Exclamations and their discourse effects in Japanese

This paper discusses two kinds of sentences with nante that can express exclamations in Japanese. I show that these two nante exclamations show the contrast observed between sentence exclamations and wh-exclamatives in English (Rett 2011). Based on the data, I propose that nante in the two types of sentences can be analyzed in a unified way: nante is like what in English in that it can range over a variety of categories. The semantic composition shows that the two nante sentences have different sentence types and hence different discourse effects (Farkas & Roelofsen 2017). One type is a marked assertion, and the other is a bonafide exclamative. I show the contrast using the discourse model in Farkas & Bruce (2010): The exclamative updates the speaker’s commitment, not the potential future common ground.

1. Introduction. Natural languages provide us with multiple ways to achieve the same goal: To ask whether it is raining, for example, one can use a polar question ('Is it raining?'), a tag question ('It's raining, isn't it?'), or a high negation polar question ('Isn't it raining?'). The goal of these questions is the same, but the speaker can convey accompanying contextual information (e.g., whether they think it is likely to be raining) by using different kinds of questions strategically.
However, this is not a unique feature of the interrogatives. English has various ways of conveying the speaker's surprise, as does Japanese. This paper investigates the discourse effects of two kinds of exclamations in Japanese that share the same expression nante. While they both convey the speaker's surprise, they are based on different speaker's expectations and behave differently toward challenge by the addressee using no. To explain these differences, this paper argues that one is a bonafide exclamative, while the other is a special assertion. The rest of the paper is structured as follows: In Section 2, exclamations with nante are shown and compared with English exclamatives and sentence exclamations. Section 3 provides an analysis of the two nante exclamations from the semantic and pragmatic perspectives. Section 4 presents the conclusions and outlines possible future directions.
2. Data. This paper focuses on two types of sentences in (1) that can express the speaker's surprise. These sentences share the same expression, nante, even though it appears in different places in a sentence. 1 In one type, nante appears at the very end of the sentence, as in (1a), while in the other nante precedes adjectives, nouns, or adverbs, as in (1b). In each case, the speaker's surprise is conveyed by uttering the sentence. 2 (1) a. Taro As the English translation of (1) shows, the differences between the two types of exclamations in Japanese are analogous to those between sentence exclamations and wh-exclamatives in English. This paper focuses on two similarities, namely, their scalarity and the behavior to the response particle.
A: I would have guessed John to be a terrible pâtissier! b. # A: I guessed he'd make delicious desserts, but these are beyond my expectation!
A: I would have guessed John to be a terrible pâtissier! b.
A: I guessed he'd make delicious desserts, but these are beyond my expectation!
Looking at the case with a sentence exclamation: (2), the continuation where the speaker explicitly states that they expected that John would bake delicious desserts is infelicitous, which is not the case when a wh-exclamative is used. The difference comes from the fact that the sentence exclamation targets a non-scalar component as the speaker's expectation. Two nante exclamations in Japanese show a similar contrast. The following two sentences state a non-scalar expectation and a scalar expectation, respectively.
(4) a. Taro After the utterance with the sentence-final nante (1a), only (4a) is a possible continuation, while the exclamation with the intra-sentential nante (1b) is compatible with both continua-tions in (4). In other words, exclamation with the sentence-final nante is non-scalar, whereas that with the intra-sentential nante is scalar.
2.2. BEHAVIOR TOWARD RESPONSE PARTICLES. In English, wh-exclamatives cannot be challenged by using no, as shown in (5b). 3 In contrast, the sentence exclamation can be challenged by no, as in (5a).
(5) a. (6) can follow (1a), which accompanies the sentence-final nante. The acceptability declines when (6) follows (1b). As in English, the challenge to a wh-exclamative by the addressee is possible without iya 'no'. To the author, to the wh-exclamative, (6) without iya sounds natural with a little pause or a filler such as etto 'well', which presumably shows some hesitation and lowers the aggressiveness of the challenging tone.
Summarizing the contrasts we have observed, we can obtain the following results shown in Table 1: Overall, exclamation with the sentence-final nante behaves like a sentence excla-Scalarity Challengeable by no (1a): Sentence-final nante (1b): Intra-sentential nante Table 1. Properties of exclamations in Japanese mation in English. In contrast, intra-sentential nante is similar to wh-exclamatives. The next step is to provide an analysis that accounts for the difference summarized in Table 1, keeping in mind that both share the same expression, nante.
3. Analysis. This section aims to account for the differences between the two kinds of exclamations in Japanese by arguing that they end up with different semantics and pragmatic effects (discourse effects), even though the base of the semantic denotation of nante is shared.
3.1. SEMANTICS. The two kinds of exclamations share the expression nante. It is possible to provide a unified account of the semantics of nante by setting its range underspecified, as in (7). This approach is also adopted by Rett (2011).
(7) The base semantics of nante nante λP τ,t λx τ .P (x), where τ is any semantic type The differences between the two nantes arise from the different ranges of the function in (7). When it appears intra-sententially, the range is set as degrees. In contrast, the range over which sentence-final nante goes is the domain of truth values. The difference is illustrated in (8) below.
These denotations of nante are used in computing the semantic denotation of these two exclamations along with other operators.
3.1.1. EXCLAMATIONS WITH THE INTRA-SENTENTIAL nante. The syntactic aspects of this exclamation are analyzed in Ono (2006). Following the analysis given there, the semantic analysis given in this section assumes that nante covertly moves to Spec CP, leaving a trace of type d (degree). Consequently, the semantic computation of (1b) is performed as in (9): What can be obtained in the last step of (9) is the degree property, not a proposition. Following Rett (2011), I assume that the unsaturated argument in (9b) is filled by the context, which makes the result a proposition that can be an argument of E-FORCE: (10) (10) E-FORCE(p), uttered by s C , is appropriate in a context C if p is salient and true in w C . When appropriate, E-FORCE(p) counts as an expression that s C had not expected that p. [Rett (2011;429)] In the end, (1b) is felicitously uttered in a context where there is a degree d such that the speaker had not expected that the desserts that Taro made would be d-delicious. This is the interpretation to be obtained as the semantics of exclamatives. Note that the range of nante used here is degrees, and accordingly, the final result involves existential quantification over degrees. This reflects the scalarity of this exclamation type, similar to the analysis of wh-exclamatives in English.

EXCLAMATIONS WITH THE SENTENCE-FINAL nante.
In the previous section, the semantic denotation of exclamations with the intra-sentential nante is calculated in the same way as that of wh-exclamatives in English, with the semantics of nante going over the range of degrees and with the help of E-FORCE. The semantics of sentence-final nante, however, arises differently from that of sentence exclamations in English. The crucial difference is that no E-FORCE operator is used, and there is a grammaticalized aspect involved.
An examination of nante sentences reveals that nante that appears at the very end of the utterance is grammaticalized so that it conveys that the truth of the sentence radical is surprising to the speaker. The sentence-final nante can be used in the embedded context, as in (11).
In this case, nante is of type tt, tt , and the first argument it takes is a function of the matrix clause, namely λp.the speaker cannot believe p.
(11) Taro When nante is used in this way, the matrix clause does not have to convey the speaker's surprise. In fact, nante can be used to convey that the truth of the proposition is made light of or is considered rather trivial, as shown in example (12) It is possible to use other complementizers, such as to 4 in place of nante. What nante contributes in (11) or (12) is to convey the speaker's attitude toward the truth of the proposition to which it is attached, and that does not have to be surprise. Given that, nante used to bridge the matrix clause and the complement clause should be treated differently from intra-sentential nante.
The nante that appears between a complement clause and a matrix clause requires a different treatment, as discussed above. However, sentence-final nante deserves a unified treatment with intra-sentential nante because the interpretations available with it are limited. For example, when someone utters a sentence with final nante, such as (1a), which is repeated here as (13), the only available interpretation is that the speaker is surprised or impressed by the fact that Taro made delicious desserts. In other words, the interpretation of (11) is available, but it is not possible to obtain the interpretation of (12). As shown in (13b), the interpretation may not be directly connected to the speaker's surprise. In this case, the fact that Taro made delicious desserts is contrary to the speaker's expectation, which is why they cannot tolerate the fact. This is more clearly shown by the unavailability of the interpretation in (13d), where the meaning of natural contradicts the fact that the speaker is trying to convey that their expectation was disappointed. One may analyze (13) as a sentence with elided material. That is, (13) is generated from (11) by deleting the matrix clause, namely sinzirarenai 'I cannot believe.' However, if this were the case, the interpretation of (12) would be one of the possible interpretations of (13), which is not the case. From this observation, it is possible to say that nante that appears at the very end of the sentence is grammaticalized to convey that the fact upsets the speaker's expec-tation. This leads to the speaker's surprise.
This grammaticalized aspect of nante can be captured if the underspecified range of nante is set as t. In other words, the sentence-final nante has the semantics in (14).
The grammaticalized part is the first argument λP t,t . It is saturated with a function such as λp.surprising(p). After the first argument is saturated, nante takes the sentence radical. The final result is a simple proposition that conveys that p is different from what the speaker expected. Here, there is no need to use E-FORCE to encode the speaker's surprise because it is encoded in the grammaticalized meaning of nante. Furthermore, unlike the intra-sentential nante, no degree property appears in the derivation, which accords with the fact that sentencefinal nante is not scalar.
3.2. PRAGMATICS. The semantic account in the previous section can at least explain why the exclamation with intra-sentential nante is necessarily scalar, while that with sentence-final nante is not. The difference is due to their different semantic calculations. However, it is not clear why one is challengable by no, and the other is not. This section aims to explain the different behaviors toward polarity particles (e.g., yes, no) between the two types of nante exclamations, claiming that their discourse effects are different: The exclamation with sentence-final nante aims to update the common ground, while that with intra-sentential nante does not.
3.2.1. BACKGROUND. I use the discourse model of Farkas & Bruce (2010) to show the differences between the two nante exclamations. An example output discourse model after an assertion "Taro made delicious desserts" by A is given as Table 2.
A There are four important components in the table. DC X is the discourse commitment of speaker X. In Table 2, it is shown that A commits to the truth of proposition p, which is Taro made delicious desserts. Table is where a speaker puts an issue or the immediate Question under Discussion (QuD) (Roberts 1996(Roberts , 2012. The pair of the syntactic form (here TMDD) and the semantic denotation is put here. The common ground is to be understood in the sense of Stalnaker (1978), which is the knowledge shared by discourse participants. Finally, the Projected Set shows possible future common ground. For example, in the case illustrated above, upon B's agreement, the proposition 'Taro made delicious desserts' is added to the common ground. In other words, the common ground is updated by making the union of the common ground at the utterance (s 1 ) and a set of possible worlds in which p is true.

DISCOURSE EFFECTS OF EXCLAMATION WITH THE FINAL nante.
Recall that the semantic denotation of the exclamation with the sentence-final nante is a proposition. Given that, I argue that this type of exclamation should be considered an assertion, albeit it is marked. Since it is a marked assertion, the discourse effect of the sentence-final nante can be built up by combining the basic discourse effects of declaratives and the extra effects tied to this specific assertion. In other words, the output table of the sentence-final nante exclamation is built upon what we have in Table 2 by adding an extra component. The results are shown in Table  3.
A  In the output discourse in Table 3, the addition is made to the discourse commitment of the speaker. As the special discourse effect of an assertion with nante, it is conveyed that the speaker is surprised at p. Meanwhile, the sentence radical part is added to the Table because this is an assertion. Consequently, the addressee (B, for example) can agree to it, and A and B can add p to the common ground. Alternatively, B can challenge it by saying no and deny the fact that the desserts were made by Taro, stating rather that they were store-bought.
This analysis argues that exclamations with sentence-final nante are marked assertions and hence have special discourse effects. However, being a "marked" assertion itself is not special in Japanese. Japanese has many discourse particles such as yo, ne, no(da) that are used sentence-finally and provide extra information on context (cf. Northrup (2014) or Hirayama (2019), among others). For instance, no(da) signals that there is contextual evidence that suggests that the sentence radical is true. The analysis presented in this paper adds sentence-final and grammaticalized nante to the list of elements that can make assertions marked.

DISCOURSE EFFECTS OF EXCLAMATION WITH THE INTRA-SENTENTIAL nante.
When an exclamation is made with intra-sentential nante, I claim that this exclamation is to be considered a bonafide exclamative, not an assertion. In other words, it is a different sentence type.
Sentence types are often distinguished by different syntactic forms in English, while Japanese uses particles or intonation. For instance, the most frequently used strategy to mark interrogatives is rising intonation. Alternatively, the question particle ka can be used. However, in the case of exclamatives, it is possible to detect a special syntax reserved for exclamatives (Ono 2006). One notable aspect is the use of the focus marker no and the mood marker daroo, as shown in (15). Without it, the sentence is ungrammatical. The syntactic property of this exclamation suggests that it has a different syntax from other sentence types, such as declaratives or interrogatives. Given that, it is safe to call this exclamation with a different sentence type a bonafide exclamative that hence has its own discourse effects.
The discourse effects of a bonafide exclamative in Japanese are as follows: It updates the discourse commitment of the speaker. They commit that they are surprised by the fact denoted by the sentence radical. Unlike declaratives, exclamatives do not put anything on the Table. As a result, the projected set remains the same because nothing is raised as an issue. 5 The discourse effects are shown schematically in Table 4. A  (Roelofsen & Farkas 2015), as shown in (16), Japanese polarity particles are known to be sensitive only to [AGREE/REVERSE] features (Pope 1972;Yabushita 1992), as shown in (17) Specifically, what is under discussion in this paper is no used to challenge the speaker, which is tied to the [REVERSE] feature, which is the only feature that Japanese no realizes. This feature crucially depends on what is on the Table because the feature shows that the answer agrees to or reverses the immediate QuD. When an exclamative is used, there is nothing on the Table, and this is why challenging the exclamation with the intra-sentential nante by saying no is infelicitous.
In sum, the most crucial difference between the two nante sentences is whether to put a proposition on the Table. A bonafide exclamative, in Japanese at least, does not update the future common ground but manifests the speaker's surprise toward the proposition denoted by the sentence radical. On the other hand, when nante is used in sentence-final position, it marks a special type of assertion and shares the basic discourse effects that other assertions have. This proposition is treated as an immediate QuD and is accessible by polarity particles. As a result, the addressee can challenge it, if necessary.
4. Conclusion and future directions. This paper has shown how two exclamations with nante in Japanese can be analyzed in terms of their different discourse effects. Their difference in scalarity is explained by the different semantic types attached to the two nantes, and I argue that one of them is grammaticalized so that it conveys the speaker's surprise. In contrast, the different behavior toward no is explained by different discourse effects, as their sentence types are different: One is a marked assertion, and the other is a bonafide exclamative. Unlike an assertion, an exclamative does not put an issue on the Table, which makes responding by saying no infelicitous.
This study explored two exclamative expressions involving nante, but they are not the only way to express surprise. One interesting example is shown in (18) This example is similar to the exclamation with the sentence-final nante in terms of the form, scalarity, and behavior toward the polarity particle. One notable feature is that this involves the topic particle wa at the end. Should this be treated as an instance of contrastive wa? If so, can we derive the discourse effect of (18) compositionally? These are fascinating open problems that need to be addressed. In the example above, wa is used sentence-finally. When it is used somewhere inside the sentence as contrastive wa, it also displays interesting behavior with exclamations. It has been pointed out that contrastive wa is incompatible with bonafide exclamatives, as shown in (19) This is a welcome result for the analysis proposed in this paper because (20) is treated as an assertion, with which contrastive wa is compatible. By adding wa to the assertion with nante, the speaker can convey implications such that they expected Taro to be a terrible cook of appetizers or main dishes, but he made surprisingly good desserts; in addition, they were surprised at the fact that Taro made delicious desserts. It seems possible to obtain this implication compositionally by combining the discourse effects of nante discussed in this paper and those of contrastive wa argued in Hirayama (2019), but another study is needed to fully discuss the issues. Finally, some discourse particles can appear with exclamations. For instance, the particle ne can be used with a bonafide exclamative, as in (21)