In this section I turn to conversational implicatures in fictional discourse. I will argue that audiences to fictions infer conversational implicatures by relying on assumptions about cooperation on the part of the narrator. We will see that conversational implicatures can be distinguished from what I will call contextual inferences. Both kinds rely on background assumptions that are often imported by default. Yet we will see that conversational implicatures are mandatorily imported but contextual inferences are not, while neither kind is mandatorily generated.
Conversational implicatures and Gricean cooperation
Consider again the example of (1)–(2).
As we noted earlier, in ordinary, non-fictional conversation the inference from (1) to (2) is a paradigmatic instance of conversational implicature. Accordingly, it is reasonable to expect that the mechanism by which (2) is inferred from (1) by an audience to a fictional story is at least analogous to how standard conversational implicatures are inferred. As I explain below, there are good reasons to think that this is so.
Here is Grice’s “general pattern for the working out of a conversational implicature,” that is, his schematic description of the way hearers infer implicatures:
He has said that p; there is no reason to suppose that he is not observing the maxims, or at least the Cooperative Principle; he could not be doing this unless he thought that q; he knows (and knows that I know that he knows) that I can see that the supposition that he thinks that q is required; he has done nothing to stop me thinking that q; he intends me to think, or is at least willing to allow me to think, that q; and so he has implicated that q. (Grice 1989, 31)
If transposed directly to the fictional case, we would commit ourselves to a number of claims concerning what audiences believe about narrators. For instance, we would be claiming that audiences believe that the narrator knows that they know that the narrator knows that they can see that the relevant inference is needed to preserve the presumption that the narrator is observing cooperative principles. Such claims could perhaps be made plausible.Footnote 25 At the same time, it is likely that some will be wary of assumptions of this kind, in particular, for cases in which the narrator is a fictional character, like Dr. Watson or Dr. Sheppard.
There is a more neutral way of explaining inferences like the one from (1) to (2). Even when the narrator is a fictional character, it is plausible to think that audiences usually believe the narrator is being cooperative, in the sense of observing the maxims or the Cooperative Principle (henceforth, CP).Footnote 26 Indeed, when reading a dialogue in a fictional story, audiences will often have to work out that a character conversationally implicated something to another character. Doing so arguably requires at least taking the former to conform to the maxims or the CP.
Given this, at least part of the explanation for the inference is that the audience think that seeing the narrator as wanting to convey (2) is a way of squaring the fact that she said (1) with the presumption that she is obeying the maxims or the CP. Specifically, in this case the relevant presumption is that the narrator is observing the Maxim of Relation, “Be relevant.” (Grice 1989, 27) So, the claim will be that taking the narrator to be conveying that Paul did not go to the party is a way of understanding her saying that Paul had to work as relevant to the information that everyone went to the party.
This kind of explanation likewise provides a way of understanding the inference triggered by (3).
Let us say that the relevant inference is from (18a) to (18b).
As I said earlier, this kind of inference is also made in ordinary conversations. If I tell you, “My parents called. They bought a new car,” you will most likely infer that my parents told me on the phone that they bought a new car.
Again, this inference is plausibly regarded as a relevance implicature. Understanding me as wanting to convey that my parents told me on the phone that they bought a new car is a way of seeing the information that they called as relevant to their having bought a new car. By contrast, imagine that I tell you, “My parents called. They finally figured out how to use their phone.” You will not infer that they told me that on the phone because there are other reasons why the information that they called is relevant to the information that they figured out how to use their phone.Footnote 27
Similarly, a plausible explanation for the audience’s inference of (18b) from (18a) is that they are assuming that the narrator is being cooperative, and in particular, is observing Relation. As before, for this reason, I will continue to speak of these inferences as conversational implicatures.Footnote 28 Next we will see that they moreover rely on the fictional record.
It is clear that the information that Paul did not go to the party is relevant to the information that everyone went. But why do audiences hit on the former information, having been told that Paul had to work? The answer is that they are assuming that there is a connection between having to work and whether or not someone is going to a party. In particular, they are assuming that having to work is usually incompatible with going to parties.Footnote 29
Accordingly, I suggest that the inference from (1) to (2) relies on it being part of the fictional record that, usually, if someone has to work, they are not going to a party. Unless the audience thinks that, according to the narrator, working (usually) precludes party-going, why would they think that this is what she wants to convey? Yet in many cases this underlying information will not have been explicitly stated by the narrator, nor imported via presupposition accommodation or contextual inference. In this case it is the result of what I will call default importation.
Similarly, the inference in (18) also relies on default importation. In particular, if it is not assumed that, according to the narrator, phone calls are made in order to tell people things, or something to that effect, there is no basis for the conclusion that the narrator is obeying Relation because she is conveying (18b) by saying (18a). Yet this information is not included in the fictional record as the result of something we have been told by Dr. Sheppard, either explicitly or implicitly.
Along similar lines, it is often suggested that fictional discourse relies on the audience and narrator sharing basic assumptions.Footnote 30 For instance, Eckardt (2015), argues that
Even at the beginning of a story, reader and author/narrator share some information. Apart from speaking the same language, the author/narrator will rely on shared information about the physical laws of the world, cultural institutions and practices, social environments, and much more. (Eckardt 2015, 66)
In terms of our framework this means that audiences take a number of things to be part of the fictional record, even though they are not explicitly stated by the narrator nor imported by presupposition accommodation or inferred from explicit content.
Why are certain things imported into story content by default and others not? We should not expect that there is a uniform principle behind default importation. One candidate source for default importation concerns genre considerations of the kind Lewis (1983 ) discussed for truth in fiction.Footnote 31 To take one example, we mentioned earlier (in 4.1) that A Study in Scarlet is the first appearance of Holmes to the reading public. Yet, it is plausible that when reading a Sherlock Holmes story published after A Study in Scarlet, in many cases audiences will import by default that Holmes is a pipe smoker, and several other things.
I will forego further discussion of this here and instead focus on the way inferences rely on default importation, regardless of the source of background information of this kind.
Contextual inferences and default importation
There are other inferences that audiences make that rely on the fictional record, but which arguably do not arise due to assumptions about the narrator being cooperative. Above I suggested that the information that Holmes was sitting down might be imported by way of presupposition accommodation as a result of the occurrence of (6).
Yet it might be said that the information that Holmes was sitting down is likely to be already included in the fictional record, even before the occurrence of (6), simply because it is reasonable to infer that Holmes was sitting down from the explicit information, quoted earlier, that “Sherlock Holmes had not yet finished his breakfast.” (Doyle (1981 ), 23) This is undeniably true. Yet this kind of importation should be distinguished from presupposition accommodation. That Holmes was eating breakfast does not presuppose that he was sitting down.
Let us say that the inference is from (19a) to (19b).
It is a clear that audiences to fictional stories routinely make this kind of inference. Here is another one:
Inferences of this kind also occur in ordinary conversational settings. If I tell you that when I got up yesterday my room-mate was still eating breakfast, you are likely to infer that she was sitting down. If I tell you that after we had dinner last night my father lit his pipe, you are likely to infer that my father is a pipe smoker.
Such inferences rely on background assumptions. In particular, for the fictional case, the inference in (19) depends on it being part of the fictional record that people usually eat breakfast sitting down. Correspondingly, the inference in (20) relies on it being part of the fictional record that, roughly, people who light pipes are usually pipe smokers. As such, like conversational implicatures, these inferences rely on default importation. Readers of A Study in Scarlet are not told explicitly that people eat breakfast sitting down or that people who light pipes are usually pipe smokers. Nor has this been presupposed or conversationally implicated. Instead, this information is part of the fictional record by default.
Yet these cases differ from the examples of conversational implicatures discussed above. For instance, we do not need to think that the narrator wanted to convey that Holmes was sitting down in order to square the fact that he said that Holmes had not yet finished breakfast with a presumption of cooperativeness. Nor do we need to infer the Holmes is a pipe smoker in order to see how saying that Holmes rose and lit his pipe is in line with observing the maxims or the CP. Rather, we infer things like (19b) and (20b) because of default assumptions about, to borrow Eckardt’s (2015, 66) formulation, “the physical laws of the world, cultural institutions and practices, social environments, and much more.”
It might be asked why we cannot say the same about, for instance, (1)–(2). That is, the narrator said that everyone went to the party and Paul had to work, the audience assume that, according to the narrator, people who have to work do not go to parties, and so they infer that Paul did not go.Footnote 32 We should agree that there can be situations of this kind. However, for many cases we will want to explain not only that audiences infer (2) from (1) but also that they conclude that the narrator wanted to convey (2) to them by saying (1). At least this will require assuming that the audience think that the narrator could see that audiences would infer (2) and did not prevent them from doing so. Yet this is arguably tantamount to assuming that the audience think the narrator is being cooperative. If the narrator does not want to convey (2) and she can see that audiences will infer (2), she is arguably violating Grice’s (1989, 26) First Maxim of Quantity, “Make your contribution as informative as is required (for the current purposes of the exchange).” Intuitively, she has said too little.
By contrast, we do not think that Watson wanted to tell us that Holmes was sitting down by telling us that he had not finished breakfast. Correspondingly, we do not judge Watson uncooperative if he did not want to convey that Holmes was sitting down. To complain that he said too little if Holmes was eating breakfast standing up is surely too draconian.
Inferences like (19) and (20) are not conversational implicatures. As I will say, they are contextual inferences. Such inferences arise because certain information is part of the fictional record, typically as a result of default importation, but do not rely on assumptions about what the narrator wanted to convey.
This conclusion can be corroborated by considering cases in which it is later revealed that what was inferred is not true in the fiction. As we said earlier, if the audience later find out that Paul in fact did go to the party, they will clearly think that the narrator had been misleading in telling them what she did. And similarly for the example from The Murder of Roger Ackroyd. However, if we find out that Holmes was in fact eating breakfast standing up, there is no parallel sense that we have been misled. This suggests that we do not take (19b) to be part of what the narrator conveyed to us by saying (19a). Even though both kinds of inference rely on the fictional record, and typically, information that results from default importation, contextual inferences are not seen as part of what the narrator wanted to communicate.Footnote 33
We should distinguish information that is part of the fictional record because it has been conveyed by the narrator from information that is part of the fictional record even though it has not been conveyed by the narrator. Information included by default interpretation and contextual inference belongs to the second category. It is clear that audiences typically keep track of this distinction.Footnote 34
So, according to this suggestion, sometimes audiences think that p is true according to the narrator, even though it is not right to say that the narrator has communicated that p. Rather, p is imported by default because taking p to be part of what the fictional world is like according to the narrator helps make sense of what she does convey and of the story more generally. But there can also be other reasons for default importation. Things may be imported by default because they are generally assumed to be true, such as that birds fly or that cars have doors, or because they are salient to the reader.Footnote 35 Such things may be imported by default even if they are not particularly necessary for making sense of the story, or of what the narrator says.
It is instructive to compare cases of default importation with cases like that of Holmes’s nostrils. We said (in 2.1) that even if one thinks that it is true in A Study in Scarlet that Holmes has exactly two nostrils, this information need not be imported as part of the story. Audiences to A Study in Scarlet need not think that it is true according to Watson that Holmes has exactly two nostrils. To be sure, they might do so, but there is no sense that cases in which they do not are deviant or outlandish.
This illustrates that something may be generated without being imported (even by default). By contrast, the cases of contextual inference just described illustrate that something may be imported even though it is not conveyed by the narrator. Moreover, as we will see below, such information may not be generated.
Inferences and importation
Like presuppositions, conversational implicatures in fictional discourse are mandatorily imported. In ordinary conversation hearers can reject conversational implicatures, as illustrated by (21).
In this case B recognizes that A wants to convey that p but prevents p from becoming common ground. By contrast, audiences to fictional stories have no such option but must include conversational implicatures in the fictional record. (As before, we acknowledge that there can be unusual cases in which a conversational implicature is not recognized, just as there can be such cases for ordinary conversation.)
To be sure, in stories like The Murder of Roger Ackroyd we later find out that certain conversational implicatures are false in the fiction. It is not true in The Murder of Roger Ackroyd that Parker said on the phone to Dr. Sheppard that Ackroyd had been murdered. But even so, we recognize that Dr. Sheppard implicated that he did. Even when we realize that the implicature is false in the fiction, we will still take it to be part of what we were told. It is a central part of the effect of such stories that we can be expected to do so.
In cases of this kind the realization that the relevant information is false in the fiction is also the result of something we are told by the narrator. After all, the whole of The Murder of Roger Ackroyd is narrated by Dr. Sheppard, and so we only find out that what was implicated earlier is false through other things that he tells us. Schematically, these are cases in which the narrator conversationally implies that p and then later explicitly says (or conveys in some other way) that not-p.
As we saw earlier (in 4.3) since we have understood fictional records to be sets of propositions, there is nothing to prevent us from agreeing that, in such cases, both p and not-p are part of the fictional record. Indeed, there is a sense that we have both been told that p and that not-p by the narrator, even though we recognize that only one of these is true in the fiction. Alternatively, it might be said that once we find out that Parker did not say on the phone that Ackroyd had been murdered, we remove this information from the fictional record. That is, we no longer take that to be true according to Dr. Sheppard. I will not adjudicate this here. I take it to be an advantage of the present account that it is compatible with either option.
Contextual inferences differ from conversational implicatures in not being mandatorily imported. For instance, when told that Holmes had not yet finished breakfast, there is no sense in which the audience is compelled to include in the fictional record that he was sitting down. To be sure, since it is later presupposed that he was sitting down, at least at that point, this will be mandatorily imported if it is not already part of the fictional record as the result of contextual inference.
Inferences and generation
We have already seen that conversational implicatures are not mandatorily generated. The implicature triggered by (3) is false in The Murder of Roger Ackroyd. Correspondingly, contextual inferences are not mandatorily generated. It might be true in some version of A Study in Scarlet that Holmes had not yet finished breakfast but false that he was sitting down. More generally, if B is a conversational implicature or contextual inference inferred from A, B may be false in a fiction, even if A is true in the fiction.
This contrasts with presuppositions. As we have seen, presuppositions in fictional discourse are mandatorily generated. The reason conversational implicatures and contextual inferences are not is that, unlike presuppositions, these inferences are not entailments. Instead, the familiar cancelability of conversational implicatures stems from the fact that conversational implicatures may be false even if the content that triggered them is true. The same holds for contextual inferences.
In other words, certain kinds of information are both mandatorily imported and mandatorily generated (presuppositions) while other kinds are mandatorily imported but not mandatorily generated (conversational implicatures), and other kinds again are neither mandatorily imported nor mandatorily generated (contextual inferences).