Evidentiality and Stance

A great deal of the literature on evidentiality has been devoted to a discussion of what “counts” as an evidential system (see the review in Dendale & Tasmowski Reference Dendale and Tasmowski2001). Some scholars (e.g., Chafe Reference Chafe, Chafe and Nichols1986, Willett Reference Willett1987, Biber & Finegan Reference Biber and Finegan1989) take a generally broad view of evidentiality, which lumps languages such as English and Spanish, with non-obligatory lexical marking of evidentiality, together with languages such as Tibetan or Quechua, which have complex, obligatory morphological marking for evidentials. Others (e.g., Aikhenvald Reference Aikhenvald2004) argue that there is a divide between strong and weak evidential systems, or in Aikhenvald’s terms, “systems” and “strategies.”

The line between strong and weak evidential systems has traditionally been drawn in terms of obligatoriness and grammaticalization (Lazard Reference Lazard2001). Obligatoriness, however, is a moving target. Evidentials are characteristically multivalent, as Dendale & Tasmowski (Reference Dendale and Tasmowski2001:345) note: “A form may have multiple meanings evoking different subdomains of evidentiality … [and] some of which are not evidential.” There are few if any languages in which evidentials do not interact with other categories such as modality, certainty, and authority. In addition, in many languages, directly attested evidence is unmarked (Lazard Reference Lazard2001). It can be difficult to determine the speaker’s rationale for using or omitting an evidential.

Curnow (Reference Curnow2003:40), a moderate voice in the debate described above, defines evidentiality as

In this article, I will adopt a definition of evidentiality as a category whose fundamental feature is reference to the perceptual source of information. However, I also consider the speaker’s construction of the source of information to be an essential component of evidentiality. It is in the process of establishing a relationship between the source and the speaker that stance is incorporated into evidential systems.

In my view, it is useful to group evidentials as a class on the basis of grammatical reference to the source of information. However, this does not erase the fact that they are deployed in a particular social context. It is in this social context, as part of a process of construction of sources, that stance is incorporated into evidentials. Evidentiality, like reported speech, is a window on the way that reality, social relationships, and authority are constructed through the deployment of linguistic markers (Hill & Irvine Reference Hill and Irvine1993). Because speakers use reportative evidentials to mediate personal relationships, evidentials (even as grammatical and/or semantic categories) are difficult to analyze without taking social considerations into account.

This observation echoes recent work by other scholars interested in evidentials and stance. Like myself, Clift asserts that “to talk exclusively in terms of what an utterance means captures less than characterizing what it does” (2006:585). Clift proposes a new typology of evidentials, distinguishing between “stand-alone” and “interactional” evidentials. She suggests that while stand-alone evidentials orient toward explicitness as a mode of accountability, a nongrammaticalized epistemic stance (in this case, reported speech in English) may orient either to accountability or to inexplicitness to avoid conflict with authority. Clift’s findings have a parallel in my own data—which show that speakers employ considerable leeway in interpreting sources of information, interpretations that are systematically related to self-presentation and audience.

Like Clift, Lempert Reference Lempert2008 looks to poetics – in particular, repetition or parallelism – as a way of accounting for the slipperiness of epistemic stancetaking, its ability to move from the propositional (evidence-based) to the interactional (interaction-oriented). However, while Clift’s work upholds some kind of distinction between types of evidential systems, Lempert’s data show considerable flexibility in the way that Tibetan speakers use epistemic stance to establish or challenge claims to authority. Since Tibetan is a language that is widely regarded as a prototypical example of a strong evidential system, this suggests that all evidential systems, whether “strong” or “weak,” are flexible enough to be sensitive indicators of speaker stance. (However, in his article Lempert takes a critical view of stance as a category. Like evidentials themselves, the concept of stance seems to lend itself to conflation of evidence and interaction.)

Du Bois (Reference Du Bois and Englebretson2007:163) suggests that stance be conceptualized as a kind of triangle, by which a stancetaker “(1) evaluates an object, (2) positions a subject (usually the self) and (3) aligns with other subjects.” Hence, epistemic or evidential stance involves a three-way relationship among speaker, proposition, and audience. I find two aspects of Du Bois’s work especially useful. First, Du Bois relates stance to the Bakhtinian notion of dialogicity, the relationship between a stancetaker’s words and those that precede the utterance in time and space, and through dialogicity to intersubjectivity, the relationship between one actor’s subjectivity and another’s (2007:140). Du Bois sees a need to “look at how the dialogic and intersubjective dimensions of stance relate to the stancetaker’s actions, and to the sociocultural frames that mediate the consequences of their actions” (2007:141). Second, Du Bois stresses the fact that these actions are “serious business, with potentially profound consequences” (2007:173). The weightiness of this responsibility bears directly on the use of evidentials as individuals position themselves in a world that includes not just the immediate past and present, but has implications reaching back into the past and, potentially, far into the future.

Working from a similar model, Kärkkäinen (Reference Kärkkäinen and Englebretson2007:185) also emphasizes the interactional component of stance, arguing that the English item I guess works “to display and project an upcoming stanced action and organize the stancetaking activity between participants,” not as an interactionally strong and salient stance itself. As my data will also show, Kärkkäinen establishes that I guess is not a “ready-made” stancetaking resource, but rather “‘surfaces’ in the sequentially constructed intersubjective weave of dialogue” (2007:212).

Throughout this recent work on epistemology and stancetaking, there is a repeated effort to deal with the tension between reference and interaction. In research on both stance and epistemology, rigid classifications and typologies may work to sort data, but they don’t really help us to understand it. Evidentials such as the one I will describe below do have propositional content; they are not empty placeholders. However, as Du Bois points out, claiming or disclaiming responsibility is a weighty matter, one that may have real consequences for speakers. Therefore, like any other signifier, an evidential can be manipulated to the highly nuanced and highly particular ends of a speaker. For example, the reportative evidential in the Spanish conversation I examine may be used to distance oneself from an unpleasant argument, to disclaim authority, or to set up a particularly juicy bit of gossip. With this in mind, I argue that in order to understand evidentials, the referential core of the evidential must be taken in the context of its use in situated interaction.

However, I also reject the assertion, most prominently made by Aikhenvald in the field of evidentials, that the study of the social aspects of language and the formal aspects of language are two fundamentally different projects. Rather, I suggest that the definition and evaluation of sources of information is a culturally specific process that is open to a wide variety of interpretations. As an example of cross-cultural differences, Hsieh (Reference Hsieh2008:224) notes that reliability assessments attached to different modes of information gathering in Chinese do not conform to typologies based on other languages. Basso (Reference Basso2008:215) suggests that “because notions of truth are not always abstract, and vary widely in world cultures, the study of both evidentiality and epistemology seems to touch very firmly on what may be a persistent Eurocentric dogma in linguistic research.” While the distinctions between perceptual sources of information, reported speech, hearsay, and other types of evidentiality may seem to be transparent, these categories and their associations are in fact widely variable. The tacit assumption that identifying information sources is a transparent, straightforward process is an assumption that is firmly rooted in our own culture and ideologies. I will argue that speakers take advantage of the flexibility involved in identifying sources of information as they calibrate the relationship among themselves, their interlocutors, and the information they wish to communicate.

Linguistic Background: Spanish, Quechua, and Portability

Most dialects of Spanish have little in the way of obligatory morphological evidential marking. Because of this tendency, Spanish, like English, is generally considered to be a language with a weak evidential system. While there is an active discussion of the evidential properties of Romance languages, these are generally a low-frequency extension of a mood or tense, the primary function of which is not related to evidentiality (Schwenter Reference Schwenter and Serrano1999, Squartini Reference Squartini2001, Hassler Reference Hassler, Guldemann and Roncador2002). However, the data that I present show systematic use of evidential features, drawn from a Spanish base but reinterpreted to form an evidential construction that speakers use when communicating reported information.

There are a number of possible explanations of why this shift toward a stronger evidential system has occurred. Although drift, independent innovation, and contact with Amazonian languages cannot be ruled out as factors, the most probable influence comes from historical contact with Quechua. The town where I work was founded by Jesuits in the early 17th century, when Quechua was well established as a lingua franca of the Andes – a policy strongly advanced not by the Inca empire, but by the Jesuits themselves (Mannheim Reference Mannheim1991:8). Toponyms and historical records indicate a longtime presence of both Spanish and Quechua speakers in the area. At present, this valley region is home to speakers of both languages, with a few Aymara speakers as well.

In the Andean region in general, Spanish shows extensive and well-documented influence from Quechua. This influence is believed to extend to all levels of linguistic structure: phonetic, phonological, syntactic, semantic, pragmatic, and social (Adelaar & Muysken Reference Adelaar and Muysken2004:585). Some examples of ways in which Andean Spanish has been claimed to differ from other world varieties of Spanish include the use of double possessive marking, irregular gender and number agreement, syntactic organization (OV vs. VO), and the semantic content of a variety of sentential particles such as pues and nomás (Escobar Reference Escobar2000, Adelaar & Muysken Reference Adelaar and Muysken2004). The Spanish spoken in the region where I work shows the presence of many of the features identified with Andean Spanish and attributed to contact with Quechua.

Many of these contact features, like the evidential markers I will discuss below, were present in some regional or historical variety of Peninsular Spanish but have since disappeared from most modern varieties outside the Andes. It has been convincingly argued that contact features are most easily transferred not only when certain social conditions are met (Thomason Reference Thomason2001:70), but also when the transfer aligns with a previously existing semantic tendency in the target language (De Granda Reference De Granda1997, Escobar Reference Escobar1997, Sánchez Reference Sánchez2004). Potential parallels between Spanish and Quechua semantics are likely sites for contact features to take hold.

It is important to keep in mind that these features are not socially neutral, nor are they evenly distributed among the population. Escobar Reference Escobar, Cole, Hermon and Martin1994 has shown that some features are more typical of L1 Quechua speakers shifting to Spanish, while others are present in the speech of L1 Spanish speakers. Vowel quality is one example of a feature that carries a strong social stigma, and in general a higher density of Quechua contact features in Spanish indexes subaltern status (Howard-Malverde Reference Howard-Malverde1995, Hornberger & Coronel-Molina Reference Hornberger and Coronel-Molina2004).

Quechua is one of the most frequently cited examples of a language with grammaticalized evidentials (Weber Reference Weber, Chafe and Nichols1986, R. Floyd Reference Floyd1999, Faller Reference Faller2002).Footnote ² Two structures are involved in the discussion of evidentiality in Quechua. The “reportative” enclitic –si contrasts with –mi, which indicates a personal assertion, and –chá, which is used for inference (Weber Reference Weber, Chafe and Nichols1986, R. Floyd Reference Floyd1999). These enclitics are used to direct focus as well as to communicate evidentiality, among other functions. Weber (Reference Weber, Chafe and Nichols1986:137) gives the following definition:

These enclitics are used in the following manner (with relevant morphemes marked in boldface):Footnote ³

• Assertion:

  Mamaymi hamushan

  Mother+1sPOSS+AFF come+PROG+3s

  My mom is coming.

• Reportative:

  Mamaysi hamushan

  Mother+1sPOSS+REP come+PROG+3s

  (I hear that) my mom is coming.

• Speculation:

  Mamaychá hamushan.

  Mother+1sPOSS+INF come+PROG+3s

  (Perhaps, it seems) it’s my mom that is coming.

These enclitics can be attached equally felicitously to verbs, adjectives, and other parts of speech.

The other structure, the past tense –sqa, is the subject of more debate. This past tense, also sometimes characterized as reportative (e.g. Cusihuaman Reference Cusihuaman1976), is used in contrast the unmarked –rqa past tense. While de Granda Reference De Granda2001, among others, asserts that –sqa uniquely co-occurs with –si and is entirely evidential, Faller Reference Faller2004 rejects this assertion, arguing instead that it is a deictic operator that places the proposition expressed outside the perceptual field of the speaker. Adelaar classifies –sqa as a “sudden discovery” or mirative tense, indicative of surprise or of a result that runs counter to expectation (Adelaar & Muysken Reference Adelaar and Muysken2004:224).

The sentences above could be modified as follows:

Here, the past tense in the left-hand column is the unmarked –rqa form, while the past tense in the right-hand column is the form that has been argued to correspond to reportative marker, deictic operator, or mirative marker.

It is important to emphasize that the interactions between –mi/–si and –rqa/sqa may be both complex and context-dependent. Feke Reference Feke2004 finds that all possible combinations of –si and –mi, –sqa and –rqa are grammatical under certain conditions. She finds that these combinations represent not only sources of information, but also degrees of certainty, degrees of speaker participation, and speaker distance (in the temporal, spatial, and psychological senses). Feke elicited judgments from native speakers using a number of methods, including role plays, felicity judgments, and observation. She concludes that “Quechua epistemics simultaneously indicate multiple meanings and are used for a variety of pragmatic functions” (2004:133). This result is compatible with the suggestion that these categories may be manipulated by speakers in the process of establishing stance.

The Spanish spoken in the region where I work, as in the Andes at large, makes use of strategies that parallel these Quechua grammatical structures. In particular, there is a body of literature related to the extension of the Spanish pluperfect tense había, which, it has been argued, has been reinterpreted by Andean Spanish speakers to correspond to the Quechua –sqa (Escobar Reference Escobar, Cole, Hermon and Martin1994, Klee Reference Klee, Roca and Benson1996, Sánchez Reference Sánchez2004). Claims regarding the exact semantic content of había and –sqa vary somewhat in this literature. One central point of agreement is the observation that the temporal function of había is secondary, or indeed nonexistent, for many speakers of Andean Spanish. Whether or not these particles are evidential markers per se, they have similar functions and perhaps a similar history to the evidential I describe.

In world varieties of Spanish, diz(que) has a range of uses (Kany Reference Kany1944, Travis Reference Travis2006). The most common uses seem to be casting doubt, mocking, and marking something as unreal or out of one’s control. This is not the sort of diz(que) that I focus on in this paper. While diz(que) clearly has roots in the Spanish particle, I am interested in its use as a true evidential, a reportative marker. Discussing the reportative function, Muysken (Reference Muysken2004:163) dismisses diz(que) as “a non-grammaticalized or at best semi-grammaticalized set of discourse particles.” De Granda Reference De Granda2001, in making a case for the back-borrowing of the Quechua nispa (lit. “it is said”) reportative marker, posits that dice ‘say’ is an impersonal reportative marker. While De Granda does not present evidence to support this supposition, Travis’s article, and my data, support the assertion that that diz(que) is used as systematic, grammaticalized evidential marker.

On the structural side, in a study of Spanish verbs, including diz(que), Company Company 2006 argues that subjectification of verbs, as a historical process, entails a move toward reduced syntax. In this article, the author places diz(que) in the context of other verbs, such as dale and ándale, which have shifted from verbs with full syntactic properties to discourse markers that have been “stripped” of their syntactic qualities. Company Company specifically links the syntactic structure of these verbs to their degree of subjectivity.

Evidentials can be easily borrowed and are an areal feature of indigenous languages spoken in much of South America, from the Andes to the Amazon. For instance, Epps Reference Epps2005 demonstrates that the Amazonian language Hup has borrowed evidential markers from two language families that are unrelated to Hup and to each other. She argues for the interaction of grammatical-structural features and cultural contact in the development of complex evidentiality systems. She concludes that it is evidentials’ “high salience in discourse that makes evidentiality particularly prone to diffusion among languages” (Epps Reference Epps2005:643).

Although no systematic investigation has been carried out, anecdotal accounts universally confirm that Quechua speakers consider evidentials to be essential to polite interaction; in fact, speech without evidentials is frequently characterized as animalistic (Hardman Reference Hardman, Chafe and Nichols1986 for Aymara; Nuckolls Reference Nuckolls1993 and Puma p.c. for Quechua). As mentioned above, it has been argued that convergence between Spanish and Quechua depends both on social environment and also on areas in which Spanish and Quechua semantics may contain parallels (Sanchez Reference Sánchez2004, Reference Sánchez2006; Escobar Reference Escobar1997). The existence of diz(que) as a reportative in historical varieties of Spanish may have provided an opportunity for bilingual speakers to express an extremely salient socio-grammatical category.

Given a large population of speakers of indigenous languages with strong evidential systems, and the apparent portability of evidentiality in other areas of the world, such as the Amazon basin and the Balkan linguistic area, it is not surprising that stronger patterns of evidential constructions have emerged in the Spanish of the Andes than in other world varieties of Spanish. These constructions take the form of existing Spanish structures, but, like the past perfect tense había, have undergone a semantic shift to more closely parallel Quechua than other varieties of Spanish.

Social Background and Methodology

The data that I examine in this paper are drawn from a corpus of some 34 hours of recordings, in which approximately 30 men, women, and children are represented, which was collected during the months of May through August 2005. In the initial review of data, I marked all instances in which diz(que) occurred in this set of recordings, then made notes on the segments of the recordings in which it appeared. On the basis of these notes, shorter segments of about one minute each were selected for transcription and analysis. The data and transcripts were checked both by myself and by a native speaker of this variety of Spanish. Finally, a number of different usages were identified, and short representative segments were selected to present examples of the use of evidentials in this body of data.

The bulk of these recordings was collected from one extended family, representing four generations and ranging in age from babies to octogenarians. During this fieldwork period, I lived with an elderly couple, Don Francisco and Doña Gerónima.Footnote ⁴ In order to gather data, I shadowed Doña Gerónima, helping her with daily chores such as shucking and shelling corn, preparing food, feeding animals and workers, and milking and herding her goats. Doña Gerónima is a skillful storyteller who delights in both traditional and personal stories and serves as a repository of knowledge for her immediate and extended family. I wore a digital tape recorder around my neck, and when I wanted to record, I would ask politely if I could turn it on, then leave it visible as long as it was recording – from 10 to 90 minutes, while we drifted in and out of stories and conversation. At times I made an appointment with another friend or acquaintance, asking her to tell me stories about her life. There was no fixed interview script; I let the speaker tell her own stories and followed the conversation wherever it led us, which was often somewhere else entirely. When asked, I usually told people that I was interested in traditional stories or personal histories. At other times I asked other people to hold the recorder for me, or left it on the table for a few minutes while I went out to do some other task. By the end of my fieldwork, I had recorded a variety of participants in a range of informal conversational settings.

Because of local gender norms, I had more access to women’s conversations than to men’s. However, because I recorded in a family context, many of my recordings came from mixed-gender groups. On some occasions, on the strength of Don Francisco’s sponsorship, my presence with the recorder was tolerated in all-male community meetings and discussions, or when friends came over to the house for a visit.

Virtually all of my recordings came from small landowners, skilled laborers, and their wives and families. In this area, as in much of the rural Andes, farming one’s own land is held in high esteem. Although my consultants were unlikely to have much formal education, or indeed great economic resources, they occupied a respectable place between landless day laborers on the one hand, and wealthier truckers, teachers, and merchants on the other. All of these recordings took place in the homes and fields surrounding a town (population 5,000) situated on the main road between the western highland city of Sucre and the eastern lowland city of Santa Cruz, Bolivia.

This geographical positioning is important because the town and its surroundings are situated on the border between the bilingual department of Cochabamba and the Spanish-dominant department of Santa Cruz, as well as being a way station for monolingual Quechua speakers from the highlands, who earn extra money in this irrigated agricultural area by working as manual laborers during the dry season. While this area has traditionally been dominated by Spanish speakers, Quechua and Spanish have been in contact in the valley region for centuries. Even as linguistic shift is occurring owing to the prestige of Spanish and its prevalence in media and education, waves of migration maintain a constantly renewed contact situation.

The speakers who participated in my research speak Spanish as a first language. However, identification as a Spanish speaker involves much more than language ability. Because of the pervasive nature of language contact in the area, all speakers have knowledge of basic Quechua vocabulary and loanwords, and even those with the least competence are familiar with Quechua sentences as punch lines in jokes or lines in popular songs. Although many have either passive or L2 competence in Quechua, they would never classify themselves as Quechua speakers; to be a Quechua speaker is not only a question of language ability, but also of race, socioeconomic class, style of dress, and place of origin (De la Cadena Reference de la Cadena2000).

In this area, it is prestigious to be a Spanish speaker. However, not all Spanish speakers speak alike, nor do they speak the same way in all situations. My consultants were primarily women; many of them were from rural backgrounds and had little education. In family contexts, men and women of all ages tend to speak a variety of Spanish with extensive Quechua contact influence. In public settings, men and educated individuals tend to shift to a more standard variety of Spanish.Footnote ⁵ Rural, uneducated women generally choose not to speak in public contexts. The term vergüenza ‘shame’ is often used to express a person’s unwillingness to expose his/her perceived ignorance, lack of education, (semi-) illiteracy, and ultimately, poverty to outside scrutiny.

Speakers of this non-prestige Spanish dialect are typically members of a historically subjugated class. Until the 1950s, they were legally serfs in the feudal hacendero system; today, they may be ridiculed by their children, whose teachers (often themselves non-prestige speakers) inculcate them with a strongly normative language ideology. Many people articulate an unwillingness to participate in recording sessions based on the assumption that outsiders consider them backward or uneducated. Quechua contact features such as evidentials are not neutral categories; they are one feature of a style of speech that is widely stigmatized.

It can be difficult for outsiders to foster mutual trust and respect in rural Bolivia. In view of the effects that this fact may have on data collection, I believe it is important to be explicit about my relationship with my consultants. Don Francisco and Doña Gerónima were not only generous hosts; they are also members of my own extended family by marriage. They had known me as a daughter-in-law and as a consistent work partner for three years at the time that I collected these data, and for seven years at the time of this writing. Their children, grandchildren, and great-grandchildren were regular participants and collaborators in my recording sessions, and many of the other individuals in my recordings shared ritual kin ties or work relations with the family.

Because stories are felt to be part of the family realm, my best recording sessions were always with the members of my own family group. In these sessions, I did not have to guide the conversation; rather, it flowed naturally; participants did not appear to be self-conscious either linguistically or physically; and conversation included interactions with participants of all ages. It is not natural, in this culture or in my own, to interview a seven-year-old child; it is natural, in the course of a conversation, to engage him through commands, questions, and stories. Adults address their own children in direct terms they would never use with another adult, often revealing normative ideologies. Likewise, while older people may be willing to engage in monologues with relative strangers, I have found a dynamic, interactive conversation with close family members to be a very productive source of data.

My requests to record naturally occurring conversation were baffling to even my best friends and were met with discomfort and suspicion in many cases. Residents of the more rural areas along the valley were among the most reluctant to be recorded, whether because of genuine modesty or because of skepticism regarding the aims of the entire process. The time that I invested in these close relationships with my consultants was essential to the kind of data I was able to collect. Because of the combined effects of the social stigma sometimes attached to nonstandard features and their relative scarcity in formal registers such as interview situations, I have found that evidentials are unlikely to appear in formal elicitation sessions in the area where I work.

Data From Valley Spanish

Diz(que) is the reportative evidential marker in this variety of Spanish.Footnote ⁶ It is derived from the 3rd person singular present form of the verb decir ‘to say’, dice, with the sentential complementizer que. I cite the evidential form as diz(que), because it may appear either as dice, diz, or with the sentential complementizer as dizque. Indeed, dizque/dice often appears as a frame, with a sentence or phrase beginning with dizque and ending with dice, as in example (1) below.Footnote ⁷

1. (1) dizque ahicito estaba el charango dice

   ‘dizque the musical instrument was right over there dice’

Because Spanish does not require an overt subject pronoun, the marker diz(que) is sometimes homophonous with the 3rd person present verb dice ‘says’. Both its function as a reportative marker and its derivation from the verb decir contribute to the overlap of the two forms. It is not always possible to tell the difference between the 3rd person present dice and the reportative dice. This is especially true of secondhand information, in which either the verb dice ‘s/he says’ or the evidential dice ‘[reported]’ would be appropriate. These cases can be disambiguated when speakers use an overt pronoun to specify the information source, as in example (2) below. Because the speaker uses her mother as the subject of the second sentence, it is clear that she is using dice as a verb, ‘she says’, rather than as an evidential, ‘[reported]’, which appears in the first sentence.

1. (2) Eso dizque era asi. Dice mi mamá, no?

   ‘That’s what dizque it was like. My mother says so, right?’

Although diz(que) has much in common with the verb decir – as one might expect, given its history – the evidential diz(que) is never inflected for person, aspect, or tense. In fact, diz(que) may appear when there is no person referent in the discourse context. In other respects, the diz(que) evidential sometimes acts like a verb; for example, when a complementizer would ordinarily be called for in the sentence, diz(que) always appears with the que complementizer. However, it also appears as a frozen form, for example appearing with the complementizer at the end of a sentence (e.g.. ha ido, dizque ‘he went, [reported]’). Diz(que) may co-occur with the verb decir, for example Dizque decía ‘[reported-that] she said’. When this marker is combined with a saying verb, the saying verb, unlike the evidential, is always inflected for tense and person and refers to some individual in the discourse context.

When a section of discourse is marked as reported speech, diz(que) appears after every phrase. In this discourse context, if a phrase is not marked with an evidential, it can be assumed to be a personal assertion; thus, within certain discourse contexts, it also satisfies the obligatoriness condition. As mentioned in the previous section, register variation in the community means that not all speakers will use these markers in all settings; however, once they appear, they are very reliable.

Diz(que) alternates with grammatical marking of other epistemic stances. Like Quechua combinations of –mi/–si reportative and –sqa/–rqa mirative marker (Feke Reference Feke2004), diz(que) reportative often co-occurs with the había mirative marker. As in other varieties of Spanish, other forms of epistemic modality are indicated primarily through extensions of verb tenses, such as the future (speculation) and the conditional (inference). Notably, inference and speculation by the speaker never appear with reportative diz(que); inference and speculation constitute a different class of information source. This class would correspond to the Quechua –chá (speculation), rather than –si (reportative). Below I present short examples of these epistemic markers from my data. In example (3), the speaker speculates that the person in question may have been dismissed because he didn’t work hard enough. In example (4), the speaker infers from the wateriness of the soup that the lamb had been boiled over and over. The relevant tense morphemes are marked in boldface.

1. (3) Future (speculation)

   Será que el no ha querido trabajar, o será....

   ‘Maybe he didn’t want to work, or maybe…’

2. (4) Conditional (inference)

   Cuantás veces lo harían hervir pues.

   ‘How many times must they have boiled it.’

At the beginning of this article, I defined evidentiality as a grammatical category whose fundamental feature is a reference to perceptual sources of information. Diz(que) acts as a reportative evidential marker in this variety of Spanish: It refers to a perceptual source of information, hearsay; unlike verbs of speaking, it is not inflected for tense, person, or aspect; it is obligatory within certain discourse contexts; and it alternates with other epistemic stances. Finally, diz(que) has extensions, such as relating myths, that are also common to Quechua evidentials (Dedenbach-Salazar Saenz Reference Dedenbach-Salazar Saenz and Howard-Malverde1997, S. Floyd Reference Floyd2005).

This discussion is not meant to establish diz(que)’s status as an evidential beyond a shadow of a doubt; rather, it is meant to clarify its place with relation to other epistemic markers, including verbs of speaking, that are employed in this variety of Spanish. Like the discussion of evidentiality in general, what is ultimately most interesting about diz(que) is how speakers use these features to distribute responsibility for information, and to manage the relationship among themselves, their audience(s), and the information they are conveying.

The transcript below shows how diz(que) evidential can be used to weaken or strengthen an authoritative claim. The four speakers use evidentials to modify their claims to authority as other participants bring new information, and new information sources, to the conversation. In the following transcript, Gerónima and Francisco, an older couple, discuss the provenance and chemical content of the “potable” tap water with their adult daughter, Juana, and adult grandson, Reymundo.

1. (5)

   

Here, G, Gerónima, begins with an assertion that there is iodine in the water, but she softens and distances the assertion by the use of the evidential and tag question (turn 1). Her husband, Francisco, F, responds with the stronger counter-assertion that chlorine is added (turn 4), but continues to distance himself from the assertion with the reportative dicen.Footnote ⁸ He is flatly contradicted by his daughter and grandson (turns 5–9), who make a strong claim that chlorine is not added to the water. Neither of these speakers includes any softeners such as evidentials. Francisco backs off from his previous assertion but continues to defend it by using a verb decían ‘they said’ to introduce a specific source of information (the Town Hall) on which he based his statement (turn 10). Finally, the speakers come back into agreement as Francisco adopts a less authoritative stance (turn 13), framing his new statement with evidential markers. Note that in turn 11, Gerónima uses ha de ser ‘must be’, an inference, which is a different kind of epistemic marker, and thus is not modified by a reportative evidential.

Prosodic features also play a part in constructing these stances. While Francisco uses a reportative marker in turn 4 to indicate that his information is not firsthand, he uses a declarative intonation that contrasts with Gerónima’s more tentative questioning intonation. Juana and Reymundo both use assertive, downward intonation in their challenge to Francisco’s assertion. His response (turn 10) is phrased and intonationally marked as a question. Finally, in turn 13, he makes another declarative statement, although he distances himself by using evidential framing.

This process of negotiation is an example of the use of evidentials to construct stance; as each participant brings his or her own evidence to the table, he or she challenges the authoritative claims of other participants (as Juana and Reymundo do by making unmitigated claims in turns 5–9, trumping Francisco’s claim to authority). None of the speakers in this exchange has personally inspected the municipal infrastructure; all their information is secondhand. Through contrasting uses of statements such as those in turns 5–9, specific sources of information as in turn 10, or the use of the reportative evidential diz(que), these speakers construct a hierarchy of claims to authoritative sources of information. Since these are competing claims, they in effect force Francisco to admit that he is wrong. At the end of the transcript, the speakers realign with each other by agreeing that the source of the tap water is a spring.

While the preceding transcript demonstrates how participants use information source to test their relationship with other participants, the following two transcripts show how reportative evidentials can be used to highlight a speaker’s relationship to the story she tells and to the characters in it. Both sections of speech were taken from the same recording session, as Juana, Ximena, and I rest in the shade during a break from potato planting. The principal speaker, Juana, is the same in both transcripts. The audience, Ximena and myself, is the same. The excerpts occur only a few minutes apart, and our physical and conversational configuration remains identical. Ximena, like Juana, is woman from a rural background, although Ximena is older; both Ximena and Juana use other Quechua contact features such as loanwords (wawa, turn 9) and the mirative interpretation of había (pronounced haya, Line 1) both in these transcripts and in the surrounding conversation.

In first transcript, Juana relates a disagreement that arose between her adult daughter’s best friend, Sosi Caravallo, and Sosi’s sister-in-law, a slightly older woman, whom I call Dora. She knows of this event because her daughter, Reynalda, told her about it sometime afterwards.

1. (6)

   

In this transcript, Juana uses diz(que) regularly after her turns, as she narrates the conflict between Reynalda’s friend, Sosi, and Sosi’s relative, Dora. As I describe above, any stretch of speech that is not marked with an evidential may be understood to be a personal assertion; for example, in turn 1 Juana states that ‘Reynalda doesn’t go to the Caravallos’ anymore’, a statement for which she would plausibly have firsthand evidence from conversations with her daughter. Likewise, in line 5, she states something that is common knowledge: Didi’s mother, who is a public officeholder in the town, isn’t home much. In lines 4 and 14, on the other hand, Juana imitates another person’s voice by changing her own voice quality and pitch (in line 4, quoting her daughter’s friend, she is calm and reasonable; in line 14, quoting Dora, she is high-pitched and aggressive). In all of the material that I examined, I never heard an evidential after a speaker used changes in pitch and prosody to perform another’s speech (acting as an “animator,” in the sense of Goffman Reference Goffman1981). Imitations such as these often indexed derisiveness or parody – another device by which stance is incorporated into the narrative.

Throughout the rest of the recording, after each turn, Juana marks reported information, information that she has heard from her daughter, with a reportative, uninflected diz(que). In lines 13, 15, and 16 this is combined with an inflected saying verb le dice and le dijo – narrating and organizing the story. Note here that while Juana narrates the events within the story with inflected verbs of saying, the story frame itself is marked by diz(que) with no inflection, as a simple reportative, although we can infer that she has heard the story from her daughter.

In the second example, the same speaker is telling a story that was related to her by her godfather’s daughter. The topic of pensions was brought up by Ximena, who told about a family that only visited their grandmother in order to get her pension money. Juana, again the principal speaker, is relating a conversation between her godfather and his daughter, Miriam. The godfather had given his pension money to his son, Mario, who took it away and spent it. Juana narrates as if she had been present for the conversation between Miriam and her father, as Miriam chides the old man for not having given the money to her. However, despite the fact that she narrates the episode without evidential marking, Juana was not present for this conversation between Miriam and her father; rather, as can be seen in line 17, Miriam told the story to Juana at a later time, possibly at her father’s wake, which Juana mentions at the beginning of the transcript.

In example 7, Juana is telling a story about a situation that she empathizes with. She relates a conversation between a woman of her own generation and the woman’s father, who is also Juana’s godfather. Juana opens the story by using the first person plural, Ay, que somos las hijas a veceees ‘Oh, we daughters are sometiiimes…’ In lines 9–17, Juana quotes a conversation between her godfather and his daughter, but rather than evidentials, she uses inflected saying verbs with personal pronouns. Unlike the previous instance, in which the evidential is omitted when speech is marked as “animated” by changes in voice quality, prosody, and pitch, the speaker narrated this segment in her normal voice.

1. (7)

   

The source of information, objectively speaking, is secondhand in both transcripts. In example 6, Juana repeats a conversation that she heard from her daughter; in example 7, she repeats a conversation that she heard from Miriam. So why does Juana use evidentials in the first excerpt, but not in the second? I suggest that Juana differentiates between these two sources because she wants to shade, to nuance her own relationship to the source and content of the information that she is repeating. Transcript 2 involves someone else’s family problems, is potentially contentious and emotionally charged, and primarily concerns members of the younger generation. Juana tells the story from the point of view of her daughter, using her voice to present an unflattering caricature of the unsympathetic character in the story; but she also distances herself from the story through the use of evidentials, as if to say, “This is not my story; these are other people’s problems.”

In the second excerpt, on the other hand, Juana aligns herself with a person of her own age, with whom she shares a ritual relationship (father/godfather). She empathizes both propositionally, in the first line of the transcript (Ay, que somos las hijas a veceeees ‘Oh, we daughters are sometiiimes….’), and implicitly, drawing the story closer to her by choosing not to use evidentials. Although the story was told to her by her godfather’s daughter, Juana construes the information to be of direct relevance to her own experience. Juana’s subjective positioning interacts with her presentation of the information, making a distinction between the natures of the two sources of information.

Here, it is useful to recall Du Bois’s remark that stance can have repercussions into the future. Juana lives in a community and must maintain her relationship with other community members; women of her own generation are especially important allies or destructive enemies. If somebody decided to make Juana’s criticisms of Dora public, Dora could very well make life difficult for her. By using evidentials, Juana avoids taking ownership of the sentiments in the story, while at the same time making her own position very clear. Notice how, in Transcript 2, as the narrative builds to a climax, diz(que) occurs more frequently. In the case of Miriam – whose family, because of their ritual kinship ties, is in a patronage relationship with her own – Juana’s positioning works to maintain and strengthen an existing relationship.

The point I wish to illustrate with these two transcripts is that the conditions for the use of the reportative evidential, and the semantic content of the evidential, do not change. Rather, Juana communicates a differing relationship to the content and source of the stories that she relates by her choice to use or not to use a reportative evidential.