You are on page 1of 31

~ltrlutl ~

ELSEVIER Journal of Pragmatics 32 (2000) 927-957

Evidentiality and deixis in narrative retelling
liana Mushin
Department of Linguistics and Applied Linguistics. University of Melbourne,
Parkville, Vic. 3052, Australia

Received 16 October 1998; revised version 2 August 1999


Evidentiality, the linguistic coding of source and reliability of information, has long been
characterized as a deictic phenomenon in language (Jakobson, 1957; Schlichter, 1986;
Woodbury, 1986). Although it is generally accepted that evidential forms function to index
information to a source and an interpreter of that source (typically the speaker), there has been
little study of the ways in which speakers make use of this property in discourse. This paper
presents an analysis of evidential use in Macedonian narrative retellings that clarifies the
deictic function of evidentiality in discourse. A version of Deictic Center Theory (Duchan et
al., 1995), a recent framework developed specifically for narrative analysis, is used to show
how evidential markers are used to build the perspective structure of narrative texts and how
they are manipulated by speakers to express information from different viewpoints. © 2000
Elsevier Science B.V. All rights reserved.

Keywords: Evidentiality; Deixis; Narrative; Macedonian; Discourse; Subjectivity

1. Introduction

Linguistic evidentiality codes the way speakers acquire knowledge of the infor-
mation they talk about.~ Evidentiality is coded in a range of lexical categories and
grammatical constructions but the focus of evidential study in linguistics has been on
its formal and semantic properties in grammaticalized systems. The Quechua tripar-

* This paper is an extended version of my 1997 ICLA paper 'Evidential variation in Macedonian nar-
rative: A perspective analysis'. The study was partially funded by the Mark Diamond Research Fund.
Many thanks to David Zubin, JeanPierre Koenig and Lesley Stifling for comments on earlier drafts.
1 Many accounts of evidentiality also include degree of speaker commitment or degree of reliability as
an 'evidential' meaning (e.g. Palmer, 1986; Chafe, 1986).

0378-2166/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved.
PII: S0378-21 66(99)00085-5

928 I. Mushin / Journal of Pragmatics 32 (2000) 927-957

tite evidential paradigm in (1) is an example of a grammaticalized evidential system.
The Quechua enclitics, -mi witness/direct experience; -shi report/hearsay; -chi con-
jecture/inference, represent the main evidential categories of Willett's (1988) typo-
logical survey of grammaticalized evidential forms.

(i) wafiu-nqa-paq-mi (a)/shi (b)/chi (c)
(a) Witness: said by the diviner '(I assert that) it will die'
(b) Report: said by someone who brings the diviner's prediction '(I was told
that) it will die'
(c) Conjecture: said in response to the diviner or to the messenger '(perhaps) it
will die'
(from Weber, 1986: 138)

This paper focuses on the deictic properties of evidentiality as a way to under-
stand its role in language use. Evidential markers function to index information to a
conceptualizing individual, typically the speaker, by virtue of how they acquired it.
The semantics of evidential markers must incorporate some notion of the conceptu-
alizer/speaker and their assessment of the source, just as the semantic characteriza-
tion of tense markers makes little sense without some notion of the speaker and
speech situation located in time relative to the time of the events described. Eviden-
tials are classified as a 'shifter', according to Jakobson (1957). More recently Floyd
(1993) characterized evidentials as deictic within a Cognitive Grammar framework.
The close relationship between evidentiality and tense was also noted in Schlichter
(1986) and Woodbury (1986). While tense marking represents primarily the coding
of temporal deixis, evidential marking can be described as the coding of epistemo-
logical deixis.
As deictic phenomena, evidentials are tools for speakers to anchor information in
particular discourse contexts not only to the source but also to their assessment of
that source. For example, use of the Quechua 'witness' clitic -mi in (1 a) indelibly
links the prediction of death 'it will die' to the predictor as witness to the event. 2
From the audience perspective, evidential markers are tools for contextualizing utter-
ances in the process of interpreting them.
Despite the general interest in linguistic evidentiality and its deictic properties
relating to its contextualizing functions, little attention has been paid to the distribu-
tion and function of grammaticalized evidential forms in discourse. 3 Recent cogni-
tive approaches to discourse study have recognized deixis as an essential component
of discourse comprehension and production. This paper adopts one of these frame-

2 The prediction is made by a diviner who presumablyhas seen the event in some vision and therefore
has legitimate authority to speak of the future event as direct witness. A person who knows of the pre-
diction but is not a diviner must use the conjectureform as in (lc).
Approaches to evidentiality within the field of discourse analysis (e.g. Biber and Finegan, 1988,
1989) have tended to focus on evidential use in English, a language which lacks a grammaticalizedevi-
dential category.

1. Mushin / Journal of Pragmatics 32 (2000) 927-957 929

works, Deictic Center Theory (Duchan et al., 1995), as a tool to characterize the
deictic nature of Macedonian evidentiality in narrative and thus to clarify its dis-
course pragmatic functions.
In the absence of much rigorous discourse study of grammaticalized evidentiality,
there is an underlying assumption in the literature that evidentials function primarily
to invariantly code an actual source of information. However discourse studies of
other deictic categories have shown that choice of category is not just a function of
its base semantics. For example, studies of tense variation in discourse (e.g.
Schiffrin, 1981; Dry, 1984; Fleischmann, 1990) have shown that while past tense is
typically defined as indexing an event to time prior to the time of speaking, past
events are not always coded with past tense forms. Tense variations in discourse
have been shown to have a range of textual functions. Based on this kind of evi-
dence, it cannot be assumed a priori that the use of evidential categories is an auto-
matic reflection of the actual source of information.
The lack of systematic study of evidential variation in the literature represents one
gap in our knowledge of evidential systems. A second gap lies in the relationship
between grammaticalized and non-grammaticalized means of coding evidential
meanings. Studies of grammaticalized evidential systems have tended to ignore non-
grammaticalized evidential strategies, even though such strategies are well docu-
mented in other languages (e.g. English see Chafe, 1986; Biber and Finegan, 1989,
and Japanese see Aoki, 1986; Ohta, 1991). There has been little detailed investiga-
tion into how different kinds of evidential markers might function as a system to
code source of information, and the attitude the speaker might have toward it.
This paper begins to address these two gaps in the evidential literature through a
study of narrative retelling in Macedonian, a language with a well documented
grammaticalized evidential contrast in past tense contexts. The formal and semantic
properties of this contrast are discussed in section 2. The text collection methodol-
ogy and subsequent corpus of narratives is detailed in section 3.
The context of narrative retelling forces speakers to talk about information they
only know by virtue of what has been told to them by a previous narrator, a canoni-
cal hearsay context. This kind of data provides the ideal environment to examine
whether indeed evidential coding is simply a function of the source of information,
or whether other factors play a role in the choice of evidential forms. Section 4 pre-
sents an analysis of variation within the grammaticalized system of Macedonian evi-
dentiality. Section 5 examines alternations between grammaticalized and non-gram-
maticalized evidential strategies, including evidential uses of direct speech
The results of this investigation demonstrate that choice of evidential coding in
discourse has a pragmatic basis that goes beyond its semantic characterization as a
marker of source of information. Here it is argued that evidential use reflects the nar-
rator's epistemological stance towards that information - her conceptualization of
the information with respect to her relationship to that source in a particular dis-
course context. The characterization of evidentiality in terms of epistemological
stance allows for mismatches between actual source of information and evidential
coding. It also clarifies motivations for the range of evidential strategies found in the

bread(L) bread(DEF) 'She baked the bread (that's what she told me/supposedly/that's what I heard)' The semantics of the Macedonian Simple Past/L-form contrast has been the sub- ject of some debate in the Macedonian descriptive literature and since explanations of its discourse functions rely somewhat on its semantic characterization.." (Lunt. but has learned about them through hearsay or infer- ence. Like other Slavic languages..the representation of narrative information from different viewpoints. Macedonian has a rich fusional verbal inflection paradigm. If the speaker has not had direct experience of events described.the relinquishment of responsibility for unwitnessed information.' (3) Taa go mesila lebot 3sg 3sg make. as in (3). Evidentiality in Macedonian Macedonian is a South Slavic language spoken predominantly in the Former Yugoslav Republic of Macedonia. Most importantly. (2) Taa go mesi lebot 3sg 3sg make. Mushin / Journal of Pragmatics 32 (2000) 927-957 retelling texts.. In contexts where the speaker has direct experience of the events described. 1952: 91) Thus the L-form is described as encoding a range of meanings.. In contrast to the L- . and the linguistic choices associated with these strategies.bread(3sgsp) bread(DEF) 'She baked the bread (I saw her do it/I vouch for it). it is worthwhile to outline some details of the debate here.930 1. Other meanings are related to the origins of the L-form as a perfect marker and do not seem to have evidential implications - the description of actions which took place in the distant past. One type of meaning is associated with the epistemological status of information . The speaker may be disclaiming responsibilityfor the accuracy of the statement by speci- fying that he was not a witness to the event. an L-form past is chosen. a simple past form is chosen. Retellers exploited the indexical function of evidential forms to develop the perspective structure of the text . This may be realized in two ways. aspect. these forms [L-forms] show an action viewed as distanced in time or reality. 2. It has long been observed that the choice of forms in Macedonian in past tense contexts is dictated to some degree by the speaker's epistemological relationship to the information. and person and number of the subject. Verbs are inflected for tense. but Macedonian speech communities also exist in northern Greece and western Bulgaria. as in (2). this paper confirms that the pragmatics of evidential coding in narrative retelling is directly related to its deictic properties. grammati- calized and non-grammaticalized. or he may be stating an action which started or took place in the past . Lunt's (1952) classic description of Literary Macedonian described the difference between Simple Past and L-form in terms of 'distancing': ".

not a semantic one. Theoretically. nor did he see him drinking alcohol.' In contrast. He 'knows' that the driver was drunk because he (the driver) was seen coming out of a bar. the simple past is described as encoding an event as 'witnessed' with no impli- cations about the 'distance' of the event from the time of speaking. it can be interpreted as representing a con- text where the speaker did not see the subject baking bread. simple pasts are used only in contexts where speakers are prepared to confirm or vouch for information. speakers may use the simple past even when they have not witnessed the event. Thus use of the L-form past connotes that the speaker infers the 4 The relationship between source of information and speaker attitude/commitment towards that infor- mation is an important area in the study of evidentiality. See Palmer (1986) for a more detailed dis- cussion of this relationship. although the most natural interpretation is that the speaker saw the subject baking bread. Mushin / Journal of Pragmatics 32 (2000) 92 7-957 931 form. the speaker uses a simple past rather than an L-form to stress that he is vouching for the validity of the assertion (with the added rhetorical punch that he was the victim of this car accident). if the speaker were just confirming a commonly held expectation that the subject was going to bake bread . . The text example in (4) illustrates the use of simple past in a context where the speaker does not have direct experience of an event. He never spoke to the driver. (4) beshe pijan znam oti beshe pijan be(3sgsP) drunk know(lsg)COMP be(3sgsP) drunk 'He was drunk. Despite the lack of direct experiential evidence. in (2) above. but is still able to vouch for the fact (for example. if they want to stress their commitment to vouch for information (cf. 1. A further impli- cation from this contrast is that the speaker did not directly perceive or experience the events described. Under these characterizations. DeBray (1980) also describes the distinction between simple past and L-form past as a witnessed/ non-witnessed distinction. while according to Friedman the L-form past does not encode 'non-con- firmable/non-witnessed' as part of its core semantics. For example. Evidential forms often carry implications of the degree to which a speaker judges information to be reliable. 4 Under Friedman's analysis.and in fact she did). but not by the speaker. its use in contrast with the sim- ple past implies that the speaker does not vouch for the information. then they must have had direct perceptual or experiential knowledge of it. 1986 'Prepared minds'). Friedman (1986) argues that the alternation between simple past and L-form rep- resents a semantic contrast in speaker attitude rather than evidence and that the wit- nessed/non-witnessed distinction is a pragmatically determined contrast. Aksu-Ko~ and Slobin. His knowledge of the driver's drunkenness is not from direct experience but from a num- ber of different sources (possibly including a police report). I know that he was drunk. the choice of Simple Past versus L-form represents a clear evidential distinction between witnessed/directly experienced events and non-directly experienced events. an implication being that if a speaker is able to vouch for information. Here the speaker is describing a car accident in which he was side-swiped by a drunk driver.

the speaker uses an L-form to imply that he did not actually witness the drunk driver opening the door of the car (although he is able to vouch for the fact that the driver ended up back in the bar). 5 To this extent the Avoidance of past tense is a relatively common strategy amongst Macedonians for evading the deci- sion whether or not to vouch for something. it can be argued that the simple past functions deictically (a) to temporally index the time of the event as prior to the speaking time. In past tense contexts. It is also wholly consistent with Friedman's (1986) characterization of the semantic differences between the simple past and the L-form past. This source is typically assumed to be what someone else has said (i. It does not necessarily imply however that the speaker knew about the door of the bar opening from hearsay.' The analysis of simple past/L-form distinction has been presented rather loosely in terms of whether the speaker is committed to vouching for the information. hearsay).) observes that historical present is used very frequently in conversationalnarrative. or assert no epistemological link to the experiencer (using L-form). and (b) to epistemo/ogically index the event to the experiencer of that event. in the absence of context. This analysis is supported by the evidence from the corpus of retellings. and (b) to epistemologically delink the event to the experiencer. the L- form functions deictically (a) to temporally index the time of the event as prior to the speaking time. How- ever. In other words. speakers are required by the language to make a choice of forms which distinguish epistemological deictic orientation. but it need not be. They might avoid past tense altogether but once past temporal deictic orientation is selected. For example.932 I. the natural interpretation of (3) is that the speaker was not present to witness the event of bread-baking and thus knows about it only from some external source. the L-form does not index the information to any specific source or type of source but rather asserts that the information is not indexed to the experi- encer. Evidentiality is a deictic category because it indexes some information to the source and/or epistemological status of that information. it is possible to give a more precise characterization of the Macedonian forms in terms of the deictic properties of evidentiality. (3) above can theoretically have a past perfect reading 'She had made the bread'. .c. While all of the examples support this analysis of the contrast. In contrast. In the case of the Mace- donian past tense contrast. speakers must either deictically link the information to the experiencer (using Simple Past). Friedman (p. In (5) below (from the same 'car accident' text as (4)). Mushin / Journal of Pragmatics 32 (2000) 927-957 information from some indirect evidence.e. as presented below. (5) toj ja otvoril vratata i pravo vo barot da ti 3sgNoM 3SgACC open(L) door(DEF) and straight LOC bar(DEF) to odi pak da pie go(3sg) again to drink 'He (had) opened the door and straight into the bar he went (goes) again to drink.

She forgot to put oil in the pastry and it came out like stone. Unlike Velika's Story. 4 first generation and 4 second generation retellings of Velika's Story were collected from women of the same generation as the original teller and who originated in the same area of Mace- donia (the villages around Bitola). The corpus thus contained texts for which the actual source of information was transparently calculable from the context. Participants were asked to listen to a pre-recorded story which was an account of the storyteller's own personal experience. . They were then asked to retell the story they heard to another participant who had not heard the story. The second story. Although they had lived in their new countries for some time. Contrary to expectations. The participants in my study all used the L-form consistently. a retelling of the personal experi- ence and a retelling of the retelling. 'Zelnik Story' (MZS) was another cooking failure story told by a woman who twenty years previously had tried to make zelnik (a type of stuffed pastry) for a dinner party. the first generation retelling of 'Zelnik Story' was told by the original audience of the story. allowing clear hypotheses concerning evidential use to be generated. but who did not know the original teller. 6 The text corpus used for the examination of reportive marking included 3 personal experience stories told by speakers from the Bitola area of southwestern Macedonia. Australia and Toronto. Canada. 'Bread Story' (MBS) was about the teller's embarrassment at baking some bad bread for her mother-in-law when she was first married and living with her hus- band's family. Friedman (1988) notes that the south-western dialects are losing the status distinction in the past tense (using simple pasts where the standard calls for an L-form). all spoke Macedonian as their home language and to each other. 6 It should be noted that there are systematic differences between the dialect spoken by my participants and the standard literary language which was used as the basis for the characterization of the tense sys- tem above. 1. The recording sessions took place at community centers in Melbourne. Mothers-in-law are notoriously strict to their daughters-in-law in the traditional Macedonian family. Mushin / Journal of Pragmatics 32 (2000) 927-957 933 distinction between Simple Past and L-form past represents a grammatical choice for Macedonian speakers. who was a m e m b e r of the same social group as the original teller and who retold her version to another member of this social group. 3. Thus for each group there was a personal experience story. this mother-in-law was very understanding. All participants originated from the Bitola area (SW) of the Former Yugoslav Republic of Macedonia. Text collection methodology The corpus of texts used in this study were all retellings of another person's nar- rative account of personal experience. This participant was given the same instructions and retold the retelling to a third participant. The original teller was a woman in her 40s who was telling the story to her own daughter as they sat around the kitchen table.

Gloss- ing abbreviations are given in the appendix. (6) Bila nekoja nevesta mlada. Mushin / Journal of Pragmatics 32 (2000) 927-957 The third personal experience story. the current speech situation. The passage in (6). the most c o m m o n inter- pretation of L-form clauses by Macedonians is that knowledge of the event was acquired through some- one else's report. Although the L-form does not exclusively code information as reportive (cf. from the retelling corpus. taking into account extranarrative evidential coding (section 5).' (MBSc2) 9 Although some retellers did consistently and frequently use the L-form past in their retelling. 7 Finite verb forms in all units were also coded as Simple Past. The retelling paradigm used here provided speakers with a context in which they were forced to talk about past time information that they did not directly experience or witness and which they knew only by virtue of hearing it from someone else. the Quechua reportive clitic -shi in (1) which does mean that information is known by virtue of hearsay). . Svekrvata poshla na be(L) some bride young knead(L)bread M-in-law(DEF) go(L) Loc pazar market 'There was (L) a young bride. the variation can be directly attributed to the deictic properties of evidentiality. The mother-in- law had gone (L) to market. Altogether. 'New Car' (MNC) was told by a young woman who had recently acquired her first car by going to an auction with her father. This is precisely the type of context where an L-form was expected. She had kneaded (L) some bread. etc. L-form past or non-past. In both cases. represents the predicted pattern of nar- rative clauses under this hypothesis. the retelling corpus consisted of 6 first generation and 6 second gen- eration retellings. a subscript letter for multiple groups of retelling. and a number representing the generation of retelling. These units were classified according to their narrative status. This text was the second gen- eration retelling in the 'c' group of retellings of the 'Macedonian Bread Story'. Furthermore. 7 These included information about the reteller's own experience as the current narrator.those that represent information within the confines of the story. Certainly the default interpretation of L-forms in the context of narrative retelling must be as a reflection of the reteller's presentation of the narrative as obtained from someone else's version. The retellers of her story were all of the older generation who had migrated to Australia as adults. Of particular importance is the distinction between narrative units . her opinion of the story she was telling. 9 Each text was assigned a label: a three letter abbreviation of the story. 8 It was therefore pre- dicted that L-forms would occur very frequently and consistently throughout the sto- rytelling. This teller had grown up in Australia.those that represent digressions from the telling of the story. Detailed analyses of two aspects of this variation are presented in the following sections: variation in evidential coding in narrative units (section 4) and variation in overall evidential strategies. there was considerable variation over the whole corpus. since simple pasts typically code information that was acquired by direct experience.934 1. and extra- narrative units . Mesila leb. These were transcribed and chunked into clause-like units. they were predicted not to occur in narrative retelling.

forms which were predicted not to occur. 1° Although no text completely lacked simple past or non-past forms. Evidential coding and narrative perspective Table 1 gives the percentage frequencies of L-forms. simple pasts and non-past forms for each individual retelling text. form clauses) . there was a certain amount of individual variation in the ratio of different tense forms to each other. Of the non-L forms. Table 2 Profile of evidential variation in total narrative retelling clauses Tense form Total (n=359) L-form 39% (139) Non-L Form 61% (220) Simple past 18% (66) Non-past 43% (154) 10 One retelling in the corpus completely lacked L-form pasts. The proportion of L-form marked clauses ranged from 19% to 68% of clauses (most texts in the 30-67% range). 18% of clauses were simple pasts (30% of non-I. Mushin / Journal of Pragmatics 32 (2000) 927-957 935 4. The explanation for this particular unex- pected result is given in section 5. Non-pasts ranged from 10% to 63% of clauses (most texts in a 32-55% range). Table 2 summarizes the distribution of L-forms and finite non-L forms over the whole corpus. Since the purpose of this section is to investigate alternations between simple past and L-form past. Only about 2/5 of clauses in the retelling corpus were coded with an L-form. the text was not included in this analysis. . no text only used L-forms and every text made some use of simple past forms. I. Simple pasts ranged from 4% to 33% of clauses (most texts in a 10-27% range). Table 1 Percentage proportion of tenses in narrative retelling Text Clauses L-form Simple past Non-past MBS~I 37 19 (7) 27 (10) 54 (20) MBSa2 35 37 (13) 31(11) 31 (11) MBSbl 31 48 (15) 10 (3) 41 (13) MBSb2 48 42 (20) 21 (10) 37 (18) MBScl 30 23 (7) 27 (8) 50 (15) MBSc2 28 57 (16) 11 (3) 32 (9) MBSdl 15 20 (3) 33 (5) 47 (7) MBSd2 46 33 (15) 4 (2) 63 (29) MZS1 37 54 (20) 14 (5) 32 (12) MZS2 19 68 (13) 21 (4) 10 (2) MNCI 33 30 (10) 15 (5) 55 (18) Total 359 39 (139) 18 (66) 43 (154) The results show that contrary to expectations.

2 54 (13) 46 (11) 100 (24) MBSbl 83 (15) 17 (3) 100 (18) MBSb2 67 (20) 33(10) 100 (30) MBScl 47 (7) 53 (8) 100 (15) MBSc2 84 (16) 16 (3) 100 (19) MBSal 88 (15) 12 (2) 100 (17) MBSa2 38 (3) 62 (5) 100 (8) MZS1 80 (20) 20 (5) 100 (25) MZS2 76 (13) 24 (4) 100 (17) MNCI 67 (10) 33 (5) 100 (15) Total 68 (139) 32 (66) 100 (205) T 80 6O 40 MBSal MBSa2 MBSbl MBSb2 MBScl MBSc2 MBSdl MBSd2 MZS1 MZS2 MNC1 Fig. 1 shows that there was a considerable amount of individual variation in the fre- quency of L-forms with respect to the total count of past tense forms in the corpus. The proportions range from nearly equal frequencies of L-forms and simple pasts (in . Fig. Individual variation in L-form frequency. This is because the evidential contrast that exists between simple pasts and L-forms is lost in non-past contexts. Fig.936 I. Table 3 presents the results of this calculation for the individual texts. Only in past tense contexts is the reteller required to make a choice between representing infor- mation as 'direct experience' (with a simple past). It is therefore necessary to filter non-past forms from the calculation to achieve a clearer picture of the ratio of simple pasts to L-forms in the retelling corpus. The per- centages represent the percentage of simple past or L-form past as a proportion of total past tense clauses. 1 charts the individual variations in L-form frequency. or as report (with an L-form past). Table 3 Percentage proportions (and numbers) of Simple pasts and L-forms Text L-form Simple past Total past MBS~I 41 (7) 59(10) 100 (17) MBS. Mushin / Journal of Pragmatics 32 (2000) 927-957 The earlier hypothesis made no predictions concerning the distribution and fre- quency of non-past forms in the corpus. 1.

1995: 132) and an object of perspective ("the content of the deictic window as it moves along its spatial. ". 1995: 132). 1995).1.. Regardless of the frequency..... I/you of face-to-face interaction.".. Mushin / Journal of Pragmatics 32 (2000) 927-957 937 3 texts) to a 4:1 ratio in favor of L-forms (in 4 texts). object of perspective rlndow:" ~sychologica] ~nate8 establishing a de: l~Lt'Jpective in the narrated w Fig.... a cognitive science based approach to the study of narrative structure. 2. space and person from which the story world is exposed to the reader/listener .. . 2. The origin of perspective however may lie anywhere within or without the story world.. ? '! ii ~ 7ili~ii~i~i i!¸ . 4. I. Deictic Center Theory. Deictic center theory The analysis of the unexpected uses of simple past forms is based on the frame- work of narrative analysis adopted in Deictic Center Theory (Duchan et al. where deixis is cut adrift from its physical moorings " Zubin and Hewitt (1995: 130) The deictically oriented 'window' onto the narrative world crucially involves the establishment of an origin of perspective ("a shifting localization in time.. into the purely textual realm of fiction.. By definition. These properties of the deictic window are illustrated in Fig. temporal and personal coordinates throught the story world". the unpredicted use of simple past forms in Macedonian narrative retelling must still be explained. where it is anchored in real world situations. the object of perspective must lie somewhere within the story world. This issue is addressed in the following subsections.. Reasons for the range of individual variations will be addressed at the end of the paper. attempts to model the consequences of shifting deixis out of the here~now.. 1995).: Object and origin of perspective (from Zubin and Hewitt.

the speaker and hearer in the actual speech situation have the 'privilege' of seeing into that character's mind. the theory assumes a storytelling event in the real world (where the teller and audience exist). Mushin / Journal of Pragmatics 32 (2000) 927-957 The degree to which the origin and the object of perspective are located in the same domain represents the degree to which information is perspectivized. 1975) and epithets (Banfield. C = 'character') Fig. In this case. Tellers can represent narrative information as it relates to the WHERE. 3 is a simple representation of this model of narration. the deictic center) in the narration. like 'objective' facts about the story world. Although Deictic Center Theory was developed to account for fictional narrative. Model of narration. In an actual telling. 1982). The story world is the object of perspective but both worlds may function as the origin of perspective (i. and a story world (where the characters of the story exist and interact with each other). A = 'audience'. Story World Real W o r l d S p e e c h Situation (T = 'teller'. WHEN and WHO of the real world speech situation.938 I.ll Deictic Center Theory is thus a practical application of the basic semantic notions of perspective and deixis to account for the use of lan- guage in fictional literature. Fig. some of which are accessible to anyone in the discourse world (including the teller and audience in the actual speech situation). their thoughts and perceptions). maintaining the 'real world' as the origin of perspective. The issue of degree of perspectivization is particularly relevant to the study of narrative where different characters know different things. . written stories that were not tied to any particular instance of telling. Both the origin and content of perspective are interpretable only through the dis- tribution of deictic forms.e. 3. irrespective of deictic center. the fundamen- tal concepts of the theory are applicable to the kinds of oral retellings used in this study. The result is a report of the characters and events in the story world that maintains the presence of the speech act participants in the real world speech situa- i~ Includingforms with "lexicalized" deictic interpretationssuch as directionalverbs (Fillmore. Other types of information are accessi- ble to only one character (e.g.

contribute to the interpretation of information as more or less expressive. independent of the actual teller and audience. Source of Linguistic Story Coding World ~ Real W o r l d S p e e c h Situation (T = 'teller'. 12 The reportive/expressivecontinuum is a discourse application of Langacker's (1990) Cognitive Grammar notions of subjectivity and subjectiflcation. C = 'character') Fig. Expressivity is coded with a variety of linguistic phenomena. a teller can represent narrative information as it relates to the WHERE. The result is a representation of characters and events in the story world that diminishes the presence of the speech act participants in the real world speech situation. Li (1992) has demonstrated the usefulness of this application in her study of reflexives in Mandarin discourse. character is deictic center). long distance reflexives. Expressiveframing in narrative (story world is origin and object of perspective. and minimizes the experience itself. and others. A = 'audience'. These devices. Banfield (1982). 4. These include use of temporal. . Under these circumstances. When narrators present infor- mation as a report that reflects the communicative intent of the current speaker and interaction with the audience. Zubin and Hewitt (1995). 1995). 4 and 5 model the expressive and reportive framing of narrative information respectively. choice of referring expressions and epithets. Wiebe (1989. the result is a more reportive mode of representation. spatial and personal deictic terms. The backgrounding or 'effacing' of the speech act participants in the real world results i n a more expressive mode of representation. many of which are dis- cussed in Cohn (1978). narrative information is deictically centered somewhere within the story world. Mushin / Journal of Pragmatics 32 (2000) 927-957 939 tion.the degree to which the experiencer/agent is coded in the linguistic signal at the sentence level. Alternatively. Use of language in narration thus can be characterized as more or less 'expressive' according to the degree to which the expe- rience itself is foregrounded as an iconic representation that diminishes the presence of the actual speaker and audience) 2 Figs. WHENand WHO of the story world itself. 1. use of def- initeness to represent the knowledge state of story characters rather than actual speech act participants.

where the words of the character are presented as the character's voice as they speak in the story world. Mushin / Journal of Pragmatics 32 (2000) 927-957 ~~ "~" "~rA'~ ] Real World Speech Situation Story World Source of Linguistic Coding Fig. 5. The most expressive type of language use in narration is the direct representation of a character's speech or thought. The second and third sentences in (a)-(d) illustrate the shift from more reportive to more expressive modes of representation coinciding with a shift in deic- tic center from external to internal to the story world. She was quite surprised. She glanced towards the door when she heard the noise. signaled by third per- son reference and past tense forms of the verbs. These linguistic signs index the nar- rative information with respect to the time of narration and to the identity of the nar- rator. or train of thought as they . She glanced towards the door when she heard the noise. The object of perspective is the character who observes her father in the doorway. Her father stood in the doorway. Without further information (e. story world is object of perspective. The shift from reportive to expressive modes of repre- sentation is illustrated in (7). The distinction between reportive and expressive language is not binary. She glanced towards the door when she heard the noise. b. The origin of perspective however is somewhere external to that character. She glanced towards the door when she heard the noise.g. she wondered. My goodness! What was Daddy doing standing in the doorway? d. She was quite surprised. c. (7) a. Utter- ances can be rated as more or less 'expressive' based on the accumulation of differ- ent types of linguistic cues. speech act participants are the deictic center).940 1. "My goodness! What is Daddy doing standing in the doorway?". In all four example passages (a)-(d). the origin of deixis is typically interpreted as centered outside of the storyworld with the exter- nal narrator. more specific deictic expressions). the first sentence is quite reportive. Reportive framing in narrative (real world is origin of perspective. Daddy was standing in the doorway.

taking into account the degree to which they are embedded in a more 'expressive' mode of representation. It is within the Deictic Center framework that the alternation between evidential L-forms and simple past can be explained. the origin of deixis may shift from the actual narrator external to the story world to any number of spatio-temporal points within the story world. which links information to an experiencer. However the ear- lier characterization assumed that the current teller was consistently the origin of deixis and the choice of past tense form was therefore a function of that teller's epis- temological relationship to the information. Evidential variation and the expressive/reportive continuum The following subsections reexamine the use of 'problematic' simple past forms in the Macedonian retelling corpus. The character is talking about some bread which she baked that turned out badly. all deixis is centered with the speaking character and the actual teller is ostensibly 'invisible'. grounding narrative information with respect to some conceptualizer.2. They also demonstrate the function of the system as an important tool for perspective shift in narrative retelling . The discussion in this section has demonstrated the relativity of the deictic center. 13 (8) is an example of direct speech from the corpus. in Macedonian retelling becomes a positive cue for the representation of narrative information from the perspective of someone who is in a position to be an experi- encer. The semantics of these forms has already been characterized as deictic: simple past deictically links information to the experi- encer. In direct representation. the use of simple past. Within this framework. Direct speech Since direct representation of characters' speech and thought involves a total deic- tic shift and a total submersion of the SELF of the actual teller. it is predicted that L- forms will not occur in direct representation.g. psychonarration. Other modes of rep- resentation may maintain some presence of an actual teller by partial deictic shifts (e. since the content is direct experience for the character.shift- ing origin of deixis from a position external to the story world to somewhere within the story world. These modes are less expressive than direct representation. It cannot be the current teller since she knew of the story events only by virtue of it being told to her. The results reinforce the analysis of the grammaticalized evidential system as deictic. 1. In narration. 4.2. the L-form past delinks information from the experiencer. Simple pasts are used. indirect speech). 13 Unless the character is talking about informationthat they themselves have not directly perceived. represented thought. 4. but they are more expressive than wholly reportive information which maintains the actual story teller and her audience as the center of deixis. In the context of renarration this experiencer can only be a character in the story world. Mushin / Journal of Pragmatics 32 (2000) 927-957 941 think (= internal monologue). .1.

well(sP) be(sP) nice 'But the bride says to her again. Table 4 presents the fre- quencies of L-form and simple past use in the clauses excluding the content of char- acter's speech. many of them did occur in conjunction with other expressive devices.942 l. as in (9). Lebot mi se godeshe. 20% of narra- tive clauses still occur in simple past. It was nice. "Why didn't it turn out for me? When I was back home."' (MBSa2) Direct speech . Yet even with this modification.2.well(so) when be(lsgsP) LOC lplACC bread(DEE) always time lSgNOM leb mesev. Beshe ubav. in Banfield's sense). even discounting content of speech.that 'Why did her bread turn out like that?' (MBSbl) The sentence in (9) is syntactically a direct question ('expressive syntax'. 4. I would always knead the bread. Table 4 Profile of tense variation in narrative retelling clauses (excludingcontent of speech) Tense form Total (n=174) L-form 80% (139) Simple past 20% (35) The exclusion of narrative information that represents the speech of story characters as they speak in the story world results in an overall ratio of L-form to simple past that is better aligned with my original prediction: L-forms constitute the vast major- ity of past tense forms in the corpus. lebot sekoj pat jas make. However. Represented thought Direct speech is one fairly obvious context where L-forms are unlikely to occur. Koga bev kaj nas. even in retelling narrative. which signals a high degree of expressivity. as Table 4 shows. The bread would turn out for me.accounts for almost half of the unexpected uses of simple past forms in my corpus.the most expressive of representations ." bread knead(lsgsP) bread(DEE) lsgOAT REEL make. However the deictic center analysis of the evidential variation can be extended to contexts which are not clearly identifiable as direct modes of representation. there are still a significant number of simple past forms which still must be accounted for. While not all instances of simple past use could be classified under 'direct repre- sentation'. The question represents what the . (9) Zoshto lebot da i se napraj taka why bread(DEE) to 3SgDAT REEL make(3sgsP)like. Mushin / Journal of Pragmatics 32 (2000) 927-957 (8) I ama nevestata mu veli pak "Zoshto ne mi se and but bride(DEE) 3SgDAT say(3sg) again why NEG lSgDAT REEL pogodi.2.

but this time the bread had turned out wrong' (MBSb2) Like direct representation.2. when the bread turned out badly.3. (10) a.) unrelated to the forms of the past tense them- selves.cook(PaRT) Svekorot mu rekol "Aren e Aren e Kje go F-in-law(DEF) 3SgDAT say(L) good be(3sg) good be(3sg) will 3sgAcc . (11) Si doshla svekrvata od pazar I prashala REFL come(L) M-in-law(DEF) tEN market and ask(L) "Kako e lebot" how be(3sg) bread(DE~:) Taa mu rekla nevestata "Aren e" reche 3sgNOM 3SgDAT say(L) bride(DEF) good be(3sg) say(3sgsP) "Ama vnatre kako zhila postana Ne dopechen" but inside how vein become(sP) NEG finish. Sega ama ovoj pat mu se zgreshi lebot now but DEM time 3SgDAT REFL be. (10) is a similar example. Mnogu se godel lebot kaj majka mu very REFLturn. the use of simple past in (9) and (10) can be explained in terms of the expressiveness of the passage. Unlike direct representation the actual teller is still a deictic center for personal deixis. Framing clauses All examples so far have associated the use of simple past with a more expressive mode of representation. Direct rep- resentation would require first person reference under these circumstances. and not the now of the actual storytelling situation. including epistemological deixis is shifted to somewhere or someone within the story world and the evidential coding reflects the knowledge of the story charac- ters and not the actual narrator. Here the shift from a more reportive to a more expres- sive stance is signalled at (b) by the temporal deictic adverb sega 'now' and by the NP ovoj pat 'this time'. use of expressive syntax. 1.wrong(sP) bread(DEF) 'The bread often turned out well at her mother's house. Although personal deixis is centered with the actual narrator (signalled by the consistent use of third person forms). either direct speech or represented thought. the actual teller's presence is (tenuously) maintained by the use of third person reference. The NOW'is the NOW of the story. This accounted for the vast majority of simple past uses in the retelling corpus. There was however a small residue of simple past tenses that did not seem to cooccur with other expressive devices. The passage in (11) is an interesting example of this 'independent' use of simple past (underlined and boldface).out(L) bread(DEF) LOC mother 3SgDAT b. all other deixis. But the expres- sive frame was established on the basis of clear independent cues (such as other deic- tic shifts. 4. Mushin / Journal of Pragmatics 32 (2000) 927-957 943 character was thinking at the time in the story (she is looking at the bread which had turned out badly). etc. Now. Thus despite the interpretation of 'Why did her bread turn out like that?' as the thoughts of a particular character in the storyworld.

So although the use of reche in medial framing clauses in (11) cannot be analyzed as fully representing the character's perspective." She said. 4. The problematic forms are the alternations between simple past and L-form pasts of the speech act verb reche 'say '14 in the framing clauses of the second and third direct speech passages. The use of simple past reche in (1 1) has neither of these functions . Up until this point in the text. It's good. Not fully anaphorically refers to the pre- vious use of reche (in expected L-form) that introduces a new reported speaker. and because they are preceded by L-form versions of the same verb in the immediately preceding framing clause. "It's good. its anaphoric medial position within a direct speech passage does not positively assert that the current narrator has direct experience of the events.the epistemological deixis is still linked with the character. the simple past form reche occurs medially in the direct speech passage ." The father-in-law said to her. In contrast. and which 14 Reche is the citation form of the verb as well as the 3sg subject simple past form. It only partially shifts the deictic ori- entation from character to narrator .it does not introduce a new character nor a new speaking event (the speaking character does not stop talking at this point). "It's not bad. all past tense verbs in narrative clauses were L-form pasts. "It is good. bride and father-in-law. consistent with the narra- tor's experience as a reteller." eat(lpl) say(3sgsP) NEG be(3sg) bad nice bake(PART) 'The mother-in-law retumed from the market and asked. Nicely cooked. Simple past as an expressive device The next two examples. Mushin / Journal of Pragmatics 32 (2000) 927-957 izedime" Reche "Nee losho Ubavo pechen. illustrate the use of simple past in narrative clauses which do not contain any other evidence of expressive framing.2." He said. Note however that the L-form of reche (rekol/rekla) introduces the new speaking character and a new speaking event in the narrative. The use of the simple past form postana 'became' in the second direct speech passage is consistent with the bride-character's perspective as an experiencer of the event. the bride. In contrast. "How is the bread?" She said.4. it was she who baked the bread. The simple past forms are problematic because they do not cooccur with other expressive devices within the framing clause (they are the only word in their respective clauses). the use of simple past reche in the middle of direct speech passages is less disruptive of the flow of repre- sented speech than use of the L-form would be. Use of the expected L-form after a passage of direct speech representation marks a dramatic shift in deictic orientation from the perspective of a speaking character as they spoke in the story world to the perspective of a current reteller who is reporting the events of a story she heard from someone else. We will eat it. "But it became vein-like inside.944 1. (12) and (13). ."' (MBSc2) The passage in (1 1) is a sequence of direct speech passages spoken by three differ- ent characters: mother-in-law. Perfective verbs (like reche) typically signal a movement forward in nar- rative time and the designation of a completed narrative event.

Kje go viknime vikniN will 3sgACC invite(lpl) d. My native speaker consultant interpreted all three passages as expressing a higher degree of 'empathy' between the current teller and the story character than in L-form passages in the same texts. Though rare in occurence. Ali toj zelnikot ne se pogodi but DEM zelnik(DEF) NEG REFL make. Otherwise she did everything well. She tried it. although they do not directly represent the speech or thoughts of story characters. (12) a. Ili mu reche da mesi leb. I. I sega svekra mu mu objasnuva kako se mesi leb.) that b. or 3DAT say(SP) to knead bread e." will make(lpl) zelnik f.well(L) 'Her husband said to her that " W e will invite the Brother-in-law to lunch. "Kje go viknime deverot na ruchek will 3sgACC invite(lpl) B-in-law LOC lunch C. bila mlada nevesta. and now M-in-law 3DAT 3DAT explain(3sg) how REFL knead d. Imala edna zhena shto have(L) one woman what b.' (MZS1) (13) a. Neka dojdi let come(3sg) e. I ovaa go napraj zelnikot and DEM 3SgACC make(sP) zelnik(DEF) g. We'll make zelnik". Mushin /Journal of Pragmatics 32 (2000) 927-957 945 also seem to signal new narrative events. these are the most problematic of the simple past forms in the corpus. Let him come.' (MBSal) This problem is resolved if the contexts in which the simple pasts are used are con- sidered carefully. kje naprajime zelnik. be(L) young bride C. Mazhot mu rekol deka man(DEF) 3SgDAT say(t. and DEM D-in-law take(sP) to knead bread f. We will invite him. That is. Go probala. 3sgAcc try(L) 'There was a woman who was a young bride. the use of simple past . I ovaa snaata zedi da mesi leb. A inako si pogodila and different REFL make.well(sP) h. And this daughter-in-law took to kneading bread. And she made the zelnik. Or (rather) she told her to knead bread. But that zelnik did not turn out well. And now her mother-in-law is explaining to her how bread is kneaded.

The use of the proximate demonstrative ovaa 'this' in (e) reinforces the interpretation of a deictic origin that lies somewhere within the story world.g. in a hearsay narrative context reflects the tendency for storytellers to use more expressive modes of representation. Similarly. In addition. even when the experience of others is being recounted. The well-documented 'historical present' (e. Despite these shortcomings. . the use of simple past marks a shift in epistemological deixis from external to the narrative to somewhere internal to the story world. the original teller). was used to describe the same event. Schiffrin 1981. Here however the shift is one of epistemological deixis rather than temporal deixis. and provides the motiva- tion for the use of simple past. illustrated in (12) and (13) is thus analogous to the use of present tense forms to talk about past time events. Both tense and adverbial forms are expressive devices that shift tempo- ral deixis to the time of the story world. The overall low frequency of L-forms. the less likely an L-form will be used. Note also that (d) and (e) occur at the beginning of the story and cannot be analyzed as representing some kind of 'climax'. Dry 1983) marks a shift in temporal deictic origin from time of narration to time of the narra- tive. the teller shifts from consistent use of L-forms in the prior text to simple past at the climax of the story (when she reports at f-g that the zelnik was made but not made well). the use of (d) and (e) in (13) continue the more expressive mode of narration established in (c) with the temporal deictic shift. Polanyi. In retelling narrative. compared with both simple past and non-past forms. 1990). Fig.the mother-in-law's instructions and the daughter-in- law's compliance with these instructions. In (13) the teller begins to tell her version of the story with L-forms to mark her reportive epistemological stance towards the narrative. Mayes. more involving than if an L-form. At (c) she shifts to imperfec- tive present tense objasnuva 'is explaining' in conjunction with the temporal adverb sega 'now'. In (f) she resumes the L-form to signal the next narrative events in the retelling (and L-forms are consistently continued in the text after this passage). This has the effect of making the narrative more 'vivid'. In (d) and (e) the teller uses simple past forms. The use of simple past to code narrative events in retelling. 6 represents the relationship between the modes of representation discussed in this section and their degree of expressivity. The narrative information is thus presented as if the teller and her audience had direct experience of the events for the rhetorical purpose of telling a more vivid and interesting story. Mushin / Journal of Pragmatics 32 (2000) 927-957 does signal a partial deictic shift from the perspective of the current teller to the per- spective of the main story character (i. Use of expressive language at text climaxes have been well attested (Labov. The use of simple past forms in (d) and (e) is problematic because neither repre- sents the speech nor thought of story characters. which auto- matically distances the teller from the events in question. To this extent. 1982. In (12).e. 1972. the more expressive the mode of representation. The teller shifts back to the L-form when she starts summarizing what else happened at (h). thus diminishing the presence of the current teller. both clauses represent new events in the narrative . the use of simple past in retelling can stand alone as an expressive device.946 1.

It is a concrete illustration of the deictic nature of evidentiality and its role in the establishment and maintenance of different discourse perspectives. 6.1. 5. The analysis presented in this section explained unexpected tokens of simple past forms in narrative retelling by reference to the perspective structure of the text - whose viewpoint is being represented (teller or character) and to what extent is the teller's presence manifested in the linguistic signal (the overlap of object and origin of perspective. Analyses of linguistic form and vari- ation in discourse (and narrative in particular) that ignore the tendency for speakers to shift in and out of expressive and reportive modes as they talk. Mushin / Journal of Pragmatics 32 (2000) 927-957 947 Simple Past L-form >1 Represented I > Framing > Narrative Thought Clause Narrative Clause Clause JMos < Fig. Reportive frames One reportive evidential strategy available to retellers was to frame their story with an initial sentence or passage which explicitly identified the source of the story as a previous teller. The analysis also clearly demonstrates the necessity for a sound theory of per- spective structure to account for the use of simple pasts in a context which violates the basic evidential meaning of these forms (i. 1.1. It was not expected to apply to extranarra- tive clauses . Title o f section? 5. .e. The analysis hinges on speakers' cognitive ability to distance themselves from their own experience and to adopt the perspective of others while they talk. Extranarrative units could however have an evidential function if they helped to construct and reinforce the representation of the narrative as a retelling of someone else's story.those which represented information that was not part of the previous telling of the story. Continuum of expressivity. ignore an intrinsic component of discourse production and comprehension without which the use of certain forms makes no sense. Narrative and extranarrative evidential strategies The prediction that L-forms would be the default past tense form used in the retellings applied to narrative units only. as in (14) and (15).1. 5. hearsay). This section discusses the extranarrative evidential strategies used by the Macedonian retellers and shows how they operated in con- junction with the grammaticalized evidential system in narrative units to reinforce the interpretation of the narrative as a retelling.

Once the retelling frame was established the L-form past was used as the 'grammatically' required past tense form in retelling contexts.' (MNC1) In (16). which introduces a reported speech frame.. This provides the context for the choice of tense/evidential forms in the narrative itself. She avoided the requirement that L-forms be used by syntactically embedding the narrative information as reported speech. after introducing the story as a retelling. introducing this narrative as a retelling. In one text. to ~EFL buy(3sg) and first 3ACC ask(3sgsP) father 3DAT 'X told me now that she wanted to buy a new car. Rather. This strategy is illustrated in (16)..' (MBSdl) (15) Ovaa lenta se rasprava za edna mlada nevesta DEM tape REEL discuss(3sg) BEN one young bride I ovaa nevestata zela leb da mesi and DEM bride(DEE) take(L) bread to knead 'This tape talks about a young bride . The informational content of the reported speech passage in (16) is coded with simple past forms rather than the predicted L-form past forms. which signals that the content of speech is embedded under the main predicate of saying.' (MBSb 1) The clauses in (14) and (15) function to index the story to a previous telling by another person..948 1. but she also syntactically represents the previous teller as the subject of a speech act verb (kazhuva 'tell').a plausible analysis based on the semantic description of the evidential contrast. (16) X mi kazhuvashe sega deka sakashe nova kola X 1DAT tell(3sgSP) now that want(3sgsp) new car da se kupe I prvo go prasha tatko i. has framed the events of the narrative as her own experience. This is different from (14) and (15) above where the objects of the main verbs of speaking in the initial clauses were simple noun phrases rather than subor- dinate clauses. the reteller did not immediately shift into L-form past after the estab- lishment of a reportive frame. the simple pasts in (16) are licensed by the syntax of reported speech in Macedonian which prefers concordance of tense in main and sub- . And this bride took on the task of kneading bread. the speaker not only makes reference to the previous teller as the source of the story. Mushin /Journal of Pragmatics 32 (2000) 927-957 (14) zhenava shto zborva ne e prikaska ami e zhiva vistina woman(DEE) what tell(3sg) NEG be(3sg) story but be(3sg) live truth koga bila mlada nevesta svekrva mu rekla da mesat leb when be(L) young bride mother-in-law 3DAT say(L) to knead(3pl) bread 'What this woman is telling is not a story but the living truth. And first she asked her father . The con- tent of the reported speech is offset by the presence of a complementizer deka 'that'. When she was a young bride her mother-in-law told her to bake bread.. or at least as something she witnessed .. I suggest that the use of simple pasts here does not signal that the teller..

The predicate kazhuva 'say/tell' shifts epistemological deixis of the subsequent sim- ple past forms from current teller-as-experiencer to reported speaker-as-experiencer. Why should a Macedonian speaker who is clearly aware that she is reporting what someone else said fail to consistently code the narrative with any kind of reportive evidential marking? One explanation for the continued lack of L-form pasts in this particular version is that the teller considered the entire narrative to be embedded under the reportive frame established as the first line of the telling. It is possible that the reteller was trying to convey an aura of confidence and authority in her version of the story based on her own prior knowledge of events. the expected default verb morphology in narrative retelling contexts. I. This was the second generation retelling of the text in (16) and (17). . Further evidence that the reteller was concerned with retelling the story as a retelling. (17) I posle tatko i rekol deka kje odat na 'auction' . (18) X reche deka saka da kupi kola X say(3sgsP) that want(3sgsP) to buy(3sg) car I ja prasha majka mu Majka mu mu reche and 3sg ask(3sgsP) mother 3sg mother 3sg 3sg say(3sgsP) 'X said that she wanted to buy a car... Her mother said .. given in part in (18). Mushin / Journal of Pragmatics 32 (2000) 927-957 949 ordinate clauses.. And she asked her mother. This analysis is plausible because the teller in (18) knew the original speaker very well and also already knew some details of her experience in getting the car.. illustrated in (17). The simple past is thus reflective of the previous teller's personal experience of the narrative events. even when simple past forms were used after the reportive frame. was the subsequent use of L-form pasts in this text. a retelling reconstructed from a retelling. The previous teller as experiencer of the events in the story is expected to use simple past forms. She maintains the use of simple pasts through- out her retelling with no further cues of reportiveness.' (MNC2) Although the teller in (18) refers to the original teller by name as the source of the story. and later father 3SgDAT say(L) that will go(3pl) LOC auction 'And later her father said that they will go to an auction .. Her total lack of other reportive cues suggests however that the choice of simple past over L-form past was a deliberate strategy to detract from the reportiveness of the narrative infor- mation and to present the story as 'vouched for'. she never subsequently backs up her stance as a reteller with the use of L-forms. thus evoking the retelling context.' (MNC1) Only one text in the corpus failed to use L-form pasts at all in retelling once a reportive frame was established. Under this analysis the reteller maintains her stance as a reteller by embedding the narrative under a predicate of saying attributed to the original teller. The passage in (17) marks the point in this version of the story where the narrator shifted to consistent L-form use.

"' (MBScl) (20) I veli "Tatko mi beshe posle na krajot" veil and say(3sg) father lSgDATbe(3sgsP) last LOC end(DEF) say(3sg) "A listatsijata pochna od pet ilijadi. Evidential direct speech Another retelling strategy was to represent the story (or parts of the story) as a quote attributed to the previous teller of the story.1. The teller in (18) was no doubt aware that she was retelling a story told to her by some- one else (in this case. Under this analysis. not the original experiencer).2. "My father was last in the end. Mushin / Journal of Pragmatics 32 (2000) 927-957 This analysis is supported by the 'mirative' analysis of the simple past/L-form dis- tinction proposed in Friedman (1995) (from work by Delancey. The fact that she departs from the plot structure in the previous telling (in her version it is the mother rather than the father character to whom the main character goes for permis- sion) is further evidence that she is not reconstructing the story as simply a retelling of the version she just heard. (19) I veli "Jas ostanav doma." she [previousteller] says. even if it had not been personally experienced. where simple pasts were more likely to be used for information that was completely 'assimilated in the speakers consciousness'. and Slobin and Aksu-K%. she chose to downplay the reportiveness of the information for pragmatic reasons that might have been compromised if L-form pasts had been used.. Go mesiv lebot. Simple pasts are less likely in contexts where the speaker talked about information that had just entered into their consciousness (e. all deixis within the content of speech is interpreted relative to the time. I nagore" veli and auction(DEF) start(3sgsP) GEN five thousand and up say(3sg) "krenavme do kaj sedum" veil "i pol ilijadi" veli rise(lplsa) DAT LOC seven say(3sg) and half thousand say(3sg) 'And she [previousteller] says. 5. the forms were not automatically chosen to match the source of information (otherwise L-forms would automatically be chosen in retellings). This differs from passages of direct speech which represented the speech of story characters. 1997. Whatever the reason.' (MNC 1) Recall that in direct speech representation. 1986). the use of the simple past is appropriate if the reteller in (18) did in fact have prior knowledge of the original teller's experience. "I stayed at home.g." she tprevioustellerl says.950 1. "and the auction started from $5000 and up" she [previousteller] says." and say(3sg) 1NOM remain(lsgsP) home 3ACCknead(lsgsP)bread(DEF) 'And she]previous tellerlsays. I kneaded the bread." She [previousteller]says. The content of speech in direct speech thus purports to be a replica of the . the complete absence of reportive marking after the first line in (18) is evidence that despite the grammaticalization of evidentiality in Macedon- ian past tense choices. However.. surprise).. "and a half thousand.. place and identity of the reported speaker as they spoke. "We raised (it) to seven. The passages in (19) and (20) are examples of the evidential use of direct speech to reinforce the inter- pretation of the text as a retelling (framing clauses are underlined).

These are there- fore examples of evidential direct speech that explicitly ground the narrative as a product of someone else's utterance. The reported speaker is not a character in the story but rather the person who told the story. and the speech act itself is an event in the story world. 16 In contrast. This includes a shift from reflecting one's own epistemological relationship to the narrative as the current narrator to representing the reported narrator's episte- mological relationship to the narrative. "Bring (us) the zelnik. In (19) the first person reference (Jas T ) and verb agreement (ostanav 'I remained'. which indexes 'he' as the speaking character. "Hey Blagoja. the direct speech passages in (19) and (20) do not represent individual events in the story. . The first person dative pronoun mi thus is interpreted as referring to the reported speaker ('she') and not the current teller. (21) is an example of what I call narrative direct speech. which are from first gener- 15 UnlikeEnglish. a conversational exchange between two characters in the story. mesiv 'I kneaded') both refer back to the reported speaker and not to the actual speaker. mu veli 'he says to her'. The second 'passage of speech' also consists of a framing clause indexing 'she' as the speaker to the following con- tent of speech. Mushin / Journal of Pragmatics 32 (2000) 927-957 951 exact words the reported speaker uttered as seen in earlier examples of direct speech representing story characters ((8) and (11)). The main difference between examples like (19) and (20) on the one hand and (21) on the other hand is the nature of the reported speaker and their relationship to the narrative. Macedonian direct speech framing clauses optionallyoccur with the complementizer deka. This is followed by an imperative clause which represents the order given by that character to bring the zelnik. but rather the prior act of narration itself. The lack of expressive devices in both (19) and (20) is attributable to the fact that the represented speech act is an act of narration. (21) Mu veli "Daj zelnikot" I taa mu veli deka 15 3SgDAT say(3sg) give(IMP) zelnik(DEF) and 3NOM 3DAT say(3sg) that "E de bre Blagoja ne mi se pogodi zelnikot. The use of 'expressive' devices such as imperative syntax and exclamatives are also associated with a direct mode of speech representation. as in (21)." EXCL EXCL EXCL Blagoja NEG lSgDAT REEL make. The reported speaker is a character in the story."' (MZS l) The passage in (21) is another canonical example of narrative direct speech from the retelling corpus. 16 See Mushin (1994) for a more detailed characterizationof direct speech as a narrative event. I. Direct speech was earlier described as the most expressive mode of representation in narrative. The first line contains a framing clause. the words that they speak are deictically centered within the story. the zelnik was not made well.well(3sgsP) zelnik(DEF) 'He says to her." And she says to him that. Direct speech involves a complete deictic shift from current speaker to reported speaker. rather than a conversational exchange between story characters. Both (19) and (20) above are also clearly canonical instances of direct speech. In (19) and (20).

Motivations for reportive evidential strategies The previous section identified four different strategies that retellers used to retell the story they heard in such a way as to make clear that they were retelling someone else's story: reportive framing (independent clause).952 l. embedded under a speech act predicate. In the context of the recording session. her epistemological stance represents the narrative information as reportive. for example. They did not. As a set of coding choices.3. Mushin / Journal of Pragmatics 32 (2000) 927-957 ation retellings originally heard as personal experience narratives. In this context L- forms would be predicted only if the original teller had represented information as deictically delinked from her own experience in the original telling. The L-form only implies in the context of renarration that the narrative information was acquired from a previous teller. Absence of reportive framing A final strategy used by retellers to establish and maintain the retelling as a retelling was simply to introduce the narrative with L-form pasts with no reportive framing. In this case. speakers used other linguis- tic devices to maintain their retelling stance towards the text. these strategies represent the range of linguistic devices involved in expressing the reteller's epistemological stance towards the text. Although theo- retically possible. once framed. Use of L-form pasts in retelling was the least 'direct' means reteller had available to represent the text as a retelling. reportive speech framing (main + subordinate clause). this never occurred in my corpus.1. However. Reportive framing and L-forms were being used together to construct the text as a retelling whereas reported speech framing and evidential direct speech alternated with the L-form in complementary distribution.2. And her mother-in-law told her to knead bread. By retelling the story as an iconic rep- resentation of the previous act of narration deictically linked to the previous teller. as in (22). (22) Si bila edna nevesta I svekra mu mu rekla da mesat RE~ be(L) one bride and M-in-law 3DAT 3OAT say(L) to knead(3pl) leb bread There once was a bride. The . where it was clear that the retelling's source was someone else's story. evidential direct speech and use of evidential L-form pasts. (MBScl) 5. they were able to maintain the shift to the representation of the previous teller's stance towards the narrative. Evidential direct speech was the most direct and explicit way that retellers pro- jected a retelling stance throughout the text. Reportive framing explicitly situates the text with reference to the current teller and their source of information. 5. the content of speech expresses the personal experience of the original narrator. produce reportive framing clauses throughout their telling. use of L-form alone was enough of a linguistic cue to establish and maintain an interpretation of the text as a retelling. since use of the L-form does not code any explicit reference to the source of information.

The differences between the evidential strategies for representing a retelling stance towards the narrative are summarized in Table 5. I izkisna . Kazhvashe "Ova e" veli "Ova prava vistina. use of the L-form can only be interpreted as representing a reportive stance.they shifted strategies from L-form to evidential direct speech and back to L-form. where both teller and audience know that the story is a retelling. tell(3sgsP) DEM be(3sg) say(3sg) DEM real truth Nee nishto laga" NEC be(3sg) nothing lies b. 'Evidential deictic shift' indicates whether there is a displacement of the evi- dential deictic center from the current teller (the reteller) to the original teller (the experiencer)." bread(DEF) and rise(3sgsP) . 'Other deictic shifts' indicates whether the reportive strategy also shifts temporal. 1. Although this doesn't specifically make reference to reportiveness per se. spatial and/or personal deixis to the original teller. Table 5 Evidential strategies for reflecting a reportive epistemological stance Evidential strategy Explicit source Evidential deictic shift Other deictic shifts Evidential Direct Speech Yes Yes Yes Reportive Frame Yes Yes No Reportive Frame + L form Yes No No L-form No No No L-forms do not code a shift in epistemological deixis from current speaker to some other specific individual. Evidence for this claim comes from the complementary use of the L-form past with evidential direct speech. I veli "Jas ostanav doma Go mesiv and say(3sg) 1NOM remain(lsgsP) home 3ACC knead(lsgsP) lebot. they maintain the actual teller as center of deixis by coding them as the 'non-experiencer'. as in (23). 'Explicit source' indicates whether the reportive strategy makes an overt reference to the original source of the story... Kako mu se sluchilo Koga se mazhila Koga bila mlada how 3DAT REFL happen(L) when REFL marry(L) when be(L) young nevestata Svekra mu oshla na pazar I ja ostajla bride(DEF) M-in-law 3DAT gO(L) LOC market and 3ACC remain(L) da mesi leb to knead(3sg) bread C. Mushin / Journal of Pragmatics 32 (2000) 927-957 953 intepretation of L-form narrative units must rely on other evidential cues and on the audience's general knowledge of the discourse context. (23) a. Tellers who used evidential direct speech as a strategy also used L-forms in their texts . in the context of narrative retelling. Rather.

Evidential direct speech is signaled with both a framing clause veli '(she) says' and by formal deictic shifts from third person to first person and from L-form past to simple past. First. more vivid and interesting way of telling a story. When she was a young bride. Conclusion The empirical analysis of reportive marking in the Macedonian corpus supports several important theoretical points about evidentiality. There are no lies. The shift to L-form thus signals a move from an expressive 'direct speech' mode to a more distant 'narrative' mode. 6. regardless of the linguistic strat- egy they chose to express it. . results in competition between the more 'distant' L-form past and less distancing reportive strategies. the regularity with which Macedonian retellers used reportive coding." she says. Mushin / Journal of Pragmatics 32 (2000) 927-957 'She was saying. is evidence of the conventionalized relationship ~7 Recallthat one of the retellers failed to take advantage of reportive coding other than mentioning the original teller as source at the start of her version of the story. At (b) the text shifts to L-form past tense as part of a total deictic shift from the previous teller as center of deixis to current reteller as center of deixis. one who can represent the motivations of characters in the story as well as accurately describe the events as they were told to her. The L-form however main- tains the current teller's epistemological relationship to the narrative as a non-expe- riencer/reteller. I kneaded the bread and it rose .954 I. And had left her to bake bread. " ' (MBScl) The passage in (23) begins with a speech act framing clause (a) which introduces a passage of evidential direct speech. . ts Mayes(1990) presents a similar analysis of direct speech in English discourse. It is interesting to observe that despite the existence of a grammaticalized sys- tem in Macedonian for coding information as reportive. "This is. "I was left at home. . And she says. It brings the audience 'closer' to the action of the story by representing the information from the perspective of the person who actually experienced the events. speakers used alternative strategies as well to achieve a reportive effect. The reteller's desire to be viewed as an interesting and coherent storyteller. At (c) the evidential strategy shifts back to evidential direct speech as the events of the story begin. This was attributed to discourse pressures that disfavoredthe consistent coding of the text as a retelling.18 Note that the teller of (23) reserved the more 'distant' L-form style of narration for description of background infor- mation about the story rather than for the events of the story themselves and reserved the more 'expressive' direct speech for the more dynamic parts of the nar- rative. This observation reflects the fact that there are many discourse pressures involved in the production of an appropri- ate text. Her mother-in-law had gone to the market. the use of evidential direct speech is a more 'expressive'. "This is the living truth. 17 For example." What happened to her when she had gotten married.

retellers could fail to adopt a reportive epistemological stance if other pragmatic factors conspired to override the need to represent infor- mation as the product of someone else's experience. a pragmatic motivation that has been built into the grammar of Mace- donian. However. Evidential direct speech is a more expressive alternative to the more reportive L-form as a strategy fbr expressing a reportive epistemological stance in retelling. . Mushin / Journal of Pragmatics 32 (2000) 927-957 955 between retelling and reportiveness in Macedonian culture. The conventionalization of this relationship is further reinforced by the grammatical status of reportiveness in Macedonian inflec- tional morphology. thus making for a more cohesive.a tool to index information in particular contexts to certain individuals based on their episte- mological stance towards that information. The fact that retellers did make use of expressive representations of character per- spective. inde- pendent of its epistemological status. The con- ceptualizer may be the actual teller. The patterns of evidential coding found in the Macedonian retelling corpus thus support the characterization of evidentiality as a deictic phenomenon . Representation of characters' speech and thoughts is a more expressive creative strategy than the reportive L-form to reconstruct the narrative as an inter- esting and cohesive story. Another important pragmatic factor that interfered with the consistent use of grammaticalized evidential marking coding was retellers' desires to be seen as skilled and interesting storytellers. despite this conventionalized relationship and the existence of gram- maticalized reportive forms. However. Evidential use relies wholly on the con- ceptualizer's constrnal of information with respect to how they acquired it. retellers could not ignore the fact that they were repeat- ing someone else's story and simply report the narrative events themselves. I. even when they were clearly committed to representing the retelling as a retelling of someone else's experience. is taken as evidence of the complex balance between a Macedonian reteller's need not to take credit for something they did not experience. or it may be 'imaginatively' represented as a character in the story. in which case the narrative information was typically coded as a retelling. in which case the narrative information is represented as the experience of that character. Perspective shifts to a more 'expressive' mode of narration in represented speech and thought help the reteller to coherently represent the intentions of the characters. to claim that epistemological stance is simply a construal of information relative to how the conceptualizer acquired it fails to recognize the systematic effects of extralinguistic context on the way information will be con- strued. and their need as storytellers to create a textured reconstruction of the events of the story that make clear the motives of the protagonists and the effects of their actions to the structure of the plot. intelligible and enjoyable story. The result of this competition is the rich variation in evidential coding found in the retelling corpus. These factors included the rela- tionship of the current teller to the original teller and any previous knowledge they may have had of the story events. When instructed to retell a story they had just heard.

Bloomington. Finegan. Evidentiality in English conversation and academic writing. Ph. N J: Ablex Banfield. NJ: Ablex. Grammar of the Macedonian literary language. London: Routledge. Nichols. Deixis in narrative: A cognitive science approach. and Albanian. Subjectification. Zeitschrift fur Balkanologie Band 24: 34-41. Ronald. Biber. Nichols. 1977. Evidentiality: The linguistic coding of epistemology. Evi- dentiality: The linguistic coding of epistemology. D. Fillmore. 203-213. Charles. 1988. Evidentiality: The linguistic coding of episte- mology. Hillsdale. Langacker. Dry.. A. liana.D.. A psychological account of the development and use of evi- dentials in turkish. Jakobson. Norwood. PA: University of Pennsylvania Press. R. Mirativity: The grammatical marking of unexpected information.F. 1982. Lunt. NJ: Ablex. Hewitt. Contact and consciousness in the Balkan Sprachbund. 1986. Evidentials in Japanese. Cognitive Linguistics 1:5-38 Li. 354-396. 1993. 1957. Labov. Patricia. Evidentiality in the Balkans: Bulgarian. In: R. 1990. Tense and narrativity. Scott. Jakobson. Boston. Columbus: Slavica. Naicong and David A. Text 9: 93-124. Columbus: Slavica. Duchan. Nichols. DeBray. Guide to the South Slavonic languages. Santa Cruz lectures on deixis. MA: Routledge. thesis. Philadelphia. Victor A. Discourse continuity and perspective taking. Labov. The transformation of experience in narrative syntax.S. The Hague: Mouton. SUNY Buffalo. Chafe. In: W. H. The function of direct speech in retelling. Duchan. 1972. The structure of wanka quechua evidential categories.E. 287-308. In: J. 159-167. eds.D. Linguistic Typology 1: 33-52. eds. Bruder and L. Dorritt. Susan. NJ: Ablex. 1978. 261-272. Chafe and J. Norwood. Chafe and J. eds. Styles of stance in English: Lexical and grammatical mark- ing of evidentiality and affect. Skopje. DeLancey. Mayes. Studies in Language 14: 325-363. Anne.. Zubin. CLS 30: 296-308. Wallace. Horace G. 1980.A. The grammatical categories of the macedonian indicative. 1975.. Ohta. In: W. 1995. Victor A. 1952. NJ: Ablex Cohn. and Johanna Nichols. Rick R. eds. Judith F. eds.. Discourse Processes 11 : 1-34. Macedonian. Nichols.. eds. Biber.A. NJ: Erlbaum. Hewitt. Paper presented at Language and Consiousness. University of California at San Diego. Chafe and J. an International Symposium.. 1997. Victor A. thesis. and Dan I. Victor A. Friedman. Aoki. Gail A.. In: W. Chafe and J. A. Ehrlich. William. Bulgaria. 1986. 1983. Norwood.. 130-147. In: W. Issues in Applied Linguistics 2: 211-238. Bruder and Lynn E. verbal categories and the Russian verb. 168-187. Shifters. Adverbial stance types in English. G... Suzanne. 1995. 1994. 1986. 1989. Varna. In: W. Evidentiality: The linguistic coding of epistemology. Friedman. In: W. 1995. 1990. eds. Perspective taking in Mandarin discourse.A. NJ: Princeton University Press. Ph.. and E. 223-239. Li. 1992.. IN: Indiana University Linguistics Club. Princeton... 1986. Nichols.. Mushin / Journal of Pragmatics 32 (2000) 927-957 References Aksu-Ko~.... 956 I. Naicong. Slobin. Wallace. Mushin. The movement of narrative time. Deixis in narrative: A cognitive sci- ence approach. NJ: Ablex DeLancey. Norwood. Douglas and Edward Finegan. Friedman. Hillsdale. Fleischman. Floyd. Evidentiality: The linguistic coding of epistemology. Evidentiality and volitionality in Tibetan.. 1991. Chafe. 1990. Quotation in spoken English. Morphological innovation and semantic shift in Macedonian. 1986. Transparent minds: Narrative modes for representing consciousness in fiction. 1990. London: Routledge. Chafe and J. Evidentiality: The linguistic coding of epistemology. Journal of Literary Semantics 12: 19-53. Norwood. NJ: Erlbaum. 1986. Friedman. Language in the inner city: Studies in the Black English vernacular. Evidentiality and politeness in Japanese. Scott. Norwood. Roman. Unspeakable sentences: Narration and representation in the language of fiction. eds. Helen D. Selected Writings: vol 2: Cambridge University 1971. Point of view. .G. 1988.

1986.. Alice. David A. Anthony. Analysing dis- course: Text and talk.A. In: J. Hillsdale. 1. 1989. She is currently an associate of the Uni- versity of Melbourne Department of Linguistics and Applied Linguistics. In: W. Information perspective. in Linguistics from the State University of New York at Buffalo. G. In: W. Mushin / Journal of Pragmatics 32 (2000) 927-957 957 Palmer. Deixis in Narrative: A Cognitive Science Approach. NJ: Ablex. Schlichter. Nichols. 263-286. Language 57: 45--62. Deborah.. and Lynn E.D. Norwood. NJ: Ablex.E. NJ: Lawrence Erlbaum. Tannen. Wiebe.E. 1982. Willett. Evidentiality: The linguistic coding of epistemology. Nichols. Wiebe.. Bruder and L. Deixis in narrative: A cognitive science approach. In: W.D. Janice M. profile. Frank. Norwood. Chafe and J. Literary complexity in everyday storytelling. 1995. .. 1986. David J. Woodbury. Hewitt. Recognizing subjective sentences in narrative: A computational investigation of narrative text. Evidentiality: The linguistic coding of epistemology. In: J. The deictic center: A theory of deixis in narrative. Dr. SUNY Buffalo. Tom.. Chafe and J. References in narrative text. Bruder and L. ed. Mood and modality. eds. 1981. Polanyi. Evidentiality: The linguistic coding of epistemology.. NJ: Ablex. C. 1986. Cambridge: Cambridge University Press. Duchan. A cross-linguistic survey of the grammaticalization of evidentiality. Duchan.F. and patterns in Quechua. 129-155. Janice M. Chafe and J. Ph. Schiffrin. 1986. eds.A. thesis.. 155-170.F. Hewitt. 46-59. 188-202. liana Mushin has a Master of Arts in Linguistics from the University of Melbourne and a Ph. eds. eds. Zubin. Weber. Nichols. Interactions of tense and evidentiality: A study of Sherpa and English. Tense variation in narrative. In: D. Livia. 137-158. Studies in Lan- guage 12: 51-97. Washington: Georgetown University Press. Hillsdale. The origin and deictic nature of Wintu evidentials. Hewitt. G. eds. 1988. Norwood... 1995. N J: Erlbaum...