A wide range of aesthetic predicates exist: ‘X is beautiful/elegant/graceful/delicate/dynamic etc.’ (Sibley 1959). Through the sincere and literal utterance of such sentences, we make aesthetic statements.1 It is unclear how to delineate the class of aesthetic predicates — which words belong on the list, and why?2 For the purposes of this paper, it is not important to answer this question; it is generally accepted that there is a wide variety of aesthetic predicates that allow for certain differentiations. One important differentiation concerns their evaluative character: some carry a strong positive or negative valence, whereas others seem only weakly evaluative or not evaluative at all (Sibley 1974; Zangwill 2001; Bonzon 2009).

In this paper, I will concentrate on beauty statements, such as

that are considered strongly evaluative. In other words, it is generally agreed that (1) conveys that the speaker maintains a positive affective attitude (e.g. liking, appreciating) regarding X — or more precisely, regarding the sensual appearance of X.3 Although (1) does not explicitly state so, we can infer from (1) that the speaker likes the way X looks or sounds. This is confirmed by the infelicity of the following statement:

Statement (2) appears contradictory: with the first conjunct, one conveys that one likes how X looks/sounds, but with the second conjunct, one negates this sentiment.4

Why is it that, from (1), we can infer that the speaker likes how X looks or sounds, even though the sentence does not explicitly state so? Let us call this phenomenon ‘the evaluativity inference’. The central question of the paper is: Why does a beauty statement give rise to the evaluativity inference?5

The explanation I prefer is based on a certain speech-act theoretical version of hybrid expressivism, which states that, by a sincere and literal utterance of (1), we perform two speech acts simultaneously — one assertive and one expressive. In the assertive act, we ascribe a certain property to X and express a belief with a certain propositional content. With the expressive act, we express a certain non-propositional affective mental attitude regarding the sensuous appearance of X. I will detail this form of hybrid expressivism and argue that, given the alternatives, it provides the best explanation for the evaluativity inference. I should note at the outset, that the theory presented here diverges from variants of hybrid expressivism discussed in meta-ethics in several significant ways. It is motivated differently, it is not characterized by analogous theses and, as a result, avoids some of the corresponding problems. While the limited space prevents a detailed discussion of the relationship between my version of hybrid expressivism and the variants in meta-ethics, I will address some of the most important differences in §7.

Defending a hybrid theory with respect to statements of beauty may initially seem surprising. In aesthetics, the observation that aesthetic statements vary in their evaluative strength often results in a classification that allows for purely evaluative statements. Frank Sibley (1974: 7–9), for example, differentiates the class of aesthetic terms by (a) solely evaluative terms, which have no descriptive but purely evaluative content (his examples include ‘beautiful’, ‘ugly’, ‘lovely’ etc.); (b) evaluation-added terms, which have both descriptive and evaluative content (his examples include ‘elegant’, ‘graceful’, ‘garish’ etc.); and (c) descriptive merit terms, which only have descriptive content but are often used to support evaluative aesthetic statements (his examples include ‘balanced’, ‘unified’, ‘vivid’ etc.). This classification is mirrored by Nick Zangwill’s (2001: 9–18) distinction at the level of aesthetic judgments. He distinguishes between verdictive aesthetic judgments, which correspond to Sibely’s solely evaluative terms, and substantive aesthetic judgments, which correspond to Sibley’s evaluation-added and descriptive merit terms.

In the context of these classifications, it might seem plausible that a hybrid theory applies to statements such as ‘X is elegant/graceful/garish, etc.’ rather than to ‘X is beautiful’. However, this may be misleading. In my view, beauty statements are more evaluative than many other aesthetic statements, but they are not purely evaluative. There is a reading of ‘beautiful’ such that the term has both descriptive and evaluative aspects; and the best way to account for both aspects is provided by the version of hybrid expressivism outlined above.6

In the next section, I continue with preliminary remarks on different readings of the word ‘beautiful’ and specify in more detail the reading I will focus on (§2). Then I make some remarks concerning the nature of the evaluative attitude conveyed by beauty statements (§3). Afterwards I discuss different attempts to explain the evaluativity inference that might seem initially plausible but will turn out to be wanting (§§4–5). Finally, I elaborate and defend the explanation based on the speech-act theoretical version of hybrid expressivism outlined above. In the course of this defence, I will specify the theory in detail (§6) and differentiate it from other forms of expressivism suggested in meta-aesthetics (§7).

Aestheticians often differentiate between a broad and a narrow reading of ‘beautiful’. In the broad reading, the term refers to aesthetic or artistic value in general; in the narrow reading, it refers to one species of aesthetic value (Beardsley 1981: 505–506; Scruton 2007: 15–16; Lopes 2018: 6).

Consider the following sentences:

Sometimes, (3) is used synonymously with (4). In such cases, the term ‘beautiful’ is used in a broad sense. In the broad sense ‘beautiful’ is used to ascribe a general value to an artwork. The function of the term in this reading is simply evaluative, used to praise the artwork without specifying any features of the work on which the praise rests (Bonzon 2009: 194). In this sense ‘is beautiful’ is used ‘synonymously’ with ‘is a good artwork’.

However, ‘beautiful’ is not always used in this broad sense. Take a look at the following conversation:

B’s reaction in (5) is not contradictory, so her use of ‘is beautiful’ is not synonymous with ‘is a good artwork’. Thus, she does not use ‘beautiful’ in the broad sense specified above, but rather in a more specific sense — one in which ‘beauty’ is not synonymous with ‘artistic goodness’. She refers to a more specific kind of aesthetic value that is traditionally taken to be closely connected to perceptual pleasure and a certain kind of well-formedness (Hume 1740/2000: 299; Kant 1790/2000: §§ 1–9; Scruton 2007: 5).8

Note that this narrow reading still gives rise to the evaluativity inference. In (5), B conveys that, even though the artwork is pleasing to the eye and that she appreciates its visual beauty, it is nevertheless a bad artwork. In what follows, I concentrate on the narrow reading of ‘beautiful.’ It is the narrow reading of (1) that I see as having a combination of descriptive and evaluative components that is best captured by a speech-act theoretical form of hybrid expressivism.9

Besides broad and narrow readings, we also need to differentiate autocentric and exocentric readings of ‘beautiful’. Take a look at the following conversation:

It is natural to assume that the painting B refers to is one that Hanna judged to be beautiful (perhaps by calling it ‘beautiful’), and that B ties her statement to Hanna’s judgment. This is commonly known as an exocentric reading of a term (Lasersohn 2005: 670). In such a reading, B’s use of the term ‘beautiful’ does not convey that B (the speaker) likes the painting, but that the person to which the statement is tied to (in the example, Hanna) does so. In what follows, I will concentrate on autocentric uses of (1) that convey that the speaker maintains a positive attitude regarding the object of predication.

In the introduction, I have described the positive affective attitude conveyed by (1) as a state of liking or appreciating. How should we understand the nature of those states? First, the states are intentionally directed towards a certain object: individuals like or appreciate something. The object about which the positive affective attitude is conveyed by (1) is the sensuous appearance of X. This is confirmed by the fact that, of the following sentences, only (7) is infelicitous:

In uttering (1), one does not convey that one likes X in all respects but rather that one likes the sensuous appearance of X. For simplicity, I will sometimes use ‘S likes X’ or ‘S likes the appearance of X’ as shorthand versions of ‘S likes the sensuous appearance of X’.10

Besides its characterisation as intentional, how else should we specify the state of liking? First, we might try to specify it as an occurrent state with a certain phenomenal character, such as an object-directed feeling of pleasure. This, however, is not correct, because we can truthfully ascribe states such as liking to individuals who are not in an occurrent phenomenal state of pleasure (e.g. sleeping persons).

A more promising option specifies the relevant state of liking as a disposition. According to this view, one likes the appearance of X if and only if one is disposed to receive a certain kind of pleasure by experiencing X (under normal conditions of perception). Note, however, that dispositions do not necessarily manifest: a glass, for example, can have the disposition of being breakable without ever actually breaking. Thus, given a dispositional understanding of the affective attitude of liking, one can like X without ever having experienced the pleasure triggered by X. This is an implausible consequence of the view. One cannot like the sensuous appearance of something — for example, the smell of strawberries, the sound of a certain guitar, or the visual appearance of Jackson Pollock’s No. 5 — without having experienced the object in question (or a copy of the object that allows for comparable experiences). The problem can be avoided by arguing that a state such as liking is not simply a disposition but a disposition that has been manifested. These special kinds of dispositions are sometimes referred to as ‘tendencies’ (Dinges & Zakkou 2021).

Another promising option is to specify liking as a functional state that is specified by a certain causal profile. Namely, it is caused by the item toward which the attitude is directed, and it will cause other attitudes as well as a certain kind of behaviour.

For the purposes of this investigation, it is not necessary to discuss the pros and cons of the tendency- and the functional-state-account of liking. It is interesting to note, however, that both accounts imply that one can like the sensory appearance of an object only if one has experienced the object (or a copy of the object that allows for comparable experiences).11

The last preliminary remark concerns the relation of the state of liking and the more general state of valuing. Not every instance of valuing is necessarily an instance of liking. It may be that one does not like the way a certain object looks, and yet one might still value the sensual appearance of the object — for example, because it exemplifies something interesting in a particularly lucid way. This is especially true with respect to artwork: it might be the case that a painting looks ugly and displeasing, such that one dislikes the way it looks in a certain sense — but one might nevertheless value the composition and general visual appearance of the painting because it exemplifies something interesting. This observation relates back to the comments regarding the different readings of ‘beautiful’. If there is a broad reading of (1) in which the speaker expresses the general mental state of valuing, I will exclude this reading in what follows.

After clarifying these preliminaries with respect to the term ‘beautiful’ (§2) and the corresponding affective attitude of liking (§3), I now turn to possible attempts to explain the evaluativity inference.

Semantic explanations of the evaluativity inference assume that it is part of the truth conditions of

that the speaker likes the appearance of X. Advocates of this approach claim that (1) entails that the speaker likes the appearance of X.

If this view is correct, then (1) is false as used by speaker S simply because S does not like X. Additionally, if we accept the standard semantics of negation — according to which a negated sentence ‘not-p’ is true if and only if the sentence ‘p’ is false — then (9) is true as used by speaker S simply because S does not like X.

This seems to be an overly chauvinistic consequence of the view. Note the following conversation:

In (10), B challenges A’s assertion that ‘This house is not beautiful’. Person A defends her statement by exclusively focusing on her own mental states, and she cites these as conclusive reasons for the truth of what she has said. This reaction is not only linguistically odd — it is also epistemically misguided to treat nothing but one’s mental state as conclusive evidence for statements such as (9). In aesthetic discourse, we point to certain features of the specified object to defend our aesthetic claims; we will not convince somebody of the truth of our aesthetic claim by merely pointing to our mental state of liking or the lack thereof. The epistemic inappropriateness of A’s reaction speaks against the semantic explanation of the evaluativity inference, because, according to that explanation, A’s reaction in (9) should be fine.

To help clarify this point, see the following comparison to explicitly relativized taste statements:

For (11) and (12), the following view is plausible: (11) is true as used in context C only if the speaker in C likes the taste of X. Thus, (11) entails that the speaker in C likes the taste of X. This explains the evaluativity inference for (11). Given the standard semantics of negation, (12) is true in C simply because the speaker of C does not like the taste of X. Note that, with respect to (12), the above-mentioned consequence does not seem problematic.

In (13), A’s reaction might be unfriendly, but it is neither linguistically odd nor epistemically inappropriate. With respect to (13), citing one’s own mental state as a conclusive reason is perfectly fine. It is the difference between aesthetic statements and explicitly relativized statements of taste that is exemplified by the difference in (10) and (13) that speaks against the semantic approach to the evaluativity inference.13

There is an additional and even more important issue to address. Remember that, according to all plausible conceptions regarding the metaphysical nature of the mental state of liking, it follows that, if one likes the sensual appearance of X, one has personal experience with X. Thus, for metaphysical reasons, advocates of the semantic approach must accept the following: if (1) entails that the speaker likes X, then it also entails that the speaker has personally experienced X. This is so because in order to like X, the speaker must be perceptually acquainted with X. Consequently, according to the semantic approach to the evaluativity inference, (1) is false simply because the speaker has not experienced X. Again, presupposing the standard semantics for negation, it follows that (9) is true, simply because the speaker has not experienced X. Thus, if the semantic explanation of the evaluativity inference were correct, then one could appropriately defend one’s assertion of (9) through the following:

This, however, is an unacceptable result. It is obviously epistemically inappropriate to regard the fact that I have not experienced an object as conclusive evidence for the truth of my claim that the object is not beautiful.

Thus, besides the problem of sanctioning the lack of personal liking as conclusive evidence for negative aesthetic claims, advocates of the semantic explanation of the evaluativity inference are confronted with a dilemma. Either they keep one of the suggested plausible analyses of liking and accept as a consequence that my statement ‘X is not beautiful’ is true solely because I have never experienced X. Or they dismiss this consequence by working out an analysis of the mental state of liking according to which it is possible to like the experience of something — for example, the smell of strawberries, the sound of a certain guitar, or the visual appearance of Jackson Pollock’s No. 5 — without ever having made the experience. Neither option is particularly attractive.

Two pragmatic explanations of the evaluativity inference suggest themselves. According to the first, it is a linguistic presupposition of (1) that the speaker likes the object of predication. Just as the term ‘quit’ in the statement ‘Jules quit smoking’ triggers the presupposition that Jules has smoked in the past, so does ‘beautiful’ in (1) trigger the presupposition that the speaker likes the experience of X. This suggestion, however, is not convincing. In contrast to the evaluativity inference, presuppositions project over a wide range of operations, such as the following:

The presupposition that Jules has smoked in the past is preserved in (15)–(17). However, the analogous sentences (18)–(20) do not trigger the evaluativity inference:

None of the sentences conveys that the speaker likes X. This difference in the projection behaviour of presuppositions and the evaluativity inference speaks against the presupposition approach with respect to the evaluativity of aesthetic statements.

The second pragmatic approach assumes that the evaluativity inference is a Griceian implicature (Grice 1989). Because implicatures are classified as conventional or conversational, this approach has two subvariants. According to the first, it is a conventional implicature that the speaker likes X. This is unconvincing because it is widely accepted that conventional implicatures exhibit a similar projection behaviour as presuppositions (Potts 2015). Thus, the objection against the presupposition approach also applies to this subvariant of the implicature account.

The second subvariant specifies the evaluativity inference as a conversational implicature. The problem with this approach is that conversational implicatures can be cancelled. If ‘q’ is a conversational implicature of ‘p’, then it can be cancelled by ‘p, but not q’ or ‘p, indeed not q’. However, as (2) has already illustrated, this is not the case with respect to the evaluativity inference triggered by ‘X is beautiful’.14 Furthermore, if the evaluativity inference were in fact a conversational implicature, there would need to be a corresponding Gricean mechanism that explains how this specific conversational implicature is generated. However, thus far, no such mechanism has been proposed and it is unclear which Gricean mechanism could account for it. This speaks against such an approach, regardless of whether the evaluativity inference can be cancelled or not.15

The previous sections have illustrated that both semantic and certain pragmatic explanations of the evaluativity inference are wanting. In what follows, I will suggest a different pragmatic explanation, namely a speech-act theoretical one. The explanation rests on the following version of hybrid expressivism: With the sincere and literal utterance of

we perform two illocutionary acts simultaneously: one assertive and one expressive. The assertive act ascribes a certain property to X and expresses the corresponding belief with a certain propositional content; in this respect, uttering (1) is comparable to uttering a declarative sentence, such as ‘X is red’. The expressive act expresses a non-doxastic, non-propositional mental state M that involves the state of liking; in this respect, uttering (1) is comparable to uttering an exclamative sentence, such as ‘How beautiful!’ or ‘Bravo!’.16

How are these two acts related? Which type of property do we ascribe to X so that the assertive act is accompanied by the expressive? I suggest the following answer: Beauty is a response-dispositional property, namely the property to evoke a certain kind of mental state M in subjects under certain (idealised) circumstances C. The mental state M is the manifestation of the disposition, and perceiving the object under ideal conditions (in the appropriate environment, with functional perceptual capacities, with the appropriate attention to certain details etc.) is the manifestation condition C. In the assertive act associated with the aesthetic statement (1), we ascribe this kind of response-dispositional property to X. Thus, ‘X is beautiful’ is true if and only if X has the dispositional property to evoke M in subjects under C. In the expressive act associated with (1), we express the mental state M that is the manifestation of the response-dispositional property, which we ascribe to the object X in the corresponding assertive act. Thus, these two acts are closely related: The mental state expressed in the expressive act is the manifestation of the dispositional property that is ascribed to an object in the assertive act.17

With respect to the relation between the two acts, I further suggest the following simple principle:

According to this principle, whenever ‘X is beautiful’ is embedded in a context wherein the assertive act is absent, the expressive act, and thus, the evaluativity inference is absent as well. This is illustrated by (18)–(20). None of those sentences conveys that the speaker likes X.

The expressive act correlated with (1) explains the evaluativity inference. As with all illocutionary acts, expressive acts are governed by sincerity conditions. The sincerity condition of an expressive act requires that one is in the non-doxastic mental state one expresses by the act. The corresponding rule is that you should utter ‘q’ only if you are in the non-doxastic mental state expressed by ‘q’. For example, utter ‘Ouch!’ only if you are in the phenomenal state of feeling pain (Searle 1969: ch. 3).

It is part of the sincerity condition of the aesthetic statement (1) that one is in the non-doxastic mental state M that one expresses by uttering the sentence. The corresponding rule is: Utter ‘X is beautiful’ only if you are in the mental state M; this mental state involves liking the appearance of X. Thus, (1) can only be uttered appropriately if the speaker likes the appearance of X. Because we usually assume that the sincerity conditions of speech acts are met, the suggested line of thought explains the evaluativity inference — namely, it explains why uttering (1) usually conveys the information that the speaker likes the appearance of X.

Note that the evaluativity inference is only explained by the expressive part of aesthetic statements. But if the expressive act corresponding to aesthetic statements is enough to explain the evaluativity inference, then why should we accept the suggested version of hybrid expressivism that claims that, in uttering an aesthetic sentence, we perform two illocutionary acts, an expressive and an assertive one? Why not just accept simple expressivism and hold that with an aesthetic statement such as ‘X is beautiful’ we perform nothing over and above an expressive act of expressing a non-doxastic, non-propositional attitude of liking?

In the following sections, I will introduce different variants of simple expressivism in meta-aesthetics and elaborate three advantages of the suggested hybrid approach. These three advantages make the hybrid approach preferable to simple expressivism. The first has to do with a problem that has been referred to as ‘Kant’s puzzle’ (7.1.), the second with certain retraction data (7.2.), and the third with the notorious Frege-Geach problem (7.3.).18

Statements of beauty carry a positive valence. By uttering

one conveys that one maintains a positive affective attitude with respect to the sensual appearance of X. Even though (1) does not explicitly state so, by hearing (1), we can infer that the speaker likes how X looks or sounds. How can we explain this evaluativity inference?

After clarifying some preliminaries (§§2–3), I have argued that semantic (§4) and certain pragmatic (§5) explanations are flawed. My positive proposal consists of a speech act-theoretical explanation based on a specific version of hybrid expressivism (§6). Namely, with the literal utterance of (1), we perform two illocutionary acts simultaneously. We perform an assertive act and thereby assert the proposition that X has the disposition to evoke a certain mental state M (that involves liking the appearance of X) in subjects S under appropriate conditions C. Additionally, we perform an expressive act, thereby expressing the non-propositional, non-doxastic mental state M, which is the manifestation of the dispositional property that we ascribed to X in the assertive act.

The expressive act correlated with (1) explains the evaluativity inference. Expressive acts are governed by sincerity conditions (Searle 1969: ch. 3); included in the sincerity condition of (1) is that the speaker is in state M. This mental state M involves liking the appearance of X, and this is why uttering (1) usually conveys that the speaker likes the appearance of X. The additional assumption that an utterance of (1) also involves an assertive act saves the suggested explanation from the serious concerns associated with simple expressivism (§7).

One problem with the suggested account is that it may appear ad hoc. If there are no other expressions that exhibit similar behaviour, it raises the question of why only aesthetic statements behave in the manner I am proposing. However, it is worth noting that in aesthetics, there is a commonly held belief that aesthetic judgments and statements belong to a highly distinctive class, which may necessitate a unique theoretical framework. Moreover, it is important to emphasize that while the proposed approach might seem ad hoc in the sense that it is only applicable to aesthetic statements, it is not ad hoc in other aspects. The suggested account not only explains the linguistic data related to aesthetic statements, but also addresses epistemic puzzles concerning aesthetic judgments or beliefs, and integrates a plausible response-dispositional conception of aesthetic properties. Furthermore, the general idea of illocutionary pluralism, which asserts that a speaker can perform multiple illocutionary acts through a single utterance token in a specific speech situation, is widely discussed and defended in speech act theory (Johnson 2019; Lewinski 2021). The considerations presented in this paper can be seen as shedding light on an overlooked instance of illocutionary pluralism, namely aesthetic statements such as ‘X is beautiful.’