Two past forms inducing conjectural or non-intrusive questions

2.1 Lack of interrogative flip

The previous studies mentioned in Section 1 surely propose different analyses for the semantics of the conjectural future, the epistemic adverb, and the epistemic auxiliary. Giannakidou and Mari (2018) claim that the Italian conjectural future is essentially synonymous with the epistemic use of the necessity modal must: it approximatively means that, as far as the epistemic agent knows, the proposition denoted by the adjacent is true. According to Ippolito and Farkas (2022), the Italian conjectural future is rather a doxastic modal and indicates that the denoted proposition is the most likely according to the agent’s subjective belief. Fălăuş and Laca (2014) argue that the Spanish conjectural future is an epistemic modal indicating that, as far as the agent knows, the probability of the truth of the denoted proposition is not high enough for her to be able to assert it. Eckardt (2020) claims that the German epistemic adverb wohl conveys that the denoted proposition is stereotypically entailed from the information known to the agent. According to Uegaki and Roelofsen (2018) and Hara (2023), the Japanese epistemic auxiliary daroo is a necessity modal whose conversational background depends on the speaker’s knowledge (in matrix cases) or the attitude holder’s one (in embedded cases). But all these analyses invoke the knowledge or the belief of the epistemic agent, who corresponds to the speaker in declarative matrix clauses.

Usually, the agent responsible for the judgment expressed by modal or evidential markers changes in interrogative sentences. Bhadra (2020, 367–8) effectively points out that “when perspective sensitive elements are embedded in [true information-seeking questions], the perspective of these elements align with the epistemic authority’s [perspective]. … In the assertion, the proposition is evaluated against the evidence possessed by the speaker; in the [true information-seeking] question, it is the addressee’s evidence that is being invoked.” This change of the responsible agent is called ‘interrogative flip’. Thus, the English epistemic modal might trigger an interrogative flip as in (7).

What characterizes the conjectural future, the epistemic adverb, and the epistemic auxiliary is the lack of interrogative flip (the fact the speaker remains the one who utters a conjecture), as suggested by the gloss ‘I wonder’ used in their English translations. In effect, all the questions in (1)–(4) may be uttered in the absence of the addressee. This fact is explicitly mentioned for the Japanese auxiliary daroo as in (4) by Nitta (1991), according to whom daroo is exceptional among Japanese modal or evidential markers in that the epistemic agent remains the speaker in interrogative sentences. Daroo has its polite form desyoo, which presupposes the presence of the addressee, as in (8). In this case, the epistemic agent is the conversational community formed by the speaker and the addressee, and the sequence desyoo-ka expresses an invitation of a joint conjecture, as shown by the English translation of (8).

Eckardt (2020) points out that the German epistemic adverb wohl triggers an interrogative flip with a verb-second order as in (9a), but doesn’t with a verb-final order, as in (9b).

In the next subsection, I define two types of non-canonical questions, conjectural and non-intrusive questions, by referring to Eckardt (2020) and Farkas (2022).

2.2 Conjectural and non-intrusive questions

According to Farkas (2022, 297), canonical questions “are characterized as ‘information seeking’ acts whereby an ignorant speaker requests an addressee she assumes to be knowledgeable to inform her of the true answer to the question she raised.” They are thus accompanied by the following four default assumptions accompanying question acts.

Different types of non-canonical questions weaken or override some of the default assumptions. Thus, quiz questions uttered by a teacher to a student, like ‘What is the capital of Morocco?’, Speaker ignorance in (10a), Addressee competence in (10b), and Issue resolution goal assumptions in (10d) are overridden. Rhetorical questions like ‘Who would do such a thing?’, whose answer is evident, override Speaker ignorance in (10a), Addressee compliance in (10c), and Issue resolution goal assumptions in (10d). In biased questions expressed by tag questions ‘Susan is coming with us, isn’t she?’ or by rising declaratives like ‘Susan is coming with us?’, Speaker ignorance in (10a) is weakened. According to Farkas (2022, 319), a non-intrusive question marker “marks the weakening of the Addressee compliance assumption […] while conjectural questions signal that the speaker does not assume Addressee competence”. Self-addressed questions uttered in a monological way are necessarily conjectural questions. Farkas (2022, 319) also points out that “a speaker may wish to weaken Addressee compliance assumption either because she cannot assume Addressee competence [in this case, we have conjectural questions] or because she judges a non-intrusive question to be more appropriate for rhetorical or other reasons.” This relation between conjectural and non-intrusive questions is clarified in terms of the four characterizations that Eckardt (2020) provides to conjectural questions.

Unlike conjectural questions where the four conditions in (11a–d) are satisfied, non-intrusive questions satisfy (11b) and (11c), but not necessarily (11a) and (11d).

Conjectural questions in the presence of the addressee are illustrated in German (12a) and Japanese (12b), each of which is uttered in the context indicated. Here, the speaker invites the addressee to a joint speculation: the epistemic agent is the conversational community.

Typical examples of non-intrusive question, where the addressee is not required to respond, are given in Spanish (13a) and Japanese (13b). Literally, these cases are conjectural questions in view of the presence of the conjectural future in (13a) or the polite form of the epistemic auxiliary desyoo in (13b). But the addressee is expected to know the answer: the condition (11a) of conjectural questions is not satisfied. The use of Spanish future or the Japanese epistemic auxiliary rather suggests that the addressee is not forced to answer the question, which makes the question more polite. Yamaguchi (1990, 95) effectively suggests that a use of desyoo often aims at weakening the request to answer, by presenting the question as self-addressed.

Eckardt (2020) observes that the German epistemic adverb wohl combined with the verb-final order does not serve to express non-intrusive questions but only conjectural ones.

Farkas (2022, 304–5) further clarifies the notion of non-intrusive questions in terms of Context structure, which consists of i) the speaker’s and the addressee’s discourse commitments, ii) the Table consisting of ‘a set of active issues awaiting resolution’, iii) the projected set, that is, ‘a set of [future] discourse commitment lists, anchored, by default, to the addressee’. Now, non-intrusive questions (which do not force the addressee to resolve the issue) are defined as invoking, in the projected set, ‘a future in which the issue is removed from the Table although it remains unresolved for both participants’ (idem, 312).

In the next section, I demonstrate that the market past in French and Japanese and the evidential past in English, German, French, and Japanese pragmatically induce conjectural or non-intrusive questions when combined with contextual factors. While French and Japanese market past forms have been extensively discussed in previous studies, the canonical or non-canonical nature of questions involving them has not yet been addressed. Recent discussions have emerged regarding questions involving the evidential past or a particle conveying a similar effect, with previous studies being divided on their canonical or non-canonical nature. My analysis will be mainly based on (i) the lack of interrogative flip (i.e., the observation that the agent of conjecture or recalling remains the speaker in questions), and (ii) the incompatibility with expressions forcing the addressee’s answer, like ‘I’m asking you. Reply me!’. I further explore the factors contributing to the derivation of conjectural or non-intrusive meaning in each case.

3.1 Market past used in polite questions uttered in customer service situations

I repeat example (5) illustrating the French market imperfective past as in (14a). A similar use of a past form in customer service situations is observed in Dutch and Spanish, as in (14b–c). But the French one has been most thoroughly discussed.

In French linguistics, roughly speaking, two lines of approaches are proposed for polite questions including this use of past form: First, it is argued that i) the use of the third person instead of the second person to refer to the customer in (14a) indicates a distance from the utterance situation: the shopkeeper does not directly talk to the customer (Lebaud 1991, 68); ii) the past form is modal in that it does not refer to a real past situation, but rather to that found in the speaker’s mental framework (De Mulder 2012, 103). De Mulder (2012, 104) also points out that the interpretation of the imperfective past as non-temporal is determined by its combination with various contextual factors. A possible extension of this approach would be to analyze (14a) as a self-addressed question parallel to a question in the free indirect speech as in (15), which may be followed by the sequence il se demandait (‘he was wondering’).

The difference between (14a) and (15) is that the latter remains unpronounced and totally self-addressed in its use while the former is pronounced in front of the customer. According to this analysis, the customer is not expressed by the second person pronoun because (14a) is presented as conveying the shopkeeper’s self-addressed question. This analysis of (14a) as related to an implicit matrix clause, je me demandais (‘I was wondering’), allows us to consider this use as parallel to other non-past uses of imperfective past form, like its evidential use (shown below) as well as its counterfactual and hypocoristic uses (for the approaches which analyze these three cases as implicitly embedded under some matrix clause, refer to the study by Barceló and Bres (2006) and Patard (2023)). This analysis of (14a), however, seems to be problematic in view of the fact that the customer may be expressed by the second person pronoun as naturally as the third person, like in (16a), and that even the plural first person nous (‘us’) may be used, as in (16b).

Criticizing the first approach as counterintuitive, Kleiber and Berthonneau (2006, 81) and Barceló and Bres (2006, 58–9), among others, claim that i) the imperfective past effectively refers to a salient past moment when the customer needed something; ii) the shopkeeper shows, by using the past form, her attentiveness to the customer’s needs by anticipating her request. Referring to Brown and Levinson’s (1987) politeness theory, Patard (2007, 350) and Patard (2023, 62) classify this use among ‘illocutory’ or ‘intersubjective’ uses, pointing out that the past form induces positive politeness in that it allows the shopkeeper to flatter the customer’s positive face by showing that she was careful to the latter’s needs even when she served other customers.

I still claim that, even if we follow the second line of approach, (14a) and (16a–b) should be interpreted, by being uttered in customer service situations, as a conjectural or non-intrusive question. Before presenting my own analysis, let’s take a look at Kleiber and Berthonneau’s (2006) analysis of the ‘imperfective past of politeness’ (imparfait de politesse in French), as in (17a), which Patard (2023, 52) classifies among intersubjective uses, together with the market imperfective past. A similar use of the past form is also observed in other languages, like Dutch (17b) and English (17c).

According to Kleiber and Berthonneau (2006, 84), the politeness comes from the fact that (17a) is interpreted as a reply to the addressee’s implicit question ‘Why did you come here?’ and that the status of the speaker is thus changed from a requester to a respondent.

Barceló and Bres (2006, 58) recognize, in the case of the market imperfective past, the pertinence of a similar causal relation, as in (18b). What is crucial is that, when the customer was waiting without saying anything, the causal relation, as in (18b), is only inferred by the shopkeeper. I claim that the market imperfective past refers to a past salient moment when the shopkeeper made a conjecture that the customer needed something, based on her observation that the latter was waiting in line. The politeness derives from implicitly communicating, by use of the past tense, the shopkeeper’s conjecture for the sake of the customer. (14a) and (16a) are thus more fully paraphrased by (19a).

The referent of it in the third interrogative sentence of (19a) only exists in the speaker’s epistemic or doxastic world, in a similar way to that of it in the second sentence of (19b), which Roberts (2021, 1880) accounts for in terms of modal subordination under attitude predicates.

I claim that, even without any explicit conjectural marker, since the use of the past form serves to communicate, in customer service situations, the shopkeeper’s past conjecture, (14a) and (16a) are interpreted as modalized questions based on the shopkeeper’s assumption. From this perspective, the plural first person nous (‘us’) in (16b) contributes to inviting a joint speculation, similar to some conjectural questions uttered in front of the addressee, as in German (12a) and Japanese (12b) (Section 2.2). More exactly, since the shopkeeper can naturally assume that the customer knows the answer to a question about her own need (i.e., the addressee's competence condition is satisfied), these interrogatives are interpreted as non-intrusive ones which do not force the addressee to answer, in a way parallel to polite questions in Spanish (13a) and Japanese (13b) (Section 2.2). In effect, if (14a) and (16a–b) were presented as canonical information-seeking questions, the speaker would take it for granted that her conjecture about the addressee’s desire is correct, which would amount to threatening the addressee’s negative face, by invading her inner territory (desire).

One evidence in favor of my hypothesis is a parallel behavior between the market imperfective past and a conjectural use of the conditional form (cf. Introduction). According to Dendale (2010, 296), the questions including a conjectural conditional may be monological or may be addressed to someone else, and in the latter case, the participation of the addressee may be solicited but not necessarily. This is confirmed by the incongruity in the continuation of the following sentences which necessitate a response from the addressee in (20a–b). As shown by (20c), if the addressee happens to know the answer, her reply is a licit reaction.

Now, let’s suppose that the shopkeeper poses a question including the market imperfective past and that the customer does not reply immediately. In this case, the shopkeeper cannot appropriately follow by uttering sentences forcing the addressee to answer, as in (21). Furthermore, when the customer replies to the shopkeeper’s question, she does not use the past form or the third person pronoun, but the present form and the first-person pronoun, as in (22). These observations suggest that the relation between the question and answer in (22) is not as direct as in the case of canonical information-seeking questions.

Interestingly, we observe, in Japanese, questions uttered in customer service situations similar to French (22) and where the past form ta and the polite form of the epistemic auxiliary desyoo cooccur, as in (23a). Watanabe (2014, 29–33) recognizes a similarity between these Japanese cases and French examples involving a market imperfective past, as in (22).

Extending Watanabe’s (2014, 32–3) idea, I claim that the past form in Japanese (23a) refers to a past salient moment when the casher made a conjecture about the customer’s possession of a fidelity card based on the observation that the latter had chosen the relevant shop, as expressed in (24). The polite form of the epistemic auxiliary serves to make the question non-intrusive (Section 2.2).

It should be noted that, in the context of (23a), desyoo seems to be necessary, as shown by the low acceptability of (23b) lacking this epistemic auxiliary: (23b) is rather interpreted as asking the addressee’s possession of a fidelity card in the past rather than in the present. This obligatory use of the epistemic auxiliary is, I assume, due to the fact that i) in Japanese, questions lacking any expression indicating the speaker’s conjecture tend to be interpreted as forcing the addressee to reply (Yamaguchi 1990, 86), and ii) in Japanese, answer-requesting questions and self-addressing questions have traditionally been explicitly distinguished, for example, by two particles ya and ka in Old Japanese or by the presence and absence of some conjectural expression in Contemporary Japanese (Yamaguchi 1990, 3–4). In French, despite the existence of some forms expressing a conjectural meaning (e.g., a conjectural use of the conditional form, illustrated in (20a–c)), they do not seem to be obligatorily used when a conjectural question is uttered.

Now let’s examine a French conditional form, put in bold letters in (25), used in customer service situations similar to those where the market imperfective past is observed.

The French conditional form is a combination of the future stem (e.g., faudr- for faudrait in (25)) and the imperfective past inflectional ending (e.g., -ait for faudrait). In Section 1, we saw that Italian and Spanish future forms may indicate a conjectural meaning. Some previous studies (cf. Dendale 2001, Azzopardi 2011) claim that the French future form too may convey a conjectural meaning. Is it conceivable to analyze (25) as a conjectural question and to decompose the conditional form into i) the future stem expressing the speaker’s conjecture, like Spanish and Italian future, and ii) the imperfective past ending corresponding to the market imperfective past? If this were the case, the two elements in the conditional form would correspond i) to the Japanese epistemic auxiliary desyoo marking the speaker’s wondering and ii) to the Japanese market past ta in (23a). Dendale (2010, 309–11) effectively mentions a possibility to analyze French conjectural conditional as a conjectural future put in the past (but later abandons this possibility).

This analysis in terms of decomposition however faces at least three problems. First, there is a difference between the imperfective past ending in (25) and the market imperfective past. As suggested by Berthonneau and Kleiber (1994, 74), the French market imperfective past is incompatible with adverbials referring to the utterance time as shown in (26a). The same incompatibility should be observed if the imperfective past ending of the conditional would correspond to the market imperfective past. But the conditional is well compatible with these adverbials, as shown by (26b).

Second, the future stem in (25) cannot be analyzed as a conjectural future. Usually, French conjectural future is only observed with verbs être (‘be’) and avoir (‘have’) (cf. Dendale 2001, 4), but not with vouloir (‘want’) and falloir (‘be necessary’). Furthermore, Mari (2016) points out an important difference between the Italian future and the French future: unlike the Italian future, the French future “can be used when the speaker has knowledge that the prejacent [i.e. propositional content] is true at the time of the utterance” (Mari 2016, 358), as in (27). The French future here “does not express a conjecture, but rather postposes the time of the verification that the shoes are on the shelf. The hearer will be in charge of the verification.” (ibid.). From this observation, Mari (2016) claims that the future form in contemporary French no longer conveys itself the speaker’s conjecture but indicates that the propositional content is rectified or confirmed by someone at a time ulterior to a given time, which usually corresponds to the utterance time (for a similar view, refer to Wilmet 2003, among others).

Moreover, according to Patard (2023, 168–70), the conjectural conditional is diachronically derived from the conditional form expressing an eventuality (e.g., counterfactual use), whose source should be distinguished from that of the conditional form expressing the future in the past.

Third, as pointed out by Azzopardi (2011, 400), the conjectural conditional is observed only in polar questions, as in (20a–b), and not in content questions including an interrogative word or phrase, as in (25). The conditional form in (25) may be more appropriately analyzed as ‘conditional of attenuation’ (conditionnel d’atténuation in French), together with the one observed, for example, in ‘Qu’est-ce qu’il faudrait inventer? (‘What would we have to invent?’)’.

In sum, I claimed that the French market imperfective past refers to a past moment when the shopkeeper made a conjecture about the customer’s need based on her observation and that the questions including this past form, as in (14a) and (16a–b), are interpreted, without any explicit conjectural marker, as conjectural or non-intrusive questions about the customer’s need. This analysis was supported by the observation that these questions are not appropriately followed by the shopkeeper’s utterance forcing the customer’s answer. It was also pointed out that, while Japanese requires, in cases including a similar market past, the presence of an epistemic auxiliary clarifying the conjectural nature, it is not the case for French: although the French conditional form has a conjectural use, the one used in similar customer service situations should be analyzed as of another use (i.e., attenuation use).

3.2 ‘Evidential’ past used in ‘recall’ questions

I next examine another use of the past form, observed in questions like (28a–c), which repeat the examples (6a–c) cited in Section 1, as well as in Japanese (28d). In these examples, the past form indicates that the speaker tries to remember the addressee’s or the third person’s name which she has heard before, but she has forgotten.

In the original example of (28d) given by Sauerland and Yatsushiro (2017), the sentence's final particle is represented as kke. I adopt here, however, the notation by ke, following Shibuya (1999, 2014). In what follows, I take up this segmentation except when citing passages from Sauerland and Yatsushiro (2017).

As regards French examples involving this type of past form, as in (28c), Barceló and Bres (2006), Patard (2012), and De Mulder (2012), among others, highlight that the origin of pertinent information is ascribed to a previous statement made by someone other than the current speaker. For example, if the information source is due to the addressee, (29a) is paraphrased by (29b) including the underlined matrix clause tu as dit ‘you said’. The same is true for the past forms of other languages, illustrated in (28a–d). Following De Mulder (2012) and Patard (2012), I call this use of the past form ‘evidential’ past.

Now, let’s get started by observing the distributional facts of the evidential past: (I) it may appear in declaratives, as in (30a–c), where the past form refers to a past moment when the speaker obtained or thought of the pertinent information (her daughter’s anniversary in (30a–c)).

(II) On the one hand, as pointed out by Sauerland and Yatsushiro (2017, 654), the English evidential past alone without the adverb again, as in (31a), conveys the meaning that the speaker tries to but cannot remember the information in question. The same is true for German and French, as in (31b–c). On the other hand, the absence of the particle ke seems to degrade the acceptability of Japanese counterpart, as in (31d): if (31d) is acceptable, it is i) when the relevant person changed his name and (31d) is interpreted as asking his previous name, or ii) when the speaker aims at testing the addressee’s memory: for example, the teacher of literature asks a student the name of the author of War and Peace which has been already taught in the class. French (31c) whose subject is the third person admits, depending on contexts, these two readings as well as the same reading as (28c) including the adverb déjà (‘already’).

(III) When the evidential past form in (28a–d) is replaced by the present form, the particles like English again, German noch mal (‘again’) and French déjà (‘already’) as well as Japanese ke may induce the meaning that the speaker tries to but cannot remember the name in question. Bhadra (2022, 32) points out that, in a similar vein, ‘the Bengali particle jyano … yield an effect of having forgotten the witness of the existential claim’, as in (32).

In the previous studies, the canonical or non-canonical nature of questions like (28a–d) and (32) has been mainly discussed with respect to the particles (e.g., again, noch mal, déjà, ke, and jyano). I then first review previous analyses of their semantic contribution, before investigating the canonical or non-canonical nature of questions including an evidential past.

Sauerland and Yatsushiro (2017, 651) claim that English again, German noch mal and Japanese ke all presuppose that ‘the answer was already made common-ground knowledge in the past’, which they call ‘remind me’ presupposition. They further observe that German noch mal can trigger the ‘remind-me’ presupposition only in content questions but not in polar questions, as in (33a). The same is true for French déjà (‘already’), as in (33b). Unlike these particles, Japanese ke can induce the same presupposition in polar questions, as in (33c).

Sauerland and Yatsushiro (2017) associate this difference between German noch mal and Japanese ke with the fact that ke lacks a repetitive reading, unlike noch mal. Assuming that these two items have the same lexical interpretation (p. 669), the two authors propose the following syntactic account: Question speech-act is decomposed into an imperative part and a common ground (make-it-known) part, both of which are represented in the syntactic structure. The repetitive reading is obtained when the particle is merged to VP. Noch mal conveys this reading since they effectively merge to VP while ke, which merges to the common ground part, does not. The ‘remind me’ reading is obtained when the particle merges with the common ground part. This condition is satisfied by ke as well as by noch mal in content questions where the latter can implicitly move to the common ground part, but not by noch mal in polar questions where the covert movement out of questions is impossible. According to this analysis, ‘Wie ist noch mal Ihr Name? (“What is your name again?”)’ is paraphrased by ‘You should make it know to me again (or Tell me again) what your name is.’, which boils down to analyzing this question as a canonical information-seeking one.

Now, let’s look at (34) including German noch mal. This example may be monologically uttered by the speaker who goes shopping and being distracted, completely forgets the shopping list. It thus shows that the information searched for in ‘remind me’ questions may be pooled not only in the common group shared by the addressee, but also in the speaker’s private knowledge.

Bhadra (2022) adopts this idea for the Bengali particle jyano and claims that “ignorance is very much a statement of lost knowledge in the mind of the speaker [rather than in the common ground shared with the addressee], unlike the ‘remind-me’ presupposition” (idem. 30). This author further argues that “in questions it leads to a ‘recall’ effect of the speaker asking for the addressee’s help in recalling information” (idem. 32), rather than requiring the addressee’s answer. From this perspective, she calls the questions including the particle jyano, like (32), ‘recall’ questions rather than ‘remind me’ questions.

A similar analysis in terms of non-canonical questions is proposed by Shibuya (1999, 2014) for the questions including the Japanese ke. According to Shibuya (1999, 209), this particle, whose origin is the past auxiliary keri in classical Japanese, essentially conveys the speaker’s search for information gained through her experience or some hearsay and stocked in her own memory. It may surely be used when she asks the addressee what she has searched in her memory but cannot remember, as in (28d). Shibuya (1999, 210), however, claims that the questions with ke are not inherently equipped with an information-requiring meaning: rather, an originally monological question, uttered in front of the addressee, is pragmatically interpreted as non-intrusive ones. At least six observations come in favor of this analysis. First, as pointed out by Shibuya (1999, 199), a question with ke is not naturally put in the complement of interrogative verbs like kiku (‘ask’) or tazuneru (‘inquire’), as in (35).

Second, unlike again, noch mal and déjà, ke may occur, combined with the evidential past, in declarative sentences and serves to search for information stocked in the speaker’s knowledge, as in (36) (see also Gendai Nihongo Bunpoo 4, 273).

Third, the particle ke has a common point with the epistemic auxiliary daroo, which is a dedicated conjectural marker in contemporary Japanese: while Japanese canonical information-seeking questions may be pronounced with a rising intonation, the questions including ke are not uttered with a rising intonation, as in (37a). If (37a) is accepted, it does not serve as an information-seeking question, but conveys a doubt about the addressee’s preceding utterance, and is naturally followed by a sentence like ‘If I remember, his name is not Kai, but John.’

Parallel to the question with ke in (37a), the conjectural questions with the sequence daroo-ka, as in (37b), cannot be pronounced with a rising intonation, as previously observed by Hara (2023).

The fourth observation suggesting the conjectural nature of the particle ke is that, as pointed out by Matahira (1996, 24), it may be always replaced by another particle kana without changing the essential meaning, as in (38).

According to Gendai Nihongo Bunpoo 4 (p. 35), the final particle kana marks a self-addressed conjectural question, like daroo-ka. In effect, in place of kana, daroo-ka is acceptable in (38). Yamaguchi (1990, 90) suggests that the part na of kana serves to stress the self-addressing nature. In the same vein, Endo (2022, 252) points out that kana ‘indicates the speaker’s wondering …without seeking information from the addressee’.

The fifth observation in favor of this analysis is that the particle ke may be followed by another particle ne which serves to express confirmation, as in (39) (Shibuya 2014, 107–8). In (39), the epistemic agent is the conversational community formed by the speaker and the addressee, but not the addressee alone. (39) is thus interpreted as ‘Based on what we heard, when in the next month was Taro’s marriage?’, but not as ‘Based on what you heard, when in the next month was Taro’s marriage?’. In other words, the interrogative flip does not occur (Section 2.1).

The sixth and last observation indicating the monological nature of the particle ke is that it is avoided when the speaker unidirectionally transmits what she knows to the addressee, as shown by its incompatibility with another sentence-final particle yo (‘I confirm you that’), as in (40): according to Endo (2022, 253), the particle yo is used ‘to elicit a sense of insistence or to get the addressee’s attention’ and the question marked by yo ‘can no longer be interpreted as the speaker’s monologue’.

These six observations indicate that the particle ke essentially serves to make questions monological. As discussed above, Japanese questions with the evidential past require the presence of ke or some other conjectural question marker, like kana. These questions are therefore naturally analyzed as basically conjectural ones and are appropriately called recall questions (Bhadra 2022) rather than remind-me questions (Sauerland and Yatsushiro 2017).

To revisit the questions involving the evidential past in English, German, and French, as in (28a–c), I propose i) that they also are basically conjectural ones even if they lack an explicit conjectural marker, like Japanese particle ke, and ii) that the answer to the question should be searched for in the speaker’s memory rather than in the common ground shared with the addressee. For this reason, they also are appropriately called recall questions. At least four observations come in favor of this analysis.

First, let’s examine the examples in (41a–c). The past form here indicates that the speaker tries to but cannot remember the director’s name of the film she adores. If the answer was searched for in the common ground, the addressee should know it. The speaker then would not be able to follow the discourse with another question, meaning ‘Do you know?’. The compatibility of such a following discourse indicates that the answer is not in the common ground.

Second, like in the questions including the French market imperfective past, and unlike the canonical information-seeking questions with a present tense, the questions including the evidential past are not naturally followed by a sentence forcing the addressee’s answer when the addressee does not immediately reply, as in (42a–c). Following the absence of an immediate response from the addressee, a more appropriate continuation would be ‘I’m sorry, I cannot remember’ rather than ‘I’m asking you. Please reply!’.

Third, as observed by Sauerland and Yatsushiro (2017, 655), the questions including an evidential past in English and German may be naturally answered with the present tense, as in (43a–b). The same is true in French (43c) and Japanese (43d). This fact suggests that the relationship between A’s question and B’s reply in (43a–d) is not as direct as in the cases of canonical information-seeking questions.

Sauerland and Yatsushiro (2017, 655) observed that B’s reply with the past tense is possible, and that “B’ first person response is more natural in the present tense. … Even in the third person in [(43b)], B’s response in the past tense is slightly odd if Kai is present in the discourse situation. We think this effect is related to the salience of time intervals: if an individual is present in a situation and is mentioned in a declarative sentence, the utterance time is salient.” Now in Japanese (43d), the past tense in B’s response is compatible with the particle ke only when B has found the relevant name after having searched for it in her own memory, as shown by the English translation of (43d). I tentatively suggest that the same is true for (43a–c): when B refers to the past tense, it shows that she found the relevant name after having looked for it in her own memory. The past tense is odd in B’s first-person response because B normally does not have to search for her own name in her memory.

Finally, in German, the questions involving an evidential past are compatible with the epistemic adverb wohl combined with the verb-final word order, as in (44). According to an anonymous reviewer, the example would improve with the addition of noch (gleich) after wohl. This compatibility suggests that these questions may be explicitly marked as conjectural ones.

In sum, I claimed that the questions involving an evidential past are basically conjectural recall questions whose answer is in the speaker’s memory rather than in the common ground. This is the case not only in Japanese where the evidential past should cooccur with a conjectural particle ke or kana and in German where the evidential past optionally cooccurs with the epistemic adverb wohl, but also in English and French where the evidential past is not combined with any explicit conjectural marker. The quasi-obligatory use of a conjectural expression in Japanese cases may be due to the fact that the Japanese questions lacking any expression indicating the speaker’s conjecture tend to be interpreted as forcing the addressee to reply: the lack of such an expression further tends to suggest that the speaker makes no effort to look for the answer in her memory, while such effort is usually a prerequisite condition to ask information which she should maintain in her memory. This analysis however cannot, for the moment, account for why English again, German noch mal and French déjà are restricted in content questions, unlike Japanese particle ke, as pointed out by Sauerland and Yatsushiro (2017). This is the subject of a future study.

By the way, the French conditional form may convey hear-say information, as in (45), which signals that the denoted information is attributed to the utterance of someone else, and therefore is qualified as ‘evidential’. With the evidential conditional, interrogative sentences most often serve to ask a confirmation of the validity of such hear-say information. These sentences are thus paraphrased by ‘Is what I heard, that is p, true?’ (Dendale 2010, 295).

Furthermore, in a context similar to recall questions with an evidential imperfective past, as in (46a), the evidential conditional is acceptable, as in (46b).

In Section 3.1, I explored the French conditional form by considering its morphological decomposition into the future stem and the imperfective past ending. My examination focused on determining whether the conditional form used in polite questions during customer service situations could be broken down into the market imperfective past and the conjectural future. The conclusion drawn from this analysis was negative. Is such a decompositional analysis conceivable for the evidential conditional, as in (46b)? In other words, is it possible to interpret questions including an evidential conditional, as in (46b), as conjectural questions, and to decompose this conditional form into the future stem and the imperfective past ending expressing the same evidential meaning as in (46a)?

At first glance, this idea seems to be promising since the origin of the evidential conditional is the future in the past use, as in (47a) (Patard 2023, 170), where the two items are transparently decomposed. But the idea that (46b) is a conjectural question and that the imperfective past ending of the conditional form in (46b) corresponds to the evidential past faces at least two problems.

First, according to Patard (2023, 169), in the first stage of the diachronic process from the future in the past use to the evidential use, the hear-say reading is only contextually induced by an indirect discourse: reported information is necessarily a hear-say one. In the next stage, when the originally embedded clause is reanalyzed as a main clause, the idea of posteriority conveyed by the future stem disappears and the notion of anterior utterance suggested by the imperfective past ending is backgrounded. Thus, the pragmatic meaning originally only contextually induced is integrated into the lexical meaning of the conditional form, which synthetically conveys a hear-say reading in the contemporary French.

Second, unlike the evidential imperfective past, the evidential conditional does not necessarily presuppose that the relevant information is stocked in the speaker’s memory. For example, although similar examples are very rare according to Azzopardi (2011, 287), we observe examples like (48a) where B hears the relevant information (the merger of the two companies) for the first time. B’s question in (48a) therefore cannot be a conjectural one about forgotten information, and is paraphrased by (48b) where the interrogative flip occurs.

The same seems true for the content question in (46b) with an evidential conditional: it may be uttered when the speaker has not yet heard the beginning time of the meeting.

In sum, just as the imperfective past ending of the French polite conditional form used in customer service situations is not associated with the market imperfective past, that of the French evidential conditional is not associated with the evidential imperfective past. Furthermore, unlike the questions involving an evidential past, those with the evidential conditional are not necessarily interpreted as conjectural.