This research was funded by the Deutsche Forschungsgemeinschaft (DFG – German Research Foundation) as part of the Collaborative Research Center 1252 “Prominence in Language”, Project-ID 281511265, which is gratefully acknowledged. An earlier version was presented as a poster at the SFB 1252 New Year’s Meeting in January 2019. We are grateful to the audience for their feedback. We thank Robert Tegethoff for proofreading an earlier draft, Mary Chambers for proofreading the final draft, and Barbara Zeyer and David Wick for practical assistance. We also thank Stefan Hinterwimmer for pointing out to us interesting FID examples in H. Mann (1918) and T. Mann (1963), of which we cite two in the present contribution. Finally, we are particularly grateful for the very helpful input from three anonymous reviewers.
- 1 An instance of FID is almost always attributed to a character different from the narrator (...)
- 2 While these elements may be taken as indicators, they also need to be processed. (...)
1Free indirect discourse (FID) is a literary phenomenon in which the speech or thought of a character is expressed without explicit lexical indications for the quote (see Lips, 1926; Banfield, 1982; Eckardt, 2014; etc.). However, readers are still easily able to recognize that it is not the narrator who is responsible for the content of the FID and, as Hinterwimmer (2019) shows, they generally know to which character within the story the referred content needs to be attributed1. This is due to the special use of deictics, evaluative expressions and other indicators (see Section 2)2. Interestingly, FID has been shown to feature a specific mix of indications which partly favor the narrator and partly a character as the speaking entity (see e.g., Maier, 2015), which is an important property distinguishing it from other forms of speech or thought representation. This constellation leads to the description, prevalent in the recent research literature, that there are two voices or contexts available which overlap in FID (see Doron, 1991; Schlenker, 2004; Eckardt, 2014; etc.).
2The resolution of FID to a character in the story is unproblematic when there is only one character available. However, when there is more than one protagonist, the resolution may be less evident. This question has only recently been approached by Hinterwimmer (2019), who shows that a disambiguation is possible and that it is always the most prominent protagonist in the context who functions as the perspectival center, i.e., to whom the FID is attributed. In example [1], Bob is the perspectival center. The FID consists of the second and third sentences.
[1] |
Bob welcomed Anna with excitement. He had found her car key! Wasn’t she lucky to have him? |
|
(Eckardt, 2014: 1, adapted) |
3In terms of the temporal characteristics of FID, the focus of the research literature is unbalanced. First, much of the literature considers tense choice (see Banfield, 1982: 104; Landeweerd & Vet, 1996; Reboul et al., 2016; etc.). Second, from a discourse perspective, temporal relations are of central interest but they have not yet received much attention (see Egetenmeyer, accepted a). Bu (2016) and, to a minor extent, Lo Cascio (2002) may be considered exceptions. Third, when the subject of temporal relations is mentioned, there is a strong tendency towards the overgeneralization of co-temporality relative to the preceding discourse. However, Egetenmeyer (accepted a) shows that a variety of temporal relations between FID events and the surrounding context is possible. A FID event is basically defined as the uttering event of an instance of FID (see Egetenmeyer, accepted a; Section 2.2 below presents further details). Finally, temporal anchoring is largely ignored in the literature (see Ehrlich, 1990: 74 ff. for an exception), even though it is a basic component of temporal relations. What is more, on the rare occasions when the relationship is mentioned, it is analyzed only with respect to an explicit verb of speech or thought (e.g. Bu, 2016), which may be realized parenthetically (see Banfield, 1982: 76). It may then parallel what we find in indirect discourse. However, cases which lack such an explicit parenthetical verb are conceptually much more interesting. As the (preceding) context may contain a potentially infinite number of time points available as candidates for anchoring, the problem is to identify which time point functions as the temporal anchor of the FID event. An important factor that reduces the number of possibilities is recency. More specifically, we may suspect that the anchor time is introduced in the sentence directly preceding the FID. However, this is not always the case, and thus recency cannot be the only relevant factor. As we will see in Section 3.2, it may be overruled. Still, as temporal relations do not seem to pose special problems to readers, we may suspect that there are well entrenched mechanisms that allow the reader to retrieve the anchor time. To understand the problem better, we therefore need to consider its structurally relevant properties.
4In sum, we seek to answer the following questions: (i) which preceding time point functions as the anchor of FID events, (ii) what properties does it have, and (iii) do these properties parallel those of the perspectival center? The hypotheses are that a FID event is anchored to a prominent time point introduced via the preceding discourse and that it thus parallels the perspectival center.
5In Section 2, we introduce the relevant properties of FID as discussed in the literature and indicate how it is delimited from other forms of speech or thought representation and perspective shift. Section 2.1 focuses on the analysis by Hinterwimmer (2019) concerning the perspectival center. This is relevant in order to show that certain properties of temporal reference parallel those of the domain of reference to individuals. Section 2.2 concentrates on the temporal relations of FID events as described in Egetenmeyer (accepted a). Section 2.3 narrows down the aim of this paper more precisely. In Section 3, we then lay out our account of the prominence value of the anchor time point and determine the relevant factors (Section 3.1). We also discuss the problem of competing time points (Section 3.2). Section 4 presents a detailed analysis of a specific set of FID instances with respect to three main categories. In the conclusion (Section 5), we compare the prominence properties of the anchor time point with those of the perspectival center.
6According to a basic definition (see also Section 1), FID is a special kind of speech or thought representation which lacks an introductory declarative verb but still expresses the conscious state of mind of a protagonist within the story (see e.g., Eckardt, 2014 for FID; and Hinterwimmer, 2017 for a distinction of FID from other forms of perspective shift; see below). In simplified terms we can state that its internal form shares characteristics both with direct and indirect discourse, while its syntactic status parallels that of direct discourse. The FID sentence or passage is not embedded under a speech verb, even though “it may (but need not) be accompanied by a parenthetical clause” (Banfield, 1982: 76), which may be positioned after (see example [2]) or within the FID sentence (see Banfield, 1982: 76).
- 3 Banfield (1982: 76) leaves out the question mark, putting a comma in its place. In the or (...)
[2] |
What message could Cam give the cook? Mrs Ramsay wondered. |
|
(Woolf, To the Lighthouse, 1974: 88; taken from Banfield, 1982: 76)3 |
7Superficially, example [2] seems to parallel what is generally assumed for indirect discourse (cf., however, Hunter, 2016; see Section 3.2). Simply put, all relevant discursive properties might be attributed to the parenthetical structure. In the case of [2], this yields a correct interpretation insofar as the content of the FID is to be attributed to the explicitly mentioned character (Mrs Ramsay) and the temporal location of the parenthetical verb is also indicative of the FID’s temporal location in discourse. However, on the one hand, FID not only shows properties of indirect speech, but combines them with characteristics of direct speech (see Sharvit, 2008: 354; Maier, 2017: 261). On the other hand, as it does not always feature a parenthetical verb, we cannot base an analysis of FID on such a verb. Before we come back to the problem of the temporal localization of FID without parenthetical verbs, we need to take a look at the central properties of FID and how the literature accounts for them. Furthermore, we will show how FID can be distinguished from competing phenomena.
- 4 We added the adverb just.
8The deictic expressions contained in FID passages differ with respect to the source of information to which they are attributed, either the narrator or a character within the story. In the recent literature, this is mostly analyzed as a distinction between two voices (see Doron, 1991) or between two contexts, in the sense of Kaplan (1989) (see Schlenker, 2004; Eckardt, 2014; etc.), which overlap in the case of FID. This can be shown nicely with our example from Section 1, which is repeated below with a further adaption4.
[3] |
Bob welcomed Anna with excitement. He had just found her car key! Wasn’t she lucky to have him? |
|
(Eckardt, 2014: 1, adapted) |
- 5 A very interesting contribution with respect to adverbials is that of Anderson (...)
9The FID passage in the example contains third person pronouns, rather than first and second person pronouns as it would in direct speech. In the example, the realization of the personal pronouns parallels indirect speech: they are chosen according to the narrator’s context, that is, the external context with respect to the FID (see Eckardt, 2014: 30). As Bu (2016) underlines, the temporal reference is typically also set relative to the reference time of the non-FID story line (see further below). By contrast, as listed by Hinterwimmer (2019: 80 f.), 1st/2nd person pronouns, temporal adverbials (like just in the example)5, evaluatives and expressives are attributed to the protagonist’s context. This also includes the illocutionary force of a clause. In the example, the exclamation and the question are attributed to the protagonist (see e.g., Eckardt, 2014: 9 f. for these properties and further indications of FID). Finally, as the discussion of tense in Landeweerd and Vet (1996) indicates, tense choice in FID is language-dependent. For example, while English uses the simple past, the Romance languages opt for imperfective tense-aspect forms (see Banfield, 1982: 104). Therefore, tense cannot receive a general attribution to one context or the other (see also Giorgi, 2015: 238).
10Its specific properties allow for FID to be singled out with respect to competing phenomena within two realms. The first one is the representation of speech and thought. The second one is perspective shifts. We have mentioned many properties already; however, we have not yet categorized them properly.
11On the one hand, FID needs to be distinguished from other forms of speech or thought representation. Relevant indicators can be found on all linguistic levels. First, there are syntactic and morphosyntactic properties which FID only partially shares with other forms of speech and thought representation. Like direct speech it is not syntactically subordinated, but, like indirect speech (which shows syntactic subordination, however), “tenses and pronouns are shifted” (Vetters, 1994: 179, with reference to Lips, 1926: 51). Above, we mentioned the specific tense choice found in the Romance languages. This parallels indirect speech (Maier, 2015: 368, footnote 39). Second, FID shares lexical and semantic properties with direct speech, insofar as the protagonist may express evaluations which are only mediated by the narrator; the style and expressions of the protagonist may also be used (see Eckardt, 2014: 9). Third, with regard to pragmatics, temporal and local deictic expressions parallel the realization of direct discourse (see Hinterwimmer, 2019: 80 f.). Finally, despite the predominant use of FID in the written medium, Maier (2015: 349 f.) even presents evidence for phonological and prosodical indicators which parallel direct discourse, as dialects may be represented in FID.
- 6 The perspective may also be shifted in the other direction, that is, from a protagonist t (...)
- 7 An example of viewpoint shift presented in Hinterwimmer (2017) is [i]:
[i] When (...)
12On the other hand, FID is a specific kind of a shifted perspective and therefore also needs to be differentiated from other perspective shifting phenomena. The unifying property of this group is that a piece of information is not presented from the point of view of the narrator but from that of a protagonist6. The most important competing subtype is called viewpoint shift by Hinterwimmer (2017) and protagonist projection by Abrusán (2020), who follows Holton (1997). Hinterwimmer (2017: 283 f., 291) lists three properties which distinguish FID from viewpoint shift: (i) only FID features a context shift, (ii) FID is always a conscious speech or thought of the protagonist, while viewpoint shift may be unconscious, and (iii) FID only occurs at the root level, while viewpoint shift is also possible sentence-internally7.
13Importantly, the lists of properties of the two realms of phenomena may be further extended and deepened. Especially with respect to other forms of speech or thought representation, the research literature is very rich and its history spans several decades. By contrast, the research on viewpoint shifting phenomena is relatively recent. Both discussions continue to be lively. As said, the distinctions themselves are not the focus of our study, although the insights we present here may and should be integrated into the more general research. Therefore, we will not go into further detail. However, we still need to determine the categories involved in the description of the temporal relations of FID.
- 8 The “assessment time” presented in Anand and Toosarvandani (2018b) can be understood as t (...)
14The reference time mentioned above is a crucial concept in the analysis of temporal discourse structure and especially in the description of narrative texts. It is the point in time to which a narrative has developed when a certain event occurs (see Kamp et al., 2011: 199). It is central when it comes to determining temporal anchoring because, within a story, the location time of an event is anchored to a preceding reference time (see Kamp & Rohrer, 1983; and Becker & Egetenmeyer, 2018a, for formal specifics). It is also relevant for FID events which form part of temporal discourse structure. When a parenthetical verb co-occurs with an instance of FID, this verb contributes the reference time. In cases where a parenthetical verb is lacking, the FID event itself contributes the reference time (see Egetenmeyer, accepted a). This is discussed in depth in Section 2.2. However, we can already emphasize that upon this conception, the FID event can be temporally localized in a narrative in a specific way. Furthermore, the account does not involve an overly extensive terminology. The localization is determined on the grounds of concepts coming from Discourse Representation Theory (DRT) (see e.g., Kamp et al., 2011). An alternative might be, for instance, the assessment time as put forth in Anand and Toosarvandani (2018a and b)8.
15As we have seen, usually a narrator and a protagonist context are distinguished in FID. Now, when the story contains more than one protagonist, the FID content may in principle be attributed to any of them. Hinterwimmer (2019) analyzes which character gets to be chosen as the one to utter or think in cases of FID. He determines the properties leading to a ranking between the available characters and shows that it is the highest-ranking character, i.e., the most prominent one, to whom the FID is attributed, that is, who becomes the perspectival center. In fact, as Hinterwimmer (2019: 81) states, being prominent is a prerequisite for being a perspectival center. He shows that two groups of prominence properties are relevant, namely, local and global ones. The following quote neatly sums this up:
[O]nly those discourse referents are available as perspectival centers that meet one of the two following conditions: They are either maximally prominent in terms of grammatical function and thematic role/agentivity features in the sentence immediately preceding the one in FID mode or they are globally prominent with respect to the entire text segment containing that sentence.
(Hinterwimmer, 2019: 86)
16Thus, the relevant local prominence properties concern the status of the expression referring to the perspectival center in the preceding clause. Generally, in nominative-accusative languages, the subject role is understood as being the part of speech yielding the highest prominence value (see Hopper & Thompson, 1980; Aissen, 2003; Næss, 2007; etc.). It often correlates with thematic role features that also contribute to a high prominence value, such as agentivity and sentience (see Dowty, 1991; Primus, 2006; etc.). With regard to the global or macro-level prominence value, Hinterwimmer (2019: 87 ff.) shows that the relevant property is that the protagonist is a non-local discourse topic.
17To substantiate his claims, Hinterwimmer (2019) presents a set of examples where he modifies certain characteristics. One of them is cited below (see Hinterwimmer, 2019: 85). We introduce a partitioning to make the discussion easier.
[4.i] |
George entered the room and looked around cautiously. Susan was sitting at a table in the corner with her best friend. |
[4.ii] |
Susan looked at George hatefully. |
[4.iii.a] |
The dumb jerk had managed to make her look like an idiot at the meeting yesterday. |
[4.iii.b] |
The mean old hag had managed to make him look like an idiot at the meeting yesterday.
(Hinterwimmer, 2019: 85) |
18To have a closer look at local prominence properties, we need to ignore [4.i] for the moment and consider [4.ii] as the beginning of our text segment. The incompatibility which surfaces supports the role of the local prominence value. Hinterwimmer (2019: 84) argues that the sentence may plausibly only be followed by [4.iii.a] as only Susan is available as a perspectival center. This is so because in [4.ii], the noun phrase Susan appears as the subject and the denoted entity may be interpreted as the agent of a viewing event, i.e., “the proper name Susan is more prominent than George” (Hinterwimmer, 2019: 84).
- 9 Hinterwimmer (2019: 87 ff.) tests and further substantiates this claim by inverting (...)
19The picture changes if we also take [4.i] into account. In these two sentences, George is introduced as a discourse topic, and therefore becomes an alternative perspectival center for the FID (see Hinterwimmer, 2019: 85). In the way the example is presented, George is actually the more probable perspectival center (Hinterwimmer, 2019: 89)9. This underlines the role of the global prominence value.
20A thorough analysis of the temporal relations in passages containing FID requires a further distinction we only applied implicitly above, namely that the FID content needs to be distinguished from the FID event (see Egetenmeyer, accepted a). So far, the research literature has generally focused on the FID content. However, this entails a restrictive view which in fact inhibits our understanding of the temporal discourse structure. Egetenmeyer (accepted a) shows that the FID event is decisive in this respect and that the temporal relations between FID and the surrounding context are much more varied than is generally acknowledged in the literature. In the following, an overview of this account is provided; for more in depth considerations the reader is referred to the original paper.
21The FID content is the propositional content of what is understood as being uttered by the perspectival center. Temporally, it is only indirectly connected to the surrounding discourse. By contrast, the FID event is a discourse entity which is a proper part of the discourse passage it occurs in. It maintains specific temporal relations to the surrounding co(n)text. We can show this distinction with resumptive pronouns. In [5.i], we repeat the example from Section 2. The second and third sentences are FID. Interestingly, [5.i] may be followed by [5.ii.a], but alternatively also by [5.ii.b].
[5.i] |
Bob welcomed Anna with excitement. He had just found her car key! Wasn’t she lucky to have him? |
|
(Eckardt, 2014: 1, adapted) |
[5.ii.a] |
But itcontent was a lie. In fact, he had simply retrieved the spare key they had been keeping in the drawer. |
[5.ii.b] |
Itevent was only a short announcement as he quickly rushed back to the TV set. |
22In both continuations, the pronoun it is used resumptively. However, in [5.ii.a], the pronoun refers back to the content of the FID which precedes it, whereas in [5.ii.b], it makes reference to the free indirect speech act itself, i.e., the FID event (see Egetenmeyer, accepted a, for a further example).
23As said, the FID event is a discourse entity. Therefore it also has a temporal extension. This can be shown with the following sentence, which is a third possible way of continuing the fragment of [5.i].
[5.ii.c] |
Then he gave it back to her. |
24As the adverbial then in [5.ii.c] indicates, the FID event may also be referred to temporally. Thus, it is part of the discourse dynamics. Due to its property of being a discourse entity, the time point introduced via the FID event receives a certain prominence value. Therefore, we will also compare it to the value of the anchor time (see Section 4.1). With this as background, we can now discuss the possible temporal relations between the FID event and the surrounding context.
25As said, tense choice in FID is language-specific (see Landeweerd & Vet, 1996). For example, while English allows for the use of the simple past, the direct equivalent is uncommon in the Romance languages (passé simple in French, préterito indefinido in Spanish, etc.), which, by contrast, prefer imperfective tense-aspect forms (e.g., the French imparfait or the Spanish imperfecto) (see Banfield, 1982: 104). The possibilities concerning the temporal relations are, by contrast, expected to be generalizable (see e.g., Bu, 2016, for a recent account concerning English). Here, we restrict our focus to the relationship between FID events and their preceding contexts. Egetenmeyer (accepted a) additionally discusses the relationship of FID to the following context and the temporal development within FID.
26The FID event is anchored to a time point which is usually provided in the preceding context. There are three main temporal relations: inclusion, sequence and (overlapping) precedence, exemplified in [6], [7] and [8], respectively.
[6] |
Bob was sitting next to Anna. How she had looked at him! |
|
(Eckardt, 2014: 48, adapted) |
[7] |
Bob opened the window. What a lovely afternoon! |
|
(Eckardt, 2014: 99, adapted) |
[8] |
Tom shuddered. The ghost in the attic was making noises again. |
|
(Eckardt, 2014: 60, adapted) |
27The first sentence in [6] introduces a reference time within which the thought is realized. Without further context, this is a case of an inclusion relation. Here we see a clear difference between the temporal location of the FID event and the FID content. As the pluperfect form underlines, the looking event referred to by means of the FID (the FID content) occurs before the FID event. In [7], the FID event is realized after the event of opening the window. In contrast to Eckardt (2014: 99), Egetenmeyer (accepted a) argues that an intervening eventuality is also probable, namely, a perception event which then leads to the expression of the emotional reaction. In [8], by contrast, the FID expresses the thought which leads to the reaction expressed in the first sentence (an explanation relation in the sense of Asher & Lascarides, 2003). Therefore, the FID event has to be temporally located before the event of the first sentence, even though partial temporal overlap is possible.
28In the discussion of these short examples, we go beyond the current research literature with respect to temporal relations. Still, an important possible problem has not yet surfaced, namely, the case of alternative temporal anchors. The following two construed examples feature more than one possible anchor time. This phenomenon is not discussed in Egetenmeyer (accepted a).
[9] |
At midnight, John went for a walk. Then he went to bed. Oh my, it was so dark at midnight! |
[10] |
John cleaned the kitchen. He had prepared a cake. The next day, his friends would come over. Awesome! The perfect party! |
[11] |
John cleaned the kitchen. Before that he had prepared a cake. Awesome! The perfect cake! |
- 10 Concerning another example, an anonymous reviewer suggested considering the role of the (...)
29What we see in [9] is that, even though the FID content concerns the time of the walk and takes it up explicitly (at midnight), the FID event is still realized at the time point introduced in the second sentence (went to bed). The preferred reading is a generalizing statement in the light of the protagonist’s specific experience. By contrast, if a pluperfect form (had been [so dark]) is used instead of the simple past (was) the reading is rendered specific. The assertion can then only concern the time of the protagonist’s walk. Importantly, the temporal anchoring to the reference time introduced in the second sentence does not change. Thus, temporal anchoring is not easy to manipulate10.
30The FID in [10] and [11] lacks an explicit verb which allows us to test the effect of the verb on the flexibility of the anchoring. Interestingly, the most probable reading of [10] is that the FID expresses John’s excited anticipation. More specifically, the main perspective time equals the reference time introduced as part of the first sentence (R1). The second sentence presents a flashback. The third sentence maintains the perspective time, but with a forward-looking perspective. The same holds true for the FID. It is temporally anchored to R1. The example is altered in [11]. We might expect that adapting John’s exclamation in such a way as to concern the quality of the cake would lead to an anchoring to the second sentence. However, this is not the case. The dominant reading is, rather, that John reflects on his cake while cleaning.
31Importantly, in these examples, the temporal anchoring cannot be directly deduced from a verb of speech or thought. Thus, they cannot be treated in the same way as indirect discourse. However, as an anonymous reviewer correctly pointed out, the temporal anchoring parallels the functioning of temporal anaphora in non-FID contexts (see especially example [10]). Therefore, even though their different conceptual status may be considered a weighty argument in favor of this treatment, it is important to show that anchoring in FID contexts may indeed diverge from other, parallel contexts. The following example pair shows this.
[12.a] |
John cleaned the kitchen. Before that he had prepared a cake. Wow! The perfect cake, but such a mess in the kitchen! |
[12.b] |
John cleaned the kitchen. Before that he had prepared a cake. He thought that the cake looked delicious but the kitchen needed to be cleaned immediately. |
32[12.a] and [12.b] present approximately the same content. However, they diverge with respect to the form of speech and thought representation. [12.a] contains an instance of FID without an explicit speech or thought verb. By contrast, [12.b] features indirect speech involving a verb of thinking (thought), which allows for a direct anchoring. Both examples are linguistically felicitous. Interestingly, they yield different temporal structures. While, in [12.a], the dominant reading is again that the FID is realized at the time of the cleaning event, in [12.b], the thinking is rather anchored to the second event or its post-state, i.e., the time at which the cake is finished. As we see, a temporally explicit verb form is not necessary (had thought). This divergence leads us to the central question of the paper at hand. Which qualities license a time point as anchor time for FID? We hypothesize that it has to be a prominent time point. In [12.a], the FID event is most probably anchored to the highest-ranking time point, which functions as the perspective time of the flashback in the second sentence. In [12.b], by contrast, a structural recency effect leads to an anchoring to the pre-past time point introduced in the second sentence. In Section 3, we determine what makes a time point prominent.
33In the preceding sub-sections, we presented the problem of the temporal anchor of FID and underlined its relevance. We showed the deficiency of the research literature, which hardly even mentions the problem. However, as we touched upon very different strands of discussions, we want to conclude this section with a brief summary of the scope of the present paper. Furthermore, we seek to delimit it from two of our other papers which also concern FID (Egetenmeyer, accepted a; Egetenmeyer, submitted a).
34We begin with two restrictions, but we will also indicate the way in which the present paper helps to solve the respective problems. The contribution at hand does not focus on the question of how to determine instances of FID. Furthermore, it does not add new insights on how to determine the character to whom the FID content is attributed. However, it does contribute to the understanding and characterization of the two problems. Firstly, its main aim is to present new insights into the functioning of FID. More specifically, it seeks to determine the quality of the anchor time of FID events. This property is central when it comes to describing FID as part of the discourse in terms of temporal structuring. Thus, this is also relevant for a general characterization of FID. Secondly, we not only describe the prominence value of the anchor time with respect to other time points in the surrounding context, but we also seek to compare this property with the characteristics of the perspectival center, that is, the character who utters the FID. Thus, by comparing the prominence values of the two different types of entities (times and individuals), the account can also deepen our understanding of the quality of the perspectival center.
35We have contrasted the present account with many other publications. However, it has not yet been properly set apart from two of our papers which also deal with FID. Egetenmeyer (accepted a) focuses on the temporal relations relevant in contexts involving FID in a general perspective. The paper introduces the distinction between the FID content and the FID event and determines the FID event as a discourse entity. It discusses the possible temporal relations of the FID event with respect to the preceding and the following discourse and also the temporal development within (longer) FID events. All three types of relationships may be instantiated as precedence, co-temporality or sequence (with subtypes involving or not involving overlap, partial or perfect co-temporality, etc.). In contrast, this paper focuses on an important property of the temporal relationship of the FID event with the preceding discourse, namely, the temporal anchoring. It determines the temporal anchor and discusses its characteristics. It shows that the temporal anchor is a time point with an elevated prominence value. It follows Hinterwimmer (2019) in considering a sentence-based local level and a macro-level (beyond the single sentence). In addition, it considers the rhetorical relations involved, which are also taken into consideration in Egetenmeyer (accepted a). In order to properly develop its specific interest, this paper presents a thorough description of the indicators of the prominence value of time points (see Section 3.1 and the data analysis in Section 4). The discussion of how to determine the anchor time is also a new contribution to the scientific research on FID (see Section 3.2).
36Egetenmeyer (submitted a) is based on the insights from the present study. Specifically, it tests one of the findings of the present paper experimentally, namely, the prominence properties of the local level. To be able to do so, it goes beyond the scope of the present paper, refining some of the relevant insights in order to operationalize them, by using a forced-choice task. In contrast to the present paper, which is not entirely restricted to French, the experiment in Egetenmeyer (submitted a) only considers French data. The experimental results not only confirm that the findings of the present paper are correct but also indicate that they are relevant for speakers/hearers when they are confronted with instances of FID.
- 11 However, habituality may sometimes be found in the preceding context. It might be worth i (...)
37The free indirect speech or thought act (i.e., the FID event) has to occur at a certain point in time. We assume that, usually, the FID event is not intended to be virtual or generic or otherwise temporally detached from the context11. Thus, the time point at which it is realized is expected to be specific. Due to the coherence requirements of a text, this time point cannot be just any time, but has to be one that is available in the preceding context. However, the preceding context may be extended and, in principle, an unlimited number of time points are possible candidates. For the selection of a time point as a temporal anchor for a FID event, the relevant time point should have specific characteristics. We hypothesize that it is one with a relatively high prominence value as compared to surrounding time points.
38For the definition of prominence, we adhere to Himmelmann and Primus (2015: 38, 42, 44 f.) and the precisions presented in Becker and Egetenmeyer (2018a: 45 ff.). Prominence is defined as a relational property of an element as part of a specific domain which holds relative to other elements of the same type and pertaining to the same domain. A prominent element stands out within the set of equal elements in its domain. Thereby, it becomes an “attentional center […]” and may “serve as anchor […] around which experience is organi[z]ed” (Himmelmann & Primus, 2015: 42, 44). This property is highly important in temporal discourse structure where the prominence values of time points and intervals prove to be a central structural determinant. For temporal structuring, the relevant domain is a semantically determined coherent text segment, e.g., an episode (see Becker & Egetenmeyer, 2018a: 45 ff. for a formalization).
39Importantly, time points may be compared in terms of the prominence value they have (see below and, more specifically, Section 3.1). In the literature on temporal discourse structure, different types of time points are considered. In this analysis, we focus on two types, namely, reference times and location times (in the sense of, e.g., Kamp & Reyle, 1993; Kamp et al., 2011). Prominence is a gradient property. Four different values of time points have been shown to be relevant for temporal discourse structure (see Becker & Egetenmeyer, 2018a; Egetenmeyer, 2020): maximal prominence within a domain, an elevated prominence value, equal prominence and minimal or low prominence. A maximally prominent time point is the most important attractor in the domain it occurs in, which may be the case, e.g., for the first time point of an episode (see Becker & Egetenmeyer, 2018a: 47 ff.). A time point with an elevated prominence value diverges from surrounding time points due to certain features, e.g., precision or explicitness, which make it stand out. Equally prominent time points are, for instance, reference times introduced via certain events as part of a narration in sequence. Their prominence value is then set by default (see Becker & Egetenmeyer, 2018a: 49), thus leveling out minor prominence distinctions which do not influence temporal discourse structure. Finally, minimally prominent time points are not available as anchors (see Egetenmeyer, 2020).
40The research literature hardly considers the role of FID within discourse structure. What is more, the prominence value of the anchor time of FID events does not seem to have been treated yet. However, in a different context, namely concerning the adverb now, Altshuler (2010) presents a relevant insight. He states that “now supplies […] the time of an event em that requires a salient antecedent ([…] perspectival event)” (Altshuler, 2010: 252 f.). On similar lines, Anderson (2019: 53) presents evidence “that tomorrow takes the time parameter of a salient perspective as its reference time”. As we will see, this is paralleled in FID.
41In the following Section 3.1, we introduce properties that mark a time point as prominent. We also propose a basic classification of these properties, which parallels the one presented in Hinterwimmer (2019) for prominent protagonists in several respects. In Section 3.2, we focus on the question of how to identify which time point functions as an anchor. There, we discuss a range of data illustrating interesting problems.
42We seek to determine the prominence value of potential anchor times of FID. Therefore, we first need to specify more precisely what it is that makes a time point prominent in general. The properties influencing the prominence value of times are quite varied. Several are already noted in Becker and Egetenmeyer (2018a). Further relevant properties are mentioned in the context of prominence scales concerning individual referents when aspectual factors are taken into account (see Hopper & Thompson, 1980; Næss, 2007). They are lexical properties of the verb phrase which may be contextually updated. In the following, we synthesize these factors. The properties we discuss are potential “prominence-lending properties” (Himmelmann & Primus, 2015: 41). Paralleling Hinterwimmer’s (2019) account concerning the perspectival center, the properties are divided into local ones and others operating on a macro-level. Both kinds may interact or may possibly be superimposed on one another. As we will see, the two underlying properties yielding prominence in the domain of time are referential precision and structural salience or weight.
43The local prominence properties pertain to the sentence-level. They may be grammatical or lexical. The decisive underlying text-referential property is referential precision. A referentially very precise time point or time span may be overtly introduced by means of an adverbial expression. As Becker and Egetenmeyer (2018a: 47 ff.) show, a temporal interval introduced by a sentence adverbial tends to be of elevated prominence. However, referential precision can also be reached by means of the verb phrase, either with a specifying adverbial expression or without it. On this level, an important factor is the marking of a verb with a perfective tense or through Aktionsart properties such as telicity and reduced or minimal temporal extension, as found in achievements. In the terms of Hopper and Thompson (1980: 251 f.), the properties of boundedness and punctuality amount to the expression of the “effectiveness with which an action takes place”. Apart from telicity in a strict sense, ingressivity attributed to an eventuality may also be relevant for the prominence status of a particular time point. As is well known, grammatical and lexical aspect interact in a clause (see e.g., De Swart, 1998). Therefore, when determining the prominence value of a certain time point, the different contributing factors cannot always be singled out well, and it is important to keep their sum in mind.
44The macro-level prominence properties concern discursive chronology, on the one hand, and “tempus relief” in the sense of Weinrich (1982: 168), on the other. The chronological discourse structure allows us to determine the first time point of an episode or sub-episode which has an elevated prominence value. As Becker and Egetenmeyer (2018a: 47 ff.; 2018b) show, its structural importance may be specified on quantitative grounds, or, more specifically, a first time point is the one to which the highest number of other time points is directly or indirectly anchored. “Tempus relief” is described by Weinrich (1982: 168 ff.) as the structuring of texts into foreground and background, which in the Romance languages is indicated by tense-aspect forms: perfective tense-aspect forms tend to mark foreground and imperfective forms background eventualities. A further possibility for foregrounding an eventuality and the time point at which it holds true is to mark it as being in progress by means of a progressive form (see Becker & Egetenmeyer, 2018a: 54). Foregrounded eventualities are expected to be more prominent than backgrounded ones, a mapping which extends to texts with diversified story lines, where eventualities and the associated time points of the main story line are also expected to be more prominent than those pertaining to lower-ranked story lines (see Becker & Egetenmeyer, 2018a: 53 f.).
45In general, the prominence value of a time point is determined relative to the value of the surrounding time points in the same domain (see Egetenmeyer, accepted b and further references therein). The analysis of the prominence value of the temporal anchor of a FID event needs to consider its relationships to two other time points or groups of time points. The first one holds between the anchor and the time point contributed by the FID event itself (see Section 4.1). The second kind of relationship is the one between the anchor time point and further surrounding time points (see Sections 4.2 and 4.3). Up to now, we have spoken of the temporal anchor in an abstract fashion. In the following section, we will explore it more specifically.
46In many examples presented in the recent research literature, the temporal anchor of a FID event is easily determined. An important reason for this is that the examples often consist of only two sentences, where the second is the FID instance and the first introduces a time point which functions as the anchor of the FID. By contrast, original literary examples are not always equally simply structured. Furthermore, as we saw already in Section 2.2, it is easy to construe short examples which do not precisely fit the pattern.
- 12 The purpose of the sentence numberings is to make the discussion easier. We do not use (...)
47A relatively frequent semantic characteristic is that the FID event may be anchored to a time point introduced by a verb expressing or implying a thought or feeling. Such a verb can function as a trigger for the FID, but the FID can also be realized without one. In [13], the context contains a verb of sentience (schämte sich, “felt ashamed”). Even though it facilitates the occurrence of the FID, the FID is temporally anchored to a time point introduced by another verb. Thus, [13] exemplifies the issue of intervening sentences between the one introducing the anchor time point and the FID12.
[13] |
(1) Im Zimmer machte er kein Licht. (2) Er schämte sich, indes sie aus dem Dunkel hinaufspähte, das Zimmer zu beleuchten, das ihr gehört hatte. (3) Es regnete. (4) Wie viele Stunden hatte sie gewartet? (5) Gewi[ss] stand sie noch immer dort, mit ihrer letzten Hoffnung. (6) Das war nicht auszuhalten! (H. Mann, Der Untertan, 1918: 99) |
|
‘(1) He was afraid to turn on the light in the room. (2) While she stood out there in the dark, looking up, he was ashamed to light up the room which had belonged to her. (3) It was raining. (4) How many hours had she been waiting? (5) She was probably still there, waiting with her last hope. (6) This was more than he could stand.’ (H. Mann, The Patrioteer, 1921: 84) |
- 13 In the English translation, sentence (6) does not maintain the marks of FID from (...)
48The sentence directly preceding the FID is a description (es regnete, “it rained”). Temporally speaking, it is not relevant for the thought event. Rather, it holds true at the time of the eventuality we are looking for, and it includes the reference time in question (see Kamp & Rohrer, 1983: 254, whose conception of temporal development in texts we follow here). The sentence before it does not push the narrative time forward either. It expresses the feelings of the protagonist which hold at a time previously introduced into the discourse. Finally, the FID event is temporally anchored to the time interval introduced in the first sentence of the extract, namely, the time point when the character enters the room. Still, the realization of the FID in (4) to (6)13 is very readily accessible due to the expression of the protagonist’s feelings in sentence (2).
49The second interesting case in point is exemplified in [14], which shows a tendency quite different from that of [13]. The example may be considered a borderline case with respect to our account, because the anchor time does not seem to have an elevated prominence value. However, as we are about to see, the FID is licensed through a so-called viewpoint shift.
[14] |
(1) Dann fing sie an, neugierig unter den Visitenkarten und Briefschaften auf dem Sekretär zu kramen… (2) Dicht beim Tintenfa[ss] lag das wohlbekannte große Schreibheft mit gepre[ss]tem Umschlag, goldenem Schnitt und verschiedenartigem Papier. (3) Es mu[ss]te noch gestern Abend gebraucht worden sein, und ein Wunder nur, da[ss] Papa es nicht wie gewöhnlich in der Ledermappe und in der besonderen Schublade dort hinten verschlossen hatte. (T. Mann, Buddenbrooks. Zerfall einer Familie, 1963: 137) |
|
‘(1) Then she began to rummage curiously among the visiting-cards and letters on the desk. (2) Close by the inkstand lay the well-known large copy-book with the stamped cover, gilt edges, and leaves of various qualities and colo[∅]rs. (3) It must have been used the evening before, and it was strange that Papa had not put it back in its leather portfolio and laid it in its special drawer.’ (T. Mann, Buddenbrooks. The Decline of a Family, 1922: 158) |
50The first sentence of [14] introduces an eventuality in which a human protagonist is involved. The second sentence literally expresses a stative situation concerning a non-human entity, Schreibheft (“notebook”). Therefore, one might expect the anchoring to function in the same way as in the preceding example. In [13], sentences (2) and (3) expressed states and therefore the FID was anchored to a time point introduced in the first sentence. However, this is not the case in [14]. Rather, in [14], the FID is anchored to a time point associated with sentence (2). This is also unexpected as it does not mention any potential conceptualizer, which, however, is necessary for a FID to be realized. Still, the constellation can be argued for as follows. Firstly, sentence (3) expresses a thought concerning the notebook (Schreibheft) from sentence (2). Secondly, sentence (1) makes reference to (the beginning of) an action, under which the situation in (2) cannot be subsumed. Importantly, the situation in (2) is not asserted for the time introduced in (1) but for a posterior time. And the FID event is anchored to this posterior time. But how can the lack of an explicit reference to a speech or thought event in (2) and to a human protagonist be resolved? Most importantly, how can a conceptualizer, which is necessary to the FID, be retrieved? The answer is that the FID expressed in (3) is licensed by an implicit perception event associated with the second sentence. A necessary concept to resolve this is viewpoint shifting, also sometimes called perspective taking, as described, for instance, in Hinterwimmer (2017). While, in (1), a narrator tells the story, (2) shifts the viewpoint to the character (in Genette’s terms, 1980: 190, a “non-focalized narrative” is changed to an “internal focalization”). This means that sentence (2) asserts that the protagonist recognizes the surprising position of the notebook, and on these grounds, she then reflects on it by means of FID in (3). Thus, the anchor time is not merely introduced by a state (with potentially low prominence). Rather, the moment of recognition functions as the anchor time, which has at least an equal prominence value to the time in (1).
51To complete the picture, we repeat the following example, already presented in Section 2.2. It pertains to a further group which lacks reference to a speech or thought event and which does not show signs of viewpoint shift to the protagonist.
- 14 The original in Eckardt (2014: 48) actually contains a non-progressive form in the firs (...)
[15] |
(1) Bob was sitting next to Anna. (2) How she had looked at him! |
|
(Eckardt, 2014: 48, adapted)14 |
- 15 In the example, the time point is underspecified, a frequent phenomenon in natural (...)
52This example does not present the complications discussed above. Sentence (1) directly concerns the protagonist to whom the FID in (2) is attributed. Importantly, in (1), the protagonist is asserted to be in a position where a speech or thought event is not excluded. Apparently, this is enough to license the occurrence of the FID. Without further context, the progressive expresses that an eventuality is ongoing at a certain point in time (see Bertinetto, 1986: 120)15. What is important for us is that, therefore, the eventuality is asserted to hold true at an identifiable time point. The FID event is anchored to this time point.
53A further issue we want to address is the lack of a bounded time point in the sentence preceding the FID. Interestingly, such examples (see [16], taken from Egetenmeyer, accepted a) are presented relatively often in the literature (see e.g., Eckardt, 2014: 88, who presents a rather similar example).
[16] |
(1) Tom was unhappy. (2) Tomorrow was Christmas and he wouldn’t be able to celebrate with his family. |
- 16 The account is interesting in the context of FID, especially due to the insights (...)
54According to a superficial reading of [16], the state expressed in sentence (1) holds within a temporal interval which is extended, or even unbounded. As we saw in Section 2.2, FID events have a specific temporal extension, namely the location time of the uttering or thinking event. Apparently, this temporal interval is relatively short in comparison to a state like the one seemingly referred to in the first sentence of the example. On these grounds, a viable interpretation would be that the FID event simply takes a subperiod of the interval as anchor. The reference time of the FID event would be underspecified in such cases. Even though this is possible, a resolution process like the following one seems more probable. In [16], the FID content is the reason for Tom’s feeling. The two eventualities overlap. They may be described as approximately sharing the initial time point, with the state expressed in (1) still holding after the FID event expressed in (2) is completed. Thus, they determine each other in terms of their temporal properties (see the concept of “cospecification”, which Pustejovsky [1991: 422] describes for verb-argument relations). A similar thought with respect to the content specifications of indirect speech verbs and indirect speech can be found in Hunter (2016: 18). However, she goes beyond such an analysis when showing that, in indirect discourse, the relationships of the parenthetical and the embedded clause with respect to the surrounding context may diverge, as “both parts of a discourse parenthetical report can be rhetorically relevant” (Hunter, 2016: 32)16.
55To verify our hypothesis that the temporal anchor of FID is prominent, and to gain insight into the functioning of the assumed categories, we analyzed a specific data set in depth. It consists of fifty instances of FID. All examples were collected from the research literature. They include both literary examples and examples made up by the scholars concerned. Most of the examples were collected by an assistant who was deliberately given only rather rudimentary instructions in order to limit any possible bias. Because the examples are used to show many different characteristics of FID in the research literature, they represent a broad range of cases. The examples were generally not altered. However, those originating from literary works were counterchecked and, where necessary, further context from the original was added, to allow for precise insights on the macro-structural level.
56The following sections cover the characteristics of the temporal anchor of FID that may contribute to an increased prominence value. First, we investigate the rhetorical relationship between the sentences introducing the temporal anchor and the FID event (Section 4.1). We then compare the temporal anchor to other surrounding time points, focusing on local (Section 4.2) and macro-structural information (Section 4.3) as described in Section 3.1. As we will see, nearly all examples show prominence effects on at least one level. However, in Section 4.4, we discuss a possible counter-example. The discussion shows that our approach is not easily falsified.
57The first important relationship is the one holding between the clause introducing the anchor time point and the FID event. Their relative prominence value can be determined on the grounds of rhetorical relations, which are the semantic or conceptual connections holding between adjacent sentences (see e.g., Asher & Lascarides, 2003). Asher and Lascarides (2003) and others distinguish between coordinating and subordinating rhetorical relations. The distinction allows us to determine a hierarchical structuring (see Asher & Lascarides, 2003). Examples of coordinating relations are parallel, narration and contrast, while elaboration and explanation are examples of subordinating relations (see Jasinskaja & Karagjosova, forthcoming). The two groups differ in the way in which content is added to a preceding clause, which is realized either on the same plane (coordination) or in a subordinating manner. The hierarchical structuring is also used to determine which propositions are available for a further discourse structural attachment: according to the so-called right frontier constraint, only those propositions are available for attachment which are “on the right” of the discourse structure at a certain point of the discourse, where we may find more than one proposition when they are localized on different hierarchical planes (see Asher & Lascarides, 2003: 11, with reference to Polanyi, 1988). In the present contribution, the basic distinction between non-subordinating and subordinating relations will suffice. The distinction shares similarities with that between foreground and background (see Weinrich, 1982: 168 ff.). Still, the description in terms of rhetorical relations is much more detailed and, not unrelatedly, may sometimes lead to diverging classifications. It thus makes available a refined distinction in terms of prominence relations (see also Jasinskaja et al., 2015: 146 ff.).
- 17 It would have been preferable to take as a point of comparison the distribution of (...)
- 18 The relatively small set of data was specifically collected for the analysi (...)
58According to our hypothesis, the clause providing the anchor time point should be more prominent than the FID event. In terms of rhetorical relations, this means that the FID instance is subordinate to the clause via which the anchor time is introduced. In our data set, 37 of the instances investigated display subordinating relations (74% of the total) as opposed to 13 cases of coordinating relations (26%). We ran a simple Chi-squared test in order to determine whether the distribution deviates from an even distribution17. It yielded a p-value of p < 0.001. With certain reservations due to the special quality of the data, we rejected the null hypothesis18.
59The group of cases with subordinating relations comprises 25 elaborations (67.6% of the subordination group) and 12 explanations (32.4%).
60The first case in point is the elaboration relation.
- 19 Vetters (1994: 180) deviates from the original and from Lips (1926: 53) by inverting th (...)
[17] |
Un doute singulier me saisit: avais-je aimé réellement Judith? (Margueritte, L’avril. La confession posthume, 1905: 102; taken from Vetters, 1994: 180, who follows Lips, 1926: 53)19 |
|
‘Doubt came upon me: had I really loved Judith?’ (Vetters, 1994: 180) |
[18] |
Il annonça sa décision. Il allait se marier. (Vetters, 1994: 179) |
|
‘He announced his decision. He was going to get married.’ (Vetters, 1994: 179) |
- 20 We prefer this translation over the one presented in Landeweerd and Vet (1996: 159), wi (...)
[19] |
Elle le regardait de ses grands yeux fixes: pourquoi donc ne mourait-il pas, puisqu’elle ne l’aimait plus, et qu’il gênait tout le monde, maintenant? (Zola, La bête humaine, 1984: 302; taken from Landeweerd & Vet, 1996: 159) |
|
‘She observed him with her great staring eyes; why on earth did he not die, since she had ceased to love him, and he was now in the way of everyone?’ (Zola, The Monomaniac, 1901: 282)20 |
- 21 An anonymous reviewer had reservations with respect to this classification. (...)
61In [17], the FID expresses what the doubt (doute) is about. This seems to be a productive pattern in the case of subordinating relations. In the example, the preceding clause at least implies a thought event. Example [18] shows that the speech or thought event may also be referred to explicitly. By contrast, [19] exemplifies that the lack of a speech or thought event in the clause introducing the anchor time does not hinder the analysis in terms of rhetorical relations, here another case of elaboration21.
62As said, about one third of the instances with a subordinating rhetorical relation are explanations. This is exemplified in [20], where the FID presents the reason for the eventuality realized according to the preceding clause.
- 22 Rascoe’s (Zola, 1922: 199) translation diverges quite strongly from the original: “Nana (...)
[20] |
Du coup, Nana perdit la tête, étranglée elle-même par des sanglots nerveux. On abusait d’elle, à la fin! Est-ce que ces histoires la regardaient? (Zola, Nana, 1880: 256; taken from Vuillaume, 2000: 117) |
|
‘As a result, Nana lost her head, trembling with sobs so that she could hardly breathe. She was being abused after all! Did these stories concern her?’ (Our translation)22 |
63The data set contains 13 cases of coordinating relations (26% of the data set). Eleven of them are narrations, while the remaining two are a result and a contrast relation. The following two examples show a narration relation.
[21] |
She fetched herself a chair. She pitched her easel with her precise old-maidish movements on the edge of the lawn, not too close to Mr. Carmichael, but close enough for his protection. Yes, it must have been precisely here that she had stood ten years ago. There was the wall; the hedge; the tree. |
|
(Woolf, To the Lighthouse, 1974: 168; taken from Doron, 1991: 52) |
64In [21], the protagonist presents thoughts about her physical position. She realizes them after having placed the easel in the position referred to in the preceding clause. Thus, the thought expressed as FID clearly follows the positioning of the easel. As an anonymous reviewer pointed out, this is underlined by the use of here, which the protagonist can only employ after her arrival at the point in question. Paralleling our discussion above, [22] shows that a narration relation can come about even when the sentence introducing the anchor time point explicitly makes reference to a speech or thought event. In the example, the preceding clause already expresses a thought event. However, the content of the thought differs in the FID. Therefore, the FID event temporally follows the anchor time.
65The last example we present in this section contains a result relation (see Kehler, 2000: 193). According to Jasinskaja and Karagjosova (forthcoming), who follow Asher and Lascarides (2003), it is a coordinating relation.
- 23 We prefer the translation by Weston-Evans (Modiano, 2015) over the one presented (...)
[23] |
J’ai sorti de ma poche le “compte rendu” que j’avais signé. Elle habitait donc square de l’Alboni. Je connaissais cet endroit pour être souvent descendu à la station de métro toute proche. Aucune importance si le numéro manquait. Avec le nom: Jacqueline Beausergent, je me débrouillerais. (Modiano, Accident nocturne, 2005: 27; taken from Reboul et al., 2016: 265) |
|
‘I took the report I had signed out of my pocket. So she lived on Square de l’Alboni. I knew the area, as I often got off at the metro station close by. It didn’t matter that the number was missing. I’d work it out with her name, Jacqueline Beausergent.’ (Modiano, Paris nocturne, 2015: 19)23 |
- 24 The second sentence of the example is another instance of FID.
- 25 An anonymous reviewer suggested that it might also be a consequence relation, which wou (...)
66In [23], the FID in question (Aucune importance…, “It didn’t matter…”) makes reference to a result following from what is stated in the sentence preceding it24. More specifically, the result relation holds between the state of knowing the place and the assessment that the lack of more precise information is unproblematic25. In the passage, the reading of the train of thought comes about easily with the first-person narrator.
- 26 The divergence is not significant. The Fisher test leaves us with a p-value of p = 0.30 (...)
67As we have seen, the rhetorical relation holding between the clause introducing the anchor time and that of the FID is predominantly of a subordinating type. This speaks in favor of the hypothesis that the anchor time point is prominent. We assume that the subordinating relation is the prototypical case. Therefore, it may be expected to occur relatively frequently in the made-up examples. This is indeed the case: among the non-literary examples, there are only 14.3% of coordinating relations, while there are 30.6% among the literary examples26.
68In Section 3.1, we discussed the ingredients which determine the prominence status of a time point on the local level. We departed from the concept that prominence is a relational property. The prominence value is determined with respect to the value of the surrounding time points. What typically endorses a prominent reading are lexical and grammatical properties which enable the relevant elements of a clause to make reference to an ideally precise and bounded temporal interval of little extension. Explicitness may be contributed by adverbial expressions. With regard to eventualities, temporal extension is, to some degree, determined by the Aktionsart of a verb phrase. In certain languages including the Romance languages, boundedness may also be contributed by grammatical aspect.
- 27 We did not consider the cases where the low prominence value was overwritte (...)
69The data also support our claim on this level. In sum, 38 cases (76%) show an elevated or equal prominence value, which also has at least some prominence to it, or where a non-prominent value is overwritten. More specifically, 9 instances clearly show an elevated prominence value on the local level (18%), 25 are of at least equal prominence (50%), and in 4 instances (8%) a low value is cancelled out. By contrast, there are only 12 cases (24%) where the preceding time point has a low prominence value at the local level. Again, we ran a basic Chi-squared test, comparing the distribution among the group which showed prominence and the one without local prominence with an even distribution (null hypothesis). It yielded a p-value of p < 0.00127. Despite maintaining our reservations (see Sections 4 and 4.1), we therefore rejected the null hypothesis.
70The following two examples show cases of elevated prominence value.
- 28 Again, we present the translation by Vizetelly (Zola, 1901: 285) and not the one from L (...)
[24] |
Puis, en revenant vers la gare […], il songea, il s’étonna de sa démarche. Avait-il donc résolu de tuer Roubaud, puisqu’il disposait déjà de sa femme et de son argent? (Zola, La bête humaine, 1984: 305; taken from Landeweerd & Vet, 1996: 158) |
|
‘Then, on his way back towards the station […], he thought the matter over, and felt astonished at what he had just done. Had he then resolved to kill Roubaud, since he was disposing of his wife and money?’ (Zola, The Monomaniac, 1901: 285)28 |
[25] |
Du coup, Nana perdit la tête, étranglée elle-même par des sanglots nerveux. On abusait d’elle, à la fin! Est-ce que ces histoires la regardaient? (Zola, Nana, 1880: 256; taken from Vuillaume, 2000: 117) |
|
‘As a result, Nana lost her head, trembling with sobs so that she could hardly breathe. She was being abused after all! Did these stories concern her?’ (Our translation) |
71In [24], the anchor time point is the reference time which is introduced via the location time corresponding to s’étonna (“felt astonished”). Superficially, we see an ingressive verb marked with the passé simple, which is preceded by another verb with the passé simple form. We may deduce, therefore, that the reference time is at least of equal prominence. However, if we have a closer look at the temporal structure contributed by the lexical constellation, we find that the anchor time is rendered prominent by means of what we may call a climax. This can be argued for as follows. The first verb in the sentence preceding the FID makes reference to an extended (though delimited) temporal interval, while the second one expresses very little temporal extension. The two pertain to the same lexical field and the sequence makes their direct semantic relationship clear. Thus, the clause containing the second verb (s’étonna, “felt astonished”) not only marks the endpoint of the first one with songea (“thought”), but also presents a consequence. Therefore, it is semantically heavy. Additionally, compared to the preceding eventuality, its extension is lower. Altogether, the reference time which it introduces is rendered prominent.
- 29 Prominence in the context of argumentation is discussed in Anscombre and Ducrot (1997) (...)
72In [25] (see also Section 4.1), the anchor time is marked as prominent through the combination of the tense-aspect form (the passé simple in perdit, “lost”) and the adverbial expression du coup (“therefore”). More specifically, the temporal update is combined with the expression of a consequence, i.e., an argumentative update, which together elevate the prominence of the proposition in question (see Egetenmeyer, submitted b)29. Lexically as well, there is clear potential for the favoring of prominence. The verb expresses a change of mental state. Therefore, the event is a good candidate to stand out in the narrative. Temporal structurally, the asserted time has little extension and high explicitness, which also contributes to the potential of a high prominence value.
73In the following, we present two variants of example [25] where we have changed specific properties to determine their effect on the prominence status more clearly.
[26.a] |
Les yeux de Muffat s’emplirent de larmes. […] Au même moment, Nana perdit la tête, étranglée elle-même par des sanglots nerveux. On abusait d’elle, à la fin! Est-ce que ces histoires la regardaient? (Zola, Nana, 1880: 256, adapted) |
|
‘Muffat’s eyes filled with tears. At the same time, Nana lost her head, trembling with sobs so that she could hardly breathe. She was being abused after all! Did these stories concern her?’ (Our translation) |
[26.b] |
Nana pleurait toute la journée, étranglée elle-même par des sanglots nerveux. On abusait d’elle, à la fin! Est-ce que ces histoires la regardaient? (Zola, Nana, 1880: 256, adapted) |
|
‘Nana cried the whole day, trembling with sobs so that she could hardly breathe. She was being abused after all! Did these stories concern her?’ (Our translation) |
- 30 However, we left out two further short sentences which intervene in the original text.
74In variant [26.a], we have added a sentence from the original text and changed the adverbial expression30. This yields a parallel relation, which we use to discard the consequential reading. In [26.b], we have changed the verb in the sentence preceding the FID to the atelic verb pleurer (“to cry”). With the aim of excluding a monofocalized progressive reading (see Bertinetto, 1986: 164), which would elevate the prominence value, we combine it with an extended adverbial expression. The adaptions have the following effects. In variant [26.a], the prominence value is reduced to the default equal value. There is a punctual verb with perfective tense-aspect marking, but no additional prominence cues. Still, it appears to be felicitous. Variant [26.b], by contrast, is linguistically a little less natural. Here, we intend to reduce the prominence value of the anchor time even further. The imparfait verb without a monofocalized reading should yield a state in temporal structural terms (see Kamp & Rohrer, 1983). However, FID events strictly prefer non-extended anchor times. Still, despite being odd, the example does not seem to be completely excluded. We hypothesize that the clash is resolved in one of the two following manners. First, if the original reading of an explanation relation is maintained, the FID event is anchored to the left boundary of the eventuality as specified by the adverbial or a pre-phase of it. That means that the thought is realized at the beginning of the day in question, which leads to the crying. The second possibility is a re-interpretation as an elaboration relation. Then the thought event is understood as occurring iteratively throughout the day. Importantly, both possibilities again involve a prominence cline in terms of rhetorical relations (see Section 4.1).
75With regard to equal prominence, a very typical constellation involves the anchor time being introduced via a verb expressing a telic event.
[27] |
Un doute singulier me saisit: avais-je aimé réellement Judith? (Margueritte, L’avril. La confession posthume, 1905: 102; taken from Vetters, 1994: 180) |
|
‘Doubt came upon me: had I really loved Judith?’ (Vetters, 1994: 180) |
76In [27], also presented in Section 4.1, the verb saisit (“seized”), which is marked by the passé simple, expresses an eventuality which is close to being punctual. As there are no further local indications which would increase the prominence value, the reference time which it introduces is of equal prominence on the local level.
77Furthermore, an equally prominent anchor time may also be associated with a state when it is rendered ingressive. This is the case of the verb phrase was off and away in [28].
[28] |
Manford’s attention, tugging at its moorings, had broken loose again and was off and away. The how-many-eth dinner did that make this winter? And no end in sight! |
|
(Wharton, Twilight Sleep, 1997: 70; taken from Juillard, 2000: 86) |
78In [28], the highlighting of the left boundary of the state makes it available as a temporal anchor for the FID event.
79The data set contains twelve cases which do not show clear prominence-lending features (see Himmelmann & Primus, 2015: 41) on the local level and where this lack is not overwritten.
[29] |
La conversation avait recommencé. Les grands vins de Bordeaux circulaient, on s’animait; Pellerin en voulait à la Révolution à cause du musée espagnol; définitivement perdu. C’était ce qui l’affligeait le plus, comme peintre. À ce mot, M. Roque l’interpella. “Ne seriez-vous pas l’auteur d’un tableau très remarquable?” (Flaubert, L’éducation sentimentale: histoire d’un jeune homme, 1870: 180; taken from Vuillaume, 2000: 117) |
|
‘The conversation had started up again. Fine clarets were served and the talk became livelier. Pellerin could not forgive the revolution for the total destruction of the Spanish museum. That was what upset him most, as a painter. Catching this last word, Monsieur Roque asked him: “Aren’t you the painter who produced a really remarkable picture?”.’ (Flaubert, Sentimental Education, 2004: 371) |
- 31 An anonymous reviewer underlined the role of the reason for the psychological state of (...)
80In [29], the anchor time is introduced via a stative expression (en voulait à, “was angry at”)31. The reader might infer that the sentence makes reference to an utterance event, but, on the one hand, it is not clear what exactly would be uttered (see Hinterwimmer, 2017, who relates this characteristic to viewpoint shifts), and, on the other hand, this reading is not unequivocal as the sentence might also be understood differently, activating common ground information. Additionally, the time at which the event holds true is underspecified. Thus, the temporal anchor is not marked locally as being prominent.
81Additionally, there are four cases with characteristics that need a proper discussion. We classify them as cases where the low prominence value is overwritten. They share the property with the group presented above that the clause which contributes the anchor time does not feature indicators of positive prominence. In contrast to what we have seen above, however, this clause makes explicit reference to a speech or thought event. Due to this lexical specificity, the same strict criteria should not be applied to this group, because, importantly, they do not share the same complication as a diverging interpretation is thus excluded. An example is [30].
[30] |
Elle se demandait s’il n’y aurait pas eu moyen, par d’autres combinaisons du hasard, de rencontrer un autre homme; et elle cherchait à imaginer quels eussent été ces événements non survenus, cette vie différente, ce mari qu’elle ne connaissait pas. Tous, en effet, ne ressemblaient pas à celui-là. Il aurait pu être beau, spirituel, distingué, attirant […]. (Flaubert, Madame Bovary, 1951: 365; taken from Weinrich, 1982: 802) |
|
‘She would wonder whether there might not have been some way, through some different set of circumstances, for her to have met a different man; and she tried to imagine what they might have been, those circumstances which had not arisen, that different life, that unknown husband. For indeed not all husbands were like this one. He might have been handsome, witty, distinguished, attractive […].’ (Flaubert, Madame Bovary: Provincial Manners, 2004: 41) |
- 32 It may be added that the data set also contains two cases where the expression of speec (...)
82In [30], the anchor time is introduced via a non-telic (though possibly ingressive) expression (cherchait à imaginer, “tried to imagine”) and the example lacks any further indicators of prominence. Additionally, it should be mentioned that the passage is intended to have a habitual reading, which may also be understood as contributing to a low prominence value (see Becker & Egetenmeyer, 2018b). However, as noted above, the verb phrase makes explicit reference to a thought event. Due to the explicitness on the content level, the FID event is readily anchored to the anchor time. Therefore, we deduce that the use of verbs of speech or thought in the sentence contributing the anchor time may overwrite the lack of temporal precision32.
83In addition to the local prominence properties, we investigated global prominence properties. As discussed in Section 3.1, two main factors are relevant in this respect: the chronological structure and the “tempus relief” (see Weinrich, 1982: 168). The chronology is important in relation to the first time point of an episode. The “tempus relief” concerns the structuring of a text passage into foreground and background. One possible means of highlighting an eventuality as part of the foreground is, for instance, via a progressive form (see Becker & Egetenmeyer, 2018a: 54; 2018b). Another relevant grounding factor may be the partitioning into story lines.
- 33 In the literary examples, we added further context wherever possible. By contrast, in t (...)
- 34 We did not include the two cases of doubt. The Chi-squared test yielded a p(...)
84Again, the data set clearly speaks in favor of our account. Concerning the macro level, 39 cases are plausibly33 at least of equal prominence value (78%), 8 cases show clear marks of an elevated prominence value (16%) and 31 may be classified as equally prominent (62%). By contrast, only 9 instances involve indicators of a low prominence value (18%). Finally, there are two examples with somewhat ambiguous properties (4%). The simple Chi-squared test we ran to compare the distribution among the groups with and without prominence as opposed to an even distribution (null hypothesis) again yielded a p-value of p < 0.00134. For this third constellation too, we rejected the null hypothesis despite the reservations we have regarding the data (see Sections 4 and 4.1).
- 35 Banfield’s (1982: 219) example starts only with Il s’y montra gai, which she (...)
85In example [31], the prominence value of the anchor time is high because it is the first-mentioned time point of a new sub-episode (see Becker & Egetenmeyer, 2018a: 47 ff.). The original quote in Banfield (1982: 219) actually lacks the necessary amount of context for this to become apparent35, which is often the case in the literature. However, by adding further material from the preceding context we are able to underline the positional role of the event introducing the anchor time as being the first time point of an episode.
- 36 We have added now to the translation to induce the FID reading more clearly.
[31] |
Frédéric passa la sienne trois jours après. Avant de partir en vacances, il eut l’idée d’un pique-nique, pour clore les réunions du samedi. Il s’y montra gai. Mme Arnoux était maintenant près de sa mère, à Chartres. Mais il la retrouverait bientôt, et finirait par être son amant. (Flaubert, L’éducation sentimentale: histoire d’un jeune homme, 1870: 152; partly taken from Banfield, 1982: 219) |
|
‘Frédéric sustained his own thesis three days later. Before leaving for his holidays, he hit on the idea of a picnic, to bring the Saturday gatherings to a close. He was in high spirits on the day. Madame Arnoux was now36 away at Chartres with her mother. But they would come together again soon, and he would be her lover in the end.’ (Flaubert, Sentimental Education, 2004: 95 f., adapted) |
86In [31], the protagonist develops the idea of a picnic which he then hosts. There is, thus, a temporal gap between the two situations, the idea and the day of the picnic. The disruption in the story line also correlates with the formal means of a paragraph break. We can determine the second situation, the one we are most interested in, as a new (sub-)episode on the grounds that the time, the location and the persons involved change (see Van Dijk, 1982: 177). Thus, the anchor time point of the FID, which is the reference time introduced via montra (“showed”), is the first-mentioned time point of a new sub-episode and has, therefore, an elevated prominence value (again, see Becker & Egetenmeyer, 2018a: 47 ff.).
87The eventualities via which the anchor time is introduced are often part of the foreground (see below). Still, this does not per se yield a high prominence value. Prominence is only elevated if the eventuality in question, which pertains to the foreground, is preceded by a backgrounded passage (see [32]), or if, as already noted, the eventuality is specifically marked as being foregrounded.
[32] |
Je pensai à mon père, que j’avais peu connu, ayant dix ans lorsqu’il mourut; je revoyais mal son visage, dont la sévérité me faisait peur […]. […] Puis je revis mon enfance, mon temps d’examens à Paris, mon affection bizarre pour Judith, mes efforts et la position acquise pour lui plaire, puis son refus […]. […] Un doute singulier me saisit: avais-je aimé réellement Judith? (Margueritte, L’avril. La confession posthume, 1905: 102; partly taken from Vetters, 1994: 180) |
|
‘I thought of my father, whom I had hardly known, as I was ten years old when he died; I didn’t recall his face well, the severity of which frightened me. […] Then I relived my childhood, the time of my exams in Paris, my strange affection for Judith, my efforts and the position acquired to please her, then her rejection. […] Doubt came upon me: had I really loved Judith?’ (The translation is ours except for the last two clauses which are taken from Vetters, 1994: 180) |
88Example [32] has already been discussed, but without the additional preceding material. With it, the overall structure can be determined. In terms of rhetorical relations, there are three summarizing statements, namely those which contain the verbs pensai (“thought”), revis (“relived”) and saisit (“took hold”), which are then elaborated on (see Section 4.1 for subordinating rhetorical relations). The verbs of these summarizing statements are all marked by the passé simple, thus indicating that they belong to the foreground (see Weinrich, 1982: 168). The passage shows strict thematic coherence. Therefore, the eventuality via which the anchor time is introduced (saisit, “took hold”) is not described as the first time point of a new episode. The passage is also coherent concerning the persons involved (the main person is clearly the first-person narrator), and therefore does not show signs of a diversified story line either. Still, and importantly, saisit (“took hold”) introduces a time point where the relief switch from the background to the foreground takes place. Therefore, its prominence value is elevated with respect to the preceding context.
- 37 Among the 36 instances are 4 cases where the anchor time is also the first (...)
89In the data set, the anchor time is frequently introduced by an eventuality in the foreground, where the grounding does not increase its prominence value further. This is the case in 36 instances (72%)37.
[33] |
She fetched herself a chair. She pitched her easel with her precise old-maidish movements on the edge of the lawn, not too close to Mr. Carmichael, but close enough for his protection. Yes, it must have been precisely here that she had stood ten years ago. There was the wall; the hedge; the tree. |
|
(Woolf, To the Lighthouse, 1974: 168; taken from Doron, 1991: 52) |
90The passage in [33], already mentioned in Section 4.1, is about an important character in the book. The narration relation holding between the first two sentences of the extract makes it perfectly clear that the eventuality introducing the anchor time (pitched) is part of the foreground.
91In literary examples, the classification of foreground material is generally not problematic. Where necessary, additional preceding material can be taken into account. However, the data set also contains non-literary examples, where further linguistic material may simply be non-existent. The question is how to deal with such examples (see, e.g., example [34]). We argue that despite the reduced preceding context, for many of them being part of the foreground is plausible, partly because of an eventive preceding clause and partly also due to the lack of alternative interpretations available in the restricted context (see also footnote 33). Therefore, we also classify such cases as being of equal prominence in this respect (see [34]).
[34] |
Mary smiled. Tomorrow she would reveal her true identity at the press conference. |
|
(Hinterwimmer, 2019: 80) |
92The final group we want to discuss is the one where the anchor time of FID does not show a positive prominence value on the macro-level. This is the case when the eventuality introducing the anchor time pertains to the background, e.g., as part of a description. Our data set contains eight instances of this kind (see [35]).
[35] |
Charlot said quickly, “I couldn’t turn him out without so much as a piece of bread. That wouldn’t have been human. I let him out the back way”. Thérèse somb[er]ly looked away from him, watching the wet world outside. They could hear the rain coming up in gusts against the house, beating against the windows and dripping from the eaves. It wasn’t a night for any human being to be abroad in, and he thought, how he must hate Chavel. |
|
(Greene, The Tenth Man, 1985: 110; partly taken from Juillard, 2000: 84) |
- 38 In fact, according to Hinterwimmer (2019: 86), Charlot should be available as a (...)
93In [35], the verb in the first sentence following the direct speech (looked away) introduces a reference time. The state expressed in the following sentence (could hear) temporally includes this reference time and does not advance the story further (see, for this kind of analysis, Kamp & Rohrer, 1983: 254). Still, the FID event is introduced relative to the state. An important argument favoring this analysis is that the thought is realized not by Thérèse but by Charlot, who is only reintroduced in the stative sentence which expresses that both can hear the rain (see Hinterwimmer, 2019, and Section 2.1)38. Additionally, it is reasonable to assume that the ability to hear the rain implies the act of listening to the rain, which has a more or less specific beginning set after the looking away of Thérèse. The FID event may be taken to stand in a temporally close relationship to this beginning. What is of most interest to us, however, is that the FID is anchored to a time point which is part of a description (background). Thus, an anchor time may also be introduced via a non-foregrounded eventuality and lack macro-level prominence. However, as we have seen, this is not overly frequent.
94To find out whether there are any examples contradicting our hypothesis, we cross-referenced our results concerning the anchor’s rhetorical relation with the FID and its local and global-level prominence value. More precisely, we checked our data set for clauses contributing anchor time points holding a coordinating rhetorical relation with the FID and which also did not show any other sign of an elevated prominence value. Such cases are very rare. There are four instances with a coordinating relation which additionally show a low prominence value on one of the two levels but an equal prominence value on the other level. Together, they make up 38.5% of the cases with coordinating rhetorical relations and 10% of the whole data set. However, there is only one example where both levels are exempt from positive prominence indicators. This one true counterexample only represents 2% of the data analyzed.
95The example was already presented in Section 4.1 (see also the footnotes there for discussion).
[36] |
J’ai sorti de ma poche le “compte rendu” que j’avais signé. Elle habitait donc square de l’Alboni. Je connaissais cet endroit pour être souvent descendu à la station de métro toute proche. Aucune importance si le numéro manquait. Avec le nom: Jacqueline Beausergent, je me débrouillerais. (Modiano, Accident nocturne, 2005: 27; taken from Reboul et al., 2016: 265) |
|
‘I took the report I had signed out of my pocket. So she lived on Square de l’Alboni. I knew the area, as I often got off at the metro station close by. It didn’t matter that the number was missing. I’d work it out with her name, Jacqueline Beausergent.’ (Modiano, Paris nocturne, 2015: 19) |
96Despite the anchor time being classified as lacking an elevated prominence value regarding all three relational properties we consider in the present contribution, the reading of the example as an instance of FID is unimpeachable. Interestingly, however, it combines at least two properties which already make it special from the outset. On the one hand, there is a first-person narrator (see e.g., Vetters [1994: 192 f.] for a discussion of this property as a possible problem and, additionally, Eckardt [2014: 51] for the possibility of referential identity of narrator and protagonist), and on the other hand, the FID occurs in an embedded passage which already shows signs of thought representation before the FID in question. First of all, there is donc (“so”) (see e.g., Eckardt, 2014: 114 ff.). Second, apart from the verb of knowing itself (connaissais, “knew”), the non-distal demonstrative cet (“this”) might be interpreted as a hint towards a focalized perspective. Thus, as the example is quite a unique case, we may expect that such true counter-examples to our theory are indeed not frequent.
97In this contribution, we presented evidence for the elevated prominence value of the anchor time of FID events. We discussed three main categories: the rhetorical relation between the clause introducing the anchor time and the FID event, and the properties of the anchor time on a local level and on a macro-level. The last two paralleled the categories applied in Hinterwimmer (2019) (see below). Our hypothesis was favored in all three respects: we found many more subordinating (74%) than coordinating relations (26%) in our data. On the local level, the majority of instances were at least of equal prominence (76%). Also, concerning the macro-level, most cases could plausibly be determined as showing at least some prominence (78%). Our data set only contains one clear counterexample, which, however, is quite a unique case in other regards as well (see Section 4.4). Therefore, we conclude that there is indeed a strong tendency towards a prominent anchor time of FID.
98On these grounds, we want to briefly compare the properties of the anchor time point with those of the perspectival center presented in Hinterwimmer (2019). This will enable us to answer the question of whether the anchor time point parallels the prominence value of the perspectival center in FID. With regard to the prominence domain, the set of available protagonists is typically quite restricted in narratives, which is less the case with regard to time points. While there are often only a few protagonists available in a narrative, there may be close to an infinite number of time points. Still, due to computational restrictions, only recently introduced time points are expected to play a role in this respect.
- 39 An anonymous reviewer suggested, however, that this could be beneficial.
99The rhetorical relation between the preceding clause and the FID does not play much of a role in the analysis of the perspectival center (see Hinterwimmer, 2019)39. By contrast, on the local level, both the analysis of the perspectival center and the anchor time are based on grammatical and semantic features indicating an elevated prominence value. Finally, the macro-level is also at play in a parallel manner in both respects. Thus, even though the precise properties are clearly different in nature, prominence can be shown to be relevant in very much a parallel fashion.
100The present article contributes to the understanding of FID by focusing on a determining factor of the temporal relations involved, namely, the status of the temporal anchor. The detailed description of the prominence value of the anchor time amounts to an advancement in the research on prominence in language. An important further contribution lies in the comparison between the domains of times and of individuals. With this article as background, future work should broaden the database. A collection of FID instances that is independent of the research literature might lead to different frequency distributions and perhaps also to additional classification challenges. A further step is to substantiate the findings with experimental evidence (see Egetenmeyer, submitted a).