LINGUIST List 13.1807

Fri Jun 28 2002

Review: Discourse Analysis:Mushin, Ilana (2001)

Editor for this issue: Dina Kapetangianni <>

What follows is another discussion note contributed to our Book Discussion Forum. We expect these discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in. If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for discussion." (This means that the publisher has sent us a review copy.) Then contact Simin Karimi at or Terry Langendoen at


  1. Grace Fielder, Mushin, Ilana. (2001) Discourse Analysis: Evidentiality and Epistemological Stance: Narrative Retelling.

Message 1: Mushin, Ilana. (2001) Discourse Analysis: Evidentiality and Epistemological Stance: Narrative Retelling.

Date: Thu, 27 Jun 2002 18:10:30 +0000
From: Grace Fielder <>
Subject: Mushin, Ilana. (2001) Discourse Analysis: Evidentiality and Epistemological Stance: Narrative Retelling.

Mushin, Ilana. (2001) Discourse Analysis: Evidentiality and Epistemological
Stance: Narrative Retelling. 
John Benjamins Publishing Company, xiv+240pp, hardback ISBN 90 272
5106 (Europe) EUR 81.68 / 1 58811 033 8 (US), $82.00. 
Pragmatics and beyond, NS 87

Book Announcement on Linguist: 

Grace E. Fielder, University of Arizona


This book examines the discourse and pragmatic conditioning of
expressions of epistemological stance assumed in the retelling of
narratives. The first half of the book is a critical discussion of
evidential semantics as a linguistic category versus epistemological
stance as a discourse pragmatic category. The second half of the book
is a cross-linguistic study of epistemological assessment in narrative
retelling using elicited data from Macedonian, Japanese and English,
languages that show three different types of evidential systems. The
various factors influencing speaker choices of reportive strategies
are then examined within the framework of Deictic Centre Theory.

Chapter 1 "Introduction" provides an introduction to the cognitive
notion of subjectivity in language, i.e. the subjective relationship
of the speaker towards the information and towards the speech
situation, which underlies both evidentiality and epistemological
stance. Thus, an objective utterance is one in which there is a
maximal distinction between the conceptualiser (subject) and the
experience (object), while in an subjective utterance the
conceptualiser is more involved in the construal of the experience.
The displacement of subjectivity, i.e. the construal of states of
affairs from a vantage point other than the here and now of the
speaker/conceptualiser, is examined using Deictic Centre Theory (DCT),
developed to account for the distribution of deictic forms in
fictional narrative but extended here to conversational narratives.
DCT models a deictically oriented 'window' onto the narrative world,
an origin of perspective (the subject/speaker) and an object of
perspective (the content of the deictic window). The degree to which
the information is perspectivized is representative of the degree to
which the origin and object of perspective are located in the same
domain. An important point made here is that subjectivity is a scalar
notion that is operative over several different domains: the
cognitive, the pragmatic and the linguistic.

Chapter 2 "Evidentiality" presents a critical overview of evidential
semantics and evidential systems. The purpose is to provide a
description of the more linguistically concrete manifestations of
evidentiality before proceeding to the more abstract, discourse
manifestations of epistemological stance. The distinction is made
between broad and narrow evidential semantics (Chafe 1986). The
narrow, or canonical, definition has to do with types of information
source, e.g., visually attested, reported, inferring, and is typically
associated with languages that have morphologically encoded
evidentials, such as Quechua. The broader interpretation espoused by
Mushin is that evidentiality reflects the speaker's attitude towards
knowledge, which can include information source, as being more
appropriate for languages with less elaborate means of encoding
evidential meaning, in particular, Macedonian and Japanese. The
author sets up a scalar relationship between the degree of
contextualization required for the interpretation of source and the
degree to which the use of the form implies a particular degree of
speaker commitment. Quechua is an example of a highly grammaticalized
category of evidentiality. The morpheme -shi explicitly codes that the
speaker heard the information from someone else and thus Quechua
represents one end of the scale where there is a context free
interpretation of source and a weak implication of the degree of the
speaker commitment. English occupies the opposite end of the scale
since it lacks any grammaticalized markers of evidentiality and
instead resorts to other means, such as the use of epistemic 'must' in
a highly contextualized interpretation of source with a strong
implication of degree of speaker commitment.

Mushin places the Balkan Slavic l-forms at a mid-way point. The
situation for standard Macedonian (Friedman 1986) is that the
semantics of the evidential distinction derives from the simple past
tense forms as marked for the grammatical category of status (in other
words, use of the simple past indicates that the speaker vouches for
the information) and arises pragmatically from the conventional
inference that if a speaker vouches for information, then s/he is
likely to have witnessed it. This half of the chapter makes a strong
contribution in that it points out the difficulties of providing a
coherent description of evidential meaning within a strictly semantic
model, thus motivating the author's consideration of the pragmatic
component. The second half is less persuasive in that it attempts to
classify the status of evidentiality in language according to concrete
criteria taken from grammaticalization theory (Lehmann 1985). The
degree of grammaticalization of any evidential marker has to do with
whether or not the forms in question are closed class items, occur in
fixed morphological slots, have 'bleached' semantics, hyperextended
use or reduced form. Thus, in Quechua, evidential forms are
considered to be highly grammaticalized since they are bound enclitic
morphemes: -mi for 'direct experience', -shi for 'reported/hearsay'
and -ch(r)a/-chi for 'inference' (Weber 1986 and Floyd 1993). These
criteria are later applied to Macedonian, Japanese and English in
Chapter 5. While this discussion of grammaticalization is
informative, it is less central to the ultimate aims of this study.

Chapter 3 "Epistemological Stance" constitutes a significant
contribution to the dialogue on evidentiality. The discussion deals
with how speakers talk about their epistemological status, i.e. it is
concerned with the relationship of function to form, rather than that
of form to function. The function to form approach is motivated by
the mismatches between the actual source of information and its
evidential coding in languages not only in English which lacks
grammaticalized evidential forms, but even in Quechua where they are
highly grammaticalized. Mushin argues that a given epistemological
stance with its concomitant linguistic choices is adopted by a speaker
based not only on the actual source of information, but also the
speaker's assessment of the actual source in conjunction with the
interactional setting, and the speaker's assessment of the
interaction. Accordingly, there is a wide range of different
epistemological stances can be adopted with respect to the same piece
of information: personal experience (both private and perceptual),
reportive, inferential, factual and imaginative. Personal experience
represents a highly subjective stance where the conceptualiser is
evoked, whereas a factual stance is highly objective with the speaker
assuming a more distanced stance. What is important here is that a
speaker can chose to position themselves anywhere along this array of
stances, although the actual adoption of a given stance in a given
context is mitigated by cultural and linguistic factors. The
remaining chapters specifically address the contextual, cultural and
conceptual factors that limit a speaker's linguistic choices.

Chapter 4 "Epistemological stance adoption in narrative retelling"
discusses the methodology by which the corpus of retellings was
elicited and recorded in Macedonian, Japanese and English. The
reteller was exposed to the original input of personal experience
either by hearing it directly from the original teller or by listening
to a recording of the original teller (who was recorded telling the
story to another native speaker of the language in question). The
reteller was then asked to retell the story to at least one native
speaker who belonged to the same speech community as the original
teller. Thus, the group of speakers was controlled for spoken dialect
and shared cultural knowledge. The retellings were then compared with
the original and with the other retellings. In analyzing the corpus,
the author makes a primary distinction between "narrative" information
(information that could be linked to information in the previous
telling of the story) and "extranarrative" information (information
that expressed a reteller's own experience). Narrative units that
directly represented the speech of characters as they spoke in the
story were coded separately as 'content' units. According to DCT,
each unit is associated with a particular origin and object of
perspective, which is in turn associated with an expected
epistemological stance. Extranarrative units represent an external
origin of perspective and an external object of perspective, hence a
personal experience epistemological stance would be expected, but not
required. Narrative units represent an external origin of perspective
but an internal object of perspective and thus a reportive stance is
expected. Content units represent an internal origin of perspective
and an internal object of perspective and thereby one would expect the
epistemological stance of the speaking character.

Chapter 5 "Reportive epistemological stance realisation in Macedonian,
Japanese and English" examines the range of linguistic devices used by
speakers in the narrative retelling tasks. Macedonian is classified as
having a grammaticalized system of evidential coding in that the
grammatical properties of the forms involved meet the criteria
established in Chapter 2. Although this would seem to put the
Macedonian evidential at the same level as languages such as Quechua,
this is clearly not the case. The issue of the grammaticalization of
the evidential in Balkan Slavic has been and still is the subject of
debate. (For the most recent round of discussions with extensive
background evidentiality in Bulgarian, see Alexander 2002, Fielder
2002 and Friedman 2002.) Mushin takes the position that in Macedonian
the past tense forms express deixis temporally and epistemologically.
The simple past functions deictically a) to temporally index the time
of the event as prior to the speaking time and b) to epistemologically
index the event to the experiencer of that event. In contrast, the
L-form functions deictically (a) to temporally index the time of the
event as prior to the speaking time, and (b) to epistemologically
delink the event to the experiencer. This position keeps Mushin's
analysis consistent with Friedman 1986, as well as with Fielder's
analysis of narrative strategies in Balkan Slavic as primarily deictic
in their semantics but pragmatic in their utilization (Fielder 1995,
1997 and 1998).

Once the speaker chooses to use the past tense, a choice must be made
to deictically link to the experiencer or assert no link. Thus, while
agreeing with Friedman's position that L-forms do not explicitly code
source of information in Macedonian, rather speaker attitude is coded
by the simple past forms, Mushin suggests that speakers do choose
simple pasts or L-forms according to epistemological stance adopted,
but that this decision is based on extralinguistic factors.
Epistemological stance then does not necessarily reflect the actual
source of information. Rather, the L- form past functions as the
'default' past tense form in retellings and are conventionally
associated with information acquired from another source. Mushin's
analysis of the Macedonian corpus reveals two other retelling
strategies: reportive framing whereby the speaker uses extranarrative
means to indicate that the retelling is a retelling and evidential
direct speech which essentially shifts the deictic perspective so that
the reteller is backgrounded and the original teller is foregrounded
(see also Fielder 1997, 1998 and 1999 for discussions of grounding as
a narrative strategy in Balkan Slavic).
In the Japanese corpus, the strategies of extranarrative reportive
framing and evidential reported speech were also found as well the use
of the sentence final particle -tte, which has many properties of a
grammatical marker of reportive evidentiality (Okamoto 1995; Suzuki
1997), and the adjectival predicate rashii, which is an lexical
inferential marker (Aoki 1986) that functions as an evidential marker
in the context of a retelling. According to the grammaticalization
criteria from Chapter 2, Japanese does not have as highly
grammaticalized evidential system as Macedonian since it is not clear
the -tte particle meets the criteria of hyperextension and reduced
form and rashii is a lexical marker.

The third language, English, has no clear grammatical markers of
evidentiality. In general, retellers adopted a reportive
epistemological stance far less frequently than in Macedonian and
Japanese. There was, however, a range of strategies, narrative and
extranarrative, that could be identified. As in Macedonian and
Japanese, some English retellers used reportive frames by introducing
their version of the story as a retelling. Mushin found no use of the
strategy of evidential direct speech, but there were some examples of
evidential indirect speech, which she regards as relatively objective
mode of representation, since it indexes the content of speech to the
previous telling of the story. The main strategy available to English
for assuming a reportive epistemological stance is to use reportive
adverbials, such as apparently and evidently, which indicate the
propositional attitude of the speaker. The conclusion of this chapter
is that the fact that each language has a variety of different
reporting strategies provides support for viewing epistemological
stance as a cognitive/pragmatic phenomenon, independent of any
linguistic realization.

Chapter 6 "Reportive strategies in narrative retelling" provides a
quantitative analysis of the distribution of reportive strategies by
using a "reportive density index" (RDI). The RDI represents the rate
of reportive coding per narrative clause. A higher RDI (up to 1.0)
corresponds to high degree of narrative information directly under the
scope of some reportive marker and therefore reflects a reportive
epistemological stance. By the same token, a lower RDI reflects a
more imaginative epistemological stance. The calculation of RDI was
applied exclusively to narrative clauses, since they represent
information acquired from the previous storyteller, while
extranarrative clauses and content clauses of direct speech and
thought were excluded. What is significant here are the large grain
generalizations that can be made for the three languages. The RDI is
relatively high for both Macedonian and Japanese, while far lower for
English. Thus, Macedonian and Japanese speakers favored the reportive
epistemological stance, while English speakers did not. Moreover,
both Macedonian and Japanese favored narrative reportive strategies
over extranarrative. Mushin attributes the high RDI of Macedonian to
the grammatical status of the L-forms and the fact that speakers must
choose an epistemological stance when using past tense verb forms.
What is interesting here is that Japanese patterns with Macedonian
with respect to the RDI, but unlike the Macedonian L-forms, Japanese
-tte is not part of an obligatory grammatical category. In this
respect, Japanese more closely resembles English. Why, then, does it
pattern with Macedonian? The explanation, according to Mushin, lies
in the pragmatic factors underlying the retelling of stories of
personal experience. Extralinguistic factors motivate the speaker to
select certain storytelling strategies, namely

1) the cultural imperatives that guide speakers to make particular
linguistic choices based on what is socially and culturally

2) the coherence imperatives that motivate retellers to reconstruct a
comprehensible and 'listenable' story, and to represent the narrative
information in a way that reflects a felicitous epistemological
interpretation of the story

Since discourse coherence pressures are assumed to be universal
across the three languages, it is the "cultural imperative" that
provides the explanation. Japanese patterns with Macedonian because
it is culturally important for Japanese speakers to indicate in whose
"Territory of Information (Kamio 1979, 1994, 1995, 1998) the
information lies. For English speakers it is apparently more
important to tell a good story, which explains the more frequent
adoption of an imaginative epistemological stance. The reason why
Macedonian has a high RDI is simply attributed to the fact that
"speakers of the language have conventionalised the relationship
between type of information and epistemological stance" because there
must have been some pragmatic imperative to do so. What this
pragmatic imperative might be, however, is not pursued (but see
Fielder 1998 and 2000 for the historical development of evidentiality
in Balkan Slavic).

Chapter 7 "Deviations from a reportive epistemological stance" is,
after Chapter 6, the most interesting chapter in that it is a hands on
discourse analysis using DCT to explain why there are deviations from
the default epistemological stance at the local level. In Macedonian
and Japanese, the use of strategies that evoke a direct experience
epistemological stance serve an expressive function. A shift from the
perspective of the reteller to the perspective of the character
results in a more dramatic and expressive story. In the English
corpus, where the overwhelming tendency was to adopt an imaginative
(i.e. direct experience) epistemological stance, shifts to a
reportive epistemological stance typically served pragmatic functions
such as indicating the climax of a story or a memory laps). An
analysis of referring expressions in the English corpus, however,
proves useful for indicating whose epistemological stance was being

The strengths of this study are the author's meticulous attention to
both the quantitative and qualitative aspects of analysis and the
important distinction she makes between evidentiality as a
semantico-grammatical category and epistemological stance as a
pragmatic and cognitive entity. The cross-linguistic results would
have more validity, however, if a similar group of Macedonian speakers
had been selected as for the Japanese and English speakers, i.e.
undergraduate and graduate students at a university. The fact that
the Macedonian speakers are significantly older speakers from a
southwestern dialect group who emigrated over 20 years ago calls into
the question the relevance of the author's claims with respect to
standard Macedonian. At the same time, it is notable that Mushin's
results are consistent with those of Fielder 1999 and forthcoming
which examines narrative strategies in the eastern dialects of
Macedonia. To conclude, I would recommend this book highly to anyone
interested in problems of evidentiality and the discourse-pragmatics
of narrative.


Alexander, R. 2002. "Bridging the descriptive chasm: the Bulgarian
"Generalized Past". Of All the Slavs My Favorites: In honor of Howard
I. Aronson, eds. V. Friedman and D. Dyer, Indiana Slavic Studies

Fielder, G. 1997. "The Discourse Properties of Verbal Categories in
Bulgarian and the Implication for Balkan Verbal Categories."
Gedenkschrift for Professor Zbigniew Golab, eds. V. Friedman,
M. Belyavski-Frank M. Pisaro and D. Testen, Balkanistika 10,
pp. 162-84.

Fielder, G. 1998. "Discourse Function of Past Tenses in Pre-Modern
Balkan Slavic Prose." Proceedings of the Twelfth International
Congress of Slavists, eds. M. Flier and A. Timberlake, pp. 344-361.

Fielder, G. 1999. "Auxiliary Use and the l-participle in Eastern
Macedonian dialects." Proceedings of the Third North
American-Macedonian Conference on Macedonian Studies, Toronto,
Ontario, June 12-14, 1997 (In Honor of Professor Horace G. Lunt, on
the Occasion of His 80th Birthday), eds. C. Kramer and B. Cook,
Indiana Slavic Studies 10:57-69.

Fielder, G. 2000. "Development of Narrative Strategies in 18th and
19th century Balkan Slavic Prose". Festschrift for Dean S. Worth,
eds. L. Ferder and J. Dingley, pp. 87- 106. Bloomington: Slavica.

Fielder, G. 2002. "Questioning the dominant paradigm: an alternative
view of the grammaticalization of the Bulgarian evidential." Of All
the Slavs My Favorites: In honor of Howard I. Aronson,
eds. V. Friedman and D. Dyer, Indiana Slavic Studies 12:171-201.

Fielder, G. forthcoming. "The Perfect in Eastern Macedonian
Dialects." [Forthcoming]. Proceedings of the Fourth
North-American-Macedonian Conference. 5-7, August 2000, Ohrid,

Friedman, V. 2002. "Hunting the elusive evidential: the third-person
auxiliary as a Boojum in Bulgarian." Of All the Slavs My Favorites: In
honor of Howard I. Aronson, eds. V. Friedman and D. Dyer, Indiana
Slavic Studies 12:203-230.

About the reviewer 
Grace E. Fielder is Professor of Balkan and Slavic Linguistics at the
University of Arizona. Primary research interests are discourse and
pragmatic analysis, language contact and issues of language and
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue