Evidently epistential adverbs are argumentative indicators: A corpus-based study

4.1.Semantic level

This level addresses the semantic types of proposition over which the adverbs have scope. As shown by recent studies, proposition types correlate with specific argumentative components [2] and constitute useful knowledge for the automatic classification of argument schemes [15]. The adopted taxonomy roughly follows that proposed by Freeman [9], based on the dichotomy between broadly logically necessary statements, whose truth/falsity is logically determinate (e.g. “One physical object cannot be in two places at the same time”) and indeterminate or contingent statements (see Fig. 1).

Fig. 1.

Taxonomy of semantic types of propositions.

Broadly logical necessary statements occur rarely in everyday prose and, when they occur, they hardly ever occur with epistentials. This is because they do not need evidence to be known: their truth does not depend on external contingent facts and is graspable through our common intuitions. Among contingent statements a distinction is made between evaluative and non evaluative, natural propositions:

Among interpretations predictions (i.e. “Evidently, he will be joining the attack”, source New York Times) have been singled out since preferably associated with certain subtypes of argument schemes: they, for example, favor the presence of relations from the cause to the effect rather than from the effect to the cause. This level of analysis is exportable to the analysis of any potential argumentative indicator having a propositional scope.

4.2.Intersubjective level

This level applies both to sententially and lexically-driven aspects. As to sententially-driven aspects, the proposition can be presented as the result of an inference of the speaker and/or as inferable by the hearer. When presenting a proposition as inferred, speakers index their own thought processes; this has two argumentatively relevant implications: the hearer may be invited to follow the same path of reasonings [26]; at the same time the hearer is warned that the proposition is a debatable standpoint. Moreover, inferred propositions hints to the available premises excluding direct access to the encoded situation as well as hearsay or report as sources of information.

The reasoning path of the speaker, however, can be followed by the hearer only when the proposition is also inferable by the hearer. Presenting a proposition as inferable by the hearer means pointing to the availability of contextual or co-textual premises. In such cases the premises constitute non-subjective evidence and express intersubjectivity as intended by Nuyts [24]. According to the scholar, the dimension of subjectivity does not rely on the quality of the speaker’s evidence (mathematically measurable or vague) in support of an epistemic evaluation, but rather on the interactive terms of the question whether the evidence is available only to the speaker or more widely known. The presence of (inter)subjective premises enhances persuasion.

Lexically driven aspects deal with the heteroglossic functions played by the adverbs, as intended in Appraisal Theory. Appraisal Theory [44,45] is a framework drawing from the main principles of Systemic Functional Linguistics [13]. Its main goal is to investigate and explain how stances are taken and managed in interaction (heteroglossia) looking at how language is used. Appraisal Theory differs from truth-functional semantic theories: according to Appraisal Theory the communicative function of the verb may in a sentence such as “Sally may be wrong” is not that of mitigating the speaker’s commitment to the truth of the proposition “Sally is wrong”, but of recognizing the presence of other conflicting beliefs (i.e. “Sally is right”, “Sally is partially wrong”).

According to the Appraisal framework heteroglossic strategies are grouped under the label of engagement, which refers to all the linguistic constructions which allow speakers to express and negotiate their positions engaging in multiple voice interactions. As shown by Vandenbergen et al. [32], certainty adverbs are privileged linguistic strategies to indicate the heteroglossic function engagement. Following White [45], the typology of heteroglossic functions is the following:

Knowledge about linguistic resources to express Engagement is crucial for the analysis of argumentative moves for at least two reasons. First, heteroglossic strategies work as rhetorical strategies in the sense that they allow the speaker to position his point of view in alignment or in disalignment with other viewpoints. Second, this negotiation of stances is at the core both of the confrontation stage and the opening stages of a critical discussion [41] where participants respectively establish their disagreement space as well as their common beliefs. According to Appraisal Theory epistentials inherently work as heteroglossic strategies: modals point to the presence of a set of alternatives opening up a dialogic space to account for any of them; evidentials point to the presence of sources of information which are communicatively relevant only if considered in an interpersonal perspective. Therefore, they are likely to prompt the recognition of argumentative moves of support or attack of opinions in the speaker/hearer interaction.

4.3.Syntactic/discourse structure level

This level applies to the occurrences in which the proposition in the adverbs’ scope is presented as inferred by the speaker and/or inferable by the hearer. It addresses whether the premises at the basis of the inferred proposition are textually manifested and what is their position among the following options:

4.4.Argumentative level

This level refers to the argument scheme that is activated in the inferential step from the premises to the proposition modalized by the adverbs. It, therefore, applies to the occurrences in which the premises in support of the truth of the proposition in the adverbs’ scope are textually expressed or easily recoverable from common ground knowledge. The adopted taxonomy of argument schemes is that proposed in the Argumentum Model of Topics (AMT) framework [27]. Compared to other taxonomies of argument schemes ([25,43]) it offers at least two advantages for annotation purposes: it is hierarchical, thus avoiding a proliferation of argument scheme. Moreover, it is based on frame semantic, non-context sensitive criteria (see Fig. 2).

Fig. 2.

AMT typology of argument schemes.

At the higher level, a tripartite distinction is made among:

At the middle level, the semantic and pragmatic connections linking premises and claims are further subclassified. We have adopted the operational definitions provided in the guidelines proposed by Musi et al. [22]:22

At the lower level, causal argument schemes are distinguished depending on the four Aristotelian cause types. Those cases where premises are not accessible have been annotated with the label no discernible to account for the impossibility of identifying an argument scheme.

5.1.Evidently

The annotation of the semantic type of proposition has received perfect agreement. The English adverb evidently has in our sample mainly scope over propositions of the interpretative type (75 occurrences), followed by evaluations and descriptions, according to the following distribution (see Fig. 3).

Fig. 3.

Distribution of semantic types of propositions in evidently’s scope.

The compatibility of the adverb with descriptive propositions could prima facie suggest that the adverb is compatible with direct evidentiality [46]. Direct evidentiality is, in fact, the most suitable evidential mode to present factual state of affairs since it points to direct experience (and testimony) as an information source. However, even in those cases, the presence of evidently entails that the speaker has not directly witnessed the state of affairs encoded by the embedded proposition:

As explained by the author himself, the use of the adverb is justified by the inaccessibility for the speaker of the singing event in its entirety: the speaker infers from the lip movements that the cowboy is singing, even though he cannot hear the song. The proposition is presented as factual due to the high reliability of the perceptual source of information. In such a context, evidently could be replaced by the epistential adverb apparently, but not by clearly, even though both the adverbs share with evidently an etymological reference to vision (apparently < Latin apparent-em < apparere, “to come into sight”; clearly < Latin clarus, “full of light”). The sentence “I say clearly because the sound was off” would sound inappropriate. Differently from “evidently” and “apparently” , “clearly” tends to have scope over events directly experienced by the speaker, i.e. “It is 11.00 p.m. and he was supposed to be here at 10.00 p.m. He is clearly/*evidently/*apparently late”. On the other hand, the adverbs evidently/apparently work as inferential evidentials, incompatible with direct evidentiality.

In the other 3 occurrences labeled as descriptive the adverb indicates experiential inference [1] since based on data perceptually available:

Therefore, evidently neutralizes the distinction between inference and hearsay or report as information sources, working as a marker of indirect evidentiality. It has to be noticed that while evidently can be replaced with apparently in example 3., it would sound less suitable in example 2. The adverb apparently is used to indicate hearsay but it is generally not attested when the information source is a third person. Occurrences where the adverb assumes an hearsay or a reportative function emerge from the annotation as those where the proposition in the verb’s scope is not the result of the speaker’s inference and, at the same time, it is not inferable by the hearer. They all encode interpretations and they include the 4 occurrences in which the interpretation is of the predictive type. These tendencies highlight the modal flavor of the adverb which indicates high speaker’s commitment: states of affairs projected in the future are the less suitable to be presented as certain. In every occurrence where evidently indicates inference there is also inferability on the part of the hearer: when the premises are textually expressed (57 occurrences) they tend to be intrasential or in a sentence textually adjacent, allowing the hearer to easily reconstruct the speaker’s inferential path. When the premises are left implicit (48 occurrences), they contain information which belongs to encyclopedic knowledge, thus recoverable even if not expressed:

The premise which allows the speaker to conclude that the passengers sit in the front of the plane have all paid the upgrade is commonsensical information about the plausible policies of the airline (i.e. “Passengers who paid the actual first-class fare have priority over passengers that just paid the upgrade”). In such occurrences the modal value of the adverb, which indicates epistemic certainty, is more prominent than its evidential function.

Turning to heteroglossic aspects, the distribution of the adverb’s functions seem to pattern with its evidential and modal values: when the evidential value of the adverb is backgrounded in favor of its modal value, evidently expresses concurrence since the truth of the embedded proposition is presented as apparent to everyone. This is the case not only when the premises are left implicit, but also when they are basic unassailable premises [10] whose factuality can be verified by the hearer:

Besides concurrence, which appears to be the most frequent rhetorical function in the considered sample (56 occurrences), the adverb is also employed in 39 occurrences to express pronouncement. This happens generally where its inferential evidential function is more prominent and where the premises, though textually expressed, are non basic premises:

The conclusion drawn by the speaker about the ideological intents’ of the author of the book is based on his personal evaluations of the writer’s narrative style. Therefore, it could potentially not meet other readers’ opinions. In such cases, the presence of textually expressed premises assumes the rhetorical function of presenting the speaker’s claim as grounded, and hence, reasonable. This intent is confirmed in example 6. by the presence of a rhetorical question which “contributes to enhancing the persuasive force of the implicated assertion” [14]. When it indicates hearsay or report, evidently matches with endorsement: the journalist does not take epistemic distance from the truth of the propositions in the adverb’s scope attributing them to somebody else, but he endorses them. It has to be remarked that, differently from evidential verbs such as demonstrate or show, the adverb does not lexically presuppose endorsement, but this rhetorical function is bound to the specific text genre: the journalist gives voice to experts or public characters whose claims mainly contribute as premises to frame the journalist’s standpoint. An exception is the following context, where the journalist’s dis-alignment with the public announcement is made explicit:

The annotation of the middle level argument schemes has received substantial agreement (overall agreement k=0.76), showing that the proposed guidelines lead to similar annotation results with expert annotators. As visualized in Fig. 4, the major aspect of disagreement concerns the label no discernible.

Fig. 4.

Distribution of argument schemes with evidently.

From the qualitative analysis of cases of disagreement between the two annotators, two different tendencies emerge: the first annotator has the tendency to not reconstruct the argument scheme at work where the premises are textually implicit; the other tries to supply the implicit premises where inferable with reasonable likelihood. From both annotations it emerges that evidently mainly selects intrinsic argument schemes, with a clear preference for those of the efficient-causal type. Moreover, it seems to impose constraints on the direction of causality which recurrently proceeds from the effect to the cause and, more specifically, from human observable behaviors to not observable intentions, values or internal states:

In example 10. the premise from which the journalist’s prise de position is drawn is the personal experience lived by Alexandra Mack who had to suffer from the security’s unreasonable request to remove her long sleeves shirt. This singular behavior is construed by the journalist as prototypical example that makes him draw generalizations about the checkpoints sexist habits. The representativity of the cited examples is stressed by the fact that the Alexandra Mack is described as the specimen of a respectable and non-suspected person.

5.2.Evidentemente

Like for the adverb evidently, the annotation of the semantic types of proposition in the scope of evidentemente has reached perfect agreement between the two annotators. However, their distribution is not equivalent to that attested with the English counterpart (see Fig. 5).

Fig. 5.

Distribution of semantic types of propositions in evidentemente’s scope.

From the comparison between Figs 3 and 5, it appears that evidentemente introduces descriptions (i.e. “E a fatica (evidentemente provato) ha ripercorso e ribadito punto per punto la sua versione”, “And putting effort (evidently in difficulty) he retraced and reiterated in detail his version”) less frequently than evidently. In contexts where experiential evidentiality is at stake, such as example 2., the Italian adverb palesemente (“patently”) would fit better than evidentemente (“evidently”). This annotation result well patterns with the different syntactic behavior of the two adverbs: taking a random sample of 100 newspaper articles from the Corriere della Sera and the New York Times it appears that evidentemente is attested in sentence initial position (21/116 occurrences) more frequently than evidently (26/193 occurrences). This evidence suggests that evidentemente works more frequently than evidently as a sentence rather than a manner adverb, having that propositional scope which is necessary for a modal-evidential value to arise. This tendency could suggest that the Italian adverb is more advanced in the process grammaticalization and subjectification encountered by Traugott [38] in her diachronic analysis of evidently. An argument in favor of this tentative explanation is that the adverb appears to be more frequently associated to evaluations, which pinpoint the speaker’s subjectivity.

As to evidential values, the adverb differs from evidently in that it cannot indicate hearsay: if we had to translate in Italian example 3. we would not use evidentemente in correspondence of evidently, but the locution a quanto pare (“as they say”) or the verb sembra (“seem”) (“Ci sono state segnalazioni che Netanyahu ha incontrato i direttori dei giornali di oggi, a quanto pare/sembra/??evidentemente per sollecitare moderazione nella copertura della vicenda”). It is however, compatible with inferential readings where the sources of information are discourses uttered by third parties:

In examples such as example 11., the rhetorical function displayed by evidentemente is genuine endorsement. Differently from cases of hearsay, in which the journalist does not directly commit to the embedded proposition (cf. example 7.), in these cases it would be difficult to conceive his later distancing from Pareto’s point of view. Besides the management of endorsement, the other main heteroglossic functions are those of concurrence and pronouncement. Differently from evidently, evidentemente shows a preference for the latter in line with its more prominent inferential meaning (40/75 occurrences). As a further proof of this semantic difference between the two adverbs, evidentemente is more frequently paired with textually expressed premises (80% vs. 56 % with evidently).

Turning to the annotation of argument schemes, the interannotator agreement is substantial (k=0.71). The distribution of the argument schemes as conceived by the two annotators is visualized in Fig. 6.

Fig. 6.

Distribution of semantic types of propositions in evidentemente’s scope.

The role of evidentemente as a stronger indicator of premises-conclusion relations than evidently is confirmed by the presence of less occurrences where an argument scheme is not discernible. Morever, evidentemente seems to be more specialized than evidently in indicating causal relations of the efficient type with an effect-cause direction:

In example 12., the relation between the premises and the inferred proposition over which evidentemente has scope is explicitly marked by the locution “a dimostrazione” which licenses the inferential rule ‘if the effect is the case, the cause is the case”. In this occurrence, where the evidential function is already signaled by locution, the meaning of evidentemente is prevalently that of expressing modal certainty.

It has to be noticed that, though presupposing premises in the vicinity, evidentemente does not impose constraints of the premises’ positions which can appear both before o after the inferred proposition:

In example 13. the high degree of epistemic commitment conveyed by the adverb is justified by the factual evidence of the sad effects that which are incompatible with effective previous infrastractural interventions.