Evidentiality in Romance Languages

1.1 The Aim of This Book and its Organisation

Some expressions in Romance languages present the reader or listener with a puzzle on closer inspection. Why, for example, is the impersonal expression for ‘say’ in the following Spanish sentence contracted with the subordinating conjunction, and why is it not in the plural as usual (Dicen que llegaron ayer)?

(1) Dizque llegaron ayer.

says-that arrived yesterday.

‘I hear they arrived yesterday.’

When we read the following sentence in French, we do not know whether it is the result of a thought process by the writer, that is, a guess, or the reproduction of an opinion heard. This uncertainty is obviously introduced into the sentence by the conditional:

(2) Cela serait une très bonne chose.

that would be a very good thing

‘That would be great news.’

The presence of such expressions is due to the fact that in our messages, we not only want to communicate facts, but also information about the ways in which we have come to these facts. However, as the French example in particular shows, we are not always precise in doing so and do not have to communicate the exact origin of our knowledge. Of course, we can do so by adding further information:

(3) Je pense que cela serait une très bonne chose.

I think that would be a very good thing

‘I think that would be great news.’

(4) Paul a affirmé que cela serait une très bonne chose.

Paul said that this would be a very good thing

‘Paul said it would be great news.’

Linguistic devices that indicate or imply the way in which the information being communicated was obtained have become a popular subject of linguistic research in recent decades, usually referred to as evidentiality.¹ However, opinions still differ when it comes to speaking of evidentiality for languages that do not have obligatory expressions for the origin of the speaker’s knowledge.

The aim of this book is to demonstrate the explanatory potential of the concept of evidentiality for the Romance languages. In addition, the state of research on the means of expressing evidentiality in these languages is to be outlined. It is only an outline because the large number of publications on evidentiality that have appeared in recent years, particularly on evidentiality in Spanish (Estrada 2013, Sentì 2020), does not allow for a complete overview. We will limit ourselves to a few lines of research and present the results under relevant headings.

The book is structured thematically according to the means of expression of evidentiality and the problems to be presented, and not primarily according to the individual languages. Spanish, French, Portuguese and Italian are considered. Corpus studies for each of these languages were also carried out. Other Romance languages will only be included in individual cases on the basis of existing studies.

In research on evidentiality, there have been many attempts to define this concept in contrast to other topics. We will take up these attempts in the first chapter, and we will also emphasise the interrelations and overlaps of evidentiality with other categories.

The second chapter will deal with the history of the concept and the term evidentiality. Of course, some earlier studies will be taken up that may seem obsolete to researchers oriented only towards contemporary scholarly literature. However, we believe that the future needs a past and that earlier work in which evidential markers were noted definitely deserves our attention, even if the evidential markers were labelled and classified differently.

The third chapter is about expressions of evidentiality, which can be traced back to lexical means. The research on adverbs, which has been carried out particularly for Spanish, will also be taken into account. In the fourth chapter, we will look at verbs and grammaticalisation processes in which some of these lexemes have already specialised in the expression of evidentiality.

In the fifth chapter, we turn to grammatical forms of verbs that have acquired the expression of evidentiality as a secondary meaning. This also includes differences between the Romance languages in the development and frequency of these forms as evidential markers.

The question of whether contact between Romance and other languages with obligatory evidentials may have contributed to the development of evidential markers or increased their frequency of use in Romance languages is considered in the sixth chapter.

We call grammaticalised elements whose main meaning is to indicate the source of communicated information or level of access to this information evidentials. An evidential is a marker of an information source as part of a closed grammatical system (Aikhenvald 2018a: 41). Linguistic signs that are not grammaticalised with this meaning and also fulfil other functions, but can indicate the origin of the speaker’s knowledge, we call evidential markers.

An evidential strategy is the use of a non-evidential category to refer to an information source (Aikhenvald 2018a: 41). These include, for example, intonational means of expression, but also textual combinations and indicators of additional evidential meanings.

Chapters 3, 4, 5 and 6 provide numerous examples to illustrate the phenomena. Idiomatic translations are provided for the examples included in the argument to improve understanding. Interlinear glosses are only given where necessary. We indicate meanings, that is, words used explicitly as concepts, indications of the meanings of words, word groups and sentences in English and therefore also the translations of the examples used, in single quotation marks.

This book is not only aimed at a specialised audience, but also at readers interested in the problems of functional linguistics and at students of Romance languages.

I thank everyone who supported me in the creation of this book. Special thanks go to James McElvenny for his help in editing the English text.

1.2 The Essence of Evidentiality

All languages have means of expressing epistemic awareness of how the content of a message was arrived at. In Romance languages and most other European languages, however, this information is not obligatory. Franz Boas (1942: 182) noted that we would read our newspapers with far more confidence if journalists were forced to indicate whether they had the content of their reports from their own experience, from hearsay or from drawing their own conclusions. We do not perceive this semantic gap in our languages, but we can always close it by making explicit statements about the origin of our knowledge. It is not uncommon for us to do so, not to precisely document the origin of our knowledge from a particular source, but to reduce responsibility for the communicated content.

To avoid misunderstandings, Alexandra Aikhenvald differentiates between the terms evidentiality and evidence: “‘Evidential’ and ‘evidentiality’ as a linguistic category does not imply ‘evidence’, as used in common parlance. ‘Evidence’ covers facts, circumstances, and objects which prove the validity or the existence of something. […] Linguistic evidentiality has nothing to do with providing proof in court or in argument, or indicating what is true and what is not, or indicating one’s belief. All evidentiality does is supply the information source” (Aikhenvald 2018a: 5–7). We do, of course, recognise this fundamental difference, but a relationship can still be established between evidentiality and evidence – the latter not understood as proof, for example, in court, but in the philosophical sense as obviousness and unquestionable insight. We will explain this relationship in more detail in Chapter 2, Section 2.3.

Evidentiality in the linguistic sense refers to knowledge and its origin, primarily on the part of the producer of an utterance. Assuming the possible ways of accessing information, the following types of evidentiality can be distinguished:

i. Visual covers information acquired through seeing.

ii. Non-visual sensory covers information through hearing, is typically extended to smell and taste, and sometimes also to touch.

iii. Inference based on visible or tangible result.

iv. Conclusion based on information other than visible results: this may include logical reasoning, assumption or simply general knowledge.

v. Reported, for reported information with no reference to who it is reported by.

vi. Quotative, for reported information with an overt reference to the authorship of the quoted source. (Aikhenvald 2018a: 12)

These distinctions are only partially confirmed by the circumstances in languages with grammaticalised evidentials. Authors such as Willett (1988: 57) therefore distinguish between direct and indirect evidentiality, where direct refers to access to knowledge through one’s own sensory experience (evidentiality types i–iii according to Aikhenvald) and indirect evidentiality refers to knowledge from directed or undirected messages or conclusions (evidentiality types iv–vi according to Aikhenvald). However, Aikhenvald objects to this view, arguing that all evidential markers in languages with evidential systems are semantically complex and that the simple binary opposition of direct and indirect is insufficient. Instead, she proposes the following classification of semantic parameters in evidential systems. The combination of characteristics results in typological groups of languages, which are listed with the letters and numbers in the second column:

Although our concern is the functional description of evidential markers and strategies² and not the typological classification of languages according to features, the use of the semantic parameters used in Table 1 seems to be useful for the description of Romance languages as well.

Recent work, in particular, emphasises that the way in which boundaries are drawn between types of evidentiality is language-specific. The semantic space of evidentiality can be divided in such a way that the distinction between direct and indirect access to the perception of the communicated facts is in the foreground, for example, in the Balkan and Caucasian languages, where quotative uses are only a pragmatic, contextual variant within a larger group of markers of indirect evidentiality. In Quechua and Latvian, on the other hand, the distinction between quotative and non-quotative forms is made systematically. Plungian (2001: 353) speaks of “different possibilities of clustering evidential values”.

Even in languages with grammaticalised evidentials, these can take on further meanings in addition to the source of information as the main meaning, which also depends on the number of elements in an evidentiality system. In systems with only two evidentials (type A in Table 1), “the firsthand evidential may have overtones of speaker’s control and participation, and the non-firsthand term may imply the opposite” (Aikhenvald 2018a: 16), while the evidential that does not express direct access to the information can also imply the unreliability of the information. In systems with multiple evidentials, the expression of visual access can be linked to certainty or the common accessibility of knowledge. A marker expressing inference can also be drowned out by doubt or speculation, and the use of a reported evidential can also imply the speaker’s doubt about the statement made.

The use of the suffixes -mi/-shi/-chi in the Huallaga dialect of Quechua is proof that, even in languages with pure evidentials, evidential markers assume the pragmatic function of indicating the speaker’s uncertainty. This use of suffixes allows the speaker to accept or reject responsibility for the content of what has been said. With -mi, the speaker assumes responsibility; with -shi, they pass it on to someone else; with -chi, they indicate that the content cannot be the subject of responsibility. According to Weber (1989: 421), the speaker expresses certainty with (5), refers to a message from someone else with (6), and a possibility with (7):

The evidentiality marker -shi used to indicate repetition goes far beyond the mere expression of second-hand information. It is also used in the so-called narrated past, for which one does not want to take responsibility. Quechua culture places great value on avoiding gullibility, which is clearly reflected in the use of this evidential marker. It seems that a blending of morphologically conceived evidentials with pragmatic facts arises as soon as one looks beyond typological issues and considers the use of these elements.

The category of evidentiality is hierarchic in the sense of typological predictability, that is, if a language has specific morphological means for expressing inferences, it also has forms for marking the source of knowledge as non-visual perception and even more so as visual perception:

A promising approach to describing this category seems to be a concept of linguistic universals that focuses on functions and allows for the fact that not all languages have grammatical means for realising the corresponding function. On the basis of communicative-pragmatic and cognitive constraints, the evidentials in some indigenous American languages could be seen as a solution, namely a grammatical one, to the problem of marking the origin of speaker’s knowledge.

Since little is known about the earlier diachronic stages of languages with grammaticalised evidentials, it cannot be ruled out that these developed from lexical expressions for sensory perception, inference, or testimony. Furthermore, it is possible that some evidentials initially had modal meanings, such as uncertainty, assertiveness, or doubt, and that the evidential meanings were added later. The results of typological studies on evidentiality systems of the world’s languages represent a synchronic finding.

These findings are relevant for the study of evidentiality in Romance languages in that overlaps with modal functions can be assumed. These will certainly be even more important in Romance languages without a grammaticalised evidentiality system.

The question of which means of expression can be adopted for evidentiality was also the subject of intensive discussion. If one determines the concept of evidentiality on the basis of findings in languages with grammaticalised evidentiality systems and relies on their specific meanings, one will only recognise grammatical means of expressing evidentiality, which in turn has implications for the determination of the concept of evidentiality itself. This approach is semasiological in that the meanings of linguistic units are taken as the starting point and basis for comparison. By contrast, our starting point is not language-specific, but conceptual. The various types of evidentiality are taken as the onomasiological basis, independently of how they are expressed in the individual languages. They are therefore independent of whether they are realised by lexical or grammatical means of expression, and also of whether their means of expression are situationally dependent or conventional. In this sense, evidentiality is not, as proposed by Anderson (1986), Bybee, Perkins & Pagliuca (1994) and Aikhenvald (2004), a purely grammatical category, but also includes non-grammatical means of expression, primarily lexical ones (cf. Wiemer 2005; 2007, Cornillie 2007a, Squartini 2001; 2008). With this onomasiological approach, we share the central assumption of functional and cognitive linguistics that linguistic structures are language-specific and therefore cannot be used as a tertium comparationis. The conceptual starting point, on the other hand, makes it possible to consider different means of expression with the same or similar functions.

1.3 Evidentiality and Modality

The relationship between the categories of evidentiality and modality is a topic that has been discussed time and again. At the beginning of the twenty-first century, Dendale & Tasmowski (2001: 241–242) distinguish three different positions in the literature:

1. Some authors assume a separation of the two categories (e.g. Bybee, Perkins & Pagliuca 1994: 320–324, Haan 1999a, Haan 1999b; Nuyts 2001: 27; Aikhenvald 2004: 6–8).
2. Others assume that there is a distinction between the two categories, but that there is also an overlap between them (e.g. Van der Auwera & Plungian 1998: 85–86, Palmer 2001: 8–9).
3. A third group of authors assumes an inclusive relationship between the two categories. Palmer (1986: 51) and Willett (1988: 52) assume epistemic modality or modality as a whole as the superordinate category and assign evidentiality as a subcategory to it, while Ifantidou (2001: 5–8) and Plungian (2001: 354) place evidentiality under epistemic modality.

These different categorisations are partly connected to the terminological question of how broadly epistemic modality is defined, but the disagreement extends beyond this question as well. One key factor contributing to the disagreement is that some evidence points to a clear distinction between expressions indicating the source of information and those reflecting the degree of certainty, while other evidence suggests a close connection between the two (cf. Boye 2012: 2).

Boye (2012) assigns his own view to a fourth group of the determination of conceptual relations between epistemic modality and evidentiality. The concept of epistemic meaning can be approached from two complementary angles: one focusing on aspects such as source of information, evidence, and justification – including epistemic justification – and the other emphasising notions like degree of certainty, speaker commitment, and the level of epistemic support conveyed. This emphasises the epistemic nature of both categories, the episteme, which in Greek (ἐπίσταμαι) already meant not only knowledge itself but also the process of acquiring knowledge. Furthermore, the character of proof is suggested again by defining evidentiality as epistemic justification, which, in its awareness and purposefulness, does not have to be attributed to the use of evidentials in language. However, with epistemicity Boye introduces a category that is superordinate to evidentiality and epistemic modality and defines it “in terms of a notion which is a generalization over the notions of epistemic justification and epistemic support: the philosophers’ notion of justificatory support (e.g. Toulmin 1958: 112)” (Boye 2012: 2–3).

However, this does not yet tell us much in terms of the determination of evidentiality as a category, apart from the fact that it is clearly assigned to the field of epistemicity. Furthermore, the question arises as to whether other concepts that have similarities and overlaps with evidentiality but still need to be distinguished from it should be assigned to epistemicity. We will pursue this question in the following sections. First, however, we will focus on the question of the relationship between evidentiality and epistemic modality.

In Romance languages, the distinction between epistemic modality and evidentiality raises difficult problems that cannot be solved easily. Even if we start from linguistic elements with the original function of marking the origin of speaker’s knowledge, this is contingent on the designation of epistemic modality. An attempt to distinguish between evidentiality and epistemic modality in Romance languages is undertaken by Squartini (2004). He points out that a distinction between the source of the communicated knowledge and the indication of the degree of its certainty seems obvious and empirically solid (Haan 1999a; Comrie 2000: 2; Comrie 1983) but can turn out to be difficult when interpreting empirical data (Squartini 2004: 874). In this context, Squartini mentions the work of Dendale & Tasmowski (2001) on the French conditional, which can be used to easily illustrate the cluster of reportative/quotative evidential meaning and low degree of assertiveness.

(8) Il y aurait de nombreuses victimes.