Compositional Semantics and Normative ‘Ought’

According to the paradigm view in linguistics and philosophical semantics, it is lexical semantics (LS) plus the principle of compositionality (PC) that allows us to compute the meaning of an arbitrary sentence. The job of LS is to assign meaning to individual expressions, whereas PC says how to combine these individual meanings into larger ones. In this paper I argue that the pair LS + PC fails to account for the discourse-relevant meaning of normative ‘ought’. If my hypothesis is tenable, then the failure of LS + CS extends to normative language in general. The reason I offer that this is so is that semantics for normative language is, in an important respect, a substantive semantics (SS). The ‘substantive’ in question means that the meaning of normative vocabulary in use is driven by metanormative views associated with a particular normative concept. SS rejects the model LS + CS and replaces it with a discourse-relevant semantics built around an interactional principle that ascribes to a particular surface syntactical form of ‘ought’ sentences a logical form that represents its discourse-salient normative content. In the paper I shall sketch how SS works and why it is worth serious consideration.


Introduction
It is a platitudinal view in linguistics and philosophical semantics that in order to compute the meaning of a sentence like 'The cat is on the mat', you must know the meanings of all the lexical items in it and recognize how the arrangement of these items contributes to the sentence's overall meaning (Janssen 1997 Westerståhl 2010a, b). So if you understand the meaning of the quoted sentence as saying that the cat is on the mat, this is said to be so because you made use of lexical semantics conjoined with the principle of compositionality. The duo LS plus CS has proven effective in interpreting the content of the sentence in question. The problem is that LS ? CS is not a universally reliable principle of interpretation of linguistic entities. It fails badly when applied to normative language. This is so, I contend, because normative vocabulary does not have substantive and discourse-relevant meaning on its own. In this paper I shall present my criticism of LS ? CS, building my case on an analysis of the paradigmatic normative word 'ought'. 'Ought' construed as an auxiliary verb has a well-elaborated semantics, under which it is treated as an ordinary modal. However, the normative meaning extractable from the modal semantics of 'ought' does not capture the meaning that this word has in everyday communicative exchanges. Things are even worse, I claim, as the normrelative sense of 'ought' is not the sense that speakers have in mind when involved in conversations. Even more intriguing is the observation that the sense which ordinary speakers assign to 'ought' in that word's normative use is in no way inspired by the context of conversational utterances. It seems that whatever meaning they associate with normative use of 'ought', that sense is gleaned neither from the lexicon, nor from the context. And if that is so, a couple of questions arises out of which the central one is this: from where does this, call it, discourse-relevant meaning of normative 'ought' come into play? Another question, no less important, is: how can this default meaning be accounted for by a theory of meaning? By ''discourse-relevant meaning of normative 'ought''' I construe the sense we attribute to the word 'ought' when this word is used in a particular grammar-in sentences of the form 'A ought to phi' in everyday communicative exchanges, and which is badly, if at all, explained by the lexical sense of 'ought'.
The paper is organized as follows. In Sect. 1, I briefly explain why compositional semantics for 'ought' fails to accommodate what I call the 'discourse-relevant sense of 'ought''. I also discuss the two most promising, to my mind, ways of 'extracting' this discourse-relevant sense of normative 'ought' from the standard compositional semantics for 'ought' and indicate why, appearances notwithstanding, these attempts are not successful. On my diagnosis the general drawback of the considered proposals has to do with conceptions, or pictures of normativity that underlie the normative meaning of 'ought' construed as a modal. I claim that as long as we assume a particular picture of normativity, here: norm-relativity, that picture will inform the semantic meaning attributable to normative expressions as such. The problem, though, is that norm-relativity accounts badly for the everyday meaning of paradigmatically normative expressions like 'ought'. And if there is a discrepancy between what a reputable theory of meaning for normative 'ought' says and how ordinary language users understand that word's normative sense, that divergence requires an explanation. It also requires, I claim, accommodation within a theory of meaning. In Sect. 2, I introduce the notion of discourse-concept and argue that normative ought as our workaday concept is a discourse-concept (DC). 1 Discourse-concepts, I stipulate, are concepts that the discourse participant posits to exist in order to get a grasp of their mental representations in the service of communicative purposes. 2 The basic idea that the notion of DC concept tries to capture is that some kind of our mental representations is responsive to the exigencies of the contextual situation. Differently put, we produce some mental representations only in relation to a particular sort of context that triggers thoughts of very concrete content which we would not probably have had, were we not systematically exposed to a particular type of experience. At least this is the view that motivates my introducing the notion of DC concept of normative ought. I will argue that the discourse participant acquires the DC concept of normative ought on the basis of her communicative experience with use of the word 'ought' whose lexical meaning badly reflects the normative sense of her own utterances as well as the sense of utterances of other discourse participants. Generally put, the notion of discourse concept is to give voice to the view that mental representations need not be fixed and immune to pragmatic considerations. Quite the opposite, as I will show, seems to be true, namely that some of our mental representations are flexible and are subject to change, depending on the context of thought that determines the purpose of use of the concept in question. The notion of discourse concept becomes useful in the situation in which alluding to a familiar concept standardly recalled as relevant, here the concept of generic ought, fails badly to cover for the key connotations associated with use of 'ought' in sentences having a particular grammatical property, i.e. sentences in which 'ought' is assigned to the grammatical subject. What is special about discourse concepts is that they are doxastic based, or theory-driven, depending on who is interested in them. They are doxastic-driven when approached from the perspective of an ordinary discourse participant and they are theory-driven when approached by a theoretician interested in modelling mechanisms of verbal comprehension. 3 By saying that DCs are ''doxastic-based'' I mean the idea that semantic content of DCs is essentially informed by some substantive views, typically of pre-theoretical character, regarding two sorts of things: (1) the nature of the phenomenon the posited DC is a mental representation of and (2) the purpose of Footnote 1 continued affinity between my approach and Kamp's resides in the motivation for developing a discourse-friendly account of meaning of certain expressions. Closer examination of our approaches is an interesting topic that I leave for another occasion. 2 The Referee in his/her comments asked me whether there are any other discourse concepts that prove themselves to be helpful in explaining other types of languages game? It is my conjecture that there are some, but unfortunately, I do not have concrete examples to support that opinion of mine. The only thing that I can say at the moment is to offer a methodological guide, which I would follow when trying to answer that question. The guide that I have in mind is to focus on those concepts that we find particularly instructive in giving accounts of various language games and which are suspected of ambiguity. In my tentative view ambiguity may prove a reliable test of insufficiency of extant concepts, and consequently be a prima facie reason for a search of more explanatory adequate discourse concepts. I leave this interesting issue raised by the Referee for another occasion. use of the kind of concept in question. 4 Doxastic ''pedigree'' of the DC of normative ought, I take it, reveals in that we assign it a particular content as a sort of answer to the question about what is special about the normativity that fuels our use of normative vocabulary under a particular grammatical guise in ordinary communicative exchanges. In Sect. 2 I offer a story of how we come to have possession of DC of normative ought, and importantly how we know that a token of a sentence, or an utterance expresses the DC of normative ought and not just the 'familiar' concept of ought. The hypothesis that I advance says that these two concepts differ in terms of the logical form of sentences that express them, and that the stipulated difference in logical forms is to be reflected in their different discourse-relevant grammatical interpretations. Section 3 builds on what has been established in previous sections and proposes a rule for interpreting the normative content of 'ought' sentences. The rule in question is a rule in theory of logical and grammatical interpretation. 5 I deem it 'substantive' because it promotes a twofold substantive view: (1) a substantive view on the source of the meaning of DS normative ought which is doxastic-based, or metanormative theory-driven and (2) a substantive view on the bearer of the relevant normative content which is not the word 'ought', but rather that word construed as situated in a particular syntactical structure. The rule under consideration is interactional because it describes a kind of interconnection between a particular surface syntactical form of 'ought' sentences and a logical form that represents its discourse-salient normative content. 6 Importantly though I call it the ''semantic rule'' for DS normative 'ought' because it does inform us what syntactical arrangement gives the discourse-relevant normative meaning. I consider my proposal as inaugurating a new direction within the research on the meaning of normative language that render syntax interpreted in light of discourse-friendly account of the meaning of the workaday concept of normative ought the bearer of normative content. In Closing Remarks I sum up what I hope to have achieved in the paper and explain what revolutionary idea in the theory of meaning substantive semantics renders attractive.
2 What is Lost in the Semantics of 'Ought'?
Consider the sentence 'Peter ought to brush his teeth at least twice a day'. The sentence in question is uncontroversially taken to have normative content. How do we know that? That is easy. Occurrence of the word 'ought' and the grammar of the 4 The idea of DC inspires the entire programme of how to account for the meaning of workaday concept of normative ought that I call ''substantive semantics'' (SS) and develop in my book (in progress). 5 I am grateful to the Referee whose insightful comments convinced me that I mistakenly thought of the proposed principle in terms of ''substantive principle of compositionality'', whereas it is more natural to think of it as of a ''normal'' rule, i.e. a rule that describes meaning-making property of syntax interpreted under the guidance of a particular substantive view about the nature of normativity expressed by utterances of sentences with agential grammar. 6 Again I am grateful to the Referee for suggesting the name ''interactional'' for the name of my principle, which made me seriously reconsider my original proposal, where I mistakenly, as I think of it now, considered the discussed principle as substantive principle of compositionality applicable to normative vocabulary. sentence-'S ought to phi' clearly suggests that. 'Ought' as the dictionary says is about ''what is the right thing to do'' (Oxford Advanced Learner's Dictionary, 7th edition), indexing ought to the grammatical subject of the sentence indicates that 'the right thing to do' which the sentence is about is addressed to Peter. Taking into account the lexical meaning of 'ought' and the structure of the sentence we come to understand the sentence's sense to be as follows: 'Brushing his teeth at least twice times a day is the right thing for Peter to do'. So far so good. We have managed to understand what is said by the sentence under consideration. Following that method, we may think, we can easily come to understand an arbitrary token of a sentence of the form 'S ought to phi'. But that assumption is false. Paying attention to what the lexicon says and what grammar suggests is deeply misleading in the case of some sentences of the discussed form. Take the sentence 'Larry ought to win the lottery' or 'Kate ought to get a promotion'. Both these sentences assign ought to the agent, Larry and Kate respectively, but neither of these sentences says what is the right thing for Larry or for Kate to do for the simple reason that the agents in question are incapable of bringing about the relevant states of affairs. Neither Larry nor Kate can be sensibly said to have influence either on the winning of the lottery, or on the getting of the promotion. The considered sentences pretend to have agentive content, whereas in fact they do not, but this trickery of grammar is explained by the fact that auxiliary 'ought' is completed by the verbal construction that has passive content. Finally, winning the lottery or getting a promotion is not in the hands of the person who is the subject of the respective actions. The lesson to be derived at this stage is that the fact that lexicon and grammar may lead us astray as to the sentence's real content is not yet a mark that LS ? CS for normative 'ought' falls flat. The real threat to the principle of compositionality when applied to normative language is that it is badly ineffective in delivering us the content that we ordinarily assign to the sentences of the form 'S ought to phi', where these sentences are taken to have obviously agentive content. However in Sect. 2 I will briefly argue for the non-intuitive claim, saying that the surface syntax of agential 'ought' sentences is generally indicative of the normative sense expressed by the sentence: norm-relative or authored-practical in my terminology. In a nutshell, according to the proposed view, originally suggested in my Klimczyk (2017) paper even sentences like the Larry-sentence are to be interpreted following the matrix of interpretation for DSought, that is, as expressing normative-authored content to some degree. What finally decides that we read the sort of mentioned agential 'ought' sentences as having non-agentive content is, on my proposal, explained by the fact of these subjects' having too little authored contribution into bringing about the respective states of affairs to count them to be full-fledged authors of actions under demand.
In this paper I argue for the view that we need a DS concept of normative ought because the standard semantics for 'ought' does not deliver the particular agentive reading of sentences of the form 'A ought to phi' (henceforth A-sentences for short), which is the reading on which A herself is to be the sole doer of the demanded action. Call the normative sense that we think the A-sentence expresses 'authored practicality'. 7 The name makes it explicit that the intuitive sense of 'ought' is the one about the requirement that the addressee be also, insomuch as it is possible, the only producer of the required action. I stipulate that authored practicality is the sort of normativity associated with 'ought to do' constructions as employed in ordinary conversations. Crudely, authored practicality, on my proposal, is the discourserelevant sense of normative 'ought'. As many authors have already observed (McNamara 2004;Broome 2013;Chrisman 2016;Finlay 2014, to mention a few prominent names in the discussion) what I have dubbed 'authored practicality' is poorly extractable from a semantics that treats 'ought' as a necessity modal. Now, one proviso is in order. The worry I am concerned with in the paper is not that modal semantics lacks formal resources to accommodate authored practicality. It certainly is capable of delivering a logical interpretation that meets the challenge in question, but the way it does so remains unsatisfactory. It is unsatisfactory either because the interpretation stumbles on the problem of ambiguity, 8 or in order to be convincing it requires a philosophical underpinning: we need to be provided with a plausible story about how it is that a norm qua norm is ever capable of having straightforward practical consequences of an authored fashion. If 'ought' sentences are primarily about what is right/good/correct in light of a norm, how does this norm-relative picture of normativity have a grasp on our behaviour? The problem that modal semantics encounters with accounting for the DR-normative meaning of 'ought' is that it is built on a picture of normativity that is agent-neutral. Normrelativity is about what ought to be done in the first place and not about personal requirement that in its normative force is weaker than personal obligation. 9 Differently put, according to a standard modal interpretation of the Peter-sentence, we retrieve the following meaning of it: Peter's brushing his teeth at least twice a day is good/correct/recommendable for Peter to do in light of some norm, principle or regulation. However, thinking in terms of accordance to norms is not what guides our ordinary use of 'ought'. When we communicate with people using 'ought' sentences with agentive content, we usually transmit to them a concrete message, not report about the norm-fittingness of the action mentioned in the sentence. Accordance to norm may background our utterance of a particular claim, but rarely is that claim's normative content. If communication as such is mainly about passing information, then normative utterances are also expected to pass some information. I stipulate that the message borne by sentences like our toy one is about what the subject of 'ought' is required to do herself. However, that very message is not readable from the sentence itself.
More disappointing is the fact that the stipulated authored practical sense of 'ought' as in the Peter-sentence is not obviously inferable from the utterance situated in a particular conversational context. Suppose we are discussing Peter's extravagant approach to life. Peter is our colleague at work. Suppose next that for 8 See Schroeder (2011) and my Klimczyk (2017). 9 Modal semantics fares well with accounting for personal obligation but it is too coarse-grained to discriminate between personal obligation and personal requirement (or 'ought to do'). The difference between the two resides in the strength of the normative commitment. Obligations are inescapable, and failure to satisfy them, by warrant, issues sanctions, whereas most 'ought to dos' are requirements due to the agent's previous commitments, values, or concerns. An interesting interpretation of 'ought to do' that tries to escape the mentioned difficulty is given by McNamara (2004). reasons unknown to us Peter hates brushing his teeth and, as he avowed to us one day, he limits his brushing routine to one time a day. Knowing Peter's intention to live long, we are surprised by his behaviour. Since one of us recently paid a visit to her dentist and was reminded that oral hygiene is extremely important to maintain good health, Peter's odd habit became the topic of our conversation. Imagine that I uttered the Peter-sentence. I think that my description of the situation in which we are leaves no hint as to the discourse-relevant sense of 'ought'. From nothing that I said in the context in which we are situated does it follow that what I have said was exactly that Peter himself is required to do the brushing the respective number of times. The only information straightforwardly delivered by my utterance is that Peter ought to make it the case that he brushes his teeth more than once. But that interpretation makes room for a reading, on which Peter himself is the author of brushing his teeth, and a reading on which someone else is. You may oppose that latter interpretation by pointing to the fact that Peter is our colleague, which implies that Peter is an adult, and it is normal to expect from adult people that they themselves take care of their oral hygiene. The suggestion you made is sensible, but not relevant. Note that the inference about Peter's maturity does not stem from the context of our conversation but from the intension of the concept ADULT that is part of your default conceptual repository. But if you base your understanding of the normative content of the Peter-sentence on the basis of sensible, but contextually unjustified material, 10 knowledge of the circumstances of the conversation does not contribute to the sentence's meaning.
Authors who agree that the practical sense of 'ought' is badly exhibited by semantics for modal 'ought' have attempted to find a way to 'instil' the 'ought-todo' sense into standard compositional semantics for modal 'ought'. One interesting proposal, recently advocated by Matthew Chrisman, was to focus on the central hallmark of 'ought to do' and next figure out how to nicely interpret that hallmark using the tools of possible world semantics. Roughly, according to Chrisman what is special about practical ought standardly expressed by sentences of the form 'S ought to phi' is its imperatival flavour. On this proposal, the Peter-sentence is given agentive interpretation, on which Peter is called to brush his teeth the relevant number of times because we interpret the sentence's normative content as an imperative addressed to Peter to the effect: Peter, brush your teeth… Ingeniously elaborating on Castañeda's original conception of practicion-non-propositional, imperative-like content graphically represented as Peter, (to) brush his teeth, Chrisman shows how this imperatival content is to be successfully accommodable by modal semantics. However much I appreciate the ingeniousness of his suggestion, overall I find his proposal unconvincing. I have two main worries about it. One worry is this: if the practical sense of 'ought' has to do with selfexecutorship of the demanded action, how this sense is to be extracted from the norm-relative interpretation of normativity assumed in the semantics of 'ought'? As has been famously observed, from the fact that some norm is in force, nothing yet follows to the effect that a particular agent is required to conform to that norm. Sill less does it follow that she is required in order-like fashion to be the sole executor of her conformance to that norm. If semantics for practicions is meant to be a standard compositional semantics, as he wants it to be, then we remain within the problematic norm-relative picture of normativity that badly fails to capture unassisted agency. This is so because the semantic rule for 'ought to do' that he proposes is treated as a sentential operator on non-propositional content that gives the semantic value true if in the top-ranked worlds the top-ranked norms (my emphasis) call for satisfaction of the embedded practicion. 11 Another worry is that it is not clear why we would think that imperatives capture the essence of practical normativity. One reason for thinking that they do, often alluded to in the literature, is that practical oughts are mainly about advice, and since it is in the nature of advice to say what is to be done, it may seem natural to assume that the person to whom the advice in question is addressed is obviously to be the doer of the advised action. But what looks 'natural' often masks something questionable. The fact that the advice is directed to YOU does not automatically imply that you yourself are to be the only person who is to realize what the advice prescribes. The only valid inference that you can make on the basis of the information that the Peter-sentence is a piece of advice is that it would be good/ sensible if Peter behaved as stipulated. But again, from saying what would be sensible for Peter to do, nothing straightforwardly follows to the effect that Peter himself (without the assistance of other agents) should be the only producer of the respective action. Differently put, my main criticism of Chrisman's proposal runs as follows: even if we account for the agentive sense of 'ought to do' in terms of its imperatival character, two things still remain to be argued for: (1) why we take imperatives to be read as issuing orders that the addressees of these orders be the agents who themselves satisfy them, and (2) why would we think that the imperatival interpretation of 'ought to do' is better than the ordinary modal interpretation if, finally, both interpretations are based on the worlds-ranking organized around the top-ranked norms which means that normativity is still about the conformance of actions to norms, and not about a particular mode of these actions' performance-in the authored fashion.
Another interesting proposal of how to expose the practical sense of 'ought to do' within the standard modal semantics for 'ought' has been delivered by Finlay and Snedegar. 12 The core of their proposal is that we construe practicality in terms of optionality of actions to be pursued and the relevant agency is settled by the context. According to Finlay and Snedegar's proposal what I called 'authored practicality' is to be retrieved by the following interpretation of the Peter-sentence: what decides that brushing the teeth is what Peter ought to do is that we consider the proposition that Peter brushes his teeth… as part of Peter's deliberative set. Note that the proposal is really ingenious and when it comes to successful accommodation of authored practicality it seems to have an obvious advantage over Chrisman's proposal. The asset I am talking about is that considering the relevant proposition as 11 See Chrisman (2016) falling within an option within the deliberative set of S makes it as clear as possible that the respective course of action be realizable by S herself. Normally we do not construe deliberative sets for agents that are not us! However, at the end of the day, the contrastivist semantics for 'ought to do' that Finlay and Snedegar advocate suffers from a variant of the same flaw that I have detected in Chrisman's proposal, which cannot be surprising if both proposals are compatible with standard compositional semantics for the modal word 'ought'. The drawback in question is that these two accounts try hard to fit their interpretation of practicality associated with 'ought-to-do' into the norm-relative picture of normativity that underlies standard compositional semantics for 'ought'. In the case of Finlay and Snedegar's proposal, this is done subtly because they concentrate on setting out the deliberative nature of 'ought to do'. Deliberation brings to mind the first-person perspective, thence nicely serves to exhibit the first-person's contribution to satisfying the ought in question. However, deliberativeness of 'ought' interpreted as proposed in contrastivist semantics for 'ought' gives results different from the intuitive ones. What actually gets represented is not so much the deliberator's sole engagement in fulfilling the required action but an optionality of courses of actions ranked against the top-ranked norms of action available in a particular agent's deliberative set. The upshot is that contrastivism manages to make explicit the optional nature of 'ought to do', and not its strictly addressed character that puts constraint on the proper fulfilment of the requirement-in an unassisted way. Consequently, it does not account for the key hallmark of normativity being associated with the concept of 'ought to do' by ordinary people. I think that it is beyond question that ordinary users of English who produce claims like the Peter-sentence do not do so primarily with the intention of exhibiting the optional status of the required action, but rather with the intention to say something specific about how that required action is to be executed by the person to whom ought is indexed.
In the next section, I shall motivate the view that our workaday normative ought concept is not the familiar concept of ought interpretable within the framework of modal semantics, but rather a discourse, default concept of normative ought with which we associate particular practical content. The content under consideration is best construed in terms of authored practicality and we assign it to the concept of ought to do because we think of actions as of intrinsic properties of agents. The sentence 'Kate runs' intuitively is about Kate's running and not about her being made to run by her coach which might be the case in some particular circumstances. The way we think of actions, I contend, inspires our way of thinking about ought to do in terms of strictly personal requirements. Ought to do is a normative property of an agent to whom it is addressed. Accounting for the nature of practical ought in terms of the agent's property makes it as plain as possible that the action to be done is to be done by the agent to whom the normative property in question belongs.

Normative Ought as a Discourse Concept and What it Means
In previous sections I have suggested that the proper interpretation of the Petersentence is that Peter himself is to be the sole producer of brushing his teeth. Differently put, I stipulated that the Peter-sentence is to be read as follows: Peter ought that Peter is such that Peter does the brushing the respective number of times. More formally put, the interpretation of the sentence in question is this: PO (PF). I claimed that my interpretation of the Peter-sentence should be considered as a paradigmatic interpretation of sentences purported to express practical normativity because it corresponds to the way in which ordinary speakers usually interpret sentences of the form 'S ought to phi'. When an ordinary user of English hears the statement 'Peter ought to brush his teeth', barring further context, she does not automatically interpret the statement in question as an imperative addressed to Peter, but rather as a description of the requirement addressed to Peter that he himself will perform the relevant course of action. In other words, I argued that normative content of agential 'ought' sentences is essentially descriptive, and thence propositional. It says what some agent is required to do and specifies the relevant manner of satisfaction of that requirement: lack of contribution of other agents. Importantly, I stressed that the normative and discourse-relevant sense of 'ought' is not to be couched in terms of norm-relevance. When we produce normative agential 'ought' sentences in conversation, we usually state what is to be done and by whom, and not that some action is permitted by some norm, rule or principle. However, that agent-centric information is not adequately delivered by a semantic analysis of 'ought' construed as a modal word. Since modal semantics falls flat in recovering the intuitive meaning of 'ought' sentences with agentive content, we need some other theory of meaning that could fill the gap and make sense of this interpretative phenomenon. My proposal is substantive semantics for normative language (SS).
The central assumption behind SS is that lexical semantics for normative vocabulary is an ineffective guide to that vocabulary's actual, discourse-relevant meaning. The name 'discourse-relevant meaning' is tricky because it can suggest two things. One is that the meaning we are interested in is the meaning that the expression in question has as it is used in ordinary communicative practices. And this is the interpretation of the name that SS assumes. The other suggestion given by the name is that the source of the meaning of that expression is given by the discourse. The latter interpretation of 'discourse-relevance' is radically contextualist and SS rejects it. The reason for that is that according to SS the discourse-relevant meaning of normative 'ought' is the meaning that a competent language user already has prior to entering communicative exchanges because she already has some pre-theoretical beliefs about what kind of content is to be associated with the concept of normative ought linguistically expressed by sentences of the form 'S ought to phi''. Now, two things require explanation. One is: what make the content of normative ought that I deem our workaday normative concept have descriptive content in the sense that it states what should be done, by whom and in what fashion? The other is: what speaks in favour of the idea that normative concepts are essentially driven by doxastic states of agents? My answer to the first question is built on a substantive view about the nature of communication. According to the view that I favour, but which I will not motivate in this paper due to the limits of space, communication is mainly about passing information. If I say to my interlocutor that the cat is on the mat, I do not only say how things are in our close vicinity, but notably pass her information that the cat is on the mat. An analogous situation takes place when we produce normative utterances. When I produce the Peter-sentence in a conversation, I may mean different things by it (e.g., expressing my frustration about Peter's habits, giving advice to him, etc.), but at the most basic level what I'm doing is sending a particular message to my hearers. The message in question is the information about what is required of Peter himself to do. How do we know that this is the information transmitted by the sentence? Again, on SS, this is so because we associate particular content with a particular normative concept, being guided by the grammar in which the content is interpreted. If normative content is borne by agential 'ought' sentence, barring special circumstances, there is good reason to think that the sentence says what S, to whom 'ought' is grammatically indexed, ought to do herself. Since communicative practice shows that people more often than not treat grammar as a guide to the proper interpretation of the meaning of their utterances, SS takes this fact as data to be accounted for in the theory it proposes. The explanation it offers says that, if speakers employ sentences like the Peter-sentence, they do it to convey some information about Peter and his relation to the ought in question. Grammatical structure may be misleading, as shown in Sect. 1, but not entirely deceptive. On SS, the normative content of an 'ought' sentence cannot clash with the content suggested by the grammar because ultimately substantive normative content is obtainable from the proper interpretation of the sentence's grammar. This observation applied to our toy sentence says what follows: whatever is the discourse-relevant sense of 'Peter ought to brush his teeth at least twice a day', that sense must say something about Peter's connection to the ought in question. Importantly though, the discourse-relevant normative sense of an 'ought' sentence is not derived from the context of conversation, but is assigned to it prior to an occurrence of any discourse situation. This is so, because SS stipulates that language users have a concept of practical ought normally expressed by sentences in which ought is indexed to S. This concept I dubbed a 'discourse concept' because this is the concept of normative ought in play in ordinary communicative exchanges. As I have already argued, whenever people get engaged into a conversation in which they produce sentences like the Peter-sentence, they are not saying that some action that the sentence is about meets a certain standard, but rather that the grammatical subject of the sentence is required to be the doer of the action in question. A worry. 13 The proposed approach suggests that there are two kinds of normativity associated with the word 'ought'. One is the familiar one expressed by the generic concept of ought and usually explained in terms of norm-relativity and another one couched in terms of authored practicality and expressed by an artificial concept called ''DS normative ought''. Let us grant plausibility to that view, but then what requires explanation is how does it come that a language user systematically and unproblematically 'switches' between these two concepts? Answer. First, let me say that proper treatment of the worry requires an argument in theory of normative content, which I do in another paper of mine (in progress). My concise answer to the question about the mechanism of mental switch between ought-concepts goes as follows: it is automatic in the sense that it goes beyond conscious processes in our mind. I suspect, though that this hypothesis requires examination in experimental psychology, that a child who learns the meaning of 'ought' from the dictionary soon starts to use and interpret this word in sentences of the form 'A ought to phi', according to my stipulation in terms of authored-practical sense. But how does it happen that the child understands agential 'ought' sentences in the way suggested by my account? Here is my suggestion. In coming to understand the normative sense of an arbitrary agential 'ought' sentence she follows the surface syntax. And it is my conjecture that she does it in a naïve (but luckily mostly successful) way by which I mean that she connects semantic significance of a particular lexical unit with its place in the structure of the sentence. According to that naïve method, as I call it, importance of semantic contribution of a lexical item is connected with the place it occupies in the sentence. If an agential 'ought' sentences begins with a name or a description of a subject of 'ought', then she 'guesses' that the sentence's overall meaning is organized about that subject. But if sentences like the Peter-sentence are interpreted with the focus on the subject of 'ought' (here Peter), it shows obvious that alluding to the lexical sense of 'ought' is of poor help to understand the Peter-sentence properly. The meaning of the word 'ought' leaves no hint as to about whom, if anybody at all, an agential 'ought' sentence is. To sum up, on my approach the language user qua discourse participant 'constructs' in an unconscious fashion the concept of 'ought to do' (DS concept of normative ought in my technical jargon) treating syntax to be the bearer of its relevant normative sense. The consequence is rejection 14 of LS ? CS in the role of operative guide to agential 'ought' utterances and replacing it by an interactional principle that assigns particular normative content-i.e. authored practicality with particular grammatical property of 'ought' sentences, which is indexing 'ought' to the grammatical subject. So far so good. But a related worry enters the stage. Worry number two. 15 If we grant the hypothesis that agential syntax is the bearer of the discourse-salient normative sense of agential 'ought' sentences, then we should be ready to accept the odd view that sentences like the Larry-sentence or the Katesentence have also authored-practical sense despite clear intuitions to the contrary. The worry is to the point, but I think controversially that we should not be afraid of accepting this oddly-sounding consequence. Here is my brief answer why. As I 14 As Referees sensibly observed what I am saying here does not demonstrate that LS ?CS falls flat in its role of content-determination (though I take sides with those researches who think that its role in the theory of meaning has been exaggerated), nevertheless it implies that at least some ancillary principle, like the interactional one that I offer, should be offered in order to accommodate the sense not captured straightforwardly by LS ?CS. The task of such an interactional principle is to deliver an instruction of how to recognize the normative sense of the expression 'ought to' and not to put into doubt the lexical sense of non-normative words occurring in the sentence and the importance of these words' syntactical arrangement. 15 Thanks to the Referee for raising this point. argued elsewhere 16 normative agential 'ought' sentences express a ''cloud of propositions'' (however my hypothesis extends to normative claims articulated with use of necessity modals in general) to borrow this nice phrase from Kai von Fintel and Anthony Gilles (2011) with different normative content, and what is especially important in the context of my discussion in this paper, this content includes some degree of authored-practical content. Very roughly speaking, the idea of cloud of propositions as I use it 17 goes in pair with the thesis of underdeterminacy of normative content and taken together they are to suggest normative multi contentness (NM): an arbitrary agential 'ought' sentence in principle expresses at least two normative propositions, where one of these propositions bears authored practical content to some degree. Applying NM to the Larry-sentence says that among available interpretations of this sentence's meaning is one, on which Larry is the sole author of the winning the lottery to some degree. Now, expression ''to some degree'' is crucial because it does not specify how much degree of authorship must be on Larry's part in order to consider him as a partial author of the respective action. Moreover and importantly, ''some'' may also in the relevant context mean ''too little'' to count. It is my view that the Larry-sentence justifies interpretation in accordance with the matrix for reading suggesting authored practicality, but because of Larry's final no control over the outcome of the lottery, the sentence under consideration is understood as having non-agentive content. However, in light of my view about underdeterminacy of normative content it is not entirely correct to say that Larry is in no sense the author of the winning of the lottery because to some degree and in an indirect way he is. He is the minimal author of the winning in the sense that the very possibility of his winning the lottery assumes at least minimal involvement of Larry into an action that is a precondition of his possible lucky fortune, which is undertaking some action to get the lottery ticket. Obviously, getting a lottery ticket does not make him the author of his winning the lottery, nevertheless is a mark of some authored activity on his part that necessarily contributes to probability of occurring the state of affairs in which his lottery ticket wins. And in light of my account argued for in my 2017 paper this is enough to assign to the Larry-sentence minimally authored-practical meaning. But such a minimally authored-practical interpretation at the end of the day shows a bad one since it misses out a great variety of ways in which Larry has no contribution to the outcome of the lottery. To sum up, my interpretation of the idea of ''cloud of propositions'' when applied to the Larry-sentence allows me to say that among multiple propositions that the speaker's utterance of the sentence in question expresses is also the elliptical one ''that Larry ought that p'' among whose interpretations is the general, template-proposition, so to speak, which is that Larry ought to make it possible that his lottery ticket will win. What makes me thing that such an elliptical proposition is included in the cloud of propositions this sentence expresses? The short answer is: syntax. When our thoughts are accurately translated into agential sentences, d-agentive (agentive to a degree) interpretation of the sentence's meaning is always in place. But proper argument for that claim would require a separate paper.
Before we move on, let us recapitulate what has already been said. Above I presented the key idea that fuels SS, which is this: semantic meaning of normative claims, here 'ought' sentences (by default) is substantive in the sense that this meaning is obtainable from somewhere other than ordinary compositional semantics. I argued that LS ? CS leads us to the normative meaning that usually we do not have in mind when we produce normative 'ought' utterances. If we tend to understand normative 'ought' claims differently from the fashion that our reputable semantic theories tell us that we do, this phenomenon requires explanation. The explanation that I propose is that this happens because normative concepts (contents) are informed by our pre-theoretical beliefs about what normativity associated with a particular concept and expressed by particular grammar is all about. Since the concept of ought to do is in English linguistically expressed by agential 'ought' sentences like the Peter-sentence, SS proposes to treat that linguistic fact as informative with respect to where to situate the essence of the respective normativity. Agential 'ought' sentences suggest agent-centric normative content. But that very content is not straightforwardly derivable from the concept of ought. The concept of ought has connotations with permissibility, which in turn is primarily associated with norm, rule or principle-relevance. Norms inform us what is required and sometimes also inform us what is required of whom, but they are silent about how the conformance to the norm in question is to be done. Precisely, norms (typically) lack constraint on executorship of the demanded action. In contrast, the concept of ought to do, construed as a DS concept is associated with personal producership of the required action. Now, if there are two conceptions of 'oughtness', one expressed by the concept of ought and the other by the concept of ought to do, 18 there is good reason to think that the unifying portrait of normativity is misguided and normativity is not a unity. SS holds that this is exactly the case. 'Normativity' is a theory-based concept: it depends on philosophical views regarding the nature of normativity and these views can vary from philosopher to philosopher. However, in order to prevent the increasing number of conceptionbased concepts of normativity appearing on the philosophical market, SS proposes that the number of admittable concepts of normativity be dictated by explanatory purposes.
Had it been so that the construal of normativity in terms of norm-relativity would satisfactorily account for all uses of normative vocabulary, there would be good reason to think that normativity is a unity (Wedgwood 2006(Wedgwood , 2007. But since there is enough empirical evidence to think to the contrary, provided by examples of our ordinary conversational exchanges, normativity is not to be considered as a uniform 18 I deny the dominant and, false to my mind, view that ''telling the agent what she ought to do'' in its genuinely practical sense can be accounted for in terms of ''norms of action' ' (cf. Cariani 2013;Dowell 2013;Chrisman 2012;Silk 2016 among others). Here I take sides with the non-normativists in the philosophy of language and mind, who hold that no action-implying consequences straightforwardly follow from the fact that some norm is correct or in force (cf. Glüer and Wikforss 2009;Hattiangadi 2007;Miller 2010). phenomenon. SS proposes distinguishing between two kinds of normativity: normnormativity and practical normativity. I speak in terms of ''kinds'' because normrelativity as such is not the source of personal requirements. Now, what is special about practical normativity is not only that it says what ought to be done instead of what is correct in light of a particular norm, but also that it stipulates the proper manner of executing this 'ought-to-do' by binding execution of the demand in question with the person to whom the demand is indexed. Importantly, the way 'practicality' is construed in SS differs essentially from how this technical term is standardly understood in philosophical literature. In philosophy, practicality has been traditionally construed in terms of motivational efficacy. A practical claim is one that has a grasp on the agent's psychological make-up, and therefore is able to incur in her the relevant action-oriented attitudes, notably intentions. On SS, practicality is a property of requirement that specifies the proper method of satisfying this requirement. Consequently, on the stipulated view, practicality is a property of content and is to be recovered from the discourse-salient logical form of the sentence. Sentences expressing different kinds of normativity have different logical forms. Sentences having norm-relative sense are logically interpreted to the effect O(p) and read as ''It ought to be (the case) that p'', whereas sentences expressing authored practicality (the DS normative sense in my terminology) have the logical form AO (A:[A F]), read in artificial and deviant grammar to the effect ''A ought that A is such that A phi-s''. Given that practicality is construed as a semantic phenomenon readable from the discourse-relevant logical interpretation of the sentence's meaning, it is clear that SS is a version of sentential semantics. The meaning of normative claims is encoded in the discourse-salient logical interpretation of the kind of normativity expressed by use of a particular normative word and is not determined by the lexical sense of the normative word in question. In other words, if we want to know the meaning of an occurrence of the Peter-sentence, we should focus on the grammar of that sentence, because the grammar is indicative of the kind of normativity expressed by the word 'ought'. If 'ought' is indexed to the subject of 'ought', there is a pro tanto reason to think that what is said by the sentence is that Peter and only he should do a particular action. However grammar can be misleading as there are a lot of agential 'ought' sentences that do not have agentive content. Recall the already discussed example 'Larry ought to win the lottery'. The Larry sentence and the Peter sentence have the same grammar, but obviously differ in their logical structure. The first does not ascribe ought to Larry, whereas the second does ascribe ought to Peter. Now the worry that SS must address is this: if grammar is to be indicative of the kind of normative content expressed by an occurrence of an 'ought' sentence, how does it happen that language users assign different kinds of normative content to sentences with the same grammar? An easy answer to that question would be that they recognize that the Larry sentence is not about Larry's standing in a normative relation to the winning of the lottery because this is what they learn from the meanings of the surrounding words. SS basically agrees with this answer, but proposes a more systematic one. The answer that SS puts forward is this: language users are not misled by the grammar of 'ought' sentences because they treat the 'official' grammar as an input to further interpretation, which they perform by assigning to the deciphered content what I call 'discourse-salient grammar representation' (DGR for short). DGR is an interpretation of the sentence's sense that is not straightforwardly given by the official grammar. Since English grammar for 'ought' admits of agential structure in case of sentences with obviously non-agentive content, but speakers rarely make mistakes as to the proper understanding of the relevant agential 'ought' sentences, it seems that they have a method of gleaning the relevant content from nondiscriminative grammar. SS holds that language users assign a unique DGR to sentences expressing either of the two normativities. That is, according to my proposal, 'ought' sentences that have practical content: say what the agent herself ought to do, are ascribed DGR to the following effect: DGR for DS normative 'ought': S ought to phi is read as follows: Ought to phi is ought FOR S that S does it. The DGR in question is an 'ungrammatical', yet the most accurate interpretation of the sentence's logical structure being SO (SF) and read as follows: S ought that S is such that S phi-s.
According to SS, the prepositional phrase ''for S'' is an interpretation of the relation of indexation between 'ought' and the grammatical subject of the sentence.
SS proposes to think of the ('official') grammar as of a vehicle through which we get to recognize the relevant normative content of an 'ought' sentence by testing how well it matches with our understanding of the sentence's meaning. Such a probing of grammar seems unavoidable because the grammar for 'ought' is not sufficiently fine-grained to discriminate between kind of normative content expressible by 'ought' sentences.
The suggestion that the preposition ''for'' is employed to mark that the respective ought is to be fulfilled by the person to whom the ought in question is indexed requires more empirical testing, but I think that treated as a tentative hypothesis it is not implausible. To support my conviction, let us consider two sentences picked out from an everyday context.
(A) 'Paul, pick up the phone. The call is for you.' (B) 'Mary, this report was left for you, so please do not involve other people into reading it'.
I think that we do not need to know the contexts in which occurrences of the sentences presented in (A) and (B) are situated in order to understand that the small word ''for'' is used to indicate that the persons who are the addressees of the utterances in question, are also expected to be the sole producers of the actions mentioned in the sentences expressed by the respective utterances. Very often though language users interpret non-normative content of utterances by use of the prepositional phrase 'for S', despite the fact that this phrase is not part of sentences expressed by the respective utterances. On my view they do this quite automatically because they associate broadly authored content with the phrase under consideration. I take it that it is clear that the sense of the utterance presented in (C) is that whatever is said in the message left on the table, it is for Tim and nobody else to be concerned about it. Analogously, I think it is obvious to interpret the sense of utterance under (D) as issuing a request for Sam that she as the boss will voice her opinion on the discussed matter.

Substantive Semantics and Compositionality
Suppose that the central view of SS is plausible, and that discourse-salient normative meaning is not a property of expression deemed to be normative, but a sentential property, encoded in the unique logical structure of 'ought' sentences meant to express the relevant kind of normative content. According to this view, when we interpret sentences like the Peter-sentence to the intuitive effect that S to whom ought is indexed is to be the sole doer of the demanded action, we do not follow the standard method prescribed by LS ? CS, but proceed otherwise. SS suggests that the way we proceed is not ad hoc but systematic: ordinary folk simply understand the sort of normative 'ought' sentences under discussion in the way SS proposes. A natural question that arises is this: how does this conformity in understanding happen if normative words, in light of SS's central assumption, are denied relevant meanings? Note that on the picture proposed by SS the interpretative situation looks as follows: we have an 'ought' sentence like 'Peter ought to brush his teeth', and we are supposed to know the lexical meanings of all the words that make up the content of the sentence in question with one important proviso being that 'ought' has some broadly normative content associated with what is to be done. However, interpreting the sentence's meaning under the guide of the lexical meaning for 'ought' is normally ineffective. So what does SS propose? Let us repeat, SS starts from the idea that language users have some pre-theoretical grasp of the concept of ought to do, and this particular grasp of the concept in question is built on some substantive view regarding the essence of being a human that manifests in agency. If being a human is notably about being an agent, that view permeates the way we think about our relationship with the world outside us. Whatever norms we learn to conform to in the social world of which we are part, we treat as guides directed to us in our capacity to act and take responsibility for our own actions. SS distinguishes between the norm-relative underpinning of semantics for normative vocabulary and the semantic meaning this vocabulary exhibits when in use. Discourse-relevant meaning of normative words like 'ought' is the product of our view about the kind of normative content assigned to sentences of agential logical structure. Grammar is a vehicle through which we reach the relevant meaning of normative 'ought' sentence. Since according to SS, the grammar of the 'ought' sentence cannot clash with the sentence's final meaning, SS considers the agential structure of 'ought' sentence as standing in need of a guide to its proper, discourse-relevant meaning. The guide in question is provided by an interactional principle that ascribes to agential 'ought' sentence a particular logical form that represents its authored-practical normative content. Crucially, this interactional principle makes up for the metasemantic principle for DS-normative 'ought' (MPDSO for short). MPDSO says the following: the discourse-relevant normative meaning of an agential 'ought' sentence is encoded in the syntax posited to express the kind of normative content being associated with the use of the particular normative word in sentences of the relevant grammatical form. Following the guidance offered by MPDSO we can propose (tentatively) a semantic rule for DS-normative 'ought'. Here is how it comes in a rough approximation.
Semantic rule for DS normative 'ought': sentences of the general form 'A ought to phi' have discourse meaning set out as ''A ought that A is such that A phi-s'' iff the agent of 'ought' who is the grammatical subject of the 'ought' sentence stands in oughtness relation to an action, construed as personal 19 property of that agent.

Concluding Remarks
In the paper I have put forward a brief overview of SS construed as a novel candidate theory of meaning of normative language. I built my argument for SS on the critical analysis of standard compositional semantics for the modal word 'ought' in its role as a satisfactory tool for bringing up the discourse-relevant normative sense of that expression. The core of my argument for SS was this: since normative vocabulary does not have standing (i.e. lexical) and discourse-relevant normative meaning, CS applied to normative language usually fails. Consequently, I argued, LS ? CS must be replaced by some other proposal that is shown to be more explanatorily effective. My proposal was substantive semantics for normative 'ought' Considered as a novel candidate for a theory of meaning of normative language, it consists of two prongs: a theory of normativity associated with the use of a particular normative concept and a theory of logical and grammatical interpretation of the respective normativity at the level of sentences. The novelty of the proposed approach lies in reverting the core assumption behind the standard semantics, namely that lexical meaning is an inevitable part of any theory of meaning. If my considerations in this paper are on the right track, that view is no longer plausible. Sometimes, and this is the case with normative vocabulary, some words derive their semantic and relevant meanings from somewhere other than the lexicon, and other than the context. They acquire meanings on the basis of our doxastically-informed (substantive) views regarding normativity associated with the use of a particular normative concept. SS suggests that we recognize what substantive view underlies our use of a particular normative concept in the practice of the understanding of live language. The novelty of SS resides in its four basic assumptions: (1) lexical meaning of normative vocabulary is not the meaning we use in ordinary conversational practices; (2) discourse-relevant meaning of normative vocabulary is informed by substantive views regarding the nature of normativity associated with a particular normative concept and expressed by sentences of a particular grammatical form; (3) discourse representation of the normative content bears on the mental representation of the understood content; (4) semantics is not a neutral business as it has been usually thought to be.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creativecommons.org/licenses/by/4.0/.Funding This work was supported by Narodowe Centrum Nauki (Grant No. 2016/23//B/HS1/02921).