1 Introduction
1.1 Previous studies
Recently, much attention has been paid to the temporal adverbial now occurring with the past tense in narrative discourses. In such a case, now does not behave like a typical indexical expressionFootnote [2] because it cannot denote the speech time (Kamp & Rohrer Reference Kamp, Rohrer, Bäuerle, Schwarze and von Stechow1983; Kamp & Reyle Reference Kamp and Reyle1993; Recanati Reference Recanati2004; Lee & Choi Reference Lee and Choi2009; Altshuler Reference Altshuler, Aloni, Bastiaanse, de Jager and Schulz2010; Hunter Reference Hunter2010, Reference Hunter, Aloni, Kimmelman, Roelofsen, Sassoon, Schulz and Westera2012). In order to give a unified semantics for its indexical and narrative uses, Kamp & Reyle (Reference Kamp and Reyle1993: 595) assume that now refers to a time interval that overlaps the time of perspective from which the described event is viewed. The temporal perspective point can be either the speech time, as in (1a), or a past reference time, as in (1b).
In the case of (1b), the perspective shifts to the reference time of the last sentence in the simple past, and the situation is viewed from that temporal location (Kamp & Rohrer Reference Kamp, Rohrer, Bäuerle, Schwarze and von Stechow1983).
Kamp & Reyle (Reference Kamp and Reyle1993) further assume that now may combine with state predicates only when it occurs with the past tense in narrative, imposing a temporal overlap relation with a situation described by a prior sentence. The intuitive reason is that “when one describes something as going on at the time of the description, one must describe it as just that, viz. as something that is then going on” (Kamp & Reyle Reference Kamp and Reyle1993: 596). Altshuler (Reference Altshuler, Aloni, Bastiaanse, de Jager and Schulz2010) makes a similar claim, requiring overlap between the time denoted by now and the time of the antecedent event. Like Kamp & Reyle, he also assumes that now only combines with state predicates.Footnote [3]
However, this claim has been empirically challenged as now can occur with telic event descriptionsFootnote [4] that temporally follow the reference event (Lee & Choi Reference Lee and Choi2009, Hunter Reference Hunter2010), as in (2), for example.
Furthermore, now occurring in a sentence which describes a state does not invariably describe a situation that overlaps with its antecedent event, either. Consider the sentences in (3).
We naturally infer from (3a) that it became pitch dark only after Jameson turned off the light. (3b) does not exclude the interpretation that Irene’s desire arose only after she went to see the speaker.
Moreover, Kamp & Reyle and Altshuler leave unexplained the question of when and why the temporal perspective shifts. Recanati (Reference Recanati2004) and Hunter (Reference Hunter2010) answer these questions by claiming that the notion of contrast is essential for the use of now with the past tense. They argue that the use of now emphasizes a certain time and contrasts it with alternative times. While Recanati (Reference Recanati2004) lays down only a very liberal constraint, such that now can refer to a past time as long as the time is contrasted with another time that is farther than the time now refers to in a relative sense, Hunter (Reference Hunter2010) requires of the lexical entry of now that the modified eventuality not hold before and after the time referred to by now.
Hunter (Reference Hunter, Aloni, Kimmelman, Roelofsen, Sassoon, Schulz and Westera2012), on the other hand, abandons the idea that a contrast between two times is a lexical requirement of now and instead argues that rhetorical relations play a crucial role in the use of now. She employs Segmented Discourse Representation Theory (SDRT; Lascarides & Asher Reference Lascarides and Asher1993, Reference Lascarides, Asher, Bunt and Muskens2007; Asher & Lascarides Reference Asher and Lascarides2003), which postulates a hierarchical discourse structure in which utterances are connected in terms of subordinating or coordinating rhetorical relations. In SDRT, a new clause 𝛼 can be attached to another clause β in discourse in a subordinating relation such as background, explanation or elaboration, or a coordinating relation such as narration or contrast.Footnote [5] Hunter (Reference Hunter, Aloni, Kimmelman, Roelofsen, Sassoon, Schulz and Westera2012: 378) claims that now triggers a presupposition that must be bound to the time of its immediately superordinate antecedent clause. She argues that, if the clause containing now and its superordinate antecedent clause are connected through a subordinating relation, the time denoted by now must be ‘as close as possible to identity’ to the time of the antecedent clause depending on a particular rhetorical relation. For example, because explanation does not allow for the times to be identical (since causes invariably start before their effects), the use of now is subject to the relation of temporal overlap where the cause begins right before the result. She argues that in certain subordinating relations, now can be omitted without a change in truth conditions. On the other hand, if a past clause modified by now has a coordinating relation with its preceding clause, the two clauses together will be subordinate to another clause that becomes the antecedent of now. In this case, the antecedent is the common topic of the two clauses, and the use of now helps structure discourse better by putting a temporal break between the two. Although Hunter’s rhetorical relation-based analysis of now has better empirical coverage than previous studies and is intuitively attractive, her claim that the relation between the time denoted by now and the time of its superordinate clause ‘must be as close as possible to identity’ is not very precise.
1.2 Main claims of this paper
1.2.1 The temporal perspective shift as a function of narrative discourse
In the previous section, we have briefly reviewed major discourse-level analyses of shifted now. The general consensus in the literature is that now combined with the past tense signals a temporal perspective shift. On the other hand, when and why such a temporal perspective shift occurs are questions that have not yet been settled.Footnote [6] The answers are necessary to explain and predict the co-occurrence of now and the past tense. Unlike most previous analyses, I argue that the temporal perspective shift is in fact a function of narrative discourse itself rather than a constraint that now imposes.
Narratives are defined as a representation or a recapitulation of a series of connected events in which a verbal sequence of clauses matches the order in which those events occurred (Labov & Waletzky Reference Labov, Waletzky and Helm1967, Labov Reference Labov1972, Hopper Reference Hopper and Givón1979, Chatman Reference Chatman1980, Reinhart Reference Reinhart1984, Abbot Reference Abbott2008, among others). According to Labov (Reference Labov1972), a minimal narrative consists of a sequence of two clauses that are temporally ordered and separated by temporal juncture. Therefore, if a discourse does not contain at least one temporal juncture/break that separates and temporally orders two event descriptions, then it is not a narrative. In narratives, changing the order of the clauses results in a change in interpretation. Compare (4) and (5) from Labov (Reference Labov1972: 360). Both recapitulate the same past events, but only (4) is classified as a narrative.
Labov (Reference Labov1972) observes that clauses containing used to, would, general present, and subordinate clauses are not narrative clauses. In these clauses, a change in the order does not alter the meaning, as shown in (6).Footnote [7]
The literature on discourse studies posits the existence of a fundamental processing difference between narrative and non-narrative discourse types with regard to the use of tense (Caenepeel & Sandström Reference Caenepeel, Sandström, Aurnague, Borillo, Borillo and Bras1992, Caenepeel & Moens Reference Caenepeel, Moens, Vet and Vetters1994, Caenepeel Reference Caenepeel1995). Non-narratives, such as face-to-face conversational exchanges, are anchored to the deictic center (the point of speech), informing the addressee of significant recent events. In this case, the tense is usedindexically, i.e. it refers to the utterance time, which is salient in the speech situation. In narrative discourse on the other hand, the construction of a narrative time line whose existence depends on the narrative itself establishes an anaphoric relation between the described events (Kamp & Rohrer Reference Kamp, Rohrer, Bäuerle, Schwarze and von Stechow1983, Partee Reference Partee1984), rendering the relationship between the referential domain of the discourse and the utterance time less prominent. In other words, since the relationship between described events and the speech time is sometimes less obvious in narratives, greater discourse-internal restrictions are imposed on temporal relations, and tense and temporal adverbials are used to mark such relations between the states of affairs being described.Footnote [8]
I claim that narrative discourse is responsible for the temporal perspective shift from the utterance time to the narrative time line.Footnote [9] I further argue that certain context-dependent lexical items, like now, are flexible in their denotation in that they can refer to coordinates of marked context like narrative times. Assuming these premises are true, the occurrence of now with the past tense in narratives, denoting the current temporal perspective in an unfolding story, is not so surprising. As expected, this combination is awkward in non-narratives, as in (7).
The clauses in (7) would most naturally feature in non-narrative, conversational discourse. The examples in (7) lack a depiction of a series of connected and temporally ordered main events, i.e. they fail to meet the requirement of containing at least one temporal juncture. For example, (7a) lacks a specific eventuality that constitutes a story telling. In (7b), the events referred to in embedded clauses, fail to advance the story (Labov Reference Labov1972).Footnote [10]
Hunter’s analysis cannot explain why the discourse in (7) is unnatural because in these examples the now clause stands in a subordinating relation with its preceding clause, providing Background. Although the temporal relation is overlap – satisfying Hunter’s requirement for the use of now – the sequences are still odd. Kamp & Reyle’s (Reference Kamp and Reyle1993) analysis does not explain (7) either; now co-occurs with a stative predicate and the preceding clause provides the temporal perspective point, leaving unexplained why the use of now is awkward in these cases. Altshuler (Reference Altshuler, Aloni, Bastiaanse, de Jager and Schulz2010) cannot explain (7b) because now has an event antecedent of coming back to Rome in 2005, but the discourse is still not felicitous.
In sum, I assert that the temporal perspective shifts from the utterance time to a real or imaginary narrative time line in narrative discourse, which organizes events to tell a story, relying on the chronological order among them on the narrative time line. In narratives, as a consequence, now denotes the shifted temporal perspective (or, say, narrative present as in literary criticism), rather than the utterance time.
1.2.2 Temporal relations between a now clause and its preceding clause(s)
In the previous section, we have observed that the skeleton of a narrative consists of a series of temporally ordered event clauses, which are called narrative clauses (Labov Reference Labov1972) or foreground (Hopper Reference Hopper and Givón1979). Any number of clauses that elaborate, evaluate, or comment on the narrated main events can come between two narrative clauses, which are called free clauses (Labov Reference Labov1972) or background (Hopper Reference Hopper and Givón1979). In other words, the bare bones narrative structure can be fleshed out with background clauses, which are typically stative, imperfective, and irrealis and stay outside of the narrative time line. An example is given in (8) below, from Labov (Reference Labov1972: 361). Because free clauses are not part of narrative time line, they hold true for the entire discourse and thus temporally overlap with the narrated events.
The foreground vs. background distinction, which Hopper (Reference Hopper and Givón1979) claims is a universal of narrative discourse, helps answer the following two question: How are the temporal relations between a now clause and its preceding clause(s) determined? Does now impose a temporal overlap relation with its preceding clause, as has been commonly assumed? I argue against this predominant view and instead claim that the temporal relation of now clauses with their preceding clauses is in fact a function of the narrative structure of foreground and background and their temporal relations of precedence and overlap, which is determined by the aspectual properties of the clauses. Since Dry’s (Reference Dry1983) seminal work, many researchers working on the phenomenon of aspect and temporal interpretation in discourse (Kamp & Rohrer Reference Kamp, Rohrer, Bäuerle, Schwarze and von Stechow1983, Partee Reference Partee1984, Dowty Reference Dowty1986, Hinrichs Reference Hinrichs1986, Webber Reference Webber1988, Kamp & Reyle Reference Kamp and Reyle1993, Kehler Reference Kehler2002) have illustrated that telic (accomplishment and achievement), perfective, and inceptive clauses, which form foreground, move the narrative time forward, whereas atelic (activity and stative), imperfective, and modal clauses, which belong to background, maintain the given reference time. Therefore, the temporal relation that a now clause establishes with its preceding clause in discourse can be independently accounted for by the principle of narrative interpretation and clausal aspect, and needs not be stipulated as an idiosyncratic lexical property of now.Footnote [11]
First of all, as we have observed in (2) above, now does not invariably overlap with a time that is already introduced in the previous verbal context. It can refer to an updated reference time introduced by a telic event description that moves the narrative time forward. (2) is a clear counterexample to the claim that a now clause must overlap with or be identical to its preceding sentence, as Kamp & Reyle require, its preceding event sentence, as Altshuler claims, or its superordinate antecedent, as Hunter argues. This paper offers an account that deals with this case by simply adopting the temporal interpretation principles in narratives. As I have mentioned, since the narrative context creates a narrative time line that is shifted from the utterance time, the temporal location of events with respect to each other on the narrative time line is crucial for an adequate understanding. The rule of thumb for temporal interpretation in narratives is that telic event clauses move the narrative time forward, forming a foregrounded skeleton of narrative, while state clauses maintain the current reference time, providing background information that elaborates, evaluates, or comments on the foregrounded events (Dry Reference Dry1983, Kamp & Reyle Reference Kamp and Reyle1993). Given this, the fact that now combined with a telic event updates the temporal context with a new reference time, viz. (2), is perfectly normal. This, in turn, entails that the use of shifted now is compatible with both foreground and background in narrative and is not restricted to background, contrary to common assumptions.
As Hunter (Reference Hunter, Aloni, Kimmelman, Roelofsen, Sassoon, Schulz and Westera2012) points out, rhetorical relations play an important role in narrative comprehension, but once again, they are not specific to the meaning of now; rather, they derive from general relevance and coherence requirements. Hunter (Reference Hunter2010) observes that Kamp & Reyle’s (Reference Kamp and Reyle1993) example in (9a) below showing the aspectual constraint sounds awkward with or without now, and that the same event predicate with now becomes felicitous with more discourse context, as in (9b). She argues that the two clauses in (9b) ‘hang together in a way that the two sentences of (9a) do not. The event described by the second sentence in (9b) concerns an entity introduced in the first sentence. By contrast, the two sentences in (9a) don’t obviously have anything to do with each other and are thus awkward even without now’ (Hunter Reference Hunter2010: 62). This further supports my claim that a new sentence must cohere in a significant way with its immediate context, and that this requirement is at work with or without now.
I have asserted that, unlike Hunter’s (Reference Hunter, Aloni, Kimmelman, Roelofsen, Sassoon, Schulz and Westera2012) claim that aspect does not determine the temporal relations, but in line with many other researchers of aspect and temporal interpretation in discourse (Kamp & Rohrer Reference Kamp, Rohrer, Bäuerle, Schwarze and von Stechow1983, Partee Reference Partee1984, Hinrichs Reference Hinrichs1986, Dowty Reference Dowty1986, Webber Reference Webber1988, Kamp & Reyle Reference Kamp and Reyle1993, Kehler Reference Kehler2002), telic event clauses move the narrative time forward, while state clauses maintain the current reference time. However, note that there was one exception to this principle, namely (3) above, repeated in (10a), in which a past state clause modified by now describes a later event.
As we observe in (10b), without now, a temporal overlap interpretation is more natural. It appears that the presence of now overrides the default narrative temporal interpretation rule of state overlap. I argue that now does not impose a temporal precedence relation with its preceding clause, but rather changes the aspectual class of the predicate to make it an event description. Assuming this, a sequential reading in (10a) is no longer an exception to the temporal update rule in narratives.
This being the case, it is important to understand why and how now changes the aspectual class of the predicate it modifies. As observed in the next section, in my corpus data, now in the past-tense clause almost always occurs with temporary state (i.e. stage-level predicate) or instantaneous event descriptions, but almost never with property-describing states (i.e. individual-level predicate, Carlson Reference Carlson1977). That is, now denotes a time at which a change of state takes place. These temporary state predicates entail the events that start the temporary states. Because of this, I argue, they can be coerced to refer to their inception event when they are modified by now if the event in the prior clause is interpreted as causing the state. In such a case, they may describe events that follow the events in the preceding sentence because cause must precede effect.
De Swart (Reference De Swart1998) defines coercion as the phenomenon of change in the aspectual type of a proposition under the influence of modifiers such as tenses, temporal adverbials and aspectual auxiliaries, which coerce the proposition to the appropriate type. For example, (11a, b) below illustrate that a state can be coerced into an event by emphasizing the starting point or endpoint of the state. (11c, d) show that events can also be coerced into states by giving the sentence an iterative or habitual reading.
Coercion happens because of a mismatch or a clash between the lexical aspect of predicates and the input constraints of their modifiers, and has been analyzed as involving an implicit coercion operator (Moens & Steedman Reference Moens and Steedman1988, Pustejovsky Reference Pustejovsky1995, Jackendoff Reference Jackendoff1997, de Swart Reference De Swart1998, Rothstein Reference Rothstein2004).Footnote [12]
Note that, unlike the aspectual coercion cases in (11), which exemplify a mismatch between argument and functor within a clause, our example of coercion involving now in (10a) takes place when a now clause combines with a preceding clause in discourse. It is triggered by a mismatch between the aspectual class of the new sentence, which is a state, and the narrative interpretation requiring that the cause precede the effect. This means that the source of coercion can be sentence-external (Egg Reference Egg2005, Dölling Reference Dölling and Robering2014). However, although the level of application is broader, I assume that the same mechanism that is responsible for coercion in (11) is at work in (10). First, the coercion operator, much the same as other aspectual operators like progressives and perfects, is an eventuality modifier, i.e. it maps a set of eventualities onto another set of eventualities (de Swart Reference De Swart1998). Second, coercion occurs when there is a mismatch or a clash in semantic composition. Once we expand the level of semantic analysis from sentence to discourse, as this paper does, the scope of coercion is not limited to the intra-sentential level. I will model the coercion process formally in Discourse Representation Theory in Section 3.
1.2.3 Change of state
I have argued so far that a shift in temporal perspective occurs in narratives and that now functions in these cases to indicate the current temporal perspective. I have further argued that the temporal precedence and overlap relation between a now clause and its preceding clause follows from a more general and independent principle of narrative interpretation that hinges on the aspectual properties of clauses. Foregrounded telic events on the narrative time line update the temporal perspective, whereas backgrounded states maintain it. These explanations leave the lexical meaning of now rather vacuous, predicting that now can freely occur with past tense as long as the discourse in question is a narrative. However, the existence of a narrative time line and the occurrence of a perspective shift in and of themselves do not license the use of now, which seems to be subject to further discourse constraints. One further constraint in the use of shifted now, I claim, is the change of state meaning. That is, now invariably indicates a change of state, denoting the turning point dividing the past and the future seen from the contextually salient narrative time.Footnote [13] This claim is based on the observation in my corpus data that now in the past-tense clause almost always occurs with temporary state (i.e. stage-level predicate) or instantaneous event descriptions, and hardly ever with property-denoting states. Moreover, as we will see in the next section, the time at which such a change of state occurs provides an important piece of information for an adequate understanding of the narrative.
As I previously mentioned, Recanati (Reference Recanati2004) and Hunter (Reference Hunter2010) also consider temporal contrast or change of state as an important component of the meaning of now. However, their notions are slightly different than the change of state meaning of now that I advocate here. Hunter’s (Reference Hunter2010) lexical entry of now imposes the condition that the described situation did not hold or will not hold at some time immediately in the past or immediately in the future of the time denoted by now.Footnote [14] However, John left the gang and was now a changed man does not imply that he will change back to a bad person again. Therefore, the condition that the situation does not hold in the future seems too strong. Moreover, in her later paper, Hunter (Reference Hunter, Aloni, Kimmelman, Roelofsen, Sassoon, Schulz and Westera2012) abandons this claim, arguing that contrast meaning is not part of lexical meaning of now but a mere pragmatic by-product, putting forward a rhetorical relation-based analysis of now instead. Recanati (Reference Recanati2004: 19) discusses a temporal comparison or contrast between two times but is not necessarily concerned with a change of state: ‘now can refer to any period, provided it stands in contrast to another, more distant period. On this view, “here” essentially contrasts with “there”, and “now” with “then”. When considering two times, or two places, if one is thought of as closer than the other, we can refer to the closer one as “now” or “here”’.
To support better my claim that the shifted now can only occur when a change of state is asserted or implied, let us look at some attested examples.
The use of now in (12) asserts a change of state that has happened to the speaker’s mother, contrasting her current vegetative state and her former normal state. Here, now combines with a state description hold no essence, and thus overlaps with an event of sitting with her. Hunter (Reference Hunter, Aloni, Kimmelman, Roelofsen, Sassoon, Schulz and Westera2012) uses this example to explain that in subordinating relations – in particular, elaboration – now can be omitted without affecting the interpretation. I argue that now still adds the meaning of contrast and change of state, indicating that such a change happened not long before and has been continuing until the event that serves as the temporal perspective point, i.e. she sat down with her mom. Without now, such a change of state is only pragmatically implicated without being asserted. When now modifies an event description that moves the narrative time forward as in (13) below, the change of state meaning of now is shadowed by the fact that events inherently entail a change of state by themselves.Footnote [15] However, there is still a clear contrast or change of state between not having caught fire and exploding, the latter of which happened only after dipping into liquid air. The explosion is consequential upon the action of dipping.
Why does now entail change of state? As pointed out in Hunter (Reference Hunter2010), since now refers to the time that is salient in context (either utterance or discourse context), it is often redundant and unnecessary unless it highlights an eventuality that holds during the contextually salient time. The eventuality becomes highlighted typically when it is contrasted with its opposite state. (14), which is a slightly modified version of (1b), provides some evidence for my claim.
(14) implies that there were more than ten people in the room before Mr. Johnson started to speak at six o’clock and the number of people shrunk to ten by that time. This implication cannot be cancelled by adding the last sentence in the discourse. Without now, however, the discourse sounds better.
In sum, now not only refers to the shifted temporal perspective in narratives but also indicates a change of state. I will elaborate on this point through the narrative examples in the next section.
1.2.4 Summary and roadmap
Let me now summarize and reiterate my main claims: (i) the temporal perspective shift is a function of narratives, reflecting the fact that narrative discourse creates a narrative time line, which serves as the current temporal perspective to which now may refer; (ii) the temporal relation of a now clause with its preceding clause is a function of the aspectual property of the VP it combines with, and can be independently accounted for using the much studied phenomenon of aspect and temporal interpretation in discourse; and (iii) the use of now signals a change of state, and the time at which such a change of state occurs provides an important piece of information for an adequate understanding of the narrative.
The rest of the paper is organized as follows. In Section 2, I examine naturally occurring narrative examples quantitatively to support the above claims. Section 3 proposes a semantic analysis of now in Discourse Representation Theory (DRT; Kamp & Reyle Reference Kamp and Reyle1993). Section 4 concludes the paper by summarizing the main points and discussing their implications.
2 A quantitative analysis of now in narrative discourse
In this section, I will present the results of my corpus analysis of now in narrative discourse data. I examined 100 randomly selected samples of narrative discourse containing now from the British National Corpus (BNC).Footnote [16] Using the source information, I only chose short stories and novels, which consist of typical narratives. I examined discourse in which a sentence with now is preceded by three to four sentences and followed by another. Table 1 summarizes the tense and aspect forms that are used with now in my corpus data. As we observe in Table 1, among the 100 examples of now in discourse, 37% occurred with the present tense, and 63% occurred with the past tense. When occurring with the past tense, 63.5% (40 out of 63) were paired with state predicates, while 38% (23 out of 63) were paired with event predicates, including 17 (28%) telic verbs (predicates that denote an event with a built-in end point, e.g. build a house, die) among the latter.
Table 1 The distribution of now sentences in narrative discourse.
2.1 Now with the present tense
When used with the present tense, a clear majority (33 out of 37 instances, 89.2%) appeared in direct speech put between quotation marks. In the present tense in both quoted direct speech and simple narratives in my corpus, now did not occur with telic event descriptions, but only with state descriptions. (15) contains an example of now occurring in a quoted direct speech in narrative.
(16) is an example of now occurring in a present-tense sentence that is not a quoted direct speech. Although a state predicate is used in this example, a temporal sequential relation can be easily inferred, although not required.
I have claimed that such a case involves a type of coercion from a stage-level state predicate (Carlson Reference Carlson1977) to an inchoative (i.e. inception/beginning) event that inaugurates the state. When the preceding event clause describes an event that causes such a change of state, now implies that the state starts to hold right after the event because the state cannot hold before the event that causes it to take place. The first clause in the previous sentence in (16) describes an event (Marie touching the narrator’s hand) that caused him to get upset. Without now, the second sentence describes an overlapping state with the preceding event.
2.2 Now with the past state sentence
As we see in Table 1, among 40 state predicates in the past tense, 14 state verbs and 27 grammatical state constructions were used.Footnote [17] One important observation in my corpus research was that 13 out of 14 state verbs were stage-level predicates but not property-denoting or individual-level predicates. These predicates include was ill, was his turn, felt not so slender, knew where he was going, looked a little less cracked, etc. There was only one example where a property-denoting predicate appears with now in my corpus, which is given in (17) below. Even here, the use of now strongly implies a change of state, i.e. Dawson City was not a ghost town but has turned into one.
Based on this observation, I assert that, because shifted now almost always occurs with temporary state descriptions in my corpus, together with the fact that it implies a change of state even when it occurs with permanent or property-denoting ones, as in (17) above, shifted now is used only when a change of state is implied. This is more important for states because events involve change of state in their lexical representation intrinsically.
Let us look at some more examples. In (18), now modifies a temporary state predicate, be cold.
What licenses the use of now here is the change of state and the temporal contrast between the current state modified by now and its prior opposite state: Helen was calm during the night but was now cold and shaking. The clause modified by now describes a situation that overlaps with a situation in the preceding telic event clause: It is inferred that Helen was cold when Anne agreed that she should be in bed. The now sentence also provides an explanation for the previous clause: Because Helen was cold and shaking with shock, Tony took her away and Anne agreed that she should be in bed. This discourse-relevance constraint, I argue, needs to be met independently, regardless of now. The only contribution now makes in discourse is that the described state, which typically overlaps the event in the preceding telic event clause, did not hold before that event, and this information is relevant for a coherent interpretation of the discourse.
(19) illustrates a case in which a state verb (modified by now) moves the narrative time forward.
It is inferred from (19) that he felt an urge to back away after she claimed she could help him, which caused him to want to flee. Here, the meaning of contrast is obvious: He did not feel the urge to back away before she said she could help, but he began to feel that way after hearing those words. Among the 40 cases in which now modifies a state predicate in the past, I found nine instances (22.5%) of coercion from a temporary state predicate to an inchoative event. The remaining 31 cases follow the general state overlap principle.
Let us next observe grammatically state constructions (such as progressives and perfects), whose temporal relationship with the preceding event clause is almost exclusively an overlap relation. That is, coercion is extremely rare in grammatically state constructions probably because of a conflict between imperfective aspectual operator and perfective coercion operator.
It is inferred from (20) that the stairs were burning when the narrator attempted to venture down them. The use of now here indicates a change of state from the stairs not burning to being consumed by flames. The discourse connective but explicitly cues the contrasting meaning.Footnote [18] The state of stairs burning began before the event of speaker trying to go down them.
(21) illustrates a case where a clause immediately preceding the now clause is also a state.
When a now clause is stative in the past tense, it describes a situation that overlaps with a situation in the preceding clause. If the preceding clause is similarly a state, as in (21), that state would also have to rely on a preceding telic event clause for its temporal location. This simply follows from the fact that background clauses, which occur between foreground telic event clauses, depend on the foreground event clauses for their temporal interpretation, i.e. they hold throughout these events. Since both the preceding pluperfect clause as well as the past progressive clause modified by now constitute background, they overlap with the foregrounded event in the first clause describing her thinking. His security escort began dishing him up to her right before she thought it was funny how things had worked out, providing an explanation for the event in the first clause. The clause immediately preceding the now clause, which is in the pluperfect form, makes explicit the contrastive meaning.
So far we have observed that past-tense state predicates modified with now either overlap with the event in a preceding clause, or are coerced to refer to a following event when the event in the preceding clause is the one causing the event in the now clause. One important difference between typical state overlap and the overlap relation in now clauses is that the former holds just because states have no beginning or end point, and they are assumed to persist unless otherwise indicated, whereas the latter clearly involves a change of state, and the inception of the changed state happens right before the event in the prior clause. In that way, it becomes relevant for the interpretation of the discourse.
2.3 Now with the past event sentence
Let us turn to now in past event sentences. Among 100 discourse examples, 23 instances of now occurred with event descriptions. I found only six examples in which now occurred with an atelic (i.e. activity verbs, which lack a built-in endpoint, e.g. run) clause. Like most state clauses, atelic event clauses do not normally move the narrative time forward, describing instead an event that overlaps with the eventuality introduced by the preceding clause. (22) is an example of now modifying an atelic clause, in which a temporal overlap relation is observed.
We most naturally infer from (22) that Tamar felt regret while she was riding from the yard. It implies that she did not harbor such bitter regrets before they rode from the yard without talking to each other. The event in the preceding clause provides an explanation for the event in the now clause.
In my corpus data, I found 17 instances in which now modifies a telic event predicate. All of the telic verbs were achievementsFootnote [19] that describe a turning point between two opposite states, such as bring, find, become, open, turn, return, take notice, come, rekindle, accept and start, some of which are given in (23) and (24).
In these examples, it is clear that the clause modified by now indicates a change of state, moving the reference time forward. In (23), the event of turning follows the event of nodding. In (24), the king’s son opened the box to take out the sun’s ray, which only became necessary when they came to the very dark islands of the Black Sea. (24) shows that the time of opening is not anaphoric at all in a strict sense. The first sentence in the discourse introduces the box for the first time, so opening it cannot overlap with a time that is already introduced in the previous discourse.
Although most of telic event clauses modified by now move the narrative time forward, there were some exceptions in my corpus, specifically those in which a now clause elaborates on its preceding clause, as shown in (25).
It is well known that telic event sequences do not always move the narrative time forward, specifically when the second clause describes an event that is included in the event mentioned in the first clause, as illustrated in (26).
Since the second event is part of the first one, the former must be temporally included in the latter. The second clause in (26) is part of background, elaborating on the foregrounded event in the first clause.
2.4 Summary
First, we have observed in the corpus data that shifted now almost exclusively occurs with stage-level state predicates or instantaneous events that result in a change of state (i.e. achievements). The fact that now combines with these predicates suggests that the perspective time denoted by now is the very moment at which a change of state occurs, i.e. a turning point that divides the past and the future. As a result, the use of now in narratives typically implies a temporal contrast between opposite states.
Secondly, we have observed that now with state sentences does not invariably describe an overlapping state but sometimes describes a later event. I have argued that a type of coercion from a stage-level state predicate to an inchoative event that starts the state is at play in such a case. The coerced inception event is instantaneous, like achievements, and now combines with it to describe a change of state that follows the given reference time.
Thirdly, the temporal inferences in discourse that we have observed follow independently from the principle of narrative progression, rather than from the meaning of now. When now combines with a state, the overlap relation follows from the general principle that state clauses overlap with the current reference time. On the other hand, when now combines with telic events, it triggers the sequential interpretation because event descriptions move the narrative time forward. Through the corpus data, we have observed that discourse containing now is subject to the same principles.
3 An analysis of now in Discourse Representation Theory
In this section, I will provide a discourse-level, formal semantic analysis of now using Discourse Representation Theory (DRT; Kamp & Reyle Reference Kamp and Reyle1993, van Eijck & Kamp Reference van Eijck, Kamp, van Benthem and ter Meulen1997). In DRT, it is the Discourse Representation Structure (DRS), not the individual sentence as such, which is semantically interpreted. Hence, the theory is useful for analyzing narrative discourses and the anaphoric relations among sentences within it. Observe how DRT analyzes sequences of sentences as in (27).
A DRS for tensed discourse includes events, states, and location times as objects in the universe of discourse, specifying relations of precedence and inclusion among them. The eventuality described by a non-initial sentence e is interpreted as related temporally to some other event eˈ introduced by the preceding discourse context. The temporal relation between e and eˈ becomes especially important when the new sentence does not contain a temporal adverbial, which typically determines the value for the variable for the location time, so that the new eventuality e can be located only in relation to the antecedent context. Although the way in which e and eˈ are related could depend on a number of different factors, the most important one is whether eˈ is an event or a state. If eˈ is an event, then it is typically understood as following the event e. This case is illustrated by the second sentence of (27a) above. When the second sentence is interpreted, the event of a man coming in is the last mentioned event. And the new event, that of him sitting down, is naturally seen as following the event of his entering. If eˈ is a state, on the other hand, the relation is typically that of overlap. This is the case for the second sentence of (27b) above. From (27b), it is inferred that the man was happy when, not after, he came in. Kamp & Reyle (Reference Kamp and Reyle1993) adopt Reichenbach’s (Reference Reichenbach1967) notion of reference time (Rpt) by introducing into the DRS a condition of the form Rpt: = 𝛼, where 𝛼 is some discourse referent which represents a time or an event already present in the DRS. (28) below is the DRS for (27). A Discourse Representation Structure K is a pair of a set of discourse referents
$\text{x}_{1}\ldots \text{x}_{\text{n}}$
(universe of K) and a set of DRS conditions C
$_{1}$
…C
$_{\text{n}}$
, which are separated by
$|$
from each other.
The first sentence of (27a) introduces an event discourse referent e
$_{1}$
of a man coming in, which is included in its location time t
$_{1}$
. The location time precedes the utterance time. The processing of the second sentence of (27a) involves adding the condition Rpt: = e
$_{1}$
, meaning that e
$_{1}$
acts as a reference point for the second sentence. The second sentence describes an event, and thus the relation between this event and the reference point is succession. On the other hand, the second sentence in (27b) is a state, and its interpretation will again require the choice of a reference time, which is the event described by the first sentence e
$_{1}$
. Since the eventuality denoted by the second sentence is a state, the processing principle entails the inclusion of the reference point. The DRS thus constructed is true iff there is an embedding function that verifies the DRS in the given model.
Kamp & Reyle (Reference Kamp and Reyle1993: 596) introduce the Temporal Perspective point (TPpt) in addition to Referent point (Rpt); they use Rpt for narrative progression, as we have observed, and TPpt for perspective, which is crucial for the interpretation of shifted now and pluperfects.
As we see in (29b), Kamp & Reyle equate TPpt with the speech time for a simple past event sentence in the present tense. When the second sentence of (29a) is interpreted, however, TPpt shifts to the time of e
$_{1}$
, providing a temporal perspective from which the event described in the second sentence is seen as past. Therefore, the TPpt is reset to e
$_{1}$
(a strikethrough on the previous TPpt, which is n, represents the resetting).
Kamp & Reyle argue that the simple past event description has the relation [TPpt at the utterance time; described eventuality before the utterance time], whereas the simple past state description has the relation [TPpt before the utterance time; described eventuality before the utterance time]. This is forced because of their assumption that now with the past tense, which signals a temporal perspective shift, can only occur with state sentences and, therefore, only state sentences involve a temporal perspective shift to the location time of a preceding event sentence. However, they do not want to claim that whenever a state sentence is processed, the TPpt shifts to the location time of the preceding event. For instance, in the example given in (30) below, the TPpt is at the speech time for the first sentence, then shifts to the time of the man entering the bar for the second sentence, because this sentence is a state description, and subsequently shifts back to the utterance time for the third and the fourth sentences.
In order to avoid this undesirable and unintuitive result, Kamp & Reyle argue that even simple past state sentences allow the relation [TPpt at the utterance time; described eventuality before the utterance time]. This allows us to keep the TPpt at the utterance time for an extended discourse like (30) above. However, as one can easily see, the price to pay is the multiple ambiguity analysis of the simple past tense in English. We also cannot predict when TPpt overlaps the state eventuality and when it does not. Kamp & Reyle seem to suggest that the occurrence of now with the past tense will clearly indicate the TPpt shift, but now does not exclusively occur with state clauses, as we have observed.
As I have argued in the previous sections, in narratives, the temporal perspective shifts to a narrative time line, which is typically in the past tense, and the utterance time becomes secondary.Footnote [20] This is the function of narrative discourse but the individual clauses and tense and aspect forms that occur in it do not share this function. I argue that TPpt coincides with the speech time in non-narratives, whereas TPpt shifts to a past narrative time in narratives and possibly gets updated as a new telic event is introduced, describing a later event. Assuming these, we can avoid the undesirable multiple ambiguity of the simple past tense in English because TPpt becomes a property of discourse rather than the lexical meaning of tense and aspect forms in isolated sentences. I articulate the discourse rules of fixing and updating TPpt in (31). I assert that now refers to TPpt.
We are ready at this point to provide the DRT analysis of sentences containing now. We have observed that now can combine both with event and state predicates. Following the general interpretation rules, when now modifies an event, the event is included in the location time; when now modifies a state, the state includes the location time. Let us first look at the case where now modifies a state sentence in the present tense. In (32), beg(s) means ‘beginning of the state s’.
As shown in (32), the condition beg(s)
${\leqslant}$
t represents the change of state meaning of now, ensuring that there is an inception event for the state and that it immediately precedes or is equated with the location time. Because of the presence of now, (32) implies that Helen has recently become cold. In the present tense, as is the case in (32), the speech time is the location time of the described event. The value of now refers to TPpt, regardless of whether it is the utterance time or a narrative time. In this case, the utterance time n is in the denotation of now because TPpt is the utterance time (I assume that (32) is a non-narrative, which is anchored to the speech time).
Let us now observe how DRT analyzes an event sentence modified by now.
As shown in (33), the event discourse referent e is included in the location time t, which in turn is equated with the utterance time n. Although the condition that the described event has just started to hold at n does not need to be specified because it is entailed by virtue of the fact that the event is temporally confined within its location time, I have added it in (33) for the sake of the uniform treatment of now. The denotation of now, as in (33), is TPpt, which is the utterance time n in this case.
Let us proceed to discourse examples in which now occurs with the past tense. (18) above, which is an example of now occurring with a past-tense state sentence, is repeated in (34a). The DRS for the simplified discourse of (34a) is given in (34b).
In (34b), now refers to the TPpt, which is t
$_{1}$
. The reason why the second sentence in which now occurs does not move the narrative time forward is due to the aspectual class of the main verb, which is a state. In this case, the location time including the event e of Anne agreeing, which serves as the reference point, is included in the state modified by now. In addition, the state described by the now clause begins right before the event of Anne agreeing.
The discourse in (25) above, which is repeated in (35a), contains an example in which now combines with an event sentence.
Since both the first and the second sentences are telic descriptions, they dynamically update the temporal context, resetting the TPpt. Furthermore, as shown in (35b), the condition e
$_{1}$
< e
$_{2}$
ensures that the event in the second sentence follows the event in the first sentence. Therefore, the observed temporal progression is a function of the aspectual class of the VP and follows from an independent condition in the DRS. In this case, now denotes the updated TPpt.
Finally, we have observed that stage-level state predicates sometimes undergo a coercion that transmutes them into an inchoative event that begins the described state. Following de Swart (Reference De Swart1998) and others, I treat coercion as an implicit operator of the same type as grammatical aspectual operators like the perfect or the progressive. That is, coercion operators C are eventuality modifiers, which map a set of eventualities onto another set of eventualities. In de Swart (Reference De Swart1998), the input and output type are represented as indices on the operator, e.g. C
$_{\text{sd}}$
. For the progressive she assumes a mapping C
$_{\text{sd}}$
from stative onto dynamic eventualities. This operator is added before the progressive operator applies, satisfying its input conditions. In her system, (36a) forms the DRS in (36b).
The coercion with now occurs when a preceding sentence describes an event that causes the change. (37a) below is a simplified example of (19) above. (37b) provides the DRS of the same example. I employ an inchoative coercion operator C inch ,Footnote [21] which changes a temporary state to its beginning point (i.e. it maps a state s to an inchoative event e that starts s). Here, coercion is triggered by a mismatch between the aspectual class of the new sentence, which is a state, and the narrative interpretation requiring that the cause precede the effect.
I provide a DRS Construction Rule for tensed sentences in Table 2 and a DRS Construction Rule specific to now in Table 3.
Table 2 DRS construction rules for tensed sentences.
Table 3 DRS construction rule for now.
4 Conclusion
It is by now well known that indexical expressions such as now are not as rigid as previously thought and can shift (e.g. now can refer to a time other than the utterance time) in some contexts (e.g. a literary style like Free Indirect Discourse in English or under report verbs in some languages). What is interesting about shifted now is that its distribution is much broader than these limited contexts. The conditions under which it can shift, however, are unclear and still under debate. Many recent proposals have tried to derive this property from the lexical meaning of now, thus treating it as a special case. Contrary to previous analyses, I argued that the temporal perspective shift and temporal relations are functions of narrative discourse itself rather than the lexical semantics of now. I further argued that the only lexical meaning of now is that its denotation is flexible, referring to a contextually salient time, whether it derives from the actual utterance context or a discourse context. In addition, now invariably indicates a change of state, denoting the turning point dividing the past and the future seen from this contextually salient location time. My claim was supported by a quantitative study of naturally occurring narrative examples from the British National Corpus, and formalized in the discourse-level formal framework of Discourse Representation Theory. I also provided a new analysis of state sentences with now triggering a narrative progression utilizing the notion of aspectual coercion.
The use of now hinges on the distinction between narrative vs. non-narrative discourse types. Understanding discourse types and their different functions is essential for interpreting even a small lexical item like now and thus needs to be explored in more detail with more formal precision informed by naturally occurring discourse data.
