Free Indirect Discourse in Non-fiction

This paper considers some uses of Free Indirect Discourse within non-fictional discourse. It is shown that these differ from ordinary uses in that they do not attribute actual thoughts or utterances. I argue that the explanation for this is that these uses of Free Indirect Discourse are not assertoric. Instead, it is argued here that they are fictional uses, that is, they are used with fictional force like utterances used to tell a fictional story. Rather than making assertions about the actual world, these occurrences of Free Indirect Discourse introduce localized fictional scenarios from which audiences are meant to learn factual information. As such, they exhibit some of the ways in which the involvement of perspective in historical fiction has been shown to facilitate learning and retention of information.


INTRODUCTION
Ways of reporting speech and thought have long been studied in both philosophy and linguistics. Familiar ways of reporting are Direct Discourse, Indirect Discourse, and Free Indirect Discourse, as in (1)-(3).
(1) Direct Discourse (DD) Ellen made a decision. "Yes! I will tell him later today, " she said/thought. (2) Indirect Discourse (ID) Ellen made a decision. She said/thought that she would tell him later that day. (3) Free Indirect Discourse (FID) Ellen made a decision. Yes! She would tell him later today(, she said/thought).
Whereas it is clear that DD and ID occur in both fictional and non-fictional discourse, FID has traditionally been associated with fictional discourse, or at least seen as a "literary style 1 ." This paper examines some uses of FID in non-fictional discourse. Three examples are given in (4)-(6).
(4) [Discussing the ancient Roman historian Livy's treatment of the Romulus and Remus myth:] Could it be that a local whore rather than a local wild beast had found and tended the twins? Whatever the identity of the lupa, a kindly herdsman or shepherd soon found the boys and took them in. Was his wife the prostitute? Livy wondered (Beard, 2015, p. 59). (5) [Discussing Julius Caesar's motivations for his invasion of Britain:] Sitting in sunlit Rome at the height of his powers, a little giddy with invincibility, Caesar must have imagined a nice little sideshow, a triumph on the cheap. Faced with the glittering armour of the legions and the eagle standards, the barbarians would simply line up to surrender. They would understand that history always fought on the side of Rome (Schama, 2000, p. 29).
(6) [Describing Darwin's reaction to Wallace's discovery of the principle of evolution:] He racked his brain to recall whether or not he had written something in a letter that tipped Wallace off. But he couldn't recall a thing. Oh, Lyell had warned him . . . Lyell had warned him. . . and now all my work, all my dreams-all my dreams-Then he caught hold of himself. He mustn't give in to this horrible feeling overwhelming his solar plexus. There wassomething more important than priority and glory and applause and universal admiration and an awesome place in history . . . namely, his honor as a Gentleman and a scholar (Wolfe, 2016, p. 30-31).
The chief aim of this paper is to account for two observations about occurrences of FID in non-fiction like those above. First, they do not convey attributions or reports of actual thoughts or utterances. For instance, (4) does not convey that Livy actually said or thought, "Can it be that a local whore rather than a local wild beast found and tended the twins?" 2 . Similarly, it would be wrong to read (5) as claiming that Caesar actually thought, "The barbarians will simply line up to surrender." Nor does (6) communicate, for instance, that Darwin actually thought, "Oh, Lyell warned me, " and so on (6) also includes DD. I return to this in 2.2 below). Second, these occurrences of FID nevertheless do impart information about actual people or events. In the case of (4) we are supposed to understand, roughly, that Livy wondered how to understand the figure of the "lupa" in the myth. (5) is arguably intended to convey something like that Caesar thought the Britons were uncivilized and would be easily conquered. Correspondingly, audiences are meant to learn from (6), roughly, that Darwin was vexed by Wallace's discovery.
I will argue that both observations can be explained by seeing such instances of FID as fictional uses. This means that they are not used with assertoric force, that is, they are not assertions about the actual world. Rather, these instances of FID are used in the same way as the utterances used to tell a fictional story, as in the text of a novel.
Consider, for example, the first sentences of A.S. Byatt's The Children's Book: (7) Two boys stood in the Prince Consort Gallery, and looked down on a third. It was June 19th, 1895. (Byatt, 2009, p. 5) In writing (7), Byatt was not asserting that on 19 June 1895 two boys were standing in the Prince Consort Gallery looking down on a third boy. (7) is not put forward as a claim about what was actually the case. Rather, (7) is used non-assertorically. While there are many ways in which sentences can be used nonassertorically, in this case, following the standard approach, (7) is used with fictional force, that is as part of telling a fictional story. Roughly, that is, (7) does not make an assertion about 2 It is well-known that DD allows reporting in a different language from the one used by the subject of the report, see e.g., Cappelen and Lepore (2007, p. 43-44). E.g., a DD report in English may report an utterance that was originally in German. I assume throughout this paper that it is clear that the same applies to FID, and hence that such a discrepancy is irrelevant to the topics discussed here. the actual world, but makes its content true in the fiction The Children's Book. Similarly, on the view I will develop here, (4) introduces a fictional scenario in which Livy thinks, "Can it be that a local whore rather than a local wild beast found and tended the twins?" Correspondingly, (6) introduces a fictional scenario in which Darwin thinks, "Oh, Lyell warned me, " and so on. In each case, the fiction is about an actual person, Livy or Darwin, just as, for example, parts of War and Peace are about Napoleon 3 . So, while these uses of FID occur in non-fictional discourse, they attribute fictional thoughts or utterances to actual people. This explains the first observation by seeing these uses of FID as non-assertoric, and hence, as not attributing actual thoughts.
Given this, the second observation will be explained as an instance of the way in which audiences routinely learn things about the actual world from fictions. Fictional uses of FID in non-fiction exemplify the way in which "manipulation of point of view" (Friend, 2007b, p. 41) can be used to facilitate audiences' ability to comprehend and retain what is being communicated. Such instances of FID allow audiences to grasp information about actual people and events in a way that would not necessarily be achieved by flatly asserting things about the relevant people or events.
Section 2 reviews some important features of FID, specifically with respect to the examples under examination. I follow the standard view in arguing that both FID and DD, when reporting thoughts, attribute inner speech. Further, I suggest that FID and DD are truth-conditionally equivalent. I then flesh out the two main observations concerning the uses of FID in non-fiction that I am interested in.
Section 3 proposes a view of these occurrences of FID on which they are used with fictional force. I suggest that fictional discourse interacts with a species of common ground information that I call a fictional record, and I show how this view captures the non-assertoric nature of the relevant uses of FID.
Section 4 reviews some findings highlighted by Friend (2007b) concerning cognitive benefits of learning from historical fiction. I then propose that the occurrences of FID in non-fictional contexts under discussion can be seen as small-scale instances of historical fiction, and I comment on the consequences of such a view concerning learning factual information from these ways of reporting the thoughts and attitudes of historical figures.

Free Indirect Discourse
The hallmark of FID is that it blends DD and ID in a relatively well-understood way. As is routinely observed, in FID tenses and person-features of pronouns are as in ID, while everything else-including indexicals, exclamations, and speaker-oriented expressions-is as in DD 4 . Take our examples in (1)-(3).
(1) Ellen made a decision. Yes! I will tell him later today, she said/thought. (2) Ellen made a decision. She said/thought that she would tell him later that day. (3) Ellen made a decision. Yes! She would tell him later today(, she said/thought).
The FID report in (3) represents Ellen's utterance or thought by referring to Ellen with the 3rd person, and by using the past tense, as in the ID report in (2). At the same time, the FID report preserves the exclamation Yes! as it occurs in DD, as in (1). Similarly, (3) uses today to refer to the day of Ellen's decision, as in DD, rather than to the day the report is made, as would be the case in ID. These features of FID are displayed by our examples. For instance, (6) refers to Darwin with the 3rd person while using the past tense as in ID. But, moreover, (6) preserves the exclamation Oh as it might appear in DD, as well as marking disfluency with "..." 5 .
(9) is parallel to (8a). (9) presents an FID report about what Livy wondered-roughly, whether the wife of the herdsman or shepherd was the prostitute. Now consider the fragment of (5) in (10).
(10) Faced with the glittering armour of the legions and the eagle standards, the barbarians would simply line up to surrender. They would understand that history always fought on the side of Rome.
It might be thought that (10) can be understood as ID, given that there are no indexicals, exclamations, or other elements the behavior of which we could cite as evidence for FID. In particular, one might suggest that (10) should be understood as an instance of the phenomenon known as Unembedded Indirect Discourse (UID), described by, among others, Bary and Maier (2014). Briefly, instances of UID are ID but without the occurrence of a matrix like "x thought/said that..." Similarly, one might think that (10) is just ID but where such a matrix is left out. However, there is evidence for FID, as opposed to UID, in (10). Consider the definite description the barbarians. As has been noted by Banfield (1982), Schlenker (2004), Bary and Maier (2014), and others, such definite descriptions behave differently in ID and FID, respectively. As Bary and Maier (2014) say, Definite descriptions in FID are protagonist-oriented (i.e., interpreted from the perspective of the reported speaker), whereas in indirect discourse they can be both protagonist-oriented and narrator-oriented (interpreted from the perspective of the actual speaker) (Bary and Maier, 2014, p. 82). This means that if (10) is ID, and in particular UID, the barbarians should permit two readings, one on which it is attributed to Caesar, and one on which it is attributed to the speaker, Schama, but not to Caesar. That is, there should be a reading of (5) on which Schama thinks of the Britons as "barbarians, " but Caesar does not. This reading is strictly speaking possible, but it is clearly neither the intended nor preferred one. Rather, the default reading of (5) is one on which Caesar, but not Schama, thinks of the Britons as "barbarians." In turn, this suggests that the default reading of (10) is the FID reading.
For these reasons, I take it that (4)-(6) do include occurrences of FID. The goal for what follows will be to sketch an account of these examples on the assumption that the relevant passages are interpreted as FID. Specifically, the aim will be to account for two main observations. First, in contrast to ordinary uses of FID in non-fiction, they do not attribute actual thoughts. Second, at the same time, these uses of FID are clearly intended to convey something about the thoughts or attitudes of the relevant people-Livy, Caesar, Darwin. In the rest of this section, I elaborate on each of these points in turn.

Fictional Uses of Direct Discourse in Non-Fiction
Before focusing exclusively on cases of FID like those in (4)-(6), I want to comment briefly on a parallel phenomenon involving DD. There are instances of DD in non-fiction that exhibit similar behavior to the occurrences of FID noted above. Here are two examples: (11) Maelzel proposed that Beethoven compose a piece of music celebrating Wellington's victory at Vitoria [. . . ]. Beethoven who was depressed and had nothing else to do said, "Yeah, sure, whatever." ("Beethoven: Wellington's Victory (1813)." Music as a Mirror of History. 2016. The Teaching Company. Audio). (12) He goes to Ireland because it's his family basically who are conquering it, and they thought, "Yippee! This is a chance to build up our own territory outside the orbit of the sort of Anglo-Welsh problems. We can now go to Ireland and become terribly powerful there." And of course then Henry II thinks, "Oh my goodness, I can't miss out on this either." ("Gerald of Wales." In Our Time. 4 October 2012. BBC. Radio).
The observations I make in this paper about the cases involving FID apply, mutatis mutandis, to these instances of DD. The latter are also plausibly regarded as fictional uses. The speaker of (11) is not asserting that Beethoven actually said, "Yeah, sure, whatever, " but is introducing a fictional scenario in which Beethoven makes that utterance. In turn, listeners are expected to learn, roughly, that Beethoven accepted but was not enthusiastic. And similarly for (12). I cannot undertake analyses of these cases of DD in this paper. Yet it should be flagged here that I take my account to apply to these, too.
To be sure, both (11) and (12) are spoken examples. Yet there are cases like these even with written DD. For instance, such occurrences of DD are found in (6), here underlined: (6) [Describing Darwin's reaction to Wallace's discovery of the principle of evolution:] He racked his brain to recall whether or not he had written something in a letter that tipped Wallace off. But he couldn't recall a thing. Oh, Lyell had warned him . . . Lyell had warned him . . . and now all my work, all my dreams-all my dreams- Then he caught hold of himself. He mustn't give in to this horrible feeling overwhelming his solar plexus. There was something more important than priority and glory and applause and universal admiration and an awesome place in history . . . namely, his honor as a Gentleman and a scholar (Wolfe, 2016, p. 30-31).
At the same time, it is not implausible to think that uses of DD of this kind in written texts may be more rare than the corresponding uses of FID, in particular, if embedded under a matrix like "x said/thought..., " which is absent from (6). I have no explanation for this contrast, if it exists, in this paper. One possible hypothesis might be that, since FID is conventionally associated with written, fictional discourse, it is more likely to be interpreted as fictional, even when appearing in non-fiction. Another suggestion is that DD presents a more verbatim or iconic representation of speech or thought and hence is harder to interpret as non-assertoric. However, these issues must be left to future work.

Free Indirect Discourse and Inner Speech
We have seen that our cases of FID do not attribute actual speech or thought. To spell this out further, note that standard uses of FID attribute what is sometimes called "occurrent, " or "conscious, " thoughts. As I will say, FID attributes inner speech 7 . In this respect, FID patterns with DD and contrasts with ID.
It is uncontroversial that ID think reports can report nonoccurrent thoughts. That is, such reports can be true even if the subject has not consciously had the particular thought picked out by the complement clause. For instance, (13) can be true even if John has never thought to himself, "There's life in other galaxies." (13) John thinks that there's life in other galaxies.
By contrast, the corresponding DD report in (14) clearly requires such an occurrent thought.
While there may be different ways of cashing this out, I will say that (14) entails that "There's life in other galaxies" occurred in John's inner speech. That is, (14) is false if John did not think to himself, "There's life in other galaxies." For instance, suppose that it can be inferred from John's behavior and other beliefs that he thinks there's life in other galaxies, although John does not know the word galaxy. In that case (14) is clearly false, while (13) is true.
As has been realized since Banfield (1982), FID is like DD in this respect 8 . Here I follow Abrusn's 2020, 10-11 summary of the evidence for this conclusion. First, in FID "It is possible to add x thought/said, as an afterthought or interjection:" (15) Tomorrow was her sixth year anniversary with Spencer, she thought, and it had been the best six years of her life (Maier, 2015).
These observations suggest that FID reports attribute inner speech, as do DD think reports 9 . A consequence of this is the following generalization:

FID-DD Equivalence
FID and DD are truth-conditionally equivalent.
In other words, an FID report is true if and only if the corresponding DD report is. For example, (19a) is true if and only if (19b) is.
(19) a. Yes! She would tell him later today, Ellen said/thought. b. "Yes! I will tell him later today, " Ellen said/thought.
With respect to the versions with thought, both report an occurrence of "Yes! I will tell him later today" in Ellen's inner speech. Hence, if Ellen did not think to herself, "Yes! I will tell him later today, " both are false. Similarly, if used with said, both are true if and only if Ellen uttered, "Yes! I will tell him later today." We are not claiming that there are no differences between FID and DD reports. Indeed, we are not claiming that there may not be differences in what such reports communicate. What we are claiming is just that there is no truth-conditional difference. This claim is analogous to the uncontroversial claim that (20a-b) are truth-conditionally equivalent.
b. It's foggy in London.
No one will deny that there are significant differences between the two, but everyone will agree that there is no truth-conditional difference in the sense that (20a) is true if and only if (20b) is. FID-DD Equivalence has consequences for our examples of FID in non-fiction. For instance, it implies that (9) is true if and only if (21) is.
I take this to be correct. If Livy did indeed think to himself, "Is his wife the prostitute, " then surely (9) is also true, and vice versa. But if Livy did not think to himself, "Is his wife the prostitute?" then clearly both are false.
This suggests that the FID reports in the examples of nonfictional discourse we have looked at are not assertoric. Given FID-DD Equivalence, if (9) is an assertion about what actually happened, then it is claiming that (9), and hence (21), are actually true. Yet this is not how (9) is interpreted, as it occurs in (4). Readers do not think that Beard is claiming that Livy actually wondered to himself, "Is his wife the prostitute?" in the same way that she might claim, for instance, that Livy wrote Ab Urbe Condita, that he was born in modern-day Padua, or that he was a friend of Augustus.
Similarly, given FID-DD Equivalence, (10) entails that Caesar thought to himself, "The barbarians will simply line up to surrender" and "They will understand that history always fights on the side of Rome." Yet audiences do not interpret these occurrences of FID as asserting that such inner speech events actually took place. You cannot object to this use of FID that Schama has no evidence that Caesar actually thought to himself, "The barbarians will simply line up to surrender, " and so on. Correspondingly, (6) does not assert that Darwin actually thought to himself, "Oh, Lyell warned me, " and so on.
We should conclude that, in all of our examples, FID is being used non-assertorically in that they do not attribute actual occurrences of inner speech. To be sure, it might be suggested instead that our examples show that FID-DD Equivalence is false, and that assertoric uses of FID sometimes do not attribute inner speech. However, the evidence summarized in (15)-(18), and earlier, disfavors this reaction. Assertoric uses of FID attribute inner speech. Yet our cases of FID do not. They are not assertoric uses.

Attributing Attitudes
The second observation we noted was that, even though they differ from standard, non-fictional uses of FID in not attributing inner speech, these occurrences of FID do convey information about the actual historical figures. At a certain level of abstraction, it is natural to say that (4) conveys that Livy wondered about the role of the she-wolf (the "lupa") in the Romulus and Remus myth, or that Caesar anticipated an easy victory over the Britons. Yet, although true, this leaves out some important aspects of the phenomenon.
(10) Faced with the glittering armour of the legions and the eagle standards, the barbarians would simply line up to surrender. They would understand that history always fought on the side of Rome.
While (10) does not assert that, for instance, Caesar thought to himself, "The barbarians will simply line up to surrender, " clearly it does convey more than just that Caesar imagined an easy victory. For example, (10) conveys that Caesar thought of the Britons as "barbarians, " that he thought the armor and the eagle standards of the Roman army would intimidate them, that Rome had a special place in history, and so on.
To illustrate further, it is useful to compare our examples with other instances of FID in non-fiction. Consider the example in (22), discussed by Fludernik (1993). ( Fludernik, 1993, 88).
As Fludernik notes, (22) clearly includes FID. But moreover, the FID in (22) is used non-fictionally. That is, the FID report in (22) makes an assertion about what Powell actually said. It asserts that Powell actually said, "Reform of the Lords is a long-pursued mirage, " and so on. This is the contrast with our examples. As we have emphasized, in our cases, FID is not used to make assertions about what the relevant individuals actually said or thought. At the same time, as we said above, these occurrences of FID do convey something about the actual individuals' attitudes. (10) conveys things about Caesar's attitudes beyond just suggesting that he thought the Britons would be easily conquered. For instance, (10) conveys that Caesar thought of the Britons as "barbarians." Similarly, (22) conveys that Powell called the reform a "mirage." Both are consequences of FID-DD Equivalence.
What we want to account for, then, is not just that (10) lacks assertoric force in that it does not attribute an actual event of inner speech to Caesar. We want to explain that, like ordinary, assertoric uses of FID such as (22), (10) (4) and (6).
In the next two sections I spell out the account I favor. I start by explaining what is meant by fictional force, as opposed to assertoric force.

Fictional Force
It is common to distinguish between different ways of using sentences pertaining to fictions. Take the example of (23) concerning the movie and play Amadeus.
(23) Salieri commissioned the Requiem. (Predelli, 2005, ch. 2) We can distinguish between three ways of using (23). A nonfictional use of (23) is an assertion about the actual world. If used in this way, (23) is false, since Salieri did not commission Mozart's Requiem in the actual world. On a fictional use (23) is part of telling a fictional story. Imagine, for example, that (23) occurs as part of a novelization of the movie, or the like. Finally, one can use (23) metafictionally, as making a claim about a particular fiction. For instance, one can use (23) to make the claim that in Amadeus Salieri commissioned the Requiem.
For our purposes, the important difference is between fictional and non-fictional uses. I adopt the standard view on this distinction, endorsed by Searle (1975), Lewis (1978), Currie (1990), Sainsbury (2010), Davies (2015), Recanati (2000), 2018, and many others. According to this view, fictional and nonfictional uses are distinguished only in terms of force 10 . In other words, there is no syntactic or semantic difference between fictional and non-fictional discourse. As Searle (1975) wrote in an often quoted passage, There is no textual property, syntactical or semantic, that will identify a text as a work of fiction. What makes it a work of fiction is, so to speak, the illocutionary stance that the author takes toward it [...] (Searle, 1975, p. 325). Similarly, Currie (1990) writes, If Doyle had been writing history instead of fiction when he wrote "It rained in London on January 1, 1895, " he would have been making an assertion. The transition from history to fiction is marked, at least, by the loss of one kind of force: assertative force (Currie, 1990, p. 6-7).
Correspondingly, (23) has the same syntactic and semantic profile when used to make an assertion about the actual world and when used as part of a fictional story 11 . That is, it is true or false (at a world w) if and only if Salieri commissioned the Requiem (in w). Instead, the difference between fictional and non-fictional, assertoric utterances is a difference in force.
On the view I favor, the FID reports in (4)-(6) are fictional uses. We have already seen that they are not non-fictional uses, since they do not assert attributions of actual inner speech. Further, our cases are not metafictional uses. They are not assertions about fictions. For instance, (6) is not making an assertion like, in such-and-such fictional story, Darwin thinks, "Oh, Lyell warned me." If it did, it would be (actually) true or false depending on whether, in the relevant fiction, Darwin thinks, "Oh, Lyell warned me." This I take to be the wrong result.
In other words, just like the sentences in a novel, (4)-(6) introduce fictional scenarios in which it is true that the relevant inner speech events take place. Similarly, for instance, Byatt's fictional use of (7) makes it true in The Children's Book that on 10 See Walton (1990, ch. 2) for opposition. 11 By contrast, the orthodox approach to metafictional uses follows Lewis (1978) in seeing them as having the same content as the corresponding sentence prefixed with an operator like "In Amadeus..." For recent, different versions of this approach to metafictional discourse, see e.g., Predelli (2008), Recanati (2018), Stokke (2020). 19 June 1895 two boys stood in the Prince Consort Gallery and looked down on a third boy 12 .
(7) Two boys stood in the Prince Consort Gallery, and looked down on a third. It was June 19th, 1895 (Byatt, 2009, p. 5).
Correspondingly, (6) makes it true in the fiction about Darwin it introduces that Darwin thought to himself, "Oh, Lyell warned me." A positive motivation for taking our examples to involve fictional uses of FID is that they are overtly indistinguishable from assertoric uses of FID, such as (22). There is no overt linguistic material in our examples that is responsible for their difference from assertoric uses of FID, such as a modal, or the like. Given this, the fact that our uses of FID do not report actual inner speech is plausibly due to the pragmatics of the way they are used, rather than to their semantic or syntactic profile. In other words, they conform straightforwardly to the observation that fictional and non-fictional uses are distinguished only in terms of force.
To be sure, one can posit covert structure for our examples, such as an unpronounced operator, and thereby claim that there is a syntactic, and hence semantic, difference 13 . The fictional account does not do so. I take that to be a point in its favor, given the overt indistinguishability. But moreover, I take it to be a prima facie plausible suggestion that the difference between the way, say, (4) represents Livy as wondering about the role of the she-wolf in the myth and the way that inner speech is standardly reported by means of FID is a pragmatic difference, and not a difference in what is said, or truth-conditional content.
If we accept that the FID reports in these cases are used fictionally, we explain the first observation noted in the last section. That is, the occurrences of FID in our examples do not convey attributions of actual inner speech because they are used fictionally to introduce fictional scenarios in which the relevant thoughts or utterances take place.

Other Examples of Free Indirect Discourse in Non-Fiction
Before moving on to spelling out the fictional account of the occurrences of FID we have looked at, it is worth commenting further on other occurrences of FID in non-fiction. Fludernik (1993, p. 88) observed that FID "occurs widely in non-literary texts." (emphasis removed) These are also cases of FID appearing in non-fictional discourse 14 . We noted above that at least some of these cases are assertoric uses, that is, occurrences of FID that attribute actual thoughts or utterances. This was the case for (22), repeated here: (22) Reform of the Lords was a long-pursued mirage, Mr Powell told a House which started to fill up as news that he was on his feet spread. 12 In this sense, fictional uses usually give rise to metafictional truths, although not always, because narrators can be unreliable. For example, given that (7) makes its content true in The Children's Book, one can subsequently use the same sentence, , to make a true, metafictional assertion about that fiction. 13 This suggestion should not be confused with accounts of FID itself that posit covert operators, like that of Sharvit (2008). 14 See also Zeman (2018, 181 fn. 6), and see Fludernik (1993) for more references.  Fludernik, 1993, p. 89).
Are these fictional uses of FID within non-fiction? That is, are they non-assertoric uses, as in our examples, or are the ordinary, non-fictional, assertoric uses, like (22)? Let us consider each in turn.
In (24) FID is used to report thoughts, that is, inner speech. So, the question is whether (24) should be taken as asserting, for instance, that Montagu actually thought to himself, "Doesn't this prove that I am on equal terms with the rulers of the land?" Most likely, this is not the right reading of (24). Rather, (24) is most naturally understood as on a par with our own examples. That is, it presents a fictional report of Montagu's thoughts, just as, for instance, (6) presents a fictional report of Darwin's thoughts.
Finally consider (25). This case is arguably different from each of the two preceding examples. It is natural to think that FID in this case reports inner speech like, "there is the indefatigable Wind, over there the stomper Briseno." Yet there is no clear indication of the subject of the report. As such, (25) is an instance of what we might call unidentified FID 15 . Here is an example from Thomas Mann's Buddenbrooks: (26) A tooth-Senator Buddenbrook had died of a toothache, that was the word around town. But, confound it all, people didn't die of that! He had been in pain, Herr Brecht had broken off the crown, and afterward he had simply collapsed on the street. Had anyone ever heard the like? (Mann, 1904, p. 666) Here FID gives a report of someone's speech. But whose? We are presented with "the word around town." But this is not a report of a particular person's speech. Instead, it is natural to think that we are being presented with the speech of an arbitrary member of the group in question, in this case the relevant townsfolk. (25) is 15 Wood (2008, 24) calls this "unidentified free indirect style." arguably parallel, except that it appears in a non-fictional context. I take the phenomenon of unidentified FID to be different from the kind of fictional uses of FID that we have looked at, and I will not attempt an analysis of such cases here.
So, we can note that, as exemplified by (22), FID can be used assertorically in non-fictional discourse. That is not surprising. Indeed we have presupposed as much here. That is, FID is a style of report which can be used to attribute thoughts (i.e., inner speech) and utterances both outside and inside fictional contexts, just like ID and DD. When FID occurs in non-fictional contexts, it standardly attributes actual thoughts or utterances, as in (22). By contrast, the cases we have examined attribute fictional thoughts or utterances to actual people, although they occur in non-fictional contexts,

Fictional Records
I have suggested that the non-assertoric uses of FID we have examined are fictional uses, that is, they are used with fictional force. To make this suggestion more concrete, it is convenient to implement it within the framework for understanding discourse and assertion familiar from the work of Stalnaker 1970Stalnaker , 1978Stalnaker , 1998Stalnaker , 2002Stalnaker , 2014. On this picture, discourse relies on a body of information, called the common ground, that is taken for granted for the purpose of the exchange. The common ground acts both as support for utterance interpretation and as storage for information communicated by the participants.
Central to this theory of communication is an understanding of assertion. To utter a sentence S with assertoric force is to propose that the propositional, or truth-conditional, content of S, given the context, become part of the common ground. By contrast, a non-assertoric utterance is one that falls short of making a proposal to increment common ground information with what it says. For instance, if uttered ironically, (27) is not an assertion, and correspondingly is not a proposal to make it common ground that The Da Vinci Code is a great novel.
(27) Oh yeah! The Da Vinci Code is a GREAT novel! Given this, to say that fictional discourse is distinguished from non-fictional discourse by not being assertoric is to say that fictional utterances do not involve proposals to increment common ground information.
Instead, I suggest that utterances made with fictional force interact with alternative bodies of information comprising what is part of the story at a given time during its unfolding 16 . Fictional discourse features presuppositions, anaphora, indexicals, and other elements that rely on contextual information. As a simple illustration, consider this sentence from Doyle's A Study in Scarlet: (28) Sherlock Holmes rose and lit his pipe (Doyle, 1887, p. 24).
(28) is the first appearance of Holmes's pipe in A Study in Scarlet, itself the first appearance of Holmes to the reading public. Accordingly, (28) has the effect of making audiences include 16 Related accounts of fictional discourse in terms of alternative contextual information have been proposed by Bonomi and Zucchi (2003), Stokke (2018, Eckardt (2015), Semeijn (2017), and Maier and Semeijn (in press).
Holmes's pipe in the story, as an instance of presupposition accommodation. Correspondingly, the pipe is available for anaphoric reference later in the text, as in the following hypothetical continuation: (29) Sherlock Holmes rose and lit his pipe. A few minutes later he put it down in an ashtray.
Observations of this kind motivate a picture according to which, during the unfolding of a fictional story, a body of information delineating what is taken to be part of the story evolves. I will call this a fictional record. As illustrated by (28), and as is apparent from run-of-the-mill fictional discourse, fictional records play a role analogous to the role played by ordinary common ground information in everyday conversation. Along these lines, we think of utterances made with fictional force as contributions to fictional records. While fictional utterances are non-assertoric in the sense that they are not directed at adding information to what is common ground, they nevertheless function to increase information that is part of fictional records. So when (28) is read by the audience, they update the fictional record with, at least, the information that Holmes has a pipe, that Holmes rose (got up), and that Holmes lit his pipe.

Adding Free Indirect Discourse Reports to Fictional Records
I suggest that when audiences engage with works like those from which we have drawn our examples, they take some of the sentences as updating the information they think of as the "official" information of the discourse (analogous to the ordinary common ground of a conversation), and others as updating fictional records. In our cases the latter being the uses of FID.
For instance, readers of (4) will be aware of what is, and has been, conveyed as official information of the book. Along the lines of what we said earlier, the official information might include things like that Livy wrote Ab Urbe Condita, that he was born in modern-day Padua, or that he was a friend of Augustus. At the same time, audiences are aware of fictional information conveyed by the text. In the case of (4), this fictional information comprises, at least, that Livy wondered to himself, "Is his wife the prostitute?" In other words, at least one effect of the fictional uses of FID is to add to a fictional record that the relevant inner speech event took place. For instance, in the case of (5), roughly, there will be a fictional record that includes the information that Caesar thought to himself, "The barbarians will simply line up to surrender." At the same time, as we said before, more is conveyed by FID reports than just the occurrence of inner speech events. For instance, it will likewise be part of the fictional record that Caesar thought of the Britons as "barbarians, " that he thought the armor and the eagle standards of the Roman army would intimidate them, that Rome had a special place in history, and so on.
Specifically, the passage in (5) has the effect, at least, of producing an information state in which, apart from the official information, there is a cache of fictional information. For ease of reference, call the fictional record to which the FID reports about Caesar in (5) are added A. We can then schematize A as follows: A A 1 Caesar thought to himself, "Faced with the glittering armor of the legions and the eagle standards, the barbarians will simply line up to surrender." A 2 Caesar thought to himself, "The barbarians will understand that history always fights on the side of Rome." A 3 Caesar thought that the Britons were barbarians. A 4 Caesar thought that the Britons would be easily conquered.
A 5 Caesar thought that Rome had a special place in history. ...
One can think of fictional records, like ordinary common grounds, as sets of propositions. Further, one can think of propositions as sets of possible worlds, in the familiar fashion. Given this, it may be useful for some purposes to represent a fictional record itself as a set of worlds, corresponding to the standard notion of a context set, that is, as the intersection of all the propositions in a particular fictional record, which will represent the possibilities compatible with what is included in the relevant record. For instance, A might be represented as the set However, for the purposes of this discussion, these further implementations will not play a role. The aim here is to give an account of the non-assertoric force of the relevant occurrences of FID, and further to suggest some factors in how they facilitate learning about actual historical events. I have argued that the former point can be explained by seeing these utterances as aimed at updating fictional records. In the next section I turn to the second point.

Coming to Know and Coming to Believe
On the view outlined in the last section, for instance, the fictional uses of FID in (5) produce a fiction about Caesar in which (at least) A 1 ...A 5 are true. I suggest that, just as one can often learn things from fictions, in these cases, audiences can learn things from the fictions introduced by the occurrences of FID.
As we use the terms here, learning that p implies coming to know that p. Hence, trivially, audiences cannot come to learn something that is actually false from any fiction. Yet, of course, they may come to believe such things based on fictions, and thereby be misled. As a simple, hackneyed example, audiences to A Study in Scarlet might come to believe that 221B Baker Street existed in 1887 when the novel was published. Yet since this is false, trivially they cannot learn that, but rather they will acquire a false belief in this case.
Considering our case of A, let us assume that A 1 -A 2 are actually false, while A 3 -A 5 are actually true. That is, while the inner speech events reported by the former did not in fact occur, the attributions of beliefs to Caesar in the latter are nevertheless accurate. In that case, we may take it that, given that other constraints on learning from fiction are satisfied, audiences may learn A 3 -A 5 from A.
To be sure, there is a significant challenge in specifying what these other constraints are. Yet it is not among my aims here to take up that challenge 17 . A few remarks are in order, however. In particular, as is standard, we should distinguish between two broad ways of learning from fiction. Gendler (2000) describes this difference as follows 18 : narrative as clearinghouse: I export things from the story that you the storyteller have intentionally and consciously imported, adding them to my stock in the way that I add knowledge gained by testimony. [...] narrative as factory: I export things from the story whose truth becomes apparent as a result of thinking about the story itself. These I add to my stock the way I add knowledge gained by modeling (Gendler, 2000, p. 76).
Given this distinction, it is plausible to think that our cases fall under the second of these general ways of learning from fiction. In particular, the process by which audiences come to believe, and learn, A 3 -A 5 from reading (5) is plausibly understood as based on these things becoming "apparent as a result of thinking about the story, " that is, in our case, as the result of understanding the FID report. Presented with the fiction in which Caesar thought to himself, "The barbarians will simply line up to surrender, " and so on, audiences are likely to learn, for instance, that he thought of the Britons as "barbarians" as a result of that truth about Caesar becoming apparent from the story.
This suggests that the process in question is far from automatic and, unquestionably, many factors are involved. For instance, the reason that an audience may come to believe A 3 -A 5 from reading (5) most likely stems from factors such as the their awareness of the genre they are reading, other evidence that they have about Caesar, and of Rome in general, their assessment of the overall plausibility of A 3 -A 5 , and more. I will not attempt a further account of this here.
Instead, I want to focus on some ways in which learning from fiction has been shown to be cognitively beneficial. More colloquially, we are good at learning things from fiction, demonstrably better than we are at learning things from "dry" expositions of factual information. As I go on to explain, there are good reasons to think that the fictional uses of FID in our examples fit this general pattern.

Perspective and Historical Fiction
Stacie Friend (2007b, 41) has reported a number of studies in cognitive psychology showing that "the manipulation of the point of view from which we learn about events" involved in historical fiction "generates numerous epistemic advantages" with respect to acquiring information about the actual events depicted in the fiction 19 .
Taking Gore Vidal's Lincoln: A Novel as an example, Friend highlights that the novel presents things to the audience as if they were eyewitnesses: Lincoln plunges us directly into the flow so that we "see" the president in action: we learn about Lincoln through the eyes and minds of people close to him, rather than from Vidal's real retrospective point of view (Friend, 2007b, p. 41).
As Friend observes, a key difference between this kind of historical fiction and conventional, non-fictional history is that it allows for direct representation of the thoughts and attitudes of historical figures 20 : Because the writer of a work of non-fiction could not possibly have such access to the minds of other people, histories and biographies standardly present the thoughts of real individuals as inferences from the evidence. And they provide information about their evidential sources. This is by contrast with Vidal's narration, which provides the reader with fictional, seemingly direct access to the thoughts of certain characters (Friend, 2007b, p. 38).
Friend summarizes a number of ways in which this kind of shift in perspective has been shown to facilitate learning and retention of information on the part of audiences. Here I want to highlight two of these. First, One advantage is that such eyewitness descriptions are more likely to be concrete, thereby generating more imagery; this in turn seems to significantly enhance memorability (Friend, 2007b, p. 41).

Second,
A related epistemic advantage of Lincoln depends on the reduction of exposition afforded by Vidal's technique. [...] It turns out that narratives display an advantage over expositions in studies of reading comprehension. Expository texts, when they treat unfamiliar topics, prompt subjects to process information as so many separate items to be memorized [...]. By contrast, narratives prompt readers to focus on the situation the text is about (Friend, 2007b, p. 42).
As I argue below, these points equally apply to the fictional uses of FID we have examined.

Concreteness and Reduction of Exposition
Our cases can be described as examples of historical fiction, albeit in the concentrated form of free-standing FID reports. If what I have argued is on the right track, they constitute smallscale pockets of historical fiction within historical non-fiction.
In particular, they are instances of precisely the kind of fictional presentation of immediate access to the thoughts of historical figures that Friend describes.
For instance, by interrupting the non-fictional discourse of the book with the FID report in (6), "Oh, Lyell had warned him... Lyell had warned him, " Wolfe gives us "fictional, seemingly direct access" to Darwin's thoughts. So, we should expect that at least some of the particular features of the ways in which we routinely learn things about actual historical events from historical fiction apply to our cases, too.
As we noted above, Friend reports that historical fiction has been seen to facilitate learning and retention of information by, among other things, being concrete and by reducing exposition. It is relatively clear that FID reports of inner speech are concrete in the relevant sense. In particular, both FID and DD contrast with ID in this respect. Rather than being presented with a description of the content of the relevant thoughts, we are presented with those thoughts themselves, so to speak. As such, it is plausible to think that FID is likely to generate imagery in the sense that has been shown to enhance memorability.
With respect to the second point we highlighted above, it is worth being clear that there are potentially two ways in which one can understand the idea of reduction of exposition-the reason being that there are two main ways of understanding the notion of exposition. First, one might think of exposition as representation of events, actions, scenes, and so on, that is, as opposed to representation of speech or thought. In this sense, exposition corresponds roughly to the traditional category of diegesis, as contrasted with mimesis. On this way of understanding exposition, it is trivial to say that FID reduces exposition, since FID is a device for representing speech or thought.
Second, however, one can understand exposition as a particular way of organizing information. This is how Friend uses the term in the quote above, where exposition is contrasted with "narrative." So, the claim is not that the mere replacement of diegesis with mimesis amounts to the kind of reduction of exposition that is cognitively beneficial for learning. Rather, it is the replacement of non-narrative presentation with a more narrative style.
As I explain below, it is arguable that FID is particularly conducive to this kind of reduction of exposition.

Narrative and Imitation
There are reasons to think that FID itself facilitates narrative presentation, whereas DD does not. Of course, DD routinely figures in narrative discourse. Yet it is not difficult to imagine a non-narrative, factual presentation of events that includes DD reports of, for example, statements by politicians, military officials, and so on. Much history is written in this way. Similarly, a government report on some issue might take that form. Yet it would be hard to imagine such a report including FID, even FID reporting speech. Indeed, the transcript of British parliamentary debate in (22) we cited earlier arguably has a narrative character, even though FID is used non-fictionally.
FID reports are typical of the kind of narrative style that counts as reduced exposition. Many studies of FID highlight this feature. In an early treatment, Cohn (1966, p. 98) called FID "narrated monologue, " and noted that it "enables the author to recount the character's silent thoughts without a break in the narrative thread." More recently, Zeman (2018, 174) has argued that FID is "restricted to narrative discourse mode only." The central motivation for this conclusion being that FID involves two kinds of perspective, the speaker's and that of the subject of the report, where this duality is seen as the key characteristic of narrative discourse.
I suggest that, first, we should understand the relevant kind of reduction of exposition as a tendency toward narrative presentation, rather than merely providing FID or DD reports (mimesis) instead of pure description of events (diegesis). And second, FID in particular, as opposed to DD, facilitates this kind of reduction of exposition because FID itself is narrative device. As argued by Zeman (2018), and others, FID is a narrative device in that it essentially involves two perspectives, that of the narrator (or speaker) and the protagonist (or subject of the report).
Given this, we should ask to what extent the double perspective involved in FID might itself play a role in the acquisition of factual information. One suggestion is that FID allows the author to communicate attitudes toward the historical figures in question in a particularly effective way. Currie (2010) has argued that FID is distinguished from DD by involving imitation, whereas DD is a way of replicating speech or thoughts: When Barkis says "Barkis is willin", I might report this by saying 'Barkis said he is willing' , or 'Barkis said "Barkis is willin"' , neither of which strikes one as particularly imitative of his odd turn of phrase, though the second certainly draws attention to it. If I say "Barkis was willin", I have injected a distinct element of imitation into the report. (Currie, 2010, p. 142) On Currie's view, this involvement of imitation in FID is particularly due to its double perspective, that is, its blending of DD and ID: With FID [...] we have the sense that it is the narrator speaking, though speaking in a way which is highly constrained by the words, the tone, the style of the character whose speech is represented: there is something theatrical about FID as a mode of reporting which makes it difficult not to think of the speaker as imitating another. (loc. cit.) For Currie, a central aspect of imitation, in this sense, is that "we have the sense of sharing with the author a way of experiencing and responding to those events, leading to a sense of guided attending on our part" (Currie, 2010, p. 106).
I argued that the use of FID in (5) allows readers to learn, for instance, that Caesar thought the Britons were uncivilized and would be easily conquered. So, one suggestion is that because of the imitative feature of FID, which stems directly from its involvement of two perspectives, readers not only learn that fact about Caesar but is allowed to share the author's understanding of it and attitudes toward it. Rather than just being told that Caesar thought the Britons were uncivilized and would be easily conquered, we are allowed to share the author's sense of Caesar's particular imperious high-mindedness, his being "giddy with invincibility, " as manifested by the representation of Caesar as thinking, "the barbarians will simply line up to surrender." Yet, at the same time, by using FID fictionally, and hence nonassertorically, the author is able to convey these attitudes toward Caesar without outright lying or making baseless claims about what Caesar's particular thoughts were.
Similar points can be said to apply to the other examples. For instance, the FID uses of direct questions in (4) convey, not just that Livy was interested in re-thinking or interpreting the myth, but that questions like how to think of the identity of the lupa character was particularly high on his mind, and that he was consciously pondering them. Similarly, the FID use of "Oh, Lyell had warned him..." in (6) may suggest that Darwin was regretful of not having considered the possibility earlier, and not having heeded Lyell's warning, and that such things were weighing on his conscience and persistently bothering him.
In other words, FID can be seen as a device for representing speech and thought that is specifically narrative in character. In particular, FID instantiates the double perspective that, as argued by Zeman (2018), is distinctive of narrative discourse. By doing so, FID involves imitation, which in turn allows audiences to learn about the author's understanding of, or attitudes toward, the people in question. More generally, it is plausible that the uses of FID in our examples are paradigmatic examples of the kind of reduction of exposition that Friend cites as one of the key cognitive benefits of historical fiction.

CONCLUSION
FID is used in both fictional and non-fictional discourse. We have examined some cases in which FID occurs within non-fictional discourse that are characterized by the fact that FID is used nonassertorically. The occurrences of FID in these cases do no assert attributions of inner speech, as FID standardly does.
I have argued that these occurrences of FID are used with fictional force. As such, they contribute to fictional records that are available to audiences alongside "official" discourse information. In turn, they introduce small-scale (historical) fictions into the otherwise non-fictional discourse. By doing so, they facilitate acquisition and retention of factual information about the relevant historical figures and events by the audience.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.