Content uploaded by Connie Dickinson
Author content
All content in this area was uploaded by Connie Dickinson on Sep 02, 2019
Content may be subject to copyright.
Studies in Language 24:2. 379–421 (2000). All rights reserved
<TARGET"dic"DOCINFO
AUTHOR"ConnieDickinson"
TITLE"MirativityinTsafiki"
SUBJECT"SL,Volume24:2"
KEYWORDS""
SIZEHEIGHT"220"
WIDTH"150"
VOFFSET"4">
MIRATIVITY IN TSAFIKI
Connie Dickinson
University of Oregon
1. Introduction
Tsafiki (Colorado), a Barbacoan language spoken in the western
lowlands of Ecuador, exhibits a complex pattern of both evidentiality and
mirativity. Mirativity codes “the status of the proposition with respect to the
speaker’s overall knowledge structure” (DeLancey 1997:3). The normal
<LINK"dic-r4">
expectation is that experienced events and states can be easily assimilated
into an existing network of assumptions and beliefs. In a mirative system,
events and states that cannot be easily assimilated are coded differently than
those that easily fall in with the speaker’s expectations. A mirative system
basically involves two different construals concerning experience. One
construal is based on the speaker’s past experiences of similar situations and
his general knowledge. This set of assumptions can range from knowledge
about purely physical interactions to assumptions based on cultural and
social norms. The second construal is based on the speaker’s immediate
experience of an event or state. If the immediate situation does not correlate
well with the speaker’s expectations, the proposition coding the event or
state receives special marking.
As a conceptual category, mirativity may be universal. All languages
probably have the means to code an event or state as occurring outside normal
expectations. Mirative morphemes have been documented in a wide range of
languages including: Hare (Athabaskan) of North America (DeLancey 1990,
<LINK"dic-r4">
1997), Turkish (Slobin and Aksu 1982; Aksu-Koç and Slobin 1986), Japa-
<LINK"dic-r12"><LINK"dic-r12">
nese (Akatsuka 1985), Korean (Lee 1985; 1993; Ko 1989), Minhe Monguor
<LINK"dic-r1"><LINK"dic-r8"><LINK"dic-r7">
(Slater 1996) and various Tibeto-Burman languages including Lhasa Tibetan
<LINK"dic-r11">
380 CONNIE DICKINSON
(DeLancey 1990c, 1992, 1997) and Kham (Watters 1998). There is now suf-
<LINK"dic-r4"><LINK"dic-r15">
ficient evidence to think that mirativity is a universal, semantic category
(DeLancey 1997). However, languages do differ in the particular mirative dis-
<LINK"dic-r4">
tinctions that are made and the degree to which those distinctions are integrat-
ed into the grammar of the language. Some languages, such as Hare, Turkish
and Kham, have a single, optional, mirative marker which may also function
as a marker of inferential evidence. In others, such as Tsafiki, mirative
markers are obligatory, occur in paradigmatic distribution and in certain
contexts are coded separately from evidentiality (see Section 1 and 3 below).
Perhaps the most well known study of mirativity is Aksu and Slobin’s
<LINK"dic-r12">
(1982; 1986) discussion of the Turkish marker mıs¸. The overlap between this
verbal suffix’s function as an inferential or hearsay evidential and a mirative
marker is illustrated by the examples below:
(1) Kemal gel-di
Kemal come-PAST
‘Kemal came.’
(2) Kemal gel-mıs¸
Kemal come-mıs¸
‘Kemal came.’
There are three contexts listed for example (2) (Slobin and Aksu 1983: 187):
<LINK"dic-r12">
a. inference: The speaker sees Kemal’s coat hanging in the front hall, but
has not yet seen Kemal.
b. hearsay: The speaker has been told that Kemal has arrived, but has not
yet seen Kemal.
c. surprise: The speaker hears someone approach, opens the door, and sees
Kemal — a totally unexpected visitor.
One of the more important aspects of Slobin and Aksu’s analysis is that the
mirative sense (c) is seen as a basic function of mıs¸and not simply an
idiosyncratic, peripheral use of an evidential marker (DeLancey 1997). In
<LINK"dic-r4">
psychological terms, its function is described as representing an experience
for which the speaker had no “premonitory awareness” (Slobin and Aksu
<LINK"dic-r12">
1982:196). When mıs¸occurs with a first person subject, it indicates lack of
conscious awareness on the part of the speaker, not simply lack of speaker
involvement. This lack of awareness is illustrated in the following example
(Aksu-Koç and Slobin 1986:160):
<LINK"dic-r12">
MIRATIVITY IN TSAFIKI 381
(3) Uyu-mus¸-um
sleep-mıs¸-1
‘I must have fallen asleep.’ (said upon awakening over one’s
books)
The lack of awareness or premonitory consciousness coded by mirative
markers is a feature which both links and differentiates mirativity from
evidentiality. An inferential evidential marker indicates physical distancing
from the event — the speaker did not see it. A mirative marker indicates
psychological distancing — the speaker did not anticipate the event or state.
But note that with the hearsay or inferential sense (a,b), the speaker may
very well have expected Kemal to arrive. The sighting of his coat in the hall
could simply be confirmation of an event that was already anticipated. A
mirative marker however, can be used despite the fact that the speaker has
witnessed the whole event. The following example, again from Turkish
(Aksu-Koç and Slobin 1986:162), could be said to a mother at the conclu-
<LINK"dic-r12">
sion of her daughter’s piano recital:
(4) kiz-iniz çok iyi piyano çal-iyor-mus¸
daughter-2. very good piano play--mıs¸
‘Your daughter plays the piano very well.’
The verbal suffixmıs¸does not indicate inference here. The speaker directly
witnessed the entire event. The speaker is simply indicating that he was not
psychologically prepared for the high quality of the performance.
Mirativity and evidentiality, along with epistemic modality are closely
related categories. All three categories essentially delineate the speaker’s
relationship, either physically or psychologically, to experienced events and
states. An evidential marker indicates the source of the speaker’s knowledge.
An epistemic modality marker indicates the speaker’s attitude towards the
proposition in terms of certainty or probability. And a mirative marker
indicates the relationship the information coded in the proposition has to the
speaker’s overall expectations and assumptions. There is a close relationship
between these three categories. An event that is witnessed is more certain
than one that occurs sight unseen, and an event that is witnessed from
beginning to end is less surprising than one that is only inferred or deduced
from its results. It is not hard to find systems where evidentiality and
epistemic modality are conflated (the English “must have” construction is a
ready example). And, as was illustrated in the Turkish examples above,
382 CONNIE DICKINSON
languages may also conflate evidentiality and mirativity. But it is equally
clear that it is possible to have an evidential system that is clearly indepen-
dent of epistemic modality and vice versa. The differentiation between
mirative and evidential systems is less evident (see DeLancey 1999 for
<LINK"dic-r4">
discussion). In Tsafiki, when the source of the information is also a primary
participant in the event, mirativity is coded separately from evidentiality,
affording an opportunity to untangle the two categories. This paper explores
the relationship between mirativity and evidentiality by first examining in
some detail the semantic and pragmatic factors that determine the distribution
of mirative markers and then examining the interaction between the two
categories.
The Tsafiki mirative system differs from the Turkish system in that
mirative markers are both obligatory and occur in paradigmatic distribution.
The only other systems of this complexity that I am aware of occur in some
of the Tibeto-Burman languages (DeLancey 1992) and Minhe Mongour
<LINK"dic-r4">
(Slater 1996). The distribution pattern of mirative markers in Tsafiki and
<LINK"dic-r11">
Lhasa Tibetan is remarkably similar and I begin the discussion by simply
illustrating the similarities between the patterns found in the two languages.
In Section Two I describe the interaction of the mirative markers with
various semantic verb classes in Tsafiki. Section Three discusses the interac-
tion and relationship between evidentiality and the mirativity. And Section
Four examines the relationship between Tsafiki future constructions and
mirative markers.
2. Complex mirative systems
The Tsafiki mirative system most closely resembles those found in
various Tibeto-Burman languages (DeLancey 1990c, 1992, 1997). The
<LINK"dic-r4">
similarities between the pattern found in Tsafiki and that found in Lhasa
Tibetan is quite startling. The similarities make it clear that this is a cross-
linguistic, typological pattern and not simply a language-specific, idiosyncrat-
ic phenomenon. Both languages exhibit a pattern that at first glance appears
to be a simple process of person marking. One verbal construction appears
with a first person subject ((5a), (6a)) and another construction with a second
or third ((5b–c), (6b)). The terms “congruent” () and “noncongruent”
() correlate with the terms “conjunct” and “disjunct,” which were
MIRATIVITY IN TSAFIKI 383
first used by Hale (1980) to describe a similar, logophoric-like system in
<LINK"dic-r5">
Newari. “Congruent” indicates that the information contained in the proposi-
tion is congruent with the speaker’s general knowledge. “Noncongruent”
indicates that the information is not congruent. I have replaced the original
glosses “conjunct” and “disjunct” with “congruent” and “noncongruent” in
the Lhasa examples.
(5) Lhasa Tibetan
a. nga bod=pa yin
ITibetan(person) be/CONGR
‘I am a Tibetan.’
b. kho bod=pa red
he Tibetan be/NCONGR
‘He is a Tibetan.’
c. khyed=rang-gis/khos byas-pa red
you-ERG/he:ERG did- NCONGR
‘You/he did it.’ (DeLancey 1992: 42–45)
<LINK"dic-r4">
(6) Tsafiki1
a. tse Tsachi joyoe2
tse Tsachi jo-yo-e
1FEM Tsachi be-CONGR-
‘I am a Tsachi.’
b. ya/nu Tsachi joe
ya/nu Tsachi jo-e
3/2 Tsachi be-
‘He/you are a Tsachi.’
But when one looks at the forms used in questions, the analysis of the
congruent/noncongruent forms as person-markers becomes confusing. In
Lhasa, the constructions appear reversed, with yin ‘’ appearing with
a second-person subject (7a) and red ‘’ appearing with a first (7b).
In Tsafiki, there is now a three-way distinction: -yo ‘’ now occurs
with a second person subject (8a), a new form -i ‘’ occurs with a
first person subject (8b), and a third person subject occurs with an unmarked
verb (8c).
384 CONNIE DICKINSON
(7) Lhasa Tibetan
a. khyed=rang bod-pa yin pas?
you Tibetan be/CONGR
‘Are you a Tibetan?’
b. nga rgya=mi red pas?
IChinese:person be/NCONGR
‘Am I a Chinese?’ (DeLancey 1992:43)
<LINK"dic-r4">
(8) Tsafiki
a. nu seke tera kiyun?
nu seke tera ki-yo-n
you good dance do-CONGR-
‘Did you dance well?’
b. la seke tera kiin?
la seke tera ki-i-n
1MASC good dance do-NCONGR-
‘Did I dance well?’
c. ya seke tera kin?
ya seke tera ki-n
he/she good dance do-
‘Did he/she dance well?’
Furthermore, in verbs of speech and cognition both languages use the
congruent form with the complement verb when the subject of both the
higher verb and the complement are the same ((9a), (10a)). When the
subjects are different, another form is used ((9b–c), (10b–c)).
(9) Lhasa Tibetan
a. khos kho bod=pa yin zer=gyis
he: he Tibetan be/CONGR say-
‘Heisays that heiis Tibetan.’
b. khos kho bod=pa red zer=gyis
he: he Tibetan be/NCONGR say-
‘Heisays that hejis Tibetan.’
c. khos nga bod=pa red zer=gyis
he: ITibetan be/NCONGR say-
‘He says that I am a Tibetan.’ (DeLancey 1992:43)
<LINK"dic-r4">
MIRATIVITY IN TSAFIKI 385
(10) Tsafiki
a. ya mantoka jiyoe tie
ya man-to=ka ji-yo-e ti-e
3 other-earth= go-CONGR- say-
‘Heisaid that heiwent to Santo Domingo.’
b. ya mantoka jie tie
ya man-to=ka ji-e ti-e
3 other-earth= go- say-
‘Heisaid that hejwent to Santo Domingo.’
c. ya la mantoka jie tie
ya la man-to=ka ji-e ti-e
31MASC other-earth= go- say-
‘He said that I went to Santo Domingo.’
Note that in the complement of the (c) examples, a first-person subject
pronoun occurs with the same verb-form as in the (b) examples, where the
subject of the complement is coded with a third-person pronoun. This is
clearly not a direct quote. Direct quotes in Tsafiki are distinguished from
indirect quotes by the use of pronouns, not by the presence or absence of the
congruent/noncongruent markers. The presence of the second person pronoun
in (11a) and the first person pronoun in (11b) indicate that these are direct
quotes. In contrast the third person possessive pronouns in (12) indicate that
these are indirect quotes. If (12a) were a direct quote the expected pronomi-
nal form would be chikechi ‘my,’ not yachi ‘her/his,’ and in (12b) a direct
quote would occur with the second person pronoun form, nuchi ‘your,’ rather
than yachi ‘his/her.’
(11) a. laka tsanti’ nu merayun?
la=ka tsan-ti-e nu mera-yo-n
1= so-say- 2hear-CONGR-
‘So he said to me, ‘Do you hear it?’’
b. la merayoetiyoe
la mera-yo-e ti-yo-e
1MASC hear-CONGR- say-CONGR-
‘I said, ‘I hear it.’’ (2.10)
386 CONNIE DICKINSON
(12) a. yachi naka fitumanyun
ya=chi na=ka fi-tu-man-yo-n
3=POSS child= eat---CONGR-
tinutie ti’
ti-nu-ti-e ti-e
say-.-- say-
‘They say he said she asked him if he ate her children.’
b. yachi fitumanyoe tinin tito,
ya=chi fi-tu-man-yo-e ti-nin ti-to
3=POSS eat---CONGR-DECL say- .say-
kirawoyoe tie ti’
kira-wo-yo-e ti-e ti-e
see-CAUS-CONGR-DECL say- say-
‘They say he said he showed (her) (his teeth) saying he
didn’t eat hers.’ (.107,110)
In Tsafiki, the congruent suffix can also occur in a simple clause with a
third person subject if the speaker can be construed as a knowing participant.
In (13a) the speaker is describing the habits of the Tsachila in the old days.
In (13b) she is describing the habits of present day Tsachila. Although the
subject in (13b) is technically third person, the speaker is a knowledgeable
member of the group and the congruent verbal suffix is used.
(13) a. jun materi tsachila pini fiminla
jun ma=te-ri tsachi=la pini fi-min=la
-day=- tsachi=PL snake eat-=
jomantie
jo-man-ti-e
be---
‘In those days they say the Tsachila used to eat snakes.’
b. amana tsachila fi-tu-min=la joyoe
amana tsachi=la fi-tu-min=la jo-yo-e
now tsachi=PL eat--= be-CONGR-
‘Nowadays (we) tsachila don’t eat snakes.’
(.20–21)
In addition, both languages exhibit minimal pairs where the use of a non-
congruent vs. congruent form results in a difference in meaning. The Lhasa
congruent construction (14a) is a simple declarative statement. The non-
MIRATIVITY IN TSAFIKI 387
congruent (14b) indicates that the speaker has just reached into his pocket
and discovered to his surprise that he has some money.
(14) a. ngar dngul tog=tsam yod
I: money some EXIST/CONGR
‘I have some money.’
b. ngar dngul tog=tsam ’dug
I: money some EXIST/NCONGR
‘I have some money!’ (DeLancey 1992:43–44).
<LINK"dic-r4">
The difference between Tsafiki examples (15a) and (15b) lies in intention.
The congruent form (15a) would be uttered in a situation where the speaker
intentionally cut someone. The noncongruent form (15b) would be produced
in a context where the speaker swung the machete and accidentally cut
someone.
(15) a. la yaka machitechi poreyoe
la ya=ka machite=chi pore-yo-e
1 3= machete= cut-CONGR-
‘I cut him (intentionally) with the machete.’
b. la yaka machitechi poreie
la ya=ka machite=chi pore-i-e
1 3= machete= cut-NCONGR-
‘I cut him (unintentionally) with the machete.’
This difference between intentional and unintentional behavior can also be
seen in Lhasa examples (16a) and (16b).
(16) a. nga-s kho dbril-pa yin
I- he knock.down- CONGR
‘I knocked him down (intentionally).’
b. nga-s kho dbril-song
I- he knock.down-PERF/NCONGR
‘I knocked him down (unintentionally).’ (DeLancey 1999: 5)
<LINK"dic-r4">
The noncongruent form (17a) was spoken in a context where I was looking
at the stars with a young Tsachi girl. I pointed at the stars and she immedi-
ately knocked my arm down, telling me never to do that. If I knew that it
was wrong and pointed at the stars anyway, the congruent form (17b) would
be appropriate.
388 CONNIE DICKINSON
(17) a. seitonke miitoto, tsaboka tedechi
seiton=ke mi-ito-to tsabo=ka tede=chi
bad= know-not.be- star= hand=
mikuwaie,
mi-kuwa-i-e
know-give-NCONGR-
‘Not knowing it was wrong, I pointed at the stars.’
b. tsaboka tedechi mikuwayoe
tsabo=ka tede=chi mi-kuwa-yo-e
star= hand= know-give-CONGR-
‘I pointed at the stars.’
Finally, (18a) was spoken in a context where a Tsachi woman was compli-
mented for her prowess in soccer. Someone said she played like a man and
she uttered (18a), which was said ironically with a shrug of the shoulders.
The congruent form (18b) is a simple statement of fact.
(18) a. unila joie
unila jo-i-e
man be-NCONGR-
‘I’m a man!!’
b. unila joyoe
unila jo-yo-e
man be-CONGR-
‘I am a man.’
In examples (14)–(18), the noncongruent form is used in situations that
contain elements that do not correlate with the speaker’s set of assumptions
and expectations. The speaker either discovers a new unanticipated situation
(14b), is surprised by the final result of an action (15b, 16b), performs an act
without an awareness of the implications of that act (17a), or finds that a
piece of information does not fit with a set of previous assumptions, i.e. a
female is demonstrating a characteristic usually associated with males (18a).
Despite differences between the two languages, both exhibit a similar
pattern:
i. The use of one construction for first person in declaratives and with
second person in questions;
ii. Logophoric-like constructions, where one form is used when the subject
of the higher verb is co-referent with the subject of the complement and
MIRATIVITY IN TSAFIKI 389
another when the subjects are different; and
iii. The occurrence of noncongruent constructions indicating surprise,
accident, ignorance or irony in complementary distribution with the
congruent constructions.
The congruent/noncongruent system clearly cannot be analyzed as a person-
marking system. The same verbal mirative suffixes can co-occur with first
person, second person or third person subjects. However the logophoric-like
constructions and the use of the congruent suffix in second-person questions
demonstrate that the congruent/noncongruent distinction is restricted to
contexts in which the source or expected source of the information is also a
primary participant in the event or state. As will be shown in Section Three
(below) a different system is used to mark mirative distinctions when the
speaker is not a primary participant.
In addition note that the mirative distinction in Tsafiki primarily codes
the degree to which the state or event is in agreement with the speaker’s
general knowledge system. The various interpretations of a mirative marker
— surprise, unintention, unconsciousness, ignorance, irony — arise from its
interaction with the overall context of the proposition and the semantics of
the verb. Note that in (18a) the noncongruent suffix is not indicating
surprise. The statement is ironical, which implies that the speaker knows, and
has reason to believe that the hearer also knows that the statement is not
true. So while surprise is the most common interpretation of a noncongruent
marker, a broader definition is needed. A noncongruent mirative marker
indicates that there is a contradiction between the speaker’s set of established
expectations and assumptions and one or more factors in the immediate
experienced situation. The contradiction is most commonly, but not necessar-
ily, surprising. The various semantic and pragmatic functions of mirative
marking can be accounted for by examining different classes of verbs and
taking into account the speaker’s experiential participation in the event or
state coded by the proposition.
3. Mirative markers and semantic verb classes
Mirative markers code the degree to which the information coded in the
proposition is congruent with the speaker’s general knowledge. The criteria
used to identify an unexpected event differ from that used to identify an
390 CONNIE DICKINSON
unexpected state. Therefore, in the following section, after first briefly
describing the Tsafiki verb class system, I discuss mirative markers in
relation to events, states, and with a special category of verbs — body
processes, emotions and desires — that code events that commonly occur
unexpectedly.
3.1 Tsafiki semantic verb classes
Tsafiki has five semantic verb classes. The class of the verb is overtly
coded by the use of a verb class marker in certain syntactic constructions.
The five verbs below all have a plural suffix that cross-references the
subject (19). Verbs with plural suffixes require a verb class marker ().
The verb class markers are derived from simple verbs.
(19) Tsafiki Verb Classes
a. ki/ke-class: from the simple verb kino ‘to do’
sonala leyahito dalakinae
sona=la leya-i-to da-la-ki-na-e
woman= gather-become- cook-PL-VCL.do--
‘Gathering together, the women cooked.’ (.49)
b. i-class: from the simple verb ino ‘to become’
tsankari wari pepa joto katsolaiti’
tsanka=ri wari pepa jo-to katso-la-i-ti-e
so= tired be- sleep-PL-VCL.become--
‘So they say being really tired (they) slept.’ (2.17)
c. ti-class: from the simple verb tino ‘to say/feel/express’
bonbo naka jori jori jori
bonbo na=ka jori jori jori
hedgehog small= jori jori jori
atilatinatie
a-ti-la-ti-na-ti-e…
shout-say-PL-VCL.say---
‘They say the little hedgehogs were shouting jori jori jori.’
(.50)
MIRATIVITY IN TSAFIKI 391
d. ra-class: from the simple verb rano ‘to be in a position’
(parala) fundan,
(para=la) fuda-n
(wild.pig=) hurt-
tsolaratie ti’
tso-la-ra-ti-e ti-e
lie-PL-VCL.be.in.a.position-- say-
‘They say he said (the wild pigs) were lying there, hurt.’
(.26)
e. jo-class: from the simple verb jono ‘to be.’
aman junbi jolajotie ti’ tsachi
aman jun=bi jo-la-jo-ti-e ti-e tsachi
now = be-PL-VCL.be-- say- tsachi
nakala
na=ka=la
small==
‘Now they say he said the little tsachi were there.’
(..19)
The ki ‘do,’ i‘become,’ and ti ‘say/feel’ classes are all active in that they
code activities, accomplishments and achievements, whereas the jo ‘be,’ and
ra ‘be in a position’ classes code states (Vendler 1957 [1967]). The ki ‘do’
<LINK"dic-r14">
and i‘become’ classes are by far the two largest classes in Tsafiki. While the
ki and iexamples above are transitive and intransitive respectively, valency
is not the distinguishing feature. Both classes have transitive and intransitive
verbs. The differences between the classes will be further discussed below.
3.2 Events
There are two related factors that determine the ease with which a
proposition coding an event occurs with a noncongruent suffix: 1) control;
and 2) degree of elaboration or complexity of the event. Propositions
portraying events in which the subject has complete control are less likely to
be portrayed with a noncongruent suffix. However, propositions that clearly
portray the initiator or initiating activity as distinct from the affected entity
or final result easily occur with a noncongruent suffix. Events that are
portrayed as less complex, either in terms of number of participants or
degree of fusion between the primary stage and final stage of the event, do
392 CONNIE DICKINSON
not as readily occur with a noncongruent suffix. In the following discussion,
I begin with constructions that readily occur with either mirative marker and
then discuss constructions that restrict their usage.
3.2.1 Ki-class transitive verbs and i-class intransitive verbs
Simple ki ‘do’ class transitive verbs and i‘become’ class intransitive
verbs whose subjects clearly undergo a change of state can occur with either
congruent or noncongruent suffixes. With both these classes of verbs, there
is an affected theme and a final result. However there is a difference in the
final interpretation of the mirative marker due to the semantics of the verb.
3.2.2 Ki-class transitive verbs
With simple, transitive verbs such as poreno ‘to cut’ in above, the
interpretation concerns intentionality. The noncongruent form indicates the
speaker volitionally performed the action of swinging the machete, but did
not intend the final result. This is illustrated with another verb keere ‘throw’
below. With the congruent form (20a), the speaker intentionally threw out the
paper. With the noncongruent form (20b), the speaker threw out some papers,
but did not intend to throw out this specific paper.
(20) a. tse pila kidoka keereyoe
tse pila kido=ka ke-ere-yo-e
1 paper skin= throw-send-CONGR-
‘I threw out the paper (intentionally).’
b. tse pila kidoka keereie
tse pila kido=ka ke-ere-i-e
1 paper skin= throw-send-NCONGR-
‘I threw out the paper (unintentionally).’
In both examples, the initiating action is intentional and volitional — the
speaker was conscious and aware of performing the action. But with the
noncongruent form, the final result does not correlate with the original
intention. The distinction between intention and non-intention in these
examples arises from the semantics of the verb and the use of the mirative
markers. Note that Tsafiki allows inanimate ‘agents’ with these verbs (21).
(21) wu pila kidoka keeree
wu pila kido=ka ke-ere-e
wind paper skin= throw-send-
‘The wind threw/blew the paper.’
MIRATIVITY IN TSAFIKI 393
These verbs do not require an intentional, volitional agent, but rather an
agent in the sense of an ultimate cause (DeLancey 1984, 1990b, 1990c), or
<LINK"dic-r4">
an effector (Van Valin and Wilkins 1996). Only with a first-person subject
<LINK"dic-r13">
or in the logophoric-like constructions is intention/non-intention concerning
the final result overtly coded through the use of mirative markers. The
restriction seems to arise from the fact that a speaker does not have direct
access to the intentions of another person. Only if the deictic center is first
placed with the primary participant by the use of the logophoric construction
can mirative suffixes be used (22).
(22) Juan pila kidoka keereie tie
Juan pila kido=ka ke-ere-i-e ti-e
Juan paper skin= throw-send-NCONGR- say-
‘Juan said he threw out the paper (unintentionally).’
Note that neither the verb alone, nor the mirative markers specifically code
intention/unintention, but rather it is the combination of the mirative markers
with this class of verbs that gives the intentional/unintentional reading. As
was illustrated in (21) these verbs can occur with inanimate, non-volitional
subjects and as will be shown in the next section, mirative markers can occur
in constructions where the subject is never construed as volitional.
3.2.2.1 Intransitive verbs. With intransitive, i-class verbs, in which the
subject clearly has no control over the event, intention and volition are
irrelevant. Here the difference between the congruent and noncongruent
forms is one of consciousness or awareness. In (23), keereino means ‘to fall.’
Both forms, congruent and noncongruent, code an accident. But with the
congruent form (23a), the speaker was in a situation where a fall was quite
likely, i.e. walking on slippery ground. The speaker was conscious of the
danger and conscious of the fall while it was occurring. The noncongruent
form (23b) indicates that the speaker had no premonitory consciousness that
he was going to fall. He just suddenly slips and falls.
(23) a. la keereiyoe
la ke-ere-i-yo-e
1 throw-send-become-CONGR-
‘I fell (expected).’
394 CONNIE DICKINSON
b. la keereiie
la ke-ere-i-i-e
1 throw-send-become-NCONGR-
‘I fell!! (unexpected).’
Other i-class verbs with theme or undergoer subjects function much like
keerei ‘to fall’ above. Katsono ‘to sleep’ is an i-class verb that can occur
with either the congruent or the noncongruent suffix. The congruent form
indicates that the speaker was well aware of falling asleep (24a). The
noncongruent form indicates that the subject fell asleep unawares (24b).
(24) a. tse katsoyoe
tse katso-yo-e
1 sleep-CONGR-
‘I slept.’
b. tse katsoie
tse katso-i-e
1 sleep-NCONGR-
‘I dozed off/fell asleep!’
3.2.2.2 Comparison of transitive ki-class and intransitive i-class verbs. With
both the transitive ki ‘do’-class and intransitive i‘become’-class verbs, the
noncongruent suffix indicates that the final state is unexpected. But with the
transitive verbs the speaker is aware, conscious and volitional up to the final
result. In the event line represented in Figure One (cf. DeLancey 1990c), the
<LINK"dic-r4">
interaction between the different mirative markers is shown with both the
transitive and intransitive forms of keere ‘throw’. With the transitive,
ki-class, congruent form (a), the speaker has a plan or intention. He moves
volitionally to carry out the action and the final result is in accordance with
his intention. With the transitive, noncongruent form (b), the speaker moves
volitionally and consciously performs the act, but the final result is not in
accordance with his original intention. With the intransitive forms (c) and
(d), intention and volition are not an issue, the speaker acts neither intention-
ally or volitionally. With (c) the speaker is aware and conscious of the
process which leads up to the result, and that result is in accordance with his
belief that those processes can lead to that result. In (d) the speaker was
totally unprepared for the event and did not anticipate the result.
MIRATIVITY IN TSAFIKI 395
intention volition action result
(a) keere eyo
(b) keere ei
(c) keerei eyo
(d) keerei ei
Figure 1. Event line interaction with congruent/noncongruent
3.2.3 Ki-class intransitive verbs
One class of verbs that does not readily occur with the noncongruent
marker is the set of ki-class, intransitive, activity verbs. When the intransitive
verb, jaya kino3‘work physically,’ occurs with the noncongruent marker,
without any context, it is odd (25a). The work is created by the action itself
and exists the moment the action begins. There is no final result. But, if a
context is given in which someone mistakenly works on the wrong piece of
land, the noncongruent marker is fine (25b). The focus of the mirative
marker shifts to the first clause and it is that action that is surprising, not the
action coded in the final clause.
(25) a. ?la jaya keie
la jaya ki-i-e
1 work do--
?‘I worked (but I was unaware of it).’
b. fetochi witaka poleto jaya keie
feto=chi wita=ka pole-to jaya ki-i-e
non.Tsachi= farm= pass- work do-NCONGR-
‘Crossing (unintentionally) over to the non-Tsachi’s farm, I
worked.’
Other verbs in this class include verbs of sound emission such as wokeno
‘bark,’ fifi kino ‘whistle,’ kanto kino ‘sing,’ and other activity verbs such as
tera kino ‘dance,’ sula kin ‘swim’ and palan kino ‘paddle a boat.’ The only
context in which a noncongruent marker easily occurs is where there is an
additional factor that receives the focus of the mirative marker. Note that in
the question example above, seke tera kiin? ‘Did I dance well?’ (8b), what
is being questioned is the quality of the dancing and not the dancing itself.
These are simple verbs in that they consist of the primary stages of the
event line, intention > volition > action. The stages are not easily distin-
guished and only if the event is elaborated with an adverb or locative do
396 CONNIE DICKINSON
these verbs easily occur with the noncongruent marker. Without the elabora-
tion, the only possible context for a noncongruent marker with these verbs is
if the subject is in a gross state of altered awareness.
There is one verb that codes an activity that is normally performed in a
state of unawareness. This is the complex predicate koron pano ‘snore.’ At
first speakers would only accept the noncongruent form (26a) with this verb.
When I suggested that perhaps you could say this in a context where you
were pretending to snore, most speakers accepted the congruent form (26b).
But one older, basically monolingual speaker pointed out that if you were
pretending, you weren’t really snoring and (26c) was actually a better form
to use. She felt that ‘real’ snoring could only occur in an unconscious state.
(26) a. katsonato koron paie
katso-na-to koron pa-i-e
sleep-- snore talk-NCONGR-
‘While I was sleeping, I snored.’
*katsonato koron payoe
katso-na-to koron pa-yo-e
sleep--to snore talk-CONGR-
b. ?nalalabe tensaanato koron
na=lala=be tensa-a-na-to koron
child== play--- snore
payoe
pa-yo-e
talk-CONGR-
?‘While I was playing with the children, I snored.’
c. nalalabe tensaanato koron
na=lala=be tensa-a-na-to koron
child== play--- snore
pamantsannin kiyoe
pa-man=tsan-nin ki-yo-e
talk-=like- do-CONGR-
‘While I was playing with the children, I pretended to snore.’
In essence, intransitive ki-class verbs can occur easily with only one of the
mirative markers. The specific marker used, congruent or noncongruent,
depends on the inherent semantics of the event. The action is depicted as
being performed in either a solely conscious or solely unconscious state.
MIRATIVITY IN TSAFIKI 397
Only if the event coded in the proposition is elaborated by the presence of
another element can the verb freely take either marker.
3.2.4 Transitive i-class verbs
Intransitive, i-class verbs with affected theme subjects easily occur with
noncongruent marking, but transitive i-class verbs do not. Both (27) and (28)
exhibit transitive verbs. They are, semantically, reflexive-like constructions
in that the instigator and the possessor of the primary affected entity are co-
referential. The distinction between the two forms concerns initiation and
control. With the ki-class form (27), the person who performs the action also
initiates the action, although as was illustrated above (15), he does not
necessarily control the final result. With the i-class form (28), the affected
participant, who also initiates the event, controls the person actually perform-
ing the action.
(27) tse a poreyoe
tse a pore-yo-e
1 hair cut-CONGR-
‘I cut (my) hair.’
(28) tse Mariabe a puraiyoe
tse Maria=be a pore-i-yo-e
1 Maria= hair cut-become-CONGR-
‘I had (my) hair cut by/with Maria.’
*tse Mariabe a puraiie
tse Maria=be a pore-i-i-e
1 Maria= hair cut-become-NCONGR-
I could not elicit the noncongruent marker with the i-class form of pore
‘cut’’ (28). Speakers found it highly unlikely that I would ‘have’ someone
accidentally cut my hair/or accidentally ‘have’ someone cut my hair.
‘Having’ someone do something indicates an intention or plan concerning
the final result. If the person does something other than what is intended, the
speaker did not ‘have’ him do that. If Maria unexpectedly cut his hair, the
speaker would use a verb with Maria as subject and the evidential/mirative
suffix used to code surprise when the speaker is not depicted as a knowing
participant in the event (29) (see Section 3.2). The noncongruent form is
possible with the ki-class transitive form (30). It indicates that the haircutting
was accidental. Note that (27), (28) and (30) are all transitive, semantically
398 CONNIE DICKINSON
reflexive constructions. But in (28) there is less elaboration of the event in
that the original intention and the final result must match. If they are not the
same, the event will be coded by a different construction.
(29) Maria lachi aka porenue
Maria la=chi a=ka pore-nu-e
Maria 1= hair= cut-EV/MIR-
‘Maria just cut my hair!.’
(30) tse aka poreie
tse a=ka pore-i-e
1 hair= cut-NCONGR-
‘I cut (my) hair (unintentionally).’
Another verb in this class occurs with the coverb aman ‘hug.’ The ki-class
form is a simple transitive verb. It can occur with either the congruent (31a)
or the noncongruent (31b). The noncongruent form indicates that the speaker
accidentally hugged Rosa (40b). Perhaps it was dark and he mistook her for
someone else. The i-class form4(32) has a sense of permission — the
speaker allowed himself to be hugged. As with purai ‘have cut,’ above, I
could not elicit a noncongruent marked form with this verb.
(31) a. la Rosaka amankiyoe
la Rosa=ka aman-ki-yo-e
1 Rosa= hug-do-CONGR-
‘I hugged Rosa (intentionally).’
b. la Rosaka amankeie
la Rosa=ka aman-ki-i-e
1 Rosa= hug-do-NCONGR-
‘I hugged Rosa (unintentionally).’
(32) la Rosabe amanhiyoe
la Rosa=be aman-i-yo-e
1 Rosa= hug-become-CONGR-
‘I let/had Rosa hug (me).’
*la Rosabe amanhiie
la Rosa=be aman-i-i-e
1 Rosa= hug-become-NCONGR-
MIRATIVITY IN TSAFIKI 399
3.2.5 Motion verbs
Motion verbs are all intransitive i-class verbs. Initially I had difficulty
eliciting a motion verb with a noncongruent suffix. Speakers thought the
construction was acceptable, but could not easily give me a context in which
you would say such a thing. Finally it was suggested you could utter the
noncongruent form (33a) in a context where someone invites you to visit his
cousin. When you get to the house you realize that you have been there
before with Carlos. You hadn’t realized that the cousin’s house and the
house you visited with Carlos were the same location. If you were already
aware that you had been there before you would use the congruent form
(33b).
(33) a. Carlosbe tse inte faie
Carlos=be tse in=te fa-i-e
Carlos= = arrive.here-NCONGR-
‘I’ve come here (before) with Carlos! (I just realized it).’
b. tse Carlosbe inte fayoe
tse Carlos=be in=te fa-yo-e
1 Carlos= = arrive.here-CONGR-
‘I’ve come here (before) with Carlos.’
Another context for the use of a motion verb with a noncongruent marker
occurred in the context of a surprise birthday party. The father, for whom the
party was planned, was invited to a soccer game by one of his sons so the
family could prepare for the party. The father volitionally attended the game
but he did not know the real reason he had been invited. Speakers say that
in the situation coded by this proposition, he is like a child that simply goes
along with Juan without fully understanding why (34).
(34) la Juanbe futbol kirabi jiie
la Juan=be futbol kira-bi ji-i-e
1 Juan= soccer see- go-NCONGR-
‘I went/was taken with/by Juan to watch soccer.’
These verbs resemble the i-class transitive verbs in that the source of the
action is also the affected participant. However, unlike the transitive i-class
verbs, these verbs do not require an intentional, volitional subject and the
original intention need not match the final location or purpose of the
movement. As with the intransitive, ki-class, activity verbs, the noncongruent
400 CONNIE DICKINSON
marker is possible if the event is elaborated by the addition of a location or
a purpose which is unexpected or unknown to the speaker.
3.3 Stative verbs: jo and ra class
The use of a noncongruent marker with a stative verb has a slightly
different connotation, as was demonstrated in (18a) above, where the woman
soccer player said ‘I am a man’ with an ironic connotation. With these verbs,
there is no process or change of state involved. States are identifiable by
clusters of features. The noncongruent marker indicates that the situation
contains a feature that is unexpected or at odds with the speaker’s general
knowledge as to what characteristics are associated with certain states.
The noncongruent form in (35a) would be said in a situation where your
actions might indicate to others that you are bad but you don’t believe
yourself to be bad. (When I presented this example to an older speaker, she
laughed and told me that la seiton joieis what people always say because no
one ever really thinks he is bad.) One speaker explained that a context for
the noncongruent form would be when a relative wants to borrow money
from you. You don’t have the money and can’t lend it to him. He, however,
has reason to believe that you do have the money. Therefore, you appear to
be bad, although you know that you are not — the loan is simply beyond
your capabilities. If you did have the money and refused to lend it, the
congruent form (35b) would be more appropriate.
(35) a. la seiton joie
la se-ito-n jo-i-e
1 good-not.be- be-NCONGR-
‘I’m bad (but I don’t believe it).’
b. la seiton joyoe
la se-ito-n jo-yo-e
1 good-not.be- be-CONGR-
‘I’m bad (and I know it).’
Chiban jono ‘to be hot,’ with the congruent form (36a), indicates that the
speaker feels hot, his body is hot, his skin is flushed — he feels hot and he
is hot. The noncongruent form (36b) is somewhat odd in that it indicates that
the speaker seems to be hot — his face is flushed, his skin is hot — but the
MIRATIVITY IN TSAFIKI 401
speaker does not feel hot. There is a contradiction between his sensation and
his body state.
(36) a. chiban joyoe
chiban jo-yo-e
hot be-CONGR-
‘I’m hot.’
b. chiban joie
chiban jo-i-e
hot be-NCONGR-
‘I’m hot!!’
The ra-class verbs have a sense of surprise when they occur with a non-
congruent marker. The speaker finds himself in a certain situation, but the
situation is unexpected. Tano ‘to have’ is a ra-class verb. The examples in
(37) are much like the Lhasa examples above, (14). In (37a) the speaker
knows he has money. In (37b) the speaker suddenly discovers he has some
money he did not think he had.
(37) a. kala tayoe
kala ta-yo-e
money have-CONGR-
‘I have money.’
b. kala taie
kala ta-i-e
money have-NCONGR-
‘I have money!!’
The stative verbs, like the intransitive i-class verbs discussed in
Section 2.2.1.2, have theme or undergoer subjects. However, these verbs
code a situation or state rather than a process. When a characteristic not
normally associated with that state is present — a woman exhibits a male
trait, someone spends all his money yet inexplicably has some, one’s skin
feels hot but there is no internal sensation of heat — the information is not
congruent with the speaker’s general knowledge and the noncongruent
marker is used.
402 CONNIE DICKINSON
3.4 Verbs of desire, emotion and body processes
There are some situations and events that regularly occur without
premonitory consciousness or obvious causes. These situations include
desires, emotions and body processes. In Tsafiki, these events and situations
are coded by two different constructions. In one construction, mirative
markers obligatorily occur with a first person subject or with the logophoric-
like constructions. The other construction is formed with a dative-subject and
cannot occur with mirative markers. Moore (1991) calls this class of verbs
<LINK"dic-r10">
that take dative subjects ‘impersonal’ because these verbs cannot take either
the congruent suffix-yo nor the noncongruent -i suffix, which he analyzes
as first-person agreement markers. While it may be true that, at least
diachronically, these constructions did not allow the mirative markers
because the dative-marked argument is not the subject, there is some
evidence from word order, plural subject agreement, switch reference
markers, placement of prefixes and relative clause constructions that the
dative-marked argument has at least some of the syntactic properties of a
subject. This will not be discussed here. What is of primary interest here is
the contrast in meaning between the dative-marked subject construction and
the construction that allows the mirative suffixes.
While the dative-subject constructions cannot take the mirative suffixes,
they show the same distribution pattern as verbs inflected with mirative
markers. Dative-subject constructions can only occur when the source of the
information is also the primary participant, i.e. either a first-person subject or
in the logophoric-like constructions. The dative-subject constructions’
inability to take mirative markers and their distribution pattern raises
questions concerning their mirativity status. I claim that these constructions
code, among other things, a third mirative distinction5. The mirative markers
indicate whether the information is or is not in accordance with the speaker’s
general knowledge. The dative-subject constructions occur with events and
sensations that commonly occur without overt, known causes or premonitory
consciousness. Body processes, emotions and desires are often fleeting
phenomena that come and go without the experiencer ever understanding
their specific causes. In essence, the unexpected occurrence of these
processes is expected.
MIRATIVITY IN TSAFIKI 403
3.4.1 Body processess
The first set of distinctions concerns verbs that code body processes.
There are two ways to form these verbs. The first uses a coverb in combina-
tion with the simple verb ki ‘make/do’ and can take mirative markers, (38a)
and (39a). The second is formed with the same coverb, a dative-subject and
the simple verb la ‘come.out,’ (38b) and (39b). This coverb class includes
achi ‘sneeze,’ are ‘burp,’ jada ‘yawn,’ jeko ‘hiccup,’ and katsa ‘vomit.’ The
distinction between the two forms is not so much one of control, but rather
between premonitory consciousness and surprise. At first speakers would not
accept katsa ‘vomit’ with the dative-subject construction (38b), because
normally one has a sensation of nausea before one vomits. Only after I
presented a scenario in which a speaker swallows a pill which immediately
comes back up would they accept the dative-subject form.. The dative-
subject form is also marginally accepted with sirapi kino ‘urinate.’ This verb
is like katsa ‘vomit,’ in that the speaker normally has some premonitory
consciousness that the event is going to occur. While control may also be a
factor, the primary criterion seems to be expectation. The dative-subject
constructions are more likely to be said immediately after the sudden,
unanticipated occurrence of the event. Achi ‘sneeze’ is acceptable with
either form (39).
(38) a. la katsa kiyoe
la katsa ki-yo-e
1MASC vomit do-CONGR-
‘I vomited.’ (lit. ‘I did a vomit.’)
b. lachi katsa lae
la=chi katsa la-e
1MASC= vomit come.out-
‘I vomited!’ (lit. ‘Vomit came out of me.’)
(39) a. la achi kiyoe
la achi ki-yo-e
1MASC sneeze do-CONGR-
‘I sneezed.’ (lit. ‘I did a sneeze.’)
b. lachi achi lae
la=chi achi la-e
1MASC=dat sneeze come.out-
‘I sneezed!’ (lit. ‘A sneeze came out of me.’)
404 CONNIE DICKINSON
3.4.2 Desire
Another set of verbs which can occur with a dative-subject are verbs of
desire. The coverb mu/muna ‘desire’ takes a variety of simple verbs to code
different kinds of desire. In (40) below, with a dative-subject and the simple
verb ki ‘do,’ mirative markers are not allowed in either the question (40a) or
in the answer (40b). This question would be appropriate in a context where
one is simply offering coffee to another person without any idea whether he
wants, needs, or even likes coffee.
(40) a. nuchi kafe mukin?
nu=chi kafe mu-ki-n
2=DAT coffee desire-do-
‘Do you want coffee?’
*nuchi kafe mukiyun?
nu=chi kafe mu-ki-yo-n
2=DAT coffee desire-do-CONGR-
b. jaan, lachi kafe mukee
jaan la=chi kafe mu-ki-e
yes 1MASC=DAT coffee desire-do-
‘Yes, I want coffee.’
*jaan, lachi kafe mukiyoe
jaan la=chi kafe mu-ki-yo-e
yes MASC=DAT coffee desire-do-CONGR-
Dative-subject constructions are always interpreted as having a first-person
subject. Only if the perspective is placed with the subject by the use of the
logophoric-like construction, is the dative-subject construction allowed with
a third person (41). This is the same pattern described above, where the
congruent/noncongruent distinction with a third-person subject can only be
used if the information source is the same as the primary participant in the
clause.
(41) Juanchi kafe mukeetie
Juan=chi kafe mu-ke-e ti-e
Juan=DAT coffee desire-do- say-
‘Juan says he wants coffee.’
In (42) below, where muna combines with the simple verb ra ‘be.in.a.posi-
tion,’ the congruent form can be used. This would appropriately be uttered
MIRATIVITY IN TSAFIKI 405
in a situation where the questioner knows that the other person is tired, he
has been up late, etc. The question is not addressing an internal desire, but
rather, a physical need and the speaker has evidence and reason to believe
that it is likely the addressee wants coffee. The subject here is not marked
with the dative nor is the ti necessary with a third-person subject (42c).
(42) a. nu kafe munarayun?
nu kafe muna-ra-yo-n
2coffee desire-be.in.a.position-CONGR-
‘Do you need coffee?’
b. jaan, la kafe munarayoe
jaan la kafe muna-ra-yo-e
yes 1MASC coffee desire-be.in.a.position-CONGR-
‘Yes, I need coffee.’
c. Juan kafe munarae
Juan kafe muna-ra-e
Juan coffee desire-be.in.a.position-
‘Juan needs coffee.’
3.4.3 Body states and emotion
The same pattern of interpretation occurs with verbs of body state and
emotion. The congruent forms (43a–b) would be said in a situation where
Juan either obviously looks sad — he is crying, his head is drooping etc. —
or when there is an obvious and culturally reasonable reason for the sadness
— someone has died or he has suffered some other misfortune. The dative-
subject construction (43c) would be used when there is neither physical
evidence from Juan’s appearance, nor an obvious reason for the sadness. The
verb ti ‘say’ is obligatory in (43c).
(43) a. Juan laki joe
Juan laki jo-e
Juan sad be-
‘Juan is sad.’
b. Juan laki ie
Juan laki i-e
Juan sad become-
‘Juan became sad.’
406 CONNIE DICKINSON
c. Juanchi laki jaetie
Juan=chi laki ja-e ti-e
Juan=DAT sad come- say-
‘Juan says he is sad.’ (lit. ‘Juan says sadness comes to
him.’)
The tripartite mirative system is demonstrated in example (44) below. The
congruent form (44a) could appropriately be said in Quito, which to a person
from the lowlands is always unreasonably cold. The air is cold and the
speaker is cold. The noncongruent form (44b) indicates that the speaker’s
body is cold. He has goosebumps or his skin is cold to the touch, but he is
not experiencing the sensation of cold. This form is a bit odd. Usually if
one’s body is cold, one feels cold. The dative-subject construction (44c)
would be said in context where there is no obvious cause for the feeling of
cold or evidence that the speaker’s body is cold. There is simply a sensation
of cold. My consultants suggested that perhaps the speaker is coming down
with a fever.
(44) a. tse jida joyoe
tse jida jo-yo-e
1FEM cold be-CONGR-
‘I am cold.’
b. la jida joie
la jida jo-i-e
1MASC cold be-NCONGR-
‘I’m cold!!’
c. tsechi jida kee
tse=chi jida ke-e
1FEM=DAT cold do-
‘I feel cold.’
In all the above examples, the speaker is aware of the cold. The difference
lies in how well the speaker’s immediate experience is matching his internal
sensations and knowledge as to what conditions engender which states. The
congruent form indicates that the speaker’s sensation of cold is congruent
with his knowledge of the world. The noncongruent form indicates there is
a contradiction between the speaker’s sensation of cold and his perception of
his body state. The dative-subject construction is essentially neutral. The
speaker experiences the sensation, but does not know the cause, his body
MIRATIVITY IN TSAFIKI 407
shows no indications of a lowered temperature and there was no previous
indication that the sensation was about to occur.
4. Evidential System
As has been noted by DeLancey (1997) and Aksu-Koç and Slobin (1986),
<LINK"dic-r4"><LINK"dic-r12">
there is often a close relationship between markers used for inferential
evidence and mirativity. In Tsafiki, when the source of the information is not
a primary participant in the event or state, an inferential evidential marker
also serves as a mirative marker. The evidential/mirative marker indicates
that the information coded in the proposition is surprising to the speaker.
As noted above, evidentiality and mirativity are closely related to
epistemic modality. But in Tsafiki, epistemic modality is not the primary
function of these constructions. Epistemic modality is clearly a separate
category.6As will be illustrated below, the evidential markers in Tsafiki
overlap with mirativity and delineate the speaker’s participatory role in the
event. They do not specifically mark epistemic distinctions.
Although his analysis differs from the one presented here, Mix (1992)
<LINK"dic-r9">
was the first to explicitly note the degree to which evidentiality is an integral
part of the grammatical structure of Tsafiki. However, the analysis presented
here is much closer to Moore (1991). Although he does not specifically
<LINK"dic-r10">
identify these markers as evidentials, he describes the hearsay evidential
(‘histórico’), the inferential (‘pasado sin participación’) and the deductive
evidential.
There are three basic evidential distinctions in Tsafiki: 1) information
obtained through direct sensory experience; 2) information inferred from
direct physical evidence; and 3) information deduced from the speaker’s
general world knowledge.
Information from directly witnessed events is coded with a morphologically
unmarked verb (45a). Information inferred from direct physical evidence is
marked with a verbal suffix-nu (45b), while inference from general know-
ledge is marked by a nominalized verb form plus a verb class marker (45c).
(45) a. Manuel ano fie
Manuel ano fi-e
Manuel food eat-
‘Manuel ate.’ (The speaker saw him.)
408 CONNIE DICKINSON
b. Manuel ano finue
Manuel ano fi-nu-e
Manuel food eat-EV/MIR-
‘Manuel has eaten.’ (The speaker sees the dirty dishes.)
c. Manuel ano finkee
Manuel ano fi-n-ki-e
Manuel food eat-NOM-VCL:do-
‘Manuel must have eaten.’ (He always eats at 8:00; it’s now
9:00.)
Tsafiki also has a hearsay evidential. This evidential is somewhat typolog-
ically unusual for an evidential system in that it may be used in congruence
with any of the other three evidentials. It indicates the type of information
the original informant had for the assertion (46).
(46) Manuel ano finutie
Manuel ano fi-nu-ti-e
Manuel food eat-EV/MIR-HS-
‘He said/they say Manuel has eaten.’ (They didn’t see him, but
they have direct physical evidence.)
An interesting aspect of this -ti is that it may be repeated to indicate up to
three sources between the speaker and the original event (47). (If more than
three sources are involved the plural form would be used.)
(47) Mantoka jitietitietie
man-to=ka ji-ti-e ti-ti-e ti-e
other-earth= go-HS- say-HS- say-
‘They say they say they say he went to Santo Domingo.’
4.1 Integrated vs. new knowledge and participation vs. non-participation
The use of these evidentials is more complicated than indicated by these
sentences elicited out of context. The direct evidential does not always
indicate that the speaker actually consciously witnessed the event. The
context for (48) below was due to a controversy as to where various mem-
bers of the family had been born. The speaker asked his mother and later
reported the information explicitly citing his mother as the source:
MIRATIVITY IN TSAFIKI 409
(48) lachi ayan la konkomabi na ie
la=chi ayan la konkoma=bi na i-e
1= mother 1 Congoma= child become-
tie
ti-e
say-
‘My mother says I was born in Congoma.’
A few days later however, he used the direct evidential form with the
congruent marker (49). When I pointed out he could not actually remember
witnessing the event he replied that this was true, but he now knew it very
well and hence could use this form.
(49) la konkomabi na iyoe
la konkoma=bi na i-yo-e
1 Congoma= child become-CONGR-
‘I was born in Congoma.’
So the use of the direct form is not simply direct, conscious experience of
the event, but also includes the degree to which the information is integrated
with one’s general knowledge. Direct forms may be used with non-witnessed
events, if the information is part of the speaker’s integrated knowledge. This
is essentially a mirative distinction.
4.2 Participation vs. non-participation
The use of evidentials also indicates participation or non-participation
in the event. The context for (50) is one in which a neighbor comes by as a
wife and her husband are eating breakfast. The husband says he is going to
Santo Domingo that day. Later the wife and neighbor go to visit the grand-
parents. The husband leaves. Neither the wife nor the neighbor actually sees
him leave. When asked where her husband went, the wife would most
probably use the direct evidence form of the verb (50a). The neighbor,
because he did not actually see the husband leave, would have to use a form
marked with either the indirect evidential or the hearsay evidential (50b). He
would be accused of being a nene pun ‘liar,’7or at the very least presumptu-
ous, if he used the direct form. The wife can use the direct form because she
is a participant in the event in a way in which the neighbor is not. But note
they are both equally certain that the husband went to Santo Domingo.
410 CONNIE DICKINSON
(50) a. The wife:
ya Mantoka jie
ya man-to=ka ji-e
3 other-earth= go-
‘He went to Santo Domingo.’
b. The neighbor:
ya Mantoka jiyoetie
ya man-to=ka ji-yo-e ti-e
3 other-earth= go-CONGR- say-
‘Heisaid that heiwent to Santo Domingo.’
The neighbor does not have an option here if he wants to speak correctly. He
must use an indirect evidential form. The wife does have the option. The use
of an evidential on the part of a wife talking about her husband’s actions
indicates either that he did not discuss the matter with her, or she is not in
agreement with him. Normally a wife is considered to be a participant in her
husband’s activities. The use of -nu in (51) indicates that she is declaring
herself a non-participant, which carries a strong implication of disapproval.8
(51) The wife:
ya mantoka jinue
ya man-to=ka ji-nu-e
he other-earth= go-EV/MIR-
‘He has gone to Santo Domingo.’ (and she does not approve)
4.3 Nu as a mirative marker
Moore (1991) analyzes the difference between participation and non-
<LINK"dic-r10">
participation as being the main function of the inferential -nu. As was shown
above, this holds in certain situations where the use of the -nu may be
interpreted as a specific declaration of non-participation. And this analysis of
-nu also makes logical sense. If one is inferring from evidence, one was
obviously not a participant. But the sense of participation/non-participation
is not strictly literal. In the three examples below, the speaker directly
experiences the event. The context in (52) is a Tsachi has gone to the jungle
to hunt. While he is going he suddenly sees another person coming towards
him on the trail. The newcomer is unexpected and a stranger to the Tsachi.
MIRATIVITY IN TSAFIKI 411
(52) junni jun jinasari tsachi jananutie
junni jun ji-na-sa-ri tsachi ja-na-nu-ti-e
then go--- person come--EV/MIR--
tie
ti-e
say-
‘Then they say he said as he was going a person (a stranger) was
coming!’
The context for (53) is one in which the speaker heard what he thought was
a car approaching. But when he saw it, he realized it was a motorcycle.
(53) moto jonue
moto jo-nu-e
motorcycle be-EV/MIR-
‘It’s a motorcycle!’
The next example, (54), was spoken in the context in which I had left my
son in Quito with friends, thinking he was returning to the U.S. the follow-
ing day. He decided not to leave and three days later unexpectedly returned.
(54) Aeson manjanue
Aeson man-ja-nu-e
Aeson again-come-EV/MIR-
‘Aeson came back!’
The above three examples seem to be a contradiction of the analysis of -nu
as an inferential evidential. The speaker is a direct witness of the event or
state. What these examples have in common is a sense of surprise. The
speaker was not psychologically prepared for the event (DeLancey 1997).
<LINK"dic-r4">
Moore’s analysis of this as indicating non-participation fits here, if non-
participation is extended to a psychological sense.
4.4 Congruent/noncongruent and the inferential evidential
The use of an inferential evidential marker with a first person is
somewhat odd. Usually one obtains knowledge about one’s actions through
direct experience. Therefore it is not surprising that in Tsafiki the use of the
-nu with a first person subject only permits the noncongruent form. The
congruent form -yo cannot co-occur with -nu. The inferential evidential
412 CONNIE DICKINSON
construction (55) could be said appropriately if the speaker went to bed and
unexpectedly woke up on the floor in the morning. As discussed above,
keerei ‘fall’ can be used with either the congruent or noncongruent form to
indicate accidental falling (23). The difference between the simple noncon-
gruent form and the evidential form concerns how the speaker knows the
action occurred. The simple noncongruent form is likely to be said immedi-
ately after the fall. The speaker knows she fell, the event was simply
unexpected. The evidential form indicates the speaker is inferring from
evidence that she fell (55).
(55) tse lowabi nechi keereiinue
tse lowa=bi ne=chi keere-i-i-nu-e
1 bed= from= throw-become-NCONGR-EV/MIR-
‘I must have fallen out of bed.’ (I’m on the floor)
In (56a) below the congruent form indicates that the behavior was intention-
al. The speaker intended to kill the pig. The noncongruent form (56b)
indicates that the speaker accidentally killed the pig. Perhaps he ran over it
with his car. In (56c) the inferential evidence form also indicates that the
behavior was unintentional, but the speaker is inferring from evidence what
must have happened. Perhaps the speaker gave the pig some medication that
inadvertently caused its death.
(56) a. la kuchika toteyoe
la kuchi=ka tote-yo-e
1 pig= kill-CONGR-
‘I killed the pig.’
b. la kuchika toteie
la kuchi=ka tote-i-e
1 pig= kill-NCONGR-
‘I killed the pig (unintentionally).’
c. la kuchika toteinue
la kuchi=ka tote-i-nu-e
1 pig= kill-NCONGR-EV/MIR-
‘I must have killed the pig (unintentionally).’
4.5 Congruent/Noncongruent and the deductive evidential
As with the inferential evidential, the deductive evidential can only
occur with the noncongruent suffix. When the source of information is also
MIRATIVITY IN TSAFIKI 413
the primary participant in the event the congruent suffix is not allowed. This
construction is somewhat unusual in that, normally a primary participant is
aware of or has direct evidence concerning his own actions. Example (57)
was said in a context where a person had unbeknownst to him, eaten
armadillo in the lunch soup the week before. Here he claimed he had never
eaten armadillo. The others present laughed and told him he had eaten it the
week before in the soup. He did not remember the meal and had no direct
physical evidence that he had eaten armadillo. But he was present at all the
meals. He had eaten soup at every meal. Therefore if armadillo was in the
soup, he had eaten armadillo. Note that this is not the expected verb form.
With a third person subject the deductive evidential form would be finkee
not fiinhie.
(57) jodoka fiinhie
jodo=ka fi-i-n-i-e
armadillo= eat-NCONGR-NOM-VCL:become-
‘I must have eaten armadillo.’
4.6 Summary
Tsafiki has three basic evidential distinctions plus a hearsay evidential.
At least two of the evidentials may also be used to indicate the status of the
information in terms of the speaker’s knowledge. The direct form may be
used for information that is not witnessed, but integrated into the speaker’s
knowledge system or it may also be used when the speaker has a participato-
ry role in the event or state. The inferential evidential may also be used to
indicate non-participation or that the information is surprising or new — a
mirative distinction. The inferential and deductive evidence forms of the verb
always require the noncongruent marker with a first person or with the
logophoric-like construction. The congruent form is not allowed with a verb
that is suffixed with an inferential or deductive evidence suffix.
5. Future constructions
There are at least four constructions that can be used to indicate future
in Tsafiki. Two of the constructions can occur with either mirative marker.
The other two can only take noncongruent markers.
414 CONNIE DICKINSON
The inceptive (58), with a congruent suffix, is used when the actor has
already begun the action. The logophoric construction is not required for a
third person (58b).
(58) a. la mantoka jichinayoe
la man-to=ka ji-chi-na-yo-e
1 other-earth= go-INCEP-PROG-CONGR-
‘I am going to Santo Domingo (I’m on my way).’
b. ya mantoka jichinae
ya man-to=ka ji-chi-na-e
he other-earth= go-INCEP-PROG-
‘He is going to Santo Domingo (He’s on his way).’
Another construction used for the future has a sense of necessity. There is an
obvious, coherent reason for doing the action. A suggested context for this
form is that the person needs to go to Santo Domingo to buy supplies. This
construction uses a nominalized verb form with the auxiliary jono ‘to be’ (59).
(59) a. la mantoka jino joyoe
la man-to=ka ji-no jo-yo-e
1 other-earth= go-INF be-CONGR-
‘I have to go to Santo Domingo.’
b. ya mantoka jino joe
ya man-to=ka ji-no jo-e
he other-earth= go-INF be-
‘He has to go to Santo Domingo.’
Both these forms can take the noncongruent suffix but the sense is a bit odd.
Example (60a) has a sense of surprise. The speaker initially believed he
would not be able to go to Santo Domingo, or perhaps he thought he was
going to another city, but suddenly he discovers that he is going to Santo
Domingo after all. (60b) would be used in a situation in which one needs
something from Santo Domingo. The speaker sent someone else to fetch the
item, but the person wasn’t able to bring it back. Here, initially, he did not
think he had to go. But it appears he does.
(60) a. la mantoka jichinaie
la man-to=ka ji-chi-na-i-e
1 other-earth= go-INCEP-PROG-NCONGR-
‘I’m going to Santo Domingo!! (after all)’
MIRATIVITY IN TSAFIKI 415
b. la mantoka jino joie
la man-to=ka ji-no jo-i-e
1 other-earth= go-INF be-NCONGR-
‘It seems I have to go to Santo Domingo.’
The next construction cannot take the congruent suffix with a first-person
subject. With a first-person (61a), but not a third-person subject (61b), the
clause has a negative connotation. The person does not want to perform the
action. The context given is that the actor has been ordered to perform the
action by another person for someone else’s benefit.
(61) a. tse mantoka jiichunae
tse man-to=ka ji-i-chu-na-e
1 other-earth= go-NCONGR-IRR-PROG-
‘I will go to Santo Domingo.’ (But I don’t want to go.)
b. ya mantoka jichunae
ya man-to=ka ji-chu-na-e
he other-earth= go-IRR-PROG-
‘He will go to Santo Domingo.’
The final form is only used with a first person subject, or with the logo-
phoric-like construction. With an active verb this construction uses the
progressive marker with a noncongruent form to indicate future (62). (The
congruent and noncongruent forms here are distinguished by an accent shift
and a glottal stop).9The context given for the noncongruent-marked con-
struction is one in which one plans and wants to go but the realization of the
action is dependent on external circumstances that are somewhat uncertain —
the bus drivers are on strike, the truck is not running and may not be
repaired in time, etc. This construction only indicates future intention. The
speaker plans and intends to go even though it may not be possible.
(62) tse mantoka ji’nae
tse man-to=ka ji-’-na-e
1 other-earth= go-NCONGR-PROG-
‘I intend to go to Santo Domingo (if possible).’
The congruent suffix can occur with the progressive marker (63), but the
default interpretation is present progressive, not future. The appropriate
context for this construction would be one in which the speaker is walking
down the road and someone asks her where she is going.
416 CONNIE DICKINSON
(63) tse mantoka jinayoe
tse man-to=ka ji-na-yo-e
1 other-earth= go-prog-congr-
‘I am going to Santo Domingo.’
What both future constructions with a noncongruent marker have in common
is that in both cases there is a conflict between the actor’s desires and
intentions and the possible realization of the event. With -ichuna the action
is due to the wishes of another person. With -‘na the speaker desires to go,
but extenuating circumstances may prevent him from going. The two future
forms taking congruent forms indicate either that the speaker has already
begun to perform the action -china, or that he needs to perform the action
for an obvious, coherent reason -no jo-. In other words, when the speaker’s
desires and intentions are congruent with extenuating circumstances the
congruent form is used. When the speaker’s desires are not in accordance
with extenuating circumstances, either he wishes to perform the action but
circumstances may prevent him, or he does not want to perform the action,
but circumstances dictate that he do so, the noncongruent form is used.
6. Conclusion
Implicit in the use of the Tsafiki mirative system is a theory of con-
scious experience. In the normal course of experience, an observer has a set
of expectations concerning events and states. When an event or state does
not meet those expectations, the Tsachi speaker must explicitly mark the
information contained in the proposition as occurring outside the normal
course of experience. The information cannot be easily assimilated into the
speaker’s set of expectations and assumptions about the world.
In this paper I have tried to elaborate some of the more important
characteristics of the complex, Tsafiki mirative system. By describing in
some detail the interaction between various semantic verb classes and the
mirative markers, I have attempted to give some sense of the range and
nature of the category. The different, final interpretations given to mirative
markers — non-intention, unawareness, irony and ignorance — arise through
the interaction of the marker and the semantics of the verb.
I have also attempted to demonstrate the distinction between evidentiality
and mirativity. While there is some overlap in Tsafiki between evidentiality
MIRATIVITY IN TSAFIKI 417
and mirativity, the ability of mirative markers to occur without evidential
markers clearly show that the two categories can be separated. The question
is not so much whether mirativity, evidentiality and epistemic modality are
universally distinct categories, but rather, how the categories are treated in
specific languages. In English epistemic modality and evidentiality are
conflated, in Turkish one single marker codes both mirative and evidential
distinctions, whereas in Tsafiki there is sufficient evidence to justify treating
the three categories separately.
Finally, as noted by Aksu-Koç and Slobin (1986:167), while all
<LINK"dic-r12">
languages probably contain implicit theories concerning conscious experi-
ence, we as yet do not know the degree to which languages embody such
theories. Tsafiki has clearly integrated a mirative system, with its underlying
theory of conscious experience, into its grammar as an obligatory, inflection-
al category. It is hoped that as more languages with complex mirative
systems are described and discussed, a full theory of experiential knowledge
and how it is expressed through language can be developed.
Author’s Address:
Connie Dickinson
Department of Linguistics
University of Oregon
Eugene, Oregon
97403 USA
ACKNOWLEDGMENTS
An earlier version of this paper was presented at the 14th Conference on American Indian
Languages, 1996, San Diego, California and at the Universidad Autonoma, Barcelona, Spain,
March, 1996. The research for this paper was completed during two field sessions in Ecuador
from October 1994-August 1995 and August 1996-March 1998. This paper could not have been
done without the support of Robert Mix and Ramón Aguavil Calazacón of the Vernacular
Language Program of the Anthropological Museum of the Central Bank of Ecuador, Guayaquil
who kindly invited me to Ecuador and provided me with materials from the collection of the
VLP during my first visit to Ecuador in 1994–95 and I wish to thank them. Linda Jacobs
provided much needed support during the my first visit — Thank you! My second stay in
Ecuador was supported by a Fulbright Fellowship and a grant from the National Science
Foundation (SBR-9614981). I would also like to thank the Governor of the Tsachila, Sr.
Manuel Calazacón Aranzona, whose interest in, and support of this project is much appreciated.
I am also grateful to Alfonso Aguavil Calazacón, Director of PIKITSA, who not only shared
418 CONNIE DICKINSON
his linguistic knowledge but also provided much needed technical support. And a sincere thank-
you to my primary language consultants, Juan Aguavil Calazacón, Primitivo Aguavil Cala-
zacón, Catalina Calazacón Aguavil, Jose Aguavil Loche, Gloria Campues Aguavil, Secundina
Aguavil Calazacón, Francisco Aguavil Alopi, Maria de los Angeles Aguavil Alopi, Franklin
Aguavil Aguavil and all the other Tsachila who shared their time and stories with me. I would
also like to thank Berta Zaracay Aguavil, Rosa Zaracay Aguavil and their families for tolerating
a stranger in their house for so many months. And I would like to thank Jose Aguavil Loche
and his extended family for making my stay in Ecuador so enjoyable. A special thanks also
goes to Montserrat Ventura i Oller for generously sharing her knowledge of the Tsachila
culture. And finally Dr. Marleen Haboud of the Universidad Pontifica Católica de Quito, and
Dr. Fernando Ortega of the Unversidad de San Francisco de Cumbaya provided support in
more ways than I can possibly delineate here. Scott DeLancey, Colette Craig, Montserrat
Ventura i Oller, Roberto Zavala, T. Givón and Eric Pederson were good enough to read and
offer comments on an earlier draft of this paper. However, any mistakes or errors are solely
my responsibility.
NOTES
1. The abbreviations used in this paper are: 1 = first-person feminine; 1 = first-
person masculine; 2 = second-person; 3 = third-person; = accusative; =
associative; = auxiliary; = causative; = congruent; = contrastive;
= dative; = declarative; = demonstrative; =different subject; =
ergative; = emphasis; / = evidential mirative; = focus; = habitual;
= hearsay; = inceptive; = infinitive; = instrument; = interrogative;
= irrealis; = locative; = numeral/adjectival classifier; = noncongruent;
= negative; = nominative; = perfective; = plural; = possesive;
= progressive; = purpose; = quotative; = reciprocal; = situation-
al; = same subject; = verb class marker.
2. Most of the Tsafiki examples cited in this paper are elicitations from my fieldnotes.
However the constructions examined were first identified in narrative texts. These texts
consist of 74 hours of recorded traditional, historical and personal narratives and
conversations. Once a construction is identified in the text, I then elicit further examples
with three bilingual consultants, who have had extensive training in linguistic analysis. I
then consult with two older, speakers, one of whom is considered an expert story-teller
and is essentially monolingual. The examples are then checked with at least 4–5 other
speakers. I then return to the texts and make sure there are no contradictions to my
analysis in the texts. For the sake of clarity, I mainly use elicited examples here, but they
have been extensively checked against the texts. The family I live with in Congoma
primarily speaks Tsafiki within the household. Therefore, many of the examples in this
paper are derived from situations encountered in daily life. Examples from texts are
marked as such, the absence of a specific reference indicates the example is elicited or
was overhead in daily conversation.
3. The majority of Tsafiki verbs (95% out of a corpus of approximately 1,000 verbs) are
complex predicates consisting of a noun or coverb in combination with one of about 30
MIRATIVITY IN TSAFIKI 419
‘simple’ or ‘light’ verbs. A coverb is a neutral element which requires a nominalizar to be
used as a noun and a simple verb to form a predicate. This process of verb derivation is
distinct from the verb class marker system. Most of the coverbs can occur with more than
one light verb. In the example below wiru ‘stand’ combines with three different light
verbs to form a transitive (a), an inchoative (b) and a stative (c). Once the complex
predicate is formed, it is then classified by the verb class markers as these plural examples
illustrate. The three simple verbs used in these examples commonly occur with coverbs
denoting positions.
a. chideka wirukarilakee
chide=ka wiru-kari-la-ke-e
tree= stand-make--:do-
‘They stood the pole up.’
b. wiruilaie
wiru-i-la-i-e
stand-become--:become-
‘They stood up.’
c. wiruralarae
wiru-ra-la-ra-e
stand-be.in.a.position--:be.in.a.position-
‘They are standing.’
4. Note that the object is deleted here. Unlike pore ‘cut’ it is understood that it is the body
that is being hugged here. Only if some other unexpected part of the body were hugged
would there be an overt object coded here.
5. I am not claiming that this is a typological, or general characteristic of dative-subject
constructions. Dative-subject constructions with these types of verbs are somewhat
ubiquitous in the world’s languages. It is doubtful that in a language that does not have
a developed mirative system, one would find the same semantic and pragmatic functions
as in Tsafiki.
6. There are a variety of ways to code epistemics in Tsafiki. The most common way of
marking doubt is to simply code the proposition as the complement of the verb tenjano
‘to think/imagine.’ In addition, the final suffix on the verb is a marker of mood. -e, -ko,
-da, and -n occur in complementary distribution as a final suffix. -e is a simple declara-
tive, -ko indicates doubt, -da is glossed by Moore as ‘as everybody knows’ and indicates
certainty, and -n is an interrogative. The progressive marker -na indicates doubt when it
occurs with a stative verb or auxiliary (much like in English: ‘He is being polite’). And
there are other syntactic constructions and lexical words that can be used to indicate
varying degrees of doubt. All these constructions are clearly separate from the mirative/
evidential system.
7. There may be some questions here as to the status of the -ti as a hearsay evidential. It
occurs as both a suffix and a finite verb, and depending on the context the subjects of the
ti may or may not be referring. My consultants insist that there is no difference between
jitieand jie tie, ‘they/he said he left.’ But the exact status of this ti is not relevant here so
it will not be discussed further. The function that is of interest here is that the hearsay
420 CONNIE DICKINSON
evidential or verb of speech locates the source of the information with someone other than
the speaker.
8. However the term nene pun ‘liar’ in Tsafiki has a broad definition. It includes misuse of
the evidential/mirative system, intentional lies, white lies and what we might consider
‘mistakes’ or ‘misinformation’ in English. Because of its broad definition it does not have
nearly as strong of a negative connotation as it does in English.
9. While this example concerns a husband and wife it should be noted that the use of direct
vs. indirect forms is not dependent on kinship relationships. Generally, people who live
in the same household use direct forms when discussing their fellow housemate’s
behavior. I can use direct forms when I speak of the activities of the family I live with,
and they can use direct forms when they speak of me.
10. The noncongruent marker for both -chuna above and -’na are listed by Moore (1962) as
<LINK"dic-r10">
-ya’chuna and -yana, although he also lists -ichuna. The speakers I work with do not use
-ya. They recognize it, but because they do not use it, I haven’t been able to determine
the difference between these forms. I have begun working with speakers of Bua, another
dialect of Tsafiki, where the use of -ya is more common and hope to soon find an
explanation for the different forms.
REFERENCES
Akatsuka, Noriko. 1985. “Conditionals and the epistemic scale.” Language
<DEST"dic-r1">
61.625–639.
Aksu-KoH, Ayhan & Dan Slobin. 1986. “A psychological account of the
<DEST"dic-r2">
development and use of evidentials in Turkish.” In: Evidentiality: The
linguistic coding of epistemology, Chafe & Nichols (eds), 159–167.
Chafe, Wallace & Joanna Nichols. (eds). 1986. Evidentiality: The linguistic
<DEST"dic-r3">
coding of epistemology. Norwood, N.J.: Ablex.
DeLancey, Scott. 1985. “Lhasa Tibetan evidentials and the semantics of
<DEST"dic-r4">
causation.” In Proceedings of the Eleventh Annual Meeting of the Berkeley
Linguistics Society, Mary Niepokuj et al. (eds.), 65–72.
DeLancey, Scott. 1986. “Evidentiality and volitionality in Tibetan.” In
Evidentiality: The linguistic coding of epistemology, Chafe & Nichols
(eds), 203–213.
DeLancey, Scott. 1990a. “A note on evidentiality in Hare.” International
Journal of American Linguistics 56.152- 158.
DeLancey, Scott. 1990b. “Crosslinguistic evidence for the structure of the
agent prototype.” Papers and Reports on Child Language Development 29.
Stanford: The Board of Trustees of the Leland Stanford Junior University.
MIRATIVITY IN TSAFIKI 421
DeLancey, Scott. 1990c. “Ergativity and the cognitive model of event
structure in Lhasa Tibetan.” Cognitive Linguistics 1:289–321.
DeLancey, Scott. 1992. “The historical status of the conjunct/disjunct pattern
in Tibeto-Burman.” Acta Linguistica Havniensia 25.39–62.
DeLancey, Scott. 1997. “Mirativity: The grammatical marking of unexpected
information.” Linguistic Typology 1.33–52. Berlin: Mouton de Gruyter.
DeLancey, Scott. 1999. “The Mirative and Evidentiality.” To appear: Journal
of Pragmatics.
Hale, Austin. 1980. “Person markers: Finite conjunct and disjunct verb forms
<DEST"dic-r5">
in Newari.” In Papers in South-East Asian Linguistics, R. Trail (ed.), No.
7.95–106. Canberra: Australian National University.
Hargreaves, David. 1991. The concept of intentional action in the grammarof
<DEST"dic-r6">
Kathmandu Newari. Ph.D. dissertation, University of Oregon.
Ko, Kyounghee. 1989. The integrated/New Knowledge Markers in Korean.
<DEST"dic-r7">
Master’s Thesis, University of Oregon.
Lee, Ho Sang. 1985. “Consciously-known but unassimilated information: A
<DEST"dic-r8">
pragmatic analysis of the epistemic modal -kun in Korean.” In Proceed-
ings of the First Pacific Linguistics Conference, 183–210. Eugene: Depart-
ment of Linguistics, University of Oregon.
Lee, Ho Sang. 1993. “Cognitive constraints on expressing newly perceived
information, with reference to epistemic modal suffixes in Korean.”
Cognitive Linguistics 4.135–167.
Mix, Robert. 1992. “Evidentiality in Tsafiki.” Unpublished manuscript,
<DEST"dic-r9">
University of Oregon.
Moore, Bruce R. 1966. Diccionario: Catellano-Colorado; Colorado-Castel-
<DEST"dic-r10">
lano. Quito, Ecuador: Insituto Lingüístico de Verano.
Moore, Bruce R. 1991. Patrones Gramaticales Del Colorado (Chibcha).
Stephen H. Levinsohn (ed.). Quito, Ecuador: Instituto Lingüístico de
Verano.
Slater, Keith. 1996. “Evidentiality in Minhe Monguor and the conjunct/
<DEST"dic-r11">
disjunct system: Language contact on the Inner Asian frontier.” Manu-
script, Department of Linguistics, University of California at Santa
Barbara.
Slobin, Dan & Ayhan Aksu. 1982. “Tense, aspect, and modality in the use
<DEST"dic-r12">
of the Turkish evidential.” In Tense–Aspect: Between Semantics and
Pragmatics, Paul Hopper (ed.), 185–200. Amsterdam: Benjamins.
422 CONNIE DICKINSON
Van Valin, Robert D.; Jr. and David P. Wilkins. 1996. “The case for ‘effec-
<DEST"dic-r13">
tor’ case roles, agents and agency revisited.” In Shibatani and Thompson,
(eds), Grammatical Constructions. Oxford University Press.
Vendler, Zeno 1957 [1967]. Linguistics in Philosophy. Ithaca: Cornell
<DEST"dic-r14">
University Press.
Watters, David. 1998. The Kham language of west-central Nepal. Ph.D.
<DEST"dic-r15">
dissertation, University of Oregon.
</TARGET"dic">
