1. Introduction

It has been largely accepted that what a sentence means is embedded in its truth condition, after Tarski’s ingenious invention of the semantic concept of truth and Davidson’s later extension of it to natural languages. In short, it is sufficient to know the meaning of a sentence if we know its truth condition; in a more formal way, this is expressed as “S is true-in-L iff p”. For instance, if John says to me, “snow is white”, the speaker knows what John’s utterance means if she knows: “‘Snow is white’ is true-in-English iff snow is white”. It seems quite convincing that one of the biggest function of languages is to communicate, and fundamentally, to communicate is to know when to agree or disagree. Therefore, some philosophers are inclined to say that at the end of every sentence there is an invisible “assent or dissent”. For sentences such as “snow is white” or “grass is green”, this theory which is generally called truth-conditional semantic comes in handy. However, not all sentences that people speak are the same. Purely extensional statements such as assertions about the physical world, are only a small part of our linguistic spectrum; many sentences, such as imperatives and metaphors, play heavy roles in our communication. This phenomenon invites us to rethink our approach to semantics—natural languages are different from formal languages into which Russell and Carnap were trying to transform languages; humans do not live in formal logic, and “truth” in its strictly logical sense does not play the most prominent role in linguistic understanding, humans have intentions, desires, wishes, and many other feelings (or intentional attitudes) which contribute heavily to what a speaker means when she says a sentence. Among many of these perspective sentences (as opposed to extensional, indicative sentences like “snow is white”), moral sentences are worth a close investigation. On the one hand, with the rise of scientism and other materialistic-inclined analytic philosophy as well as the dismantling of metaphysics, moral philosophy is in peril, being threatened to become nothing more than convenient and instrumental norms for social order, which might be called as naturalism; on the other hand, diving into the semantic of moral sentences helps with better disentanglement of some misconceptions about moral philosophy. This article shall adopt a semantic and analytic method to investigate some aspects of moral sentences. At the end of this paper, two conclusions shall be drawn: 1) Ethics cannot be naturalized; 2) truth condition itself is not sufficient to explain moral sentences and perspective sentences in general.

2. Rejecting Naturalism

One of the most noticeable movements in modern moral philosophy is naturalism. Its main goal is to attack ancient teaching of ethics, mainly Platonic ethics. Suppose someone says, “telling the truth is good”, what she might mean is that the action, telling the truth, shares some aspects of the entity (or the Platonic form), Good; paraphrasing what she said a little bit, the sentence means: there is some goodness in the action of telling the truth. So, when someone says the sentence “telling the truth is good”, the speaker is referring to something which has certain properties on it; in doing so she makes an ontological commitment of taking Good as an existing entity. Here is another interpretation of “Telling the truth is good”, the sentence describes the action as a good action. In this way, the speaker does not commit herself to the ontological position that Good exists, but only good things exist. This interpretation, influenced by Frege-Russell’s tradition, treats Good as not something actual, but as a predicate that binds a variable to a function. W.V. Quine argued that the ontological commitment of treating predicate in a sentence as something actually existing is a mistake [1]. Since treating Good as something actual which exists, mayhap, in a spiritual world seems ridiculous and backward, the second interpretation is much more preferred. If someone asks what “Good” is, the answer could be either a) Good does not exist, or b) it is a set of characteristics of actual beings. This article is not to deny that the second interpretation is more appealing. In fact, without any empirical observation, it is hard to confirm that Good exists and is actual. However, even if Good (and other ethical concepts) do not refer to real things, it does not mean naturalism is correct.

Let us continue to investigate the second interpretation and withdraw ourselves from moral contexts for now. Suppose someone agrees that Good is a set of characteristics of things. For example, when someone says, “Christ is a good teacher”, what she means is that there exists a man who is named Christ, and who is a teacher, and who has certain characteristics. Now she says, “that chair is a good one”, then she means there exists a chair which has certain characteristics. The characteristics of Christ and the characteristics of the good chair are not the same, then how is someone able to attribute the same word to two different things? This is the argument G.E. Moore took when he tried to refuse naturalism in “Principia Ethica” [2]. But it does not answer the question fully, for a naturalist can of course accept that there is no universal understanding of “good”, and humans can only gradually develop the concepts along with encountering more and more instances with certain characteristics, one simply needs to agree that humans have the capacity to generalize a concept through individual instances. It is R. M. Hare later who gave us a more explicit and powerful objection to naturalism.

His objection attacks not toward the absurdity of generalizing concepts through vast instances, but toward the distinction between value words (such as “good”) and descriptive words, such as “red”. In his mind experiment, someone is set to teach a foreigner who had no understanding of English to speak English. If the teacher is to teach the foreigner the word “red”, he can show him various red objects such as red apples, red chairs, red flowers, etc., as long as their mentality is not drastically different from one and another, at some point the foreigner will get the sense when he should use the word “red” to describe an object. However, it would become very difficult to teach him the word “good”, because the characteristics to which “good” is attached cannot be shown empirically. The teacher could demonstrate a good fire extinguisher, a good chair, a good mayor, and many other good things, the foreigner could still fail to comprehend how he should use the word “good”. In Hare’s argument, it is when the teacher starts to ask what the foreigner would choose among many other similar things, he started to make sense of “good”. What Hare emphasizes here is that value words are tied to people’s mentality of making choices and evaluating [3].

He later explains how this would refuse naturalism. A naturalistic definition of “good” is: x is a good A iff x is A which is C; but it becomes circular when saying good A’s are C which are A’s. For example, The Mona Lisa is a good painting, and by “good painting”, it means paintings that have characteristics such as having the tendency to arouse some recognizable feelings in people who are looking at them; But to answer why good paintings are good, then by the naturalistic definition, it is only appropriate to say: Good paintings are good because paintings with certain characteristics such as arousing certain recognizable feelings have certain characteristics, which is tautologous. It would be a mistake to deny the logical validity of A=A, since in actual communications people say sentences such as “bachelors are unmarried men” all the time, and it will not become a problem if someone simply changes it to “unmarried men are unmarried men”. It is assumed that people have known what “bachelor”, “unmarried”, and “man” mean before understanding the sentence “bachelors are unmarried men” as an analytical truth. However, what Hare says is that value words are not to describe things in the same way as “red” does. It has certain commending power which is tied to speaker’s choices and it would be canceled out when treating them as descriptive words.

What Hare has in mind, in some aspects, can be explained by Franz Brentano’s intentionality; intentionality in different philosopher’s work does not usually mean exactly the same, its basic concept, however, is that a person’s mental states are about something, and the process of intentional states is “seeing” something as something. This notion can also be applied to many other intensional sentences, which are referentially opaque. When someone says, “Mona Lisa is a good painting” it is not a description in the sense that the speaker is describing some characteristics of The Mona Lisa; the speaker projects some of his intentional attitudes (Hare calls them evaluation) into what the sentence means. Therefore, the meaning of “Mona Lisa is a good painting” on the one hand might refer to certain characteristics of the painting, such as arousing some recognizable feelings in the speaker, on the other hand, it also means that the speaker feels in certain ways. If the word “good” is naturalized, then the sentence can no longer convey how the speaker feels. Now, someone might point out so far, the discussion is only about “good” in a non-moral context. However, the boundary of moral and non-moral contexts is a matter of degree. There is no sharp distinction between a moral “good” and a non-moral “good”. The passages have discussed how naturalism fails to capture the intentionality of value words in non-moral contexts; the same critique can also be applied to moral contexts.

It is clear then, that moral sentences are not a set of descriptions about the world; in other words, moral sentences cannot be treated as indictive sentences, they are essentially perspective sentences. To naturalize ethics (if we agree there is something called ethics at all), is to treat moral sentences as a set of sentences which can be inferred from worldly facts. However, perspective sentences cannot be inferred from indicative premises alone. This conclusion has been partially seen from the investigation above, but further clarification is required. For instance:

Jumping out of the window kills you

Therefore, you should not jump out of the window.

This argument is “incomplete”. There is no way to infer from “jumping out of the window kills you” to “you should not jump out the window” without adding the premise “you should not kill yourself”. A speaker does not have to say this premise out loud because it is commonly accepted and it requires little explication; but it has to be presupposed by the hearer if she understands the sentence at all. Sentences about worldly facts are, in a naturalist view, extensional, and indicative; but no indicative premises alone can have a perspective conclusion. Moral sentences are perspective sentences and cannot be treated as indicative ones; then it is apparent that moral principles cannot be inferred from worldly facts alone. Therefore, ethics cannot be naturalized.

3. Refusing Truth-conditional Semantics

It should be noticed that some more radical approaches from naturalism would commit themselves to reductionism, which in a way eliminate intentionality all together. Whether these approaches are successful or acceptable is not the main interest in this paper, if someone agrees with reductionism, it is all welcome for her to throw the entire moral philosophy into the meaningless hole; Russell, Carnap, and early Wittgenstein all agreed that moral propositions are meaningless, they only express emotional attitudes. But should someone feel slightly suspicious about the reductionist method, further investigation into how and why “truth” is not enough for understanding what moral sentences mean is required.

Radical naturalists will suggest that moral sentences are meaningless. In a sense, they agree that moral sentences convey the speaker’s intentionality. However, intentionality cannot be expressed formally. If someone understands the meaning in terms of the truth condition, then she has to admit that moral sentences are meaningless, because moral sentences do not have truth conditions. The following passages should explain the above argument step by step. Intentionality in language is usually associated with intensional sentences, and intensional sentences cannot be expressed formally. At the surface level, formal language should, by itself, extensional; this means that they are referentially transparent, this allows replacing co-referential terms without changing the truth value. But the truth value of (if there is any) of “Monalisa is a good painting” would certainly change, should the sentence be translated it into:

∃x [(x = (Dy∧ Fy))∧ Gx)]

“x” stands for Monalisa, “y” stands for a painting, “D” “F” and “G” are predicates that mean “painted by da Vinci in 1503-1517”, “hung in Louvre in France” and “a set of characteristics which arouses certain emotional states in the viewer” respectively (The formula breaks the original sentence into a definite description of a certain painting).

If Christ says, “Monalisa is a good painting” he might not know that the Monalisa is a painting which hangs in the Louvre and is painted by da Vinci during 1503-1517. However, as some might suggest, Mona Lisa can be treated as a rigid designator and the name itself cannot be expressed through definite description; in other words, when Christ refers to Mona Lisa, all he can do is referring to that painting. But this approach only emphasizes the problem, for to rigidly designate Mona Lisa is determined by the speaker’s epistemic capacity of managing the stored information about the said painting. On a deeper level, intentionality as a “subjective” aspect of mentality is not only about the physical world which is transparent to the hearer, it also “ontologically private”.

When Tarksi published his “The Semantic Conception of Truth”, he explicitly said that his definition of truth can only be applied in formal systems [4]. Not only because only formal languages have explicit rules of syntax, but also only in strictly defined language can we avoid the absurdity of self-reference dilemma. In other words, if someone is to apply Tarskian truth into a semantic theory, she has to translate natural sentences into formal sentences (of course some leeway is allowed here, Davidson for example thinks self-reference dilemma can be tolerated). Some philosophers and linguists, however, attempt to modify the form of the theory. For instance, formal theorists can invent a symbol for intentional attitudes; “Mona Lisa is a good painting” can be translated as “A ψ that p (Mona Lisa is a good painting)”. But again, this simply emphasizes intentionality one more time, unless the theory reduces the mental states, when the speaker is committed to some intentional attitudes, into brain states, which can be mapped out by some physical methods. But in doing so, the theory has to answer more serious questions in the theory of mind, as it might have to admit that humans are nothing more than meat computers. Also, possible-world semantic is a popular modification to traditional truth-semantic; it utilizes counterfactuals and set theory to explain moral sentences. However, this approach again falls into naturalism; it translates perspective sentences into indicative sentences about different set of possible worlds. In doing so, the theory tries to assign truth condition to perspective sentences, but it again ignores the intentional aspects of the speaker.

Philosophers who are disinclined to diminish moral sentences all together usually say they express certain emotions, and there is nothing true or false about them. For example, John Searle uses speech acts to explain semantics [5]. These philosophers are partially correct, moral sentences are about intentional attitudes of the speaker, but they indeed have meaning. When someone says the imperative “do not jump out of the window”, the hearer understands what the sentence means by showing assent, which is not jumping out of the window. Whenever a speaker says something, she wants to communicate and this is also the starting point of truth- conditional semantic. The speaker says a sentence, and the hearer can either assent of dissent. But truth condition is too narrow to really appreciate the art of communication. Wittgenstein in his later years, in “Philosophical Investigations”, explains language as playing games, someone understands what a sentence means when he knows how to play the language game with that sentence [6]. It is ridiculous to play chess with tennis rules, and it is equally ridiculous to play moral language game with a set of rules designed for physical language game. It is a common affliction among philosophers to simplify and generalize each phenomenon they encounter, but such generalization is not always correct.

It is a mistake to believe that there exists a set of rules that gives answers to every linguistic practice; one of Wittgenstein’s wisdoms is that understanding is indetermined by rules. Professor Brandom interprets it as an anti-normativism position. He states that if our practices have to follow a set of rules, then a set of meta-rules would also have to be presupposed for following the rules, and so on so forth. Brandom does not believe semantic is to be reduced to subjective whims; what matters is that an implicit understanding comes before any explicit rules [7]. He and the pragmaticist tradition believe that formal truth comes second to the implicit understanding in linguistic practices, and it is usually referred as “use theory” as a counterpart to truth-conditional semantic.

However, it is worth noticing that use theory is not untouchable, as one can always ask what the implicit understanding is. Professor Brandom explains it as speaker and hearer playing checkboard games, as the conversation goes on, the participants make material inferences and give each other scores to see if this conversation is successful. However, the problem remains, as it remains ambiguous on what criteria the participants give scores. Some examples might help us understand. Imagine Christ and I are chatting, and I say, “It is raining outside.”, Christ goes to the window and looks out, then says, “yes, indeed.”; and I might continue saying “we should stay home, the rain is not going to stop any time soon”, and Christ might say “but I really want to watch the movie in Bruno theater”, then I will, if he really has to watch that movie tonight, and the conversation goes on; during the conversation what Christ and I do behind our utterance is seeking and giving reasons, if the reasons for a sentence are good, we assent, otherwise, we dissent. For pragmaticism, although the speaker and hearer have to seek and give reason based on some criteria, it is good enough that they know how to apply their reasoning in actual linguistic practices. The criteria of good reasoning, even if they for now remain implicit, are not subjective and can be objectively understood. However, giving and asking for reasons is too broad to function as a guideline of semantic understanding; when inferring, not only should the speaker and hearer ask and give reasons, they should ask and give good reasons, and the appreciation of good reasoning is associated with the conception of truth and falsehood. Assigning giving and asking for reasons as what gives sentences meaning, on the one hand, leaves semantic understanding still in shadow, on the other hand, undermines one of the central interests of philosophy—asking for truth.

Ernst Tugendhat, also influenced by Wittgenstein, has a very similar approach to semantic. Although he does not go down all the way to pragmaticism, he states that to explain the meaning of a sentence is to show the open move of a language game; in other words, he also interprets language as playing a game, and to explain the semantic is to show the opponents when the game starts—namely showing its verification rule [8]. For example, if someone says, “It is raining outside”, the hearer can either assent or dissent, to explain the meaning of my sentence is to tell the hearer “The game starts, now it is your move, go check the window outside to see if it is raining”. If the speaker says, “Monalisa is a good painting” the verification rule might become something such as “look at Monalisa and see if it is a good painting to you”. The verification rules can change in different circumstances, and it is related to how familiar the participants are with the language game. Of course, neither Brandom or Tugendhat gives a straight answer to how semantics is actually understood; in Wittgenstein’s words, we can only explain a thing by showing what it is, not as something else. Nevertheless, truth conditions are too narrow to be universally applied to every semantic situation. It is unfair to treat moral sentences and physical, descriptive sentences the same.

4. Possible Solutions

Though the article has discussed some negative accounts for semantic of moral sentences—it is indetermined by truth condition, a positive explanation of moral sentences remains in question. However, it is a very difficult process; one of the reasons is that moral sentences as well as other perspective/intensional sentences are heavily related to intentionality. In order to give a sophisticated explanation, a thorough understanding of mental states is required. Usually, language and mind are seen as two sides of the same coin, one cannot be fully explained without the other. To explain moral sentences, we have to consider explanations for desire, choice, wish and other intentional attitudes. The holistic relationship of mental states, as stated by Davidson, makes it nearly impossible to explain how our mind comes to be [9]. We are able to describe our mind, either through mental languages or physical languages, but it is extremely difficult to nail down how and why it happens at the first place. The fine details of the debates in the philosophy of mind are not this article’s concern at the moment, it is sufficient to say, however, that it is also difficult to find the “source” and “essence” of moral semantic. Given the previous discussion, one of the possible candidates is phenomenology. Its inventor, Husserl, adopted his teacher Brentano’s intentionality marvelously into phenomenology, and Heidegger, as Husserl’s successor talks heavily on the ontological state of human beings as well as how we live in terms of our intentional relationship toward our surroundings. It is possible that some phenomenological insights can help us better understand the process of our intentional attitudes which are intertwined with moral sentences. However, a phenomenological approach might result in a form of transcendentalist account of moral semantic. The metaphysical orientation of transcendentalism must be treated with care, a moral semantic theory can on the one hand accept phenomenology’s insight about a meaningful world, on the other hand, however, it should not accept metaphysics as foundation of semantics. It is better to understand metaphysical concept (including moral concepts) in language, not the other way around. Another candidate could be virtue theory, as Sharpe and Schwartz state, what matters for moral decisions and moral semantics is the practical wisdom a person make when being in his circumstance [10]. The practical wisdom from the ancient time is not taught, the old generation could only teach the young principles and give them the opportunities to practice on their own; in other words, one can only learn the wisdom through self-perfection, and this might render the entire discussion, from a contemporary perspective, quite mystical. A reconciliation of virtue theory and phenomenology can be the third candidate, Heidegger’s ontological description of Dasein can be nicely transferred into the structure of virtue theory, as the semantics of moral sentences can be explained through speaker’s choices, which can be further explained in terms of the speaker’s relation to the world. But how to clean up the metaphysical orientation and how to express the speaker’s choice more explicitly demand further work.

5. Conclusion

Although the article denies naturalistic account of Ethics and truth-conditional semantics in terms of moral sentences, a clear explanation of moral semantics remains to be clarified; one paper is hardly enough to discuss the matter thoroughly. Some possible candidates of further research have been noted, but none of them can articulate the delicacy of moral language without intense investigation and modification. The spirit of this article is to clean up some misconceptions concerning moral semantics, and in so doing to broaden new horizons of moral philosophy. It is my hope that philosophers can see Ethics with humane respect and clarify the ambiguity.