How do ‘rumours’ and reportative evidentiality match? A comparative study of the French conditional, Dutch zou + inf, and German sollen + inf

Abstract

In this paper, I will compare three reportative constructions: the French reportative conditional, Dutch zou + inf, and German sollen + inf. Although these markers share the reportative function as one of their established meanings, they clearly differ in how this reportative meaning actually functions. One of the most important differences pertains to the fact that the French conditional (and to a lesser extent Dutch zou + inf) often combines reportative meaning with epistemic denial, i.e. the speaker distances him- or herself from the content of what he or she reports. German reportative sollen also allows for such distancing interpretations but to a much smaller extent. Specifically for this paper, I will look at the behaviour of the three markers in the immediate context of the noun ‘rumours’ (French rumeurs, Dutch geruchten, and German Gerüchte), a context which – at least in theory – is strongly compatible with reportative marking, on the one hand, and with epistemic denial, on the other. On the basis of a self-compiled corpus of recent newspaper language, I will show that the French conditional occurs with a relatively high frequency in this specific context, especially in contrast to German sollen, and that the conditional often combines reportative semantics with epistemic denial, which again especially contrasts with German sollen + inf. Dutch zou + inf takes up an intermediate position in both respects.

1. Introduction

French, German, and Dutch feature verbal reportative markers, i.e. evidential markers that express that the speaker has obtained the information from another information source (another speaker, rumours, folk tales, or any written document). In French, the conditional, which is inflectionally marked, functions as a reportative (see example (1a)), Dutch has an auxiliary construction, which consists of zou – an originally past tense form of the modal/future auxiliary zullen ‘shall’ – combined with an infinitive (1b), whereas German uses a present indicative form of the modal auxiliary verb sollen ‘shall, should’ (in the remainder of this paper: soll _ind ) with an infinitive in this function (1c).

French

Dutch

German

In the abundant literature on reportative evidentiality, this category is often linked to epistemic overtones of doubt and denial. The notion of ‘reportative exceptionality’ (coined by AnderBois 2014) refers to the fact that of all the evidential markers, it is only the reportative ones that may associate with epistemic denial.

In summary, we find that cross-linguistically it is (at least) nearly universal that an evidential-marked claim can be felicitously denied by the same speaker only if its evidence type is reportative. (AnderBois 2014: 240)

AnderBois (2014: 238) illustrates the concept of reportative exceptionality by a number of examples, the following (2) from Estonian being one of them:

Estonian

Interestingly, it has also been shown in the literature that reportative markers may differ with respect to their actual propensity towards distancing (i.e. doubt and denial) interpretations (Wiemer & Socka 2017, Wiemer 2018). For the three markers addressed in this paper, Mortelmans (2024) has suggested that they differ with respect to the frequency with which they actually evoke a distancing interpretation: the French reportative conditional is most often used in contexts in which the speaker doubts or even explicitly denies the content of what is reported, whereas German soll _ind + inf is hardly used in such contexts. Dutch reportative zou + inf takes up an intermediate position. The present study aims to explore this issue in greater depth: if there are indeed frequency differences with respect to distancing interpretations of these three reportative markers, then it can be hypothesised that the three markers show up a different distribution in the context of the noun ‘rumours’ as well. Rumours are typically unconfirmed and often untrustworthy pieces of second-hand information and can therefore be expected to be compatible with reportative marking, on the one hand, and epistemic distancing, on the other. The following ‘classic’ definitions of the noun ‘rumour’ demonstrate both aspects:

an unofficial interesting story or piece of news that might be true or invented, and quickly spreads from person to person (https://dictionary.cambridge.org/dictionary/english/rumour)

unsubstantiated, often false stories that spread through a community person to person by word of mouth in a manner rather similar to the spread of communicable diseases. Rumor can affect the behavior of crowds for good or ill—unfortunately often for ill, e.g., by provoking riots. (https://www.oxfordreference.com/display/10.1093/oi/authority.20110803100433179)

An additional question that this paper will address, is whether the notion of ‘rumours’ itself might be conceptualised differently in French, Dutch, and German, such that in one language, different aspects connected to rumours might be foregrounded more strongly than in another.

The article is structured as follows. In Section 2, I will present an overview of the semantics of the three markers in present-day French, Dutch, and German, which will reveal, among other things, the strong semantic affinity between the French conditional and the Dutch zou + inf construction. In Section 3, I will first go into the notion of reportative evidentiality and describe its link with modal notions like commitment and epistemic denial. I will then zoom in on the reportative use of the three markers, as it has been described in the relevant literature. In Section 4, I will address the corpus selection and its annotation; a number of relevant examples will be discussed as well. Section 5 then builds the core of the present article. The results of the corpus analysis will be presented, whereby three aspects will be in focus: the frequency of the use of the respective reportative markers in the subclause after Gerüchte, dass, geruchten dat, and rumeurs selon lesquelles ‘rumours that’; the wider context in which the noun ‘rumours’ ¹ appears (e.g. the main verb it patterns with); and the possible association between the use of a specific reportative marker and an explicit distancing interpretation. Finally, in Section 6, I will summarise and try to account for the main findings.

2. The semantics of the French conditional, Dutch zou + inf, and German soll _ind + inf

The French conditional (see Dendale & Tasmowski (eds.) 2001; Dendale 2001, 2018; Bres, Azzopardi & Sarrazin 2012; Patard 2017; Bres 2018; Van De Weerd 2021) and the Dutch auxiliary construction zou + infinitive (see Harmes 2017; Mortelmans 2022) are remarkably alike from a semantic point of view, despite the fact that they have developed independently from one another and belong to different morphosyntactic categories. The conditional is inflectionally marked (by means of the suffix /ʁɛ/ for the singular and third-person plural paradigm, the first- and second-person plural endings are -rions and -riez) ², whereas zou + inf is a temporal/modal auxiliary construction. For both constructions, four main meanings can be discerned in present-day French and Dutch, respectively, which are highly similar.

First, both the conditional and zou + inf occur in hypothetical contexts, often in conditional sentences, in which they express that the described event is hypothetical or counterfactual (Dendale 2001; Harmes 2017: 152; Patard 2017; Bres 2018: 12). ³

French

Dutch

Second, both constructions function as tense markers to express ulteriority in the past, often in contexts of indirect speech like (4a, b). In (4a), it is the prior speech event (a dit ‘has said’), which is the reference point from which the future action is projected. Note that what was promised eventually did not take place (‘and he forgot me’). In the Dutch example (4b), the conversation between the young woman and the beggar took place in the past (fluisterde de jonge vrouw de bedelaar toe ‘the young women whispered to the beggar’), but what the young woman said pertains to a future moment with respect to the past reference point. Note again that the projected state of affairs does not take place – the context makes it clear that the child was not picked up by the mother. ⁴

French

Dutch

Third, both the conditional and zou + inf can be used to soften the illocutionary force of an utterance, as in (5a, b). For authors like Patard (2017), this use does not count as a separate meaning of the conditional but is merely a ‘meaning effect’ (‘effet de sens’) of the hypothetical one, which is tied to particular verbs – often modal ones – in present-day French. For Dutch, we observe that this mitigating use often combines with modal verbs (like moeten ‘must’ in (5b)) as well.

French

Dutch

Finally, both markers occur with evidential meaning, of which the reportative is the most prominent ⁵ one (reportative examples have been given above, see examples (1a, b)).

The first two meanings – hypotheticality and ulteriority in the past – can be viewed as central or core meanings in both languages: they are already present in Old French occurrences of the conditional (see Patard & De Mulder 2012) and Old Dutch occurrences of zou (Harmes 2017: 156) and they still occur with considerable frequency in present-day French (see Mortelmans 2024). The reportative meaning, however, is a more recent one – Van de Weerd (2021: 234) finds the first occurrence of the conditional’s reportative use in 1507; in Harmes’s corpus material, the reportative reading of zou only pops up in Early New Dutch, although Harmes notes that Dutch historical dictionaries ‘attest evidential meanings for all the language stages’ (Harmes 2017: 158). The reportative reading of both markers is considerably less frequent in present-day French and Dutch than the hypothetical and ulteriority-in-the-past uses of the constructions (see Mortelmans 2024: 703): in a newspaper language corpus, reportative uses account for 21.5% (French) and 18.5% (Dutch) of all uses of the respective constructions.

Turning to German soll _ind + inf, we find that the present indicative form of sollen ‘shall, should’ – the one denoting reportative evidentiality ⁶ – mainly functions as a modal verb expressing (some kind of) deontic modality (see e.g. Diewald 1999, Baumann 2017). A typical feature of deontic sollen is the fact that it evokes an external source of obligation, i.e. it is a third instance (generally neither speaker nor subject) that obliges the subject to act (cf. Baumann 2017: 136: ‘eine Notwendigkeit mit einer dritten Person als modaler Quelle’ [‘a necessity with a third person as modal source’, my translation]. A representative example is presented in (6), in which it is the main clause subject of sagte ‘said’ (er ‘he’), which functions as the modal source of the directive, as he demands to be stabbed.

German

In actual present-day use, and especially in journalistic prose, soll _ind is mostly used not so much to refer to a strong obligation but rather to denote a third person’s intentions, goals, or plans (cf. Baumann 2017: 140), as in (7). As such, deontic soll _ind has a clear future-time orientation.

German

In contrast to the French conditional and Dutch zou + inf, one could argue that soll _ind + inf has a rather low affinity to modal contexts of hypotheticality or counterfactuality. It typically predicates over actions and situations that are to or should be realised.

The reportative meaning of soll _ind is said to have developed in Middle High German (Diewald 1999, Zeman 2013) but qualifies as ‘stabil dokumentiert’ (Gloning 2001: 187) only from 1700 onwards. In present-day German, the reportative use is clearly less dominant than the non-reportative, deontic one; in a corpus of recent newspaper language from Germany, the reportative reading accounts for about 18.5% of all instances of soll _ind (see Mortelmans 2024: 703) – which is remarkably similar to the relative proportion of reportative uses of the reportative conditional and zou + inf (see above). In the following subsection, we will zoom in on these reportative uses and try to provide a first comparison. Before we do this, however, a short introduction into reportative evidentiality and its link with particular epistemic overtones seems warranted.

3. Reportative evidentiality and the reportative uses of the French conditional, Dutch zou + inf, and German soll _ind + inf

3.1. Reportative evidentiality, epistemic commitment, and epistemic distancing

I take reportative ⁷ evidentiality to signal that the speaker is reporting information that they did not witness firsthand but via hearsay. Reportative evidentials, just like other evidentials, primarily aim at providing epistemic justification for a piece of information (i.e. a proposition), in the sense of Boye 2012 (see also Wiemer 2018). Like epistemic modality, the notional category of evidentiality is related to the speaker’s knowledge and belief state; hence, evidential markers can be defined as deictic, i.e. speaker-oriented markers. Importantly, despite referring to hearsay, reportative markers do not necessarily evoke a concrete previous speech act, nor do they obligatorily refer to a concrete or well-defined speaker.

Report[at]ives use the reference to the existence of a source only as a means to an end, namely the justification of a proposition, therefore they will often omit overt reference to this source. […] They are different from other (indirect) evidentials in being the only ones that evoke an entirely separate consciousness as source of information. (Vanderbiesen 2015: 25)

In the literature, a finer distinction is often made within the broader category of reportatives or hearsay markers between ‘reported’ and ‘quotative’ markers (see Aikhenvald 2004: 177; Aikhenvald 2018: 12): The former refer to cases in which the speaker obtained their information from an unspecified source, whereas the term ‘quotative’ is reserved for hearsay with an explicit source. In this paper, I will use the broader term ‘reportative’, in view of the fact that the relevant markers differ with respect to their preference to express reported and/or quotative meanings.

Highly debated in the literature on evidentiality is the issue of speaker commitment (for a good overview, see e.g. Cornillie 2018). For reportative evidentials, it has often been remarked that they do not express any commitment of the reporting speaker to the truth of what they are reporting (Mélac 2014: 56–59, Cornillie 2018). In this regard, the French conditional has been analysed as a non-commitment marker (Abouda 2001, Coltier et al. 2009) and so has reportative soll _ind + inf in German (Faller 2012: 300). Of course, if a particular marker does not signal any speaker commitment to the proposition, it is also compatible with a reading in which the speaker distances him- or herself from the content he or she acquired through second-hand information. Since such denial readings are more or less typical of reportative evidentials (but do not or hardly occur with inferentials, for instance), they are coined under the term ‘reportative exceptionality’ (AnderBois 2014). Following Wiemer (2018) in his in-depth study on the relationship between epistemic modality and evidentiality, such epistemic extensions or overtones – i.e. the explicit signalling of doubt or non-belief – often arise with reportative evidentials, but they can also be suppressed, such that the reportative marker simply signals non-commitment (‘the actual speaker can remain agnostic with regard to their own epistemic attitude’, Wiemer 2018: 92). As they can be easily cancelled or suppressed, Wiemer considers epistemic overtones with reportative evidentials (i.e. distancing interpretations) in terms of Generalised Conversational Implicatures. At the same time, however, cases can be found in which the reportative marker does not easily lose its epistemic overtones (in the case of the Russian reportative marker jakoby, for instance, which keeps ‘its strong connotation of doubt’, Wiemer 2018: 103; see also De la Mora & Maldonado 2015 on Mexican Spanish dizque), i.e. reportative markers may differ with regard to the extent to which they actually evoke distancing interpretations (see in this respect, see also Wiemer & Socka 2017). This issue has been addressed by Mortelmans (2024) for the French reportative conditional, Dutch zou + inf, and German soll _ind + inf, whereby it was suggested that in journalistic discourse, the French reportative conditional is indeed more prone to expressing epistemic distancing than its Dutch and German counterparts. The present study aims to build on this earlier finding in that it addresses the use of the three markers in a reportative-dubitative context (rumours); note that Mortelmans (2024) studied these markers in general ⁸ and not in a specific context that potentially favours reportative readings. In the following Section 3.2, we will take a closer look at the literature on the reportative use of the French conditional, reportative zou + inf, and reportative soll _ind + inf, before embarking on the actual analysis of the three reportative markers in the context of rumours.

3.2 Reportative uses of the French conditional, Dutch zou + inf, and German soll _ind + inf

3.2.1. The French conditional

A long-standing controversy in the literature on the reportative use of the French conditional (see Van de Weerd 2021: 46-51for an overview) pertains to the fact whether it is first and foremost an evidential marker (a position defended by Dendale 2001, 2018); a modal-epistemic marker, signalling a lack of commitment with respect to the proposition (see e.g. Abouda 2001) or denoting the uncertainty of the information (see e.g. Merle 2004); or a combination of both, such that the conditional is said to signal both non-commitment – as a modal meaning – and second-hand information – as an evidential meaning (Kronning 2002, 2012). Scholars like Dendale and Kronning seem to agree on the basic evidential nature of the reportative conditional, whereby the marking of ‘second-hand information’ is a definitional part of its semantics. A similar position is found in the publications of Patard (2017) and Bres (2022), among others. Patard (2017), for instance, views the conditional in terms of a network of three different generic constructions: a hypothetical conditional construction, an ulteriority-in-the-past conditional construction, and an evidential conditional construction, which is characterised as ‘un authentique marqueur évidentiel exprimant l’information empruntée’ (‘a genuine evidential marker expressing second-hand information’ [my translation], Patard 2017: 119). Similarly, in an interesting comparison of the evidential functions of the conditional and the modal verb devoir in French, Bres (2022) also assumes that the conditional is mainly an evidential marker.

A notable exception to this theoretical position – which can be regarded as more or less dominant in French scholarship – is provided by Celle (2020). Celle (2020) questions the evidential nature of the ‘reportative’ conditional and argues that the marking of second-hand information is not a necessary component of this use. Rather, Celle considers the conditional to be a marker of epistemic distancing in that it encodes ‘the non-factual status of an utterance’ in cases in which the speaker has only access to unreliable or unexpected information with respect to the proposition (Celle 2020: 84). Celle does not deny the fact that the conditional is often interpreted reportatively (especially in journalistic discourse) but tries to give a unified account of all its evidential uses (remember that the conditional also has inferential meaning, mainly in questions), which are regarded as ‘evidential extensions’ of a basic epistemic category that expresses epistemic possibility (Celle 2020: 86). For the use of the conditional in reportative contexts, Celle argues that it is used by the speaker ‘to disclaim responsibility’ (Celle 2020: 98). In earlier publications (Celle 2006, 2007, 2009), Celle also stressed the distancing function of the ‘reportative’ conditional: in her view, the conditional not only introduces a different point of view (i.e. the reported speaker) but also questions the trustworthiness of this alternative point of view and thus casts doubt on the proposition (Celle 2009: 284). Similar remarks regarding supposedly inherent distancing features of the French conditional are found in Merle (2004) and Haillet (1998), among others.

Le conditionnel journalistique est conditionnel de reprise : il s’emploie pour reprendre des propos tout en manifestant une prise de distance à l’égard de ces propos, le critère non vérifiable / non vérifié étant toujours pertinent. [‘The journalistic conditional is a conditional of recapitulation: it is used to recapitulate statements while expressing a certain degree of distance from these statements, with the criterion of non-verifiability/non-verified always remaining relevant’, my translation and emphasis] (Merle 2004: 248)

To summarise, there is a remarkably strong attention for distancing uses of the French reportative conditional in French scholarship on this topic. The long-standing controversy regarding its status as a mainly modal or a mainly evidential marker bears witness to this. In fact, the findings in Mortelmans (2024) also suggest that the French reportative conditional indeed occurs with a distancing interpretation more often than its Dutch and (especially) its German counterparts (although neutral readings are the most frequent ones with all three reportative markers). To account for this difference, Mortelmans stresses the multi-perspectivity ⁹ associated with the French conditional: it often combines with direct speech fragments, on the one hand, and frequently occurs in embedded complement contexts, on the other – both of which typically bring another speaker’s perspective into the linguistic scene. ¹⁰ Consider in this respect the following example (8), in which multiple perspectives are present – apart from the writer’s: the author (l’ auteur), who claims that the commission of inquiry (la commission d’ enquête; another perspective) refutes someone else’s thesis (again another perspective) about a barbarian struggle. Note that there are also direct speech fragments – in quotation marks – in this example.

French

3.2.2. Reportative soll _ind in German

Turning to reportative soll _ind in German, it is generally argued to be a straightforward evidential marker, which ‘by default does not carry any epistemic overtones’ (Wiemer & Socka 2017: 51). Also, according to Diewald & Smirnova (2013), soll _ind is a deictic evidential marker – fully oriented towards the reporting speaker (see also Mortelmans 2000); Diewald & Smirnova also point out that soll _ind does not accord any prominence to a preceding speech event or a potential original speaker.

Mithilfe von sollen markiert der aktuelle Sprecher, dass der geäußerte Inhalt aus einer oder mehreren anderen Äußerungssituationen stammt. Es ist dabei unerheblich, wer genau die Äußerung tatsächlich getätigt hat. Wichtig ist, dass der aktuelle Sprecher die Original-Äußerung gehört, gelesen oder in irgendeiner anderen Weise wahrgenommen hat, d.h. dass er die Information über den beschriebenen Sachverhalt aus einer anderen Kommunikationssituation bezieht. [The current speaker marks by means of sollen that the uttered content originates from one or more other utterance situations. It is irrelevant who exactly actually made the utterance. What is important is that the current speaker has heard, read, or in some other way perceived the original utterance, i.e., that he or she obtains the information about the described facts from another communication situation] (Diewald & Smirnova 2013: 454, my translation)

As such, reportative soll _IND can be contrasted with the German present subjunctive (Konjunktiv I), which as a marker of indirect speech integrates the perspective of the reported speaker and the original speech act to which it refers more strongly in the discourse than reportative soll _IND does (see Mortelmans 2009; Diewald & Smirnova 2013; Vanderbiesen 2015, 2016). Consider in this respect the two following examples involving the use of the present subjunctive (gekommen sei), on the one hand, and a reportative soll _ind -construction (soll gekommen sein), on the other. In the first example (9a), reference is made to the original information source which is explicitly addressed (Sie ‘you’, werter Herr Minister ‘dear Minister’) and the original context in which the speech act (in this case more likely: the act of writing) took place (in Ihrer Pressemitteilung ‘in your press statement’). The present subjunctive occurs in an embedded complement clause introduced by the verbum dicendi gaben Sie […] bekannt ‘you announced’. By contrast, the use of reportative soll _ind in (9b) does not refer to an original speaker source nor to an original speech act. The writer solely signals that they acquired the information via hearsay.

German

The above pair (9a, b) also illustrates another observation concerning the difference between the present subjunctive and reportative soll _ind : in contrast to the present subjunctive, reportative soll _ind hardly occurs in subordinate contexts, and when it does, it typically evokes its reportative meaning, i.e. soll _ind is normally not used to attribute a particular assertion to a concrete source (see Diewald & Smirnova 2013: 459).

3.2.3. Dutch reportative zou

Interestingly, Dutch reportative zou has been argued (Mortelmans 2009) to be functionally somewhere in between German soll _ind and the German present subjunctive ¹¹ used to mark indirect speech. On the one hand, Dutch zou can be used as an equivalent of German reportative soll _ind in main clause contexts (see examples (10a, b) from the europarl parallel corpus), without evoking an original speech act nor referring to a specified speaker source. On the other hand, however, since Dutch zou + inf is more strongly oriented towards the reported speaker than German soll _ind , it can also occur in contexts in which German typically uses the present subjunctive as a marker of indirect speech. Examples (10c, d) are based on the same press statement of Buckingham Palace: there is a clear information source (the statement) written by a more or less identifiable author (Buckingham Palace), which can be quoted (as happens in the German text). Dutch uses zou here, as it does not possess any dedicated means to render indirect speech.

Dutch

Let us take stock. The literature overview has revealed that in French, the reportative conditional is often associated with epistemic distance alongside its evidential meaning. By contrast, the German reportative marker can be characterised as a pure evidential marker that generally does not evoke epistemic overtones, i.e. distancing interpretations. In German, reportative soll _ind can be contrasted to the present subjunctive. The latter marks indirect speech, which typically introduces another point of reference – the reported speaker – into the linguistic scene. The Dutch reportative marker zou is compatible with both general hearsay readings (like German sollen _ind ) and uses in which the ‘author’ of what is being reported is more specific and identifiable. In the (relatively scarce) literature on reportative zou (De Haan 2001, Mortelmans 2022), there is no consequent association of zou with distancing interpretations.

4. Corpus analysis: methodology and some examples

For the study reported on in this paper, I collected 200 instances of the German noun + complementiser combination Gerüchte, dass ‘rumours that’, 200 instances of the similar Dutch combination geruchten dat ‘rumours that’, and 200 instances of the French construction rumeurs selon lesquelles ‘rumours according to which’ ¹² from different newspaper corpora. The German instances were collected (a) via the Cosmas IIplatform of the Institut für deutsche Sprache (see https://cosmas2.ids-mannheim.de/cosmas2-web), whereby the search was restricted to instances from the Süddeutsche Zeitung, n = 100) and (b) via the DWDS website (https://www.dwds.de/r), whereby only instances from the daily newspaper Tagesspiegel were taken into account, n = 100). For the French data, online editions of the French newspaper Le Monde and the Belgian newspaper Le Soir were manually searched to collect 100 instances per newspaper. For the Dutch data, the Corpus Hedendaags Nederlands (Instituut voor de Nederlandse Taal (ivdnt.org) was used, again with a restriction to newspaper language, whereby 100 instances were randomly collected from the NRC, a daily newspaper from the Netherlands, and another 100 instances from various Flemish, i.e. Belgian Dutch, newspapers.

Each instance was annotated for a number of parameters: formal characteristics of the finite verb in the subclause (mainly tense, mood (when relevant), and whether it is a modal verb), the wider context of the (main) clause in which the noun Gerüchte/geruchten/rumeurs ‘rumours’ appears (whether it is accompanied by verbs that easily give rise to epistemic overtones like Dutch geruchten weglachen ‘to laugh off rumours’ or French démentir les rumeurs ‘deny rumours’), and whether a clear epistemic overtone (i.e. distance or denial with respect to the proposition in the subclause) can be assumed. Let us consider the following French instances to illustrate the data and the annotations made.

French

In (11), the finite verbs in the subclause are forms of the past conditional (aurait été admis, aurait subi). The noun rumeurs is the direct object of the verb a démenti ‘has denied’ and can as such be said to give rise to an epistemic overtone of doubt with respect to the validity of the proposition (as the Vatican has denied it). The conditional is by no means the only possible form in the relative subclause, however. In (12), the finite verb a rejoint ‘has joined’ is a perfect tense (so-called passé composé in French), which generally does not have a modal flavour – in contrast to the conditional in (11).

French

For (12), there are no reasons to assume that the speaker should somehow express epistemic distance with respect to the proposition in the subclause: the rumours that General Tlass has joined his family in Paris are officially confirmed (cf. il confirme les rumeurs ‘he confirms the rumours’), in a written and undersigned statement by general Tlass himself; there is even – as the previous context makes clear – a confirmation by the French president Hollande.

In the German instance (13), we find present tense indicative forms (liegt ‘lies’, will ‘wants’) in the complement clause. The main clause context is one that does not give rise to doubt or denial: speculations and rumours are ‘fuelled’ (i.e. strengthened) by the fact that the Pope does not plan to visit Argentina, his country of birth.

German

In the following Dutch instance (14), we find a simple present tense form in the complement clause (is ‘is’). The wider context seems to be a neutral one: er zijn geruchten dat ‘there are rumours that’. The following sentence seems to corroborate the rumours in that more concrete information about the number of infected persons is given, albeit again rather vaguely: Her en der wordt gesproken over negen getroffen personen in en rond de club ‘there is talk here and there’. Again, there is no reason to assume a distancing interpretation.

Dutch

The main research questions this study tries to answer concern the distribution of reportative markers in the subclause introduced by dass, dat, and lesquelles; the proportion of clear denial readings with such reportative markers; and whether denial is explicitly marked. These issues will be addressed in Section 5.

5. Main results

5.1. Which verb forms occur in subclauses after ‘rumours that’?

In the following subsections, we will see that a broad variety of verb forms occurs in the subclause introduced by rumours that: reportative markers, on the one hand, but also tense and mood markers, whereby each language seems to have its own systematic preferences and restrictions.

5.1.1. Reportative markers after ‘rumours that’

A first striking observation is that the French conditional is used with a significantly higher frequency in the context of the noun rumeurs than both Dutch zou and (especially) German soll _ind . Figure 1 presents an overview of the distribution. It shows that the conditional appears in a majority of cases (n = 114, 57%), whereas zou is the finite verb in about one quarter of the Dutch data (n = 54, 27%), while German soll _ind occurs in only 12 instances (6%).

Figure 1. German soll, Dutch zou, and French conditional in subclauses after ‘rumours’.

Of course, as we have seen in Section 2, not all instances of these three markers can automatically be classified as reportative. In journalistic prose, only a minority normally functions with reportative meaning (see Mortelmans 2024). Which distribution do we find in this particular corpus? It turns out that in the French data, the conditional has reportative meaning in a clear majority (n = 89/114, 78%) of its occurrences in this particular linguistic environment, i.e. the conditional does not signal ulteriority in the past nor hypotheticality/counterfactuality. An example is provided in (15); a similar one is provided in (11) above.

French

Cases in which the French conditional does not (unambiguously) express reportative meaning occur but are clearly less frequent (n = 25/114, 22%) than in the routine use of the marker. In fact, two types can be discerned here. A number of instances (n = 10) refer to states of affairs that can be regarded as lying in the future from a past perspective and can thus be interpreted as uses in which the conditional expresses its well-described meaning of future in the past, as in (16). Note that a reportative meaning can be argued to be present as well but it combines with a future-in-the-past reading.

French

The remaining cases (n = 15) feature the conditional of pouvoir ‘can’ (pourrait) in the relative subclause, which is best interpreted as expressing epistemic possibility, i.e. the rumours concern a state of affairs that might be realised in the future or at this moment. The conditional could be interpreted here as weakening the epistemic force of the base form peut (cf. the distinction between may vs. might). It is revealing, though, that almost every instance featuring the verb pouvoir in the subclause occurs in the conditional, ¹³ again pointing to the fact that the conditional can be regarded as a default form in this particular context.

French

Summarizing for French, first, the conditional is the most natural form in this context (occurring in 57% of all instances), and second, a reportative interpretation is the most common one when the conditional is used (78% of all conditionals are unambiguous reportative conditionals). And note that even in those cases in which the conditional expresses ulteriority in the past, a reportative reading cannot be ruled out completely.

If we turn to Dutch and German, it becomes clear that neither language fully matches the situation in French. In Dutch, zou(den) appears in 54 instances, of which 20 occurrences evoke a purely reportative reading. ¹⁴ This means that reportative zou occurs in 10% (20/200) of all instances in the corpus, which is considerably less compared to the French reportative conditional, which occurs in 44.5% of all instances (89/200).

Dutch

Turning to German soll _ind , it appears that reportative sollen is marginal in this particular environment: of the 12 soll _ind -instances in the corpus, only 5 (2.5%) occur with a reportative reading. Example (19) is one of them.

German

We could describe this situation in terms of a cline, with French having the strongest preference for a reportative marker in this context, Dutch being in between and German clearly not showing a preference for reportative soll _ind at all. Given these observations, the question arises which (other) verb forms are used – and to some extent preferred – in this particular environment.

5.1.2. Indicative tense forms after ‘rumours that’

The following Table 1 presents a general overview of the verb forms found in the subclauses introduced by rumeurs/geruchten/Gerüchte. It shows that indicative tense forms can be regarded as the default choice in Dutch (n = 146). They are also the most frequent category in German (n = 100), although present subjunctive forms (n = 56) provide an important option as well.

Table 1. Finite verb forms in subclauses after ‘rumours that’.

If we zoom in on the (indicative) tense forms (see Table 2), we find a remarkable similarity with respect to the relative frequency of (most of) the tense categories: past tense markers occur more frequently than (plu)perfect markers, which are more frequent than future markers. The most frequent tense category in Dutch and German, however, is present tense, which accounts for only 15.2% of the French tense markers. We can account for this imbalance if we assume that the conditional is the default tense category in this environment and is used in cases in which Dutch and German have a present tense.

Table 2. Distribution of (indicative) tense forms in subclauses after ‘rumours’.

The high frequency of indicative tense forms in German and especially Dutch suggests that speakers and writers of both languages do not mark the special epistemic status of what is presented as a rumour: the indicative can be taken to express neutral epistemic stance towards the content of the rumour.

5.1.3. Present and past subjunctive forms in German

The relatively frequent use of the present subjunctive in German (see Table 1, n = 56/200) – as a marker of indirect speech – is interesting. It shows that rumours can also be conceptualised as referring to speech events, despite the fact that often neither clear sources nor a concrete act of speaking can be identified. So, in (20), the subclause contains a present subjunctive form (stehe), that can be taken to ‘simulate’ the presence of a previous speech-event.

German

Corroborating this analysis is the fact that we can even detect direct speech fragments marked with quotation marks (‘einen jungen Theatermacher aus dem Osten’ in (21)) in present subjunctive subclauses, again hinting at the presence of a preceding concrete speech event, pieces of which are being quoted.

German

The use of the past subjunctive in German (either as a synthetic form or expressed periphrastically by means of würde + infinitive or hätte/wäre + past participle, n = 32) can be accounted for on two grounds. The past subjunctive is either used as a so-called Ersatzform for the present subjunctive, in cases in which a present subjunctive formally coincides with an indicative (see example (22), n = 11), or we are dealing with epistemic uses of past subjunctive könnte ‘could’, which frequently occur in this context (n = 21). With respect to the former use, let us consider example (22): geschrieben/geholfen hätten is a past subjunctive form (in fact, it is a pluperfect, a so-called Konjunktiv Plusquamperfekt), which – following standard grammar rules – ‘replaces’ the expected present subjunctive geholfen haben, as the latter is formally identical with an indicative (indicative perfect geholfen haben) and as such cannot unambiguously signal indirect speech. Again, uses of this type indicate that rumours in German are easily conceptualised as referring to concrete speech acts.

German

Interestingly, the relatively frequent use of past subjunctive könnte (n = 21) – with epistemic meaning – matches the use of French pourrait (n = 15), which was discussed above. Both verbs typically refer to either possible states of affairs that might materialise in the future (see examples (23a) and (17)) or to a situation that might be the case right now (23b). Rumours can thus be argued to be modally ‘harmonic’ with possibilities.

German

Interestingly, Dutch does not often feature forms of kunnen ‘can’ in this environment (n = 4), and when they occur, they prefer a present tense (kan), which – as we have seen – is the default tense in Dutch in this specific context. We can only speculate on why kunnen is clearly less frequent than pourrait and könnte. Both pourrait and könnte also function outside this context with a relatively well-established epistemic meaning, whereas the epistemic potential of kunnen in general is considerably lower (Nuyts 2001: 188). Moreover, non-reportative zou-instances – expressing future-in-the-past – also code meanings that are very close to epistemic possibility and might thus preclude the use of kunnen in Dutch. So, in the following Dutch example, kan could be replaced by zou without altering the meaning of the sentence.

Dutch

To summarise: after rumeurs, French uses the conditional in 57% of all cases (n = 114), of which a clear majority (78%, n = 89) has ‘purely’ reportative meaning. Dutch has reportative zou in about 10% of all the cases in the sample, whereas reportative soll _ind in German is marginal. In Dutch, simple present (n = 73) and – to a somewhat lesser extent – simple past tense forms (n = 49) are most often found in the complement subclauses after geruchten, suggesting that there is no special marking for the epistemic status of what is being rumoured. In German, we find either indicative tenses (again with simple present (n = 56) and simple past (n = 20) being the most frequent options) or forms of the present subjunctive (n = 56) or past subjunctive (n = 32). The use of present subjunctive (and some of the uses of the past subjunctive) point to the fact that rumours in German are often conceptualised as referring to actual speech events, even allowing for direct quotation.

5.2. Distancing interpretations in French, Dutch, and German

The second part of the analysis addresses the epistemic stance towards the proposition in the subclause: Is there a positive or neutral attitude towards the content of the rumour, or is its content doubted or denied? The main basis for classification is provided by the wider context in which the noun Gerüchte/geruchten/rumeurs appears: When rumours are the object of verbs like Dutch bevestigen, German bestätigen, or French confirmer ‘confirm’, a positive stance is expressed, whereas combinations with German dementieren/züruckweisen ‘deny’, French démentir ‘deny’, or Dutch niet kloppen ‘be false’ are classified as denoting a negative attitude. Neutral stance is expressed in constructions like German es gibt Gerüchte ‘there are rumours’, French on a entendu des rumeurs ‘one has heard rumours’, or Dutch er gaan geruchten ‘rumours circulate’. Note that the attitude expressed is not necessarily the writer’s attitude but may well be a third person’s attitude in cases like ‘she denied the rumours that she had been contacted’ (see example (25)).

The analysis reveals a remarkable difference between the three languages with regard to the frequency of negative stance: whereas in French, cases of doubt and denial abound (n = 91, 45.5%), these are much less frequent in Dutch (n = 32, 16%) and in German (n = 22, 11%). In the French data, explicitly negative verbs and constructions with distancing meaning are highly frequent. The most frequent main verb with which rumeurs combines as a direct object is the verb démentir ‘to deny’ (n = 46). Apart from démentir, rumeurs is the object of semantically similar verbs like balayer ‘to sweep aside’ (n = 4), rejeter ‘to reject’ (n = 4), réfuter ‘to refute’ (n = 4), nier ‘to deny’ (n = 3), infirmer ‘to invalidate’ (n = 2), or dénoncer ‘to denounce’ (n = 2). Highly prevalent in the corpus are hence instances like the following:

French

As table 3 shows, these verbs are more or less evenly distributed in the French (Le Monde) and Belgian French (Le Soir) sample, which lends support to the thesis that the frequent combination of rumeurs with verbs of negative epistemic stance is not a random observation.

Table 3. Verbs of ‘negative’ epistemic attitude in the French sample.

Moreover, we find rumeurs as the subject of predicative constructions that clearly express the speaker’s (or protagonist’s) negative epistemic stance regarding the validity of the rumours: [les rumeurs selon lesquelles y] sont infondées ‘are unfounded’ (n = 3) / fausses ‘false’ (n = 1) / incorrectes ‘incorrect’ (n = 1) / ridicules ‘ridiculous’ (n = 1) / insenseés ‘nonsense’ (n = 1) / exaggérees ‘exaggerated’ (n = 1).

French

Other constructions may also evoke a distancing interpretation: in the following example (27), the actor Jean-Claude Van Damme asks his fans not to believe particular rumours about his health; as such, serious doubt is cast on the content of the rumours.

French

By contrast, verbs of negative epistemic attitude are relatively rare in the Dutch and German sample. The counterparts of French démentir and its synonyms – Dutch ontkennen ‘deny’, ontkrachten ‘refute’, ontzenuwen ‘refute’, smoren ‘smother’, weerleggen ‘refute’, weglachen ‘laugh away’, and tegenspreken ‘deny, oppose’; German dementieren ‘deny’, zurückweisen ‘reject’, von sich weisen ‘dismiss’, widersprechen ‘oppose’, and entgegentreten ‘oppose’ – account for only 14 (Dutch) and 9 instances (German) in the respective samples. In the French data, however, we find no less than 65 instances of these verbs.

In German and Dutch, a neutral attitude towards the content of the rumours prevails: their validity is neither endorsed (e.g. in a context like ‘x confirms the rumours that’) nor doubted. Such neutral non-commitment interpretations account for 80% (160/200) of the Dutch cases and even 87.5% (175/200) of the German ones. In line with this observation is the fact that the most frequent construction used in the main clause in combination with German Gerüchte is the existential construction es gibt Gerüchte ‘there are rumours’ (n = 59, see example (28)), while the equally neutral Gerüchte kursieren ‘rumours circulate’ takes up second position (n = 24).

German

Similarly, in Dutch the neutral collocation er gaan geruchten ‘rumours go/circulate’ is the most frequent combination (n = 47), followed by the existential construction er zijn geruchten ‘there are rumours’ (n = 33).

Dutch

Note that the corresponding existential construction in French (il y a des rumeurs ‘there are rumours’) occurs only six times in the sample, whereas ‘neutral’ combinations with verbs of movement (circuler ‘circulate’, courir ‘run’) are only found twice. The only instance with the verb circuler is the following one (30).

French

Note that cases of neutral stance are not infrequent in French (n = 93, 45.5%), but they are clearly less prevalent than in Dutch (n = 160, 80%) and especially German (n = 175, 87.5%).

Let us summarise the main findings: this subsection has shown that expressions of negative epistemic attitude are significantly more prevalent in French than in German and Dutch. This is also indicated by the verbs with which the noun ‘rumours’ co-occurs most frequently in the respective languages: whereas in French, the verb démentir ‘deny’ is the most frequent full verb (n = 46), Dutch and German show an outspoken preference for neutral collocations that can be associated with non-commitment (er gaan geruchten ‘rumours go’; es gibt Gerüchte ‘there are rumours).

5.3. Is negative epistemic attitude formally marked?

A final question this study intends to answer is whether cases of negative epistemic stance – in French, Dutch, and German – show up a preference for particular markers. For French, we have seen that the conditional is the most frequent marker in the context of rumours. Given the relatively high frequency of negative attitude instances in French, it might be the case that the conditional is more often used when the speaker or a third person wants to express their doubt with respect to the content of the rumour or a straightforward denial of its validity. Interestingly, however, this does not seem to be the case, as Table 4 shows. If we focus on those cases in which an explicit verb of denial (démentir and its synonyms, n = 65) is used in the main clause and look at the proportion of conditional uses in the subclause, we do not find a significantly increased use of the conditional compared to its use in neutral contexts (chi-square = 0.1426, p value = .705725, not significant at p < .05) or its use in positive contexts (chi-square = 0.5272, p value = .467804, not significant at p < .05). The same holds if one compares all the negative instances with the neutral and positive ones: again, there is no significant decrease of conditional use in the neutral (chi-square = 0.1227, p value = .726153, not significant at p < .05) nor in the positive contexts (chi-square = 0.5081, p value = .47595, not significant at p < .05). So, although there is a slight tendency towards more conditional use in negative contexts (59% to 60%) and less conditional use in positive contexts (50%), this tendency is not significant on the basis of my data.

Table 4. Conditional and non-conditional use in French in negative, neutral, and positive stance contexts.

This means that in all three context types, we find conditional and non-conditional tenses, with the former being generally more pervasive, without there being a clear semantic difference between conditional and non-conditional use. Let us consider the following pair (31a, b), which exemplifies the use of different markers in a context of clearly positive stance. In (31a), we find the verb confirmer ‘confirm’ as main verb, of which rumeurs functions as a direct object. Also, for (31b), we can assume that the context in which the rumours occur is a positive one (the police – as a generally credible source – have reasons to believe particular rumours). In these positive contexts, we find a present tense sont en cours ‘are in progress’ in (31a), whereas the verb form in (31b) (aurait quitté) is a conditional.

French

The following pair (32a, b) contains the complex preposition à la suite de ‘following’, which introduces the rumours at stake in a neutral, non-committal way. In (32a), a pluperfect is used (avait truqué) and in (32b) a past conditional (serait venue), again without there being any obvious difference regarding the epistemic stance towards the rumours.

French

Finally, we also find instances with present tense (projette in (33a)) and imperfective past tense and pluperfect (était, avait eu in (33b)) after the negative verb démentir (for the more frequent cases with conditional marking, see examples discussed earlier, (25), (26a), and (27)).

French

The findings for French can be interpreted in more than one way. Although no significant difference can be established between the marking of negative, neutral, and positive stance – which seems to suggest that tense selection is more or less random – we should not forget that the French conditional is the dominant tense form in this context and that the frequent use of the conditional goes hand in hand with an increased frequency of negative stance contexts in the context of rumeurs, a feature that is absent in Dutch and German.

To conclude this subsection, let us have a look at the Dutch and German data with respect to whether and how negative stance is coded. In both Dutch and German, negative stance instances are relatively infrequent: the Dutch sample contains 31 negative instances and the German one 22 instances. Interestingly, though, the Dutch sample shows up a significantly increased preference for zou, as 17 out of 31 negative instances (55%) have zou in the subclause, i.e. one third of all zou-instances in the corpus (n = 54) occur in a negative environment, as Table 5 shows.

Table 5. Dutch zou in positive, neutral, and negative stance contexts.

By contrast, the instances of neutral epistemic stance only feature zou in 22.5% (36/160) of all cases, whereas the positive ones (n = 9) exhibit only one occurrence of zou. The association of zou with negative contexts is statistically significant (chi-square = 12.3685, p value = .000437, significant at p < .05). In the three following examples, we find negative contexts (‘rumours are very unlikely’ in (34a), they are ‘refuted’ in (34b), and ‘put to an end’ in (34c)) combined with the use of reportative zou.

Dutch

For Dutch, therefore, we can posit a link between negative epistemic stance and an increased use of reportative zou. The situation is quite different in German. For one thing, reportative soll _ind hardly occurs in combination with Gerüchte (n = 5). Of these five instances, four occur in a neutral context and one in a negative one. There does not seem to be an association between negative epistemic overtones and the use of reportative soll _ind in German. This raises the question whether negative epistemic overtones are marked in German at all. This does not seem to be the case. The following Table 6 presents an overview of the forms found in the complement clauses introduced by dass in distancing contexts. Of course, the absolute numbers are small, but the high formal variation is nevertheless striking. We find no less than seven different tense or mood types in these 22 instances: present indicative (ind.prs), past indicative (ind.pst), pluperfect indicative (ind.pst.prf), indicative sollen (soll _ind ), indicative future tense (ind.fut, with the auxiliary werden), present subjunctive (sbjv.prs), and past subjunctive (sbjv.pst), without one type being clearly dominant. Remarkably, indicative forms are just as frequent (n = 11) as subjunctive ones (n = 11).

Table 6. Verb forms in negative contexts (German).

This finding corroborates the conclusion that epistemic distance is not explicitly marked in German in this particular context. Indicative forms and subjunctive ones – the latter mainly to be interpreted as markers of indirect speech – can be used here, depending on what aspects of the rumours the writer wants to emphasise. The following instances exemplify some of the variation found in contexts of negative epistemic stance: a present indicative ist in (35a), a pluperfect indicative (aufgehoben hatte) in (35b), and a present subjunctive (unterstütze) in (35c).

German

6. Conclusions

The study presented in this paper has shown that French, Dutch, and German clearly differ with respect to the use of their reportative markers in subclauses after ‘rumours’. Whereas in French, a clear preference for the use of the (reportative) conditional could be established (44.5% of all instances in the corpus have the reportative conditional in the subclause), this tendency is less prominent in Dutch – with reportative zou + inf occurring in about 10% of all instances – and completely absent in German. The frequent use of the reportative conditional in French combines with the observation that in the French data, the combination rumeurs selon lesquelles occurs remarkably often in contexts of negative epistemic stance, for instance, in combination with the main verb démentir ‘deny’ (or one of its synonyms). The most frequent verb in the French sample is indeed démentir, whereas in Dutch and German, neutral collocations prevail (e.g. ‘rumours go’, ‘rumours circulate’, and ‘there are rumours’). This lends credit to the hypothesis that in French journalistic prose, the general epistemic stance towards rumeurs is more negative than in Dutch and German. The fact that rumeurs strongly associates with the conditional – which has been described as a marker of epistemic distancing (Celle 2020), of uncertainty (Merle 2004), or of contestation (Haillet 1998) – points to the same direction.

As a reportative marker, Dutch zou + inf is found to strongly associate with contexts of negative epistemic stance, but such contexts are not as dominant in Dutch (neutral contexts clearly prevail, accounting for 80% of all cases) as they are in French. For German soll _ind + inf, no association whatsoever between reportative soll _ind and negative epistemic stance could be detected. In fact, it seems that negative epistemic stance is not formally marked in German at all. An interesting observation for German pertains to the fact that rumeurs are also conceptualised as referring to speech acts, parts of which can even be quoted. This is at least suggested by the highly common use of the present (and sometimes also past) subjunctive – as a marker of indirect speech also giving prominence to the reported speaker and the original speech act – in subclauses after Gerüchte.

That soll _ind patterns completely differently than the French conditional (and to a lesser extent Dutch zou) can be accounted for on multiple grounds. First, whereas the conditional and zou are tightly connected to the expression of negative epistemic stance in their main use (note that the hypothetical/counterfactual use of both markers appears to be the most frequent one in present-day journalistic prose, see Mortelmans 2024), sollen more strongly associates with a (positive) inclination towards realisation of what is being planned or intended. Second, both the conditional and zou contain past tense morphology and hence often evoke a past temporal point of reference (which is clearly the case in their function of signalling ulteriority in the past), which is distinct from the present speaker’s point of reference. By contrast, reportative soll _ind is closely tied to the present indicative in German and does not evoke an alternative point of reference. More difficult to account for are the (mainly quantitative) differences between the French and Dutch reportative markers: while it is clear that both associate with negative epistemic stance, the French conditional in general occurs more frequently in negative stance contexts.

Remarkably similar are French and German in the relatively frequent occurrence of the epistemic modal pourrait/könnte in the subclause following rumeurs/Gerüchte (pourrait: n = 15/200; könnte: n = 21/200). This hints at another meaning aspect connected to rumours: that what is being rumoured about is conceptualised as possible by the speaker.

Finally, I hope to have shown that a comparative, strongly empirical approach to linguistic data like the one presented in this paper can provide important insights regarding the various language systems. The French conditional is an inherently different reportative from the German one, which is an insight that cannot be won by studying the French conditional and the Dutch soll _ind + inf construction in isolation.