This is something that Geach and his followers recognize and his critics do not reject. What [OI] implies then is that the antecedent of the conditional in Premise 1 as well as Premise 2 possesses propositional status. Otherwise, either Premise 1 would be ill-formed, since conditionals require propositions as their antecedents, or else the argument could not go through since only propositions (such as Premise 2) count as steps in inferential chains.
Nevertheless, the validity of the ‘is-true’ argument can satisfactorily be explained by a minimalist, non-descriptivist account of the truth predicate, in which the predicate is treated as semantically irrelevant, as much as by a descriptivist, realist approach to truth, in which ‘is-true’ is interpreted as a contentful predicate. Both explanations are compatible with the basic intuition represented by [OI]. Let us see how both competing approaches would proceed.
Truth (‘is true’) is a function of propositions. By ‘functions of propositions’ I mean predicative notions that have propositions as arguments. Truth is one of them, further examples being the class of propositional-attitude verbs (‘know’, ‘consider’, ‘doubt’, ‘believe’, etc.) and the class of logical constants (‘if’, ‘no’, ‘and’, ‘or’). Commenting on the category of arguments of the truth concept is an alternative way of introducing the classical topic of truth bearers. Whether propositions are the primary truth bearers is an issue that has stirred substantial controversy during the rise of analytic philosophy in the first half of the past century. Nevertheless, what is uncontroversial is that truth is said of items that can be linguistically represented by declarative sentences, and that speakers attribute truth to what other people or they themselves say in particular assertive acts. Philosophers call these items ‘propositions’.Footnote 31
There is no standard linguistic form that functions of propositions systematically adopt, and in fact, functions of propositions can occur in sentences as ordinary predicates, operators, adverbs, or loose sentential clauses, as Urmson showed concerning parenthetical verbs.Footnote 32 Truth provides a good illustration of the linguistic adaptability of functions of propositions. The expressions ‘is true’, ‘it is true that’, ‘truly’, and ‘truth’, are all linguistic representations of the concept truth. Grammar is thus misleading. As Geach remarked, albeit with a different purpose, ‘here as elsewhere people have not learned from [Frege's] work as much as they should’.Footnote 33
If, as in our example, what she said is that Theresa May resigned as PM, then (1) and (2) convey the same content,
(1) What she said is true,
(2) It is true that Theresa May resigned as PM.
When functions of propositions adopt the superficial form of sentential operators, as in (2), the propositional status of the argument is explicitly displayed. By contrast, when functions of propositions occur in the form of ordinary predicables, as in (1), the propositional nature of the grammatical subject is hidden under the grammatical surface. As Frege taught us,Footnote 34 the grammatical categories of subject and predicate lack logical significance, something that in the analysis of normative concepts is particularly important to keep in mind.
Definite descriptions, e.g. expressions of the form ‘What she said’, are from a grammatical viewpoint singular terms. GeachFootnote 35 and RyleFootnote 36 stressed one of their characteristic semantic features by calling them ‘namely-riders’. According to these authors, they are expressions that call for an instantiation in the form of a namely-clause. Some examples:
(3) What she said, namely, that Theresa May resigned as PM,
(4) What he wore, namely, just a few drops of Channel No. 5,
(5) What she bought, namely, a Kindle Oasis.
As these examples show, the nature of the entity that satisfies the description may vary. In (3), the namely-clause introduces a reference to a proposition whereas in (4) and (5) the clauses introduce a description of non-countable stuff and the name of an object, respectively. In the case of truth, the description of the item to which truth is ascribed has to ‘ride on’ the propositional content. Otherwise, the sentence-like entity that results will be ill-formed, as in (6) and (7),
(6) What he wore is true*,
(7) What she bought is true*.
This feature of some descriptions as namely-riders with propositional content explains the grain of truth in the central claim of ascriptivism as Geach characterizes it, i.e. ‘that to say “what the policeman said is true” is not to describe or characterize what the policeman said but to corroborate it’. The explanation goes as follows: according to descriptivism, a sentence such as (8),
(8) The policeman stopped the car,
depicts a situation or a set of situations in which a policeman and a car are involved, and where an action by the policeman has the effect of stopping the car. This is quite uncontroversial. But it is equally uncontroversial that not all declarative English sentences depict or describe in the way that (8) arguably does. A sentence such as (9),
(9) What the policeman said is true,
cannot be said to describe (see section 4 below) any specific type of situation. If the policeman had uttered any of the following sentences (10), (11), (12),
(10) The suspect ran away,
(11) It's a lovely day,
(12) The European Commission encourages politics of austerity,
the corresponding use of (9) would have conveyed a different descriptive content. None of the situations allegedly depicted by (10), (11), or (12) would be in a better position to be the descriptive content of (9). In this specific sense, it is natural to say that truth ascriptions do not describe. Nevertheless, that these ascriptions do not describe or even that their function is to ‘corroborate’ what somebody has said does not imply that they lack propositional content. By uttering sentence (10), the policeman says that the suspect ran way. In an appropriate context, a speaker who utters (9) corroborates what the policeman had said and thereby says herself that the suspect ran away. What is crucial for my point is that the further act of corroboration does not annihilate the available propositional content. If this is so, the antecedent of Premise 1 and the free-standing truth ascription, Premise 2, share their content and thus modus ponens can be applied. A fallacy would appear if corroboration destroyed the propositional content, an assumption that has not been defended by anybody so far, and that would go against the logical grammar of ‘corroborate’.
With these analytic tools in hand, let us go back to Geach's ‘is-true’ argument.
[‘is true’ argument]
[Premise 0: She said that Theresa May resigned as PM]
Premise 1: If what she said is true, then Theresa May resigned as PM
Premise 2: What she said is true
Conclusion: Theresa May resigned as PM.
In the antecedent of Premise 1, and in Premise 2, full propositions are required to move the inferential mill. Grammatically, sentences are the linguistic items that can occur after ‘if’ and that can stand alone as steps in inferences. The description ‘what she said’ cannot do either of these jobs, since as a singular term it does not belong to the appropriate grammatical category. Now, there are two alternative ways of explaining the role of the truth predicate that accompanies this description, a deflationist explanation and a ‘substantive’ explanation. According to the deflationist explanation, the grammatical predicate ‘is true’ does not add anything to the content of the description, which is already fully propositional. It is semantically irrelevant. The role of ‘is true’ is merely instrumental, helping to represent the propositional content of the description in a sentential way. The dummy predicate ‘is true’ would be a de-nominalizor.Footnote 37 The crucial point here is that while the truth predicate adds no new information to the proposition that satisfies the description, it does not take anything away either, and hence the truth ascription expresses a proposition which is identical to the propositional content of the definite description.
Let us go now to the substantive, descriptivist explanation. The substantive explanation of the meaning of truth does not need to be defended on this account, because it does not challenge the claim that truth ascriptions possess propositional content. Truth ascriptions behave as indicative sentences do, in which some property is predicated of some entity, or some entities are said to be related in some way. As the substantive explanation goes, in keeping with standard versions of descriptivism in which grammatical predicates represent properties that qualify objects, the propositional content of truth ascriptions is not jeopardized, and the validity of the ‘is-true’ argument needs no specific analysis. The semantic and metaphysical difficulties that a substantive account of the truth predicate faces are present though, and relate to the kind of objects that propositions are, and which substantive information the truth-predicate adds to the proposition expressed by truth-ascriptions.
Both explanations, the deflationist one and the substantive one, are compatible with the validity of the ‘is-true’ argument, which is not in question. Thus, FGA does not provide an argument against a non-descriptivist approach to truth or in favour of a representationalist semantic view.
3.2 The ‘is-bad’ argument
Let us now turn to Geach's ‘is-bad’ argument. Here is an instance:
[‘is-bad' argument 1]
Premise 1: If what she did is bad, then she should be punished
Premise 2: What she did is bad
Conclusion: She should be punished.
As in the ‘is-true’ argument, the predicate ‘is bad’ has a definite description as its grammatical subject, which is also a namely-rider − what she did, namely torturing animals. This [‘is-bad' argument 1] is a general argument that can be implemented in various ways. A possible instance is [‘is-bad’ argument 2]:
[‘is-bad’ argument 2]
Premise 0: Torturing animals is what she did
Premise 1: If what she did, namely torturing animals, is bad, then she should be punished
Premise 2: What she did is bad
Conclusion: She should be punished.
As it happens in the case of truth, ‘What she did is bad’ expresses a proposition because the ‘is-bad’ argument is in fact valid. Nevertheless, unlike ‘is true’, ‘is bad’ is a function of concepts, properties, or actions,Footnote 38 i.e. a function of n-adic concepts (n > 0) instead of a function of propositions (0-adic concepts). In some cases, ‘is bad’ might occur as a property of objects as well.Footnote 39 The dissimilarity in logical form between the ‘is-true’ argument and the ‘is-bad’ argument precludes the extrapolation of the deflationist strategy for ‘is true’ to the ‘is-bad’ argument. An explanation of the propositional status of ‘what she did is bad’ cannot rely on any alleged semantic irrelevance of ‘is bad’. The reason for this is that the appropriate arguments for ‘is bad’, on their own, are not the kind of items that can play the role of premise and conclusion in inferences. Thus, if the argument is valid, the alternative to the substantive/representationalist explanation of ‘is bad’ cannot be the semantic irrelevance of the predicable, as in the deflationist explanation of ‘is true’. The two arguments are different in logical form and require different logico-semantic explanations.
Certainly, one might reject the dissimilarity of the two arguments and assume that they are analogous in structure.Footnote 40 Then the explanation might proceed by arguing that the predicate ‘is bad’ can also occur as an operator, ‘it is bad that’. In that case, its argument should be a sentence. Here, (13) and (14) illustrate this move,
(13) Torturing animals is bad,
(14) It is bad that you torture animals.
Nevertheless we should bear in mind that sentential forms do not guarantee propositional contents. Even if propositions are often represented linguistically by sentences, belonging to the category of sentences is neither necessary nor sufficient for an expression to have propositional content. Singular terms, such as descriptions and pronouns, can stand for propositions and some sentences are unable to bear full propositionality. The grammatical subject of (1) above, i. e. ‘what she said’, is an example of the first case, and the sentential argument of (14), on its more natural interpretation, is an example of the second case. A further example of the second case is the argument of the that-clause in Castañeda's famous example,Footnote 41 (15),
(15) The editor of Soul knows that he is a billionaire.
Whereas the following example, (16), offers a clarifying contrast,
(16) The editor of Soul knows that Queen Elizabeth is a billionaire
In (16) the that-clause is detachable. (16) entails (17),
(17) Queen Elizabeth is a millionaire.
In (15) the argument of the that-clause, although sentential in form, is non-detachable.Footnote 42
The move from (13) to (14) is a formal transformation without deeper significance. If the subject in (13) is non-propositional, the transformation of (13) into (14) is not enough to confer propositionality to the operator's argument. The operator ‘it is bad that’ opens up, as a merely syntactic issue, a sentential slot that the predicate ‘torture animals’ cannot fill. The syntactic solution is to convert the predicable ‘torture animals’ into a sentence ‘in form’,Footnote 43 i.e. ‘x torture animals’ / ‘you torture animals’, whose syntactic status is now appropriate.
Torturing animals, helping refugees, or cheating on partners—that is, objects of moral evaluation—can be said to be disgusting, admirable, shameful, fun, or boring, but none of these can be said to be true or false.
The argument of the that-clause in (14), i. e. (18),
(18) You torture animals,
could express a proposition if the pronoun were used referentially and the context provided a content. It might be the case that, by uttering (18), somebody meant something like (19),
(19) You, Peter, torture animals.
In that case, (14’),
(14’) It is bad that you, Peter, torture animals,
would be a suitable candidate for the deflationary explanation given above for the case of truth. The predicable ‘is bad’ would not add new content to (19) and would be a mere expressive device. Ayer accepted this possibility:
The presence of an ethical symbol in a proposition adds nothing to its factual content. Thus if I say to someone, ‘You acted wrongly in stealing that money’, I am not stating anything more than if I had simply said, ‘You stole that money’. In adding that this action is wrong I am not making any further statement about it. I am simply evincing my moral disapproval of it.Footnote 44
The deflationary explanation clarifies the propositional nature of functions of propositions plus their arguments by showing that these functions do not add new concepts to their propositional arguments. This explanation works only for functions with propositional arguments.Footnote 45 As the most natural interpretation of (13) is, as Ayer saw, a generalization in which ‘is bad’ applies to courses of action, behaviour, or actions, and not to singular propositions such as (19), (13) does not meet the conditions required for the deflationary explanation.
The standard, substantive explanation implies a realist view of values, whose consequences are known by every ethical theorist and philosopher of language. Nevertheless, even if the deflationary explanation is off the table, there is still an alternative to the substantive view. InferentialismFootnote 46 is the non-descriptivist view that better explains the validity of the ‘is-bad’ argument and represents a natural development of [OI]. According to inferentialism, propositions are those entities capable of performing the roles of premises and conclusions in arguments. Propositions are individuated by their inferential relations with other propositions and are essentially the bricks with which inferences are built.Footnote 47 If a semantic item can be asserted (as happens with Premise 2), and its assertion has the significance of a move in an inferential game, it possesses a propositional nature.Footnote 48 Biting Geach's bullet and acknowledging, directly and immediately, the propositional status of instances of ‘x is bad’ is what follows from [OI] and what inferentialism, as a semantic theory, develops.
The criticism of non-descriptivism via FGA assumes that ‘torturing animals is bad’ expresses a proposition because ‘is bad’ represents an ingredient of state-of-affairs. This might be correct, but it does not follow from FGA. The alternative explanation, that ‘torturing animals is bad’ has propositional content because it can be inferentially articulated as a premise or a conclusion, is also compatible with the validity of the argument.
In summary, the validity of Geach's argument patterns require that the propositional status of normative propositions be acknowledged, and for this it is enough to accept the propositional criterion formulated in [OI]. Both the ‘is-bad’ argument and the ‘is-true’ argument are valid, as Geach correctly states. Nevertheless, the set of possible non-descriptivist explanations available to us in order to account for their validity is not identical. The truth operator can be assumed to be semantically irrelevant, but ‘is bad’ cannot be so explained because this would imply the invalidity of the ‘is-bad’ argument. Not all non-descriptivist approaches that work for ‘is true’ are appropriate for ‘is bad’, even if the converse might well be true. The challenge of FGA for the analysis of moral discourse requires more radical moves than the relatively harmless modifications that are capable of explaining, in a deflationist fashion, the behaviour of functions of propositions.Footnote 49 If propositions are interpreted as 0-adic concepts, any general explanation given for concepts whose arguments are also concepts would cover both the case of ‘is true’ and the case of ‘is bad’. Moreover, any explanation that attended to semantics by relying on inferential relations and not on compositional procedures, could similarly explain the validity of both arguments.
4. Answering some criticisms
Inferentialism is a semantic theory, and as such stands at the same level as expressivism, descriptivism, and representationalism, and it is a particular development of the [OI], albeit a very natural, and straightforward one, because it also individuates propositions ‘from above’: if what appears to be two different propositional contents possess the same properties and stand in the same relations, then they are one and the same. This is an instance of the Leibniz Law. But although [OI] and inferentialism stand at different levels, some criticisms of inferentialism apply also to [OI]. So it is advisable to say something about them.
The two worries that seem to be more deeply rooted in the minds of critics of non-descriptivism are, first, the vanishing of a limpid separation between the descriptive and normative realms and, second, the danger of a contagious generalization of non-descriptivism to any areas of discourse. Both qualms are related and I will comment on them in turn.
The first concern is the collapse of the descriptive/normative distinction. Chrisman and Dreier,Footnote 50 among others, have given voice to the suspicion that non-descriptivism has the effect of blurring the distinction between descriptive and normative uses of language, between ethical realism and anti-realism, between i ses and o ughts. An initial step towards answering this worry is to point out that the lack of any radical divide between these two realms is not as consequential as it might seem. The distinction may very well be gradual rather than radical, and context-dependent rather than absolute.
In any case, [OI] shows only that the distinction (or the gradation) does not belong to the meta-theory of propositions and concepts, but rather to some particular substantive theories that implement [OI]. The hardly debatable fact that normative claims can take part in inferences, as illustrated by FGA, shows that the distinction between descriptive and normative claims, at most, classifies kinds of propositions rather than demarcates the class of propositions from the class of non-propositional entities. In particular, a clear-cut distinction between normative and descriptive propositions and concepts is a vestige of empiricist semantics,Footnote 51 but does not need to belong to the general intuition that defines propositions and concepts. Other, less hierarchical, semantic implementations cannot draw the distinction as easily as empiricist proposals.Footnote 52 In an inferentialist, rationalist approach to concepts, there is no a priori, general, criterion to distinguish normative from descriptive concepts and contents.
The intuition that we all feel that there is a difference in use between ‘The cat is on the mat’ and ‘She did the right thing’ may touch upon semantic features that are worth distinguishing. Other sentences such as ‘Two times two is four’, ‘Boris Johnson insists on a hard Brexit’, ‘I love you so much’, or ‘Children usually resemble their parents’ surely present differences as marked as the differences we see in the first two examples. The point made by [OI] is simply that all these differences do not affect the propositional nature of these contents. Thus, the general answer to this first concern, provided by the view I am defending, is that the distinction between descriptive and normative claims surely marks differences in use that deserve to be acknowledged, but this acknowledgement does not require exiling normative claims from the realm of propositions. Pragmatist approaches would welcome gradual and context-dependent distinctions, and semanticist approaches would favour clear-cut distinctions.
The second concern is what Dreier and Chrisman have called ‘creeping minimalism’.Footnote 53 The charge of creeping minimalism is the suspicion that a non-representationalist (expressivist, minimalist, deflationist, and the like) account of some notion (e.g. truth, proposition, assertion, and the like) forces the corresponding non-representationalist accounts upon all members of its semantic family. Belonging to a semantic family means that the definition of any of its members necessarily involves reference, explicit or implicit, to some of the others. In other words, the criticism is that non-representationalist approaches to any one of the notions of truth, proposition, knowledge, and assertion contaminate the semantics of the other notions as well. This concern is completely justified. Here too we should bite the bullet, since the semantic views that account for truth, proposition, knowledge, and assertion have to be mutually coherent. In the range of semantic proposals for truth, proposition, knowledge, and assertion, the result of creeping minimalism stands at one end. At the opposite end stands the result of what we might call ‘creeping descriptivism’ or ‘creeping representationalism’, i.e. the view that all assertions have to be descriptions and represent states of affairs. Sentences such as ‘What the policeman said is true’, ‘Joan trusts Victoria’, ‘Torturing animals is wrong’, and ‘The European Commission encourages austerity politics’ would be classified by creeping descriptivism/representationalism as descriptions of some kind. Creeping descriptivism/representationalism also implies that, if these sentences are true, there must be some combinations of entities that act as their truth-makers.Footnote 54 The risk of creeping descriptivism/representationalism places the objection of creeping minimalism in its right perspective. If descriptivists were justified in accusing non-descriptivists of having drained notions such as truth and proposition of their content, non-descriptivists would be justified in accusing descriptivists of having drained notions such as description and representation of their contents. In terms of semantic drainage, no party is in a better position. The options are now clearly stated. Nevertheless, the fact that undoubtedly all parties have still a long way to go in order to present proposals free of semantic difficulties does not necessarily imply that both views have similar credentials. Representationalism/descriptivism seems to have exhausted its options and presents itself as a declining paradigm, whereas the potential of its alternatives remains to be seen. Moreover, the uncontaminated intuitions of speakers and theorists, those that belong to the set of background assumptions and that have conferred FGA the philosophical status that it currently enjoys, tip the balance in favour of a minimalist development, if only for methodological reasons. For, in order to understand the import of FGA, it is not necessary to involve oneself in the endless difficulties derived from the realist and representationalist views of normative notions.
Let us take stock and draw some concluding remarks. FGA undoubtedly refutes the interjection view on normative claims, but it does not affect semantic approaches to the meaning of some terms that are truth-conditionally irrelevant,Footnote 55 nor the approaches to propositions that individuate them by non-representationalist criteria. The organic intuition that propositions are the entities that behave as such is a basic presupposition of FGA and a presupposition that has not been challenged by anti-representationalists, who have not felt the need to argue for it either. The organic intuition seems thus be a very basic intuition about propositions shared by all parties and should not be used to support either a representationalist account of normative terms or a realist account of values.
According to [OI], propositions are the basic entities able to bear logical properties. If we philosophers consider that normative and descriptive claims manifest essentially distinct properties, and require different semantic and metaphysical treatments, this needs not be done at the price of denying the former their status as genuine propositions. There is nothing to gain and, as the endless discussions on FGA and the semantics of ethical discourse show, very much to lose.Footnote 56
