Content uploaded by Connie Dickinson
Author content
All content in this area was uploaded by Connie Dickinson on Sep 06, 2014
Content may be subject to copyright.
(In , 2011. Evans, Nicholas, Alice Gaby,, Stephen C.
Levinson, Asifa Majid (eds.). Amsterdam/Philadelphia: John Benjamins Publishing
Company, pp. 277-313.)
Chapter 17
Reciprocal Constructions in Tsafiki
Dr. Connie Dickinson
University of Oregon/FLACSO, Ecuador
csd@uoregon.edu or conniedickinson7@yahoo.com
Author's contact details:
Dr. Connie Dickinson
Adjunct Research Faculty
University of Oregon/FLACSO
Chimborazo 700 y Pampite
Cumbaya, Ecuador EC170157
ABSTRACT
Tsafiki reciprocal constructions have fairly unique characteristics due to the nature of the
constructions from which they arise and the overall grammatical structure of Tsafiki.
Reciprocals are coded by elements that are already grammaticalized for other functions. < <
Symmetrical positional reciprocal constructions consist of a subset of positionals <that
inherently code reciprocity. There are two basic types of active reciprocal constructions. The
semantic distinction between the two concerns mirative notions such as the degree to which
the event concurs with the speaker’s expectations and general knowledge. The function of
each reciprocal element is explored by examining its role in other constructions.
1. INTRODUCTION
Tsafiki is a Barbacoan language spoken by around 2,000 people, known as Tsachila
(Colorados), in the western lowlands of Ecuador. Tsafiki has a variety of reciprocal
constructions which are used to code symmetrical relationships. As shown clearly in
utterances elicited with a series of video clips and supported by natural discourse, there are
two clear categories—one for positional symmetry and another for action symmetry.
Positional notions of symmetry such as ‘alongside each other’, ‘on top of each other’
and motion verbs indicating positions such as ‘follow’ are coded in Tsafiki by a positional
element and/or the collective enclitic =:
(1) sili-ka=tala
INTJ three-NCL woman long.flexible-CL=COL stand-BE.POSITION-DCL
‘Umm three women are standing next to each other in a line.’ (V #47)
(2) bene
3P
2
man again=go.out-INF because again behind GO-PRG-DCL
‘After coming out again, this man is following (her) again.’ (M #27)
The two most common and productive strategies for coding symmetrical actions as
opposed to symmetrical positions, are differentiated by the choice of an inflecting verb in a
complex predicate construction. This system most closely resembles the ‘auxiliary strategy’
for coding reciprocity (Evans 2006). The first construction utilises the predicating element
‘RR’ (reflexive/reciprocal) and the second, the verb ‘become’. The difference between these
constructions concerns mirative notions such as spontaneity and the degree to which the event
was anticipated by the speaker. In (3) the speaker provides a typical reason for why men fight.
Given the circumstances the fight is expected. In (4) two groups of Tsachila, normally on
friendly terms, get drunk and suddenly begin fighting. The fight was not anticipated given
their normal amicable relationship.
(3)
other=PL defend HAVE defend EXPRESS-IMPF.P=SMBL
a
EXPRESS:VCL-SR hit-INTS-RR-COL-DO:VCL-DCL
‘Others, jealous ones, getting jealous (over a woman) fight each other.’
(18Jun0404.1465.198)
(4) i
person=PL 3D2=SMBL hit-INTS-BECOME-COL-BECOME:VCL-EV-RP-DCL
intoxicate BECOME-SR
‘(They) say that apparently these guys suddenly just fought each other, getting
drunk.’ (23Sep0401.2819.751)
This paper has two primary concerns. The first is to describe the Tsafiki reciprocal
constructions with particular attention to action symmetry and the meaning distinctions
expressed by the above examples; the second is to compare the elicited data with data found
in natural Tsafiki discourse. The elicitation was done with the series of video clips described
in Chapter Two (this volume), recorded so as to portray different (a)symmetrical relationships
such as simultaneous vs. sequential and notions such as strong reciprocals ‘Bob and Bill hit
each other’, asymmetrical ‘She hit her’, adjacent, ‘next to each other’, and chains ‘follow each
other’. As might be expected given the types of distinctions made in Tsafiki, while the elicited
data did clearly demonstrate that Tsafiki distinguishes symmetrical positions from actions it
otherwise did not correlate well with the distinctions portrayed in the video clips. In
particular, the elicited data was strongly biased in favor of the reciprocal construction formed
with - ‘RR’ illustrated in (3) above.
As pointed out by Frajzyngier and Curl (1999:viii) reciprocal functions are often
coded by elements that have already become grammaticalised for other functions. This is
certainly true in Tsafiki. Reciprocal constructions are composed of elements used in other
constructions. The reciprocal constructions are not just about symmetrical or reciprocal
relations but also code other meanings that are a core part of Tsafiki grammar. The
constructions coding symmetrical positions comprise a subset of positional constructions in
Tsafiki and hence carry aspects of this system with them into the reciprocal construction.
Active reciprocal constructions are complex predicates, as are most predicates in
Tsafiki. In addition, the active reciprocal construction using the verb ‘become’ (2 above) has
a relationship to 'middle voice' constructions and to a mirative suffix which indicates the event
is non-congruent with the speaker's expectations. The relationship between mirativity and
middle voice has been noted by both Maldonado (1993) and Shibatani (2006). The Tsafiki
reciprocal construction falls somewhere in the middle of a grammaticalisation chain in which
the function and use of ‘become’ ranges from a simple verb coding a change of state to the
non-congruent verbal suffix which has lost most of its predicating functions and serves
solely as a marker of non-congruent or unexpected information. While the verb in the
reciprocal construction still functions as a predicate, semantically it codes a notion of
unexpectedness similar to that coded by the non-congruent suffix -The coding of mirativity,
or the degree to which a state or event is anticipated, is an obligatory and core aspect of
Tsafiki grammar.
After examining the formation of reciprocal constructions forming (a)symmetric
positional relationships, I will discuss reciprocal constructions coding active symmetry and
discuss their relationship with other constructions and the grammaticalisation chain
mentioned above. I will then look at the elicited data and compare this data with reciprocal
constructions found in natural discourse. However, because there is little published data
available on Barbacoan1 languages in general or Tsafiki in particular, I will begin with a brief
sketch of Tsafiki grammar.
2. AN OVERVIEW OF TSAFIKI GRAMMAR
Tsafiki exhibits a fairly consistent SOV word order and strong SOV traits including:
postpositions, possessors precede possessed, and dependents precede heads. It has three basic
patterns of case-marking for the arguments of bivalent predicates: (1) nominative/accusative
(active predicates); (2) nominative/locative (stative predicates); and (3) dative/accusative or
dative/nominative (a small set of complex predicates). Nominative case is unmarked.
Accusative postpositions exhibit a differential pattern in that the accusative marker almost
always occurs with humans or specific objects and rarely with non-humans or non-specific
objects. However Tsafiki allows rampant ellipsis and almost half of all clauses in natural
discourse occur with no expression of core arguments (Dickinson 2002). The only trace of
argument cross-referencing in the verb is a single optional marker which usually (but not
always) indicates a collective subject.
Three grammatical subsystems in Tsafiki—the complex predicate system, the verb
class system, and the mirative system—have important consequences for the encoding of
reciprocity and will now be briefly discussed.
2.1 Complex predicates, verb class system and auxiliaries
Tsafiki has a system of complex predicate formation which closely resembles those
found in northern Australian languages (Schultze-Berndt 2000; McGregor 2002). There are
only some thirty fully inflecting verbs (generic verbs). The majority of predicates are formed
by combining one of these inflecting generic verbs with an element from a large open class
(coverbs), which cannot directly take finite verbal morphology. Both elements of the complex
predicate can contribute semantic participants to the clause and affect the final valency. In the
following example (5) there are two complex predicates, one containing the coverb
‘squeeze’ and the generic verb ‘PUT’ and the second the coverb ‘breast’ and ‘EAT’.
(5) bete po ku
small-NCL food squeeze PUT-SR again=breast
fi
EAT-NMLZR.OBL BE:AUX-DCL
‘(You) have to feed babies, (by) mashing up the food.’ (19Sep0301.1664.378)
The predicate template is further complicated in that a subset of the generic verbs can
also occur as second-order semantic verb class markers or as auxiliaries. This limited set of
predicating elements can therefore occur as: (1) simple verbs (the sole predicator in a clause);
(2) generic verbs (in complex predicate constructions); (3) second-order semantic verb
classifiers; and (4) auxiliaries.
An example of the same element ‘do’ occurring as a generic verb, a verb class
marker and an auxiliary is given in (6) below. In the examples, when an element functions as
a simple verb or coverb it is glossed with lower case letters. Generic verbs are glossed with
small capital letters. Auxiliaries and verb class markers are also written in small capital letters
but are then followed by AUX or VCL respectively to indicate their function.
(6) ki ki ki
SP:tell DO-CNTR DO:VCL-INF DO:AUX-DCL 2=COM
‘(I) just want to talk with you.’ (10Oct0304S1.0009)
The generic verbs, verb class markers and auxiliaries can be clearly differentiated by
syntactic position and function. The verbal template is quite complicated with the order of the
elements varying depending on the characteristics of the individual morpheme and what other
elements are present. However, the relative order of the coverb, the generic verb, the verb
class marker, and the auxiliary does not vary regardless of what other elements might be
present (6).
The second-order semantic verb class marker system assigns each of the thirty or so
inflecting verbs to one and only one of five semantic classes exemplified in certain
morphosyntactic constructions (see Appendix A for a list of verbs and the class to which they
belong). The five semantic verb class markers (7) are identical in form and meaning to five of
the generic verbs: ‘do’; become’; ‘say/express’; ‘be in a position’; and be’. The
verb class markers follow certain verb suffixes2. When a predicate suffixed with one of these
elements is not followed by a verb class marker it forms a subordinate clause or nominal
(Dickinson, 2002). The verb class marker is necessary to form an active finite clause 3. In the
examples below the verb class markers follow the non-obligatory collective marker 4.
(7) a. ki ( , ) ‘They ate.’
b. i ( , ) ‘They went.’
c. ti ( , ) ‘They spoke.’
d. -ra ( , ) ‘They are sitting.’
e. jo ( , ) ‘They don’t exist.’
The verb class markers are redundant in that they “copy” the argument structure of the
generic verb. For example all ‘do’ class verbs take an actor as the nominative argument; all
‘become’ class verbs take an undergoer as the nominative argument; ‘express’ class verbs
take an actor nominative argument; ‘be/exist’ class verbs and ‘be in a position’ class
verbs take an entity which is in a state or location. An example is given in (8) here the coverb
‘desire’ is followed by the generic verb ‘express’ which is followed by the collective
suffix and then the verb class marker -. The final ‘express’ in the sentence is a simple
verb creating the matrix clause, ‘they said’. Hence occurs in this sentence as a generic verb,
a verb class marker and a simple verb. The next two examples illustrate that it is the generic
verb and not the coverb that determines the semantic class. ! ‘anger’ first occurs with
‘BECOME’ in (9) and then with ‘CAUSE.BECOME’ in (10). Each generic verb is followed
by a collective marker and then the verb class marker with which it invariably co-occurs (see
Appendix A). Note that there can be more than one verb class marker in a single clause as in
(10) where there are two. The verb phrases are delineated with brackets (8-10).
(8) "muna titi#
then 3=PL [also=desire SAY/EXPRESS-COL-EXPRESS:VCL-PRG-DCL]
"ti#
[express-DCL]
‘Then they said that they kissed (him) (LIT:they expressed desire for (him)).’
(06JUN0501S1.0493.166)
(9) "pa i-i-#
but 1M=ACC [anger BECOME-COL-BECOME:VCL-DCL]
‘Nonetheless they got angry at me.’ (05Dec0502.373)
(10)
3P2 two-NCL 1M=ACC
"pa suwa--ki-ki#
[anger CAUSE.BECOME-INTS-DO:VCL-COL-DO:VCL-PRG-DCL]
‘The two of them are really really bothering me.’ (29Mar9701.668)
I reserve the term auxiliary for five predicating elements that occur after verbs
suffixed with ‘imperfective participle’, ‘perfective participle’, ‘infinitive’, or
‘oblique nominalizer’. If these verb forms are not followed by an auxiliary, they function
as nominals. As with the verb class markers, the auxiliary is necessary to form an active finite
clause. However these five auxiliaries—‘do’, ‘become’, ‘be in a position’, ‘be’, and
‘not be/not exist’—differ from the semantic verb class markers in that they can occur with
any verb that carries one of the above suffixes. In (11) below the verbs are suffixed with
‘imperfective participle’ followed by the auxiliary ‘be’, regardless of the verb’s underlying
semantic class. This contrasts with the examples in (7) above, where only the verb class
marker appropriate for the class to which the verb belongs can be used.
(11) a. min jo ‘He eats.’
b. min jo ‘He goes.’
c. min jo ‘He talks.’
d. -min jo ‘He sits.’
e. -min jo ‘It doesn’t exist.’
$
% $ $
&$
'%$%(
% ) % *+,- $
(12) a. no ki ‘He wants to go.’
b. no i ‘He can/has permission to go.’
c. -no ra ‘He is ready to go.’
d. -no jo ‘He has to go.’
e. -no ito ‘He doesn’t have to go/he will never go.’
An important trait of the predicate template in Tsafiki, is that while both the coverbs
and the generic verbs can contribute semantic participants to the clause and affect valency, the
final predicating element determines which case-marking pattern will appear on core
arguments and determines which participant will receive nominative case5. However while the
verb class markers and auxiliaries control the case-marking pattern, they differ from the
generic verbs and coverbs in that they do not affect valency and carry schematic rather than
detailed semantic information concerning the event. The schematic information concerns the
type of event, i.e. an act of ‘doing’, ‘becoming’, ‘expressing’, ‘being in a position’, ‘being’ or
‘not existing’, but does not include semantic detail such as path or manner of movement, or
type of ‘doing’ or ‘being’ such as ‘seeing’, ‘eating’, ‘putting’, ‘causing’, ‘sitting’, ‘lying’ etc.
The generic verb is the strongest predicating element in the construction in that it can affect
valency, control case-marking, determine semantic class and add specific semantic detail as
well as schematic information concerning the event. The coverbs, verb class markers and
auxiliaries each have a subset of these properties, but only the generic verb has the potential to
carry them all (Table 1).
Table 1: Characteristics of Tsafiki predicators
Coverbs Simple/generic
verbs
Verb Class
Markers
Auxiliaries
Can affect valency Yes Yes No No
Can control case-
marking pattern
No Yes Yes Yes
Can determine second-
order semantic verb
class
No Yes N/A Yes
Can contribute detailed
semantic information
Yes Yes No No
Can contribute
schematic semantic
information
No Yes Yes Yes
There are two types of macro-roles that play a role in Tsafiki argument realisation.
The traditional macro-roles, actor and undergoer determined by the individual predicators,
and what I call event-roles concerning the initiation and termination of the event. Nominative
case is always assigned to a participant coded by the generic verb, with actors taking
precedence over undergoers if both are present in the semantic structure of the generic verb.
However, nominative case does not necessarily code an actor. Nominative and accusative
cases in Tsafiki are essentially markers of semantic event-roles rather than grammatical
relations. However these event-roles are tied to the semantics of the unfolding event rather
than more traditional macro-roles such as actor or undergoer. In active clauses the nominative
argument is associated with the initiating point of the event and the accusative with the focus
or endpoint of the action. Regardless of the specific semantic role of the participant in terms
of actor or undergoer, the nominative argument will represent the initiating phase of the event
as it is presented in the discourse. This will be discussed further below in Section 4.3. There is
no real passive in Tsafiki or other constructions that ‘promote’ a participant to subject or
nominative position over a participant positioned earlier on the event line (see Dickinson
2002). Voice distinctions, including reciprocal constructions, are coded by the use of different
generic verbs in the complex predicate construction.
2.1 Mirativity
A final characteristic of Tsafiki is its obligatory system of evidential and mirative
marking. Mirative coding has to do with the degree to which the information coded in the
clause is congruent with the speaker's expectations and general knowledge concerning the
event or state. Much as one could say English speakers are obsessed with espistemic modality
(certainty, doubt, probability) in that English has a large inventory of items (modals, verbs,
adverbs) and a variety of constructions for coding these distinctions, Tsafiki speakers are
obsessed with the coding of mirative distinctions. The most grammaticalised system consists
of two categories. One set of markers is used when the source of the information is a primary
participant, either socially (he is a member of the group) or physically6. A primary participant
in an event has an epistemologically privileged position with access to information
concerning the event that a non-participant does not have. Another set of markers are used
generally, i.e. either by participants or non-participants. Both categories have independent,
although notionally related, sets of markers coding mirative values. Events which are
congruent with the speaker's expectations are coded with congruent markers and events which
are not congruent with the speaker's expectations are coded with non-congruent (DeLancey
19977, 1999; Dickinson 2000). The marker that is of interest in this paper is the recently
grammaticalised, non-congruent marker used in the participant condition (Column two,
Table 2). This marker has a relationship both semantically and diachronically with the middle
voice and one of the reciprocal constructions.
,./$
General
(source of information may or may not be
a participant)
Participant
(source of information = participant)
Æ neutral
-non-congruent (mirative) -/inon-congruent
-congruent (situational) - congruent
-speaker knows/hearer may be
surprised or skeptical of information, i.e.
the “You won’t believe this but...”
morpheme
-congruent remote past, habitual
An important aspect of these markers is that they do not code person. There is a
default pattern in which the participant markers are more likely to occur with first person
subjects in both declaratives and questions and second person subjects in questions, whereas
the non-participant markers most commonly occur with third person subjects. However, a
speaker is free to declare himself either a primary participant or a general or non-participant at
any time and hence any of the markers in either the general or the participant paradigms can
occur with any person. For example, in (13) the clause has a third person subject but the
congruent participant marker is suffixed to the verb. This indicates that the speaker
considers himself a Tsachi and the information in the utterance is congruent with his
expectations and knowledge. A non-Tsachi could say this and the result would not be
ungrammatical, simply highly presumptuous since the speaker is not a Tsachi. On the other
hand in (14) the speaker uses the general congruent form ‘-SITU’ with a first person
subject in a situation where he was clearly the primary participant. This is often done when
speaking about things one did in childhood. Here the mirative stance of the adult speaker
contrasts with the mirative stance of the same speaker as a child, i.e. the child didn’t know he
couldn’t comprehend. So the speaker distances himself from his childhood perspective by
using a general form, coded with the mirative perspective appropriate for the adult. The suffix
-‘situational’ indicates that given the situation, the information coded in the proposition
is expected (Moore, 1991) , i.e. it is normal for a child to lack comprehension.
(13)
Tsachi=PL non-Tsachi=ACC marry-BECOME-NEG-IMPF.P=PL
yo
BE:AUX-CNGR-DCL
‘The Tsachila don't marry mestizos.’ (appropriate for a Tsachi to say)
(14) *-
but 1M=ACC be-SR=FOC know-BECOME-CNTR
man
know-NOT.BE-PF.P BECOME:AUX-SITU-DUB
‘Being a child of course, (I) just couldn't comprehend.’ (what it meant
to study shamanism). (10Oct0303S1.1449.39)
The interpretation of the mirative markers can differ depending on the semantics of
the verb and the context. The difference between congruent and non-congruent forms can
indicate accidental actions with highly transitive verbs, i.e. verbs which code actions with a
volitional agent and an affected patient (15). However this is not the case with other types of
verbs where the difference can indicate the degree of awareness or consciousness on the part
of the participant (16), or simply a discrepancy between what the speaker thinks and what the
situation indicates (17).
(15) a. yo
1F 3P2=ACC hit-CNGR-DCL
‘(I) hit (him).’ (on purpose)
b. i
1F 3P2=ACC hit-NCNGR-DCL
‘(I) hit (him).’ (accidentally)
(16) a. yo
1Fhit send-RR-BECOME-CNGR-DCL
‘(I) fell.’ (expected—the ground is slippery)
b. i
1Fhit send-RR-BECOME-NCNGR-DCL
‘(I) fell.’ (unexpectedly—for no obvious reason)
(17) a. yo
heart-NCL NOT.BE-CNGR-DCL
‘(I) am stupid.’ (and I know it)
b. i
heart-NCL heart-NCL NOT.BE-NCNGR-DCL
‘(I) must be stupid.’ (although I didn’t think I was)
Mirative coding is complex in that it has to do with each speaker’s expectations and
knowledge not only concerning the event but also the participants, his relationship to the
event, the context and the type of event and state.
The mirative, evidential and verb class markers usually do not occur with non-finite
clauses. Tsafiki has a rich inventory of clause types including: complement, purpose clauses,
serial verb constructions, conditional clauses, associated motion clauses, secondary
predicates, switch reference clauses, counter-factual and several different kinds of adverbial
clauses, all of which can be identified by their lack of finite verb morphology.
3. SYMMETRICAL POSITIONS
Tsafiki has a huge number of positional coverbs that are similar to positionals in
Mayan languages (Haviland 1992; Grinevald 2006). The positionals in Tsafiki code
configuration and position in space as well as giving other kinds of information such as shape
and texture. While the positional coverbs can code configurations concerning a single object
and asymmetrical relationships, some of the coverbs that code position are inherently
symmetrical. In general, Tsafiki uses positional coverbs for the conditions which Darlymple
et. al. (1998) call intermediate reciprocity, (e.g. ‘alongside’) and intermediate alternative
reciprocity (e.g. ‘on top of each other’) and inclusive alternate ordering (e.g. ‘stacked like
sardines in a can’).Tsafiki has a large number of positional coverbs which denote different
configurations in space to which these ‘reciprocal’ coverbs belong. A few are listed below:
(18) Positional coverbs
‘two objects side by side’ ‘stand close together’
‘parellel lengths together w.
space between’
‘stack 3-dim. object’
‘lengths together touching’ ‘stack flat object’
‘close together, dense’ ‘randomly mixed’
‘string-like objects tangled up’ ‘spread apart’
‘separated’ ‘points together’
‘flat surfaces put together
vertically’
‘together in bunch’
In the data elicited with video clips, two situations were coded with ‘lean
against each other (rigid objects)’. In both situations the objects were supporting each other
(19). The phrase is repeated twice to indicate that there are two pairs of sticks. In a situation in
which only one of the objects is supporting the other either ‘lean’ or ‘prop up’
would be used (20).
(19)
now 3D1 late=in/on=in.region 3P2 four tree:NCL=ACC
tulipa-le tulipa-le
lean-SUF lean-SUF stand CAUSE-PF.P=PL BE:AUX-DCL
‘Now later, they have stood the four sticks up leaning against each other,
leaning against each other.’ (F #34)
(20) tili 0
3P1=ACC tree=INSTR prop.up DO-IMP
‘Prop this (leaning fence) up with a board!’
Other positional coverbs can also be used to code notions such as ‘lead’, or ‘follow’
when they occur in conjunction with motion verbs:
(21) kake
SP:corridor=through one-NCL front spring.up GO-PRG-IMPF.P=ACC
bene
behind=in.region foot=in.region get-ASSOC.M
bene
behind have-ST-GO-PRG-DCL
‘He chases from behind grabbing at the other one who is running through the
corridor in front (of him).’ (M #64)
A common strategy for stative relationships is to use a nominalised positional coverb
and the collective But let’s begin by considering its basic use. When combined with a
nominal or demonstrative pronoun, clearly has a collective sense as in (22).
(22) =tala
all house=ACC do-PF.P-IMPF.P-DCL 3P2=PL=COL
‘They had all been building the house together.’ (15Oct0303.1270.291)
The collective = can combine with nominalised positional coverbs, nouns or
nominalised verbs. Its basic function appears to be to group a set of entities or actions
together. The relationship between the grouped entities or the kind of entity grouped is then
elaborated by the element to which the collective encliticises. In (23) and (24) the enclitic
occurs with a nominalised positional coverb to indicate positional symmetry.
(23) =tala bete-n-de=tala
3P2=PL=COL behind-ST-long.rigid:NCL=COL
sit-COL-BE.POSITION:VCL-DCL
‘They sit together back to back.’ (V#12)
(24) kake-n=tala
two=NCL woman=PL two-NCL man=PL front-ST=COL
stand-BE.POSITION-COL-BE.POSITION:VCL-DCL
‘Two men and two women are standing together face to face.’ (F#44)
When occurs with a nouns coding kinship, it can indicate a symmetrical dyad
relationship (Evans 2006). I have no evidence that it can be used to indicate asymmetrical
dyad relationships. Again it groups the entities and then delineates the kind of relationship, in
this case one of brotherhood (25).
(25) tala
brother.of.male=COL be-IMPF.P-RP-DCL
‘(They) say (they) are brothers.’ (CA.1549.338)
3.1 Elicited positional reciprocal constructions
As noted above it is the specific coverb rather than the construction itself that encodes
symmetry vs. asymmetry. In the data elicited with the video clips described in Chapter Two,
the same construction was used for both asymmetrical and symmetrical positional
relationships. The speakers used the positional construction in the clips which demonstrated
either positions such as ‘lean against each other’ or ‘sit next to each other’ (clips 8, 12, 15, 18,
35 and 53) or notions such as ‘follow’ or ‘chase’ (clips 14, 24, 27, 36, 39 and 64). This
positional construction was used regardless of whether the clip concerned simultaneous vs.
sequential actions, or depicted strong, chaining, adjacent, asymmetrical or pair relationships.
The elicited data clearly demonstrates that Tsafiki uses this construction for (a)symmetrical
positions as opposed to the construction used for symmetrical actions which will be discussed
in the following section.
4. ACTIVE RECIPROCAL CONSTRUCTIONS
In the following section I describe Tsafiki active reciprocal constructions in general,
including the case-marking of reciprocant arguments. I will then look at other constructions in
which the constituent elements of the reciprocal constructions occur, namely, ‘intensive’,
‘collective’, ‘become’ and ‘RR’ occur. These constructions will then be compared with
the reciprocal constructions.
4.1 General characteristics of active reciprocals
As noted above Tsafiki has two basic complex predicate constructions that code
symmetrical actions. One construction is based on ‘become’ and the other on ‘RR’. These
can be further divided into two categories depending on whether the construction is simple or
utilizes the intensifying suffix .
Table 3: Four predicate reciprocal constructions
Simple reciprocal Reciprocal with intensifier
i 'become' (coverb) (verb)-BECOME-COL-
BECOME:VCL
(coverb) (verb)-INTS-BECOME-
COL-BECOME:VCL
-a ' RFLX/RCP'(coverb) (verb)-RR-COL-DO:VCL (coverb) (verb)-INTS-RR-COL-
DO:VCL
The simple reciprocal construction only occurs with a limited set of predicates. Many,
but not all, of these predicates fall into the category of what have been called ‘natural
reciprocals’. Kemmer (1993) used this term to semantically delineate events which are either
necessarily symmetrical (‘meet’) or else very often code a symmetrical relationship (‘hug’,
‘compete’).
(26) amani-la-i
three-NCL non.Tsachi =PL hug-BECOME-COL-BECOME:VCL-PRG-DCL
‘Three gringos are hugging each other.’ (Maria: #20)
(27)
man =PL=COM snake woman=ACC
isa-ta-a-la-ki
defend-HAVE-RFLX/RCP-COL-DO:VCL-PRG-SITU-RP-DCL
‘(They) say the men were competing with each other to get the snake
woman.’ (19Mar9701S3.042).
A list of predicates which have been found to occur with the simple reciprocal
construction, including some Spanish borrowings, is given below (28). Note that some of
these predicates can occur with either ‘BECOME’ or ‘RR’. The meaning differences
between the two forms will be discussed further below.
(28) a. Simple reciprocal predicates with 'BECOME'
‘gather’
‘group’
‘mix’
‘hug’
‘touch/have’
‘desire’
‘have sex’
‘marry’ (SP)
‘meet’
‘greet’ (SP)
‘tell/converse’
‘defend’
‘negotiate’ (SP)
‘exchange’
b. Simple reciprocal predicates with - 'RR'
‘kill’
‘defend jealously’
‘divide’
‘ask’
‘ask a favor’
c. Simple reciprocal predicates with either 'BECOME' or - 'RR'
‘play’
‘crash into
‘argue’
‘beg’ (SP)
The second reciprocal construction, formed with the intensifier - and either
‘BECOME’ or ‘RR’ is much more productive and can occur with any predicate, provided
there is some kind of symmetrical relationship between participants. This construction is not
dependent on the transitivity of the predicate or the grammatical roles of the reciprocants. The
predicate can be intransitive as in (29), where the symmetry is between possessed locations. It
can indicate symmetry between possessed objects as in (30), benefactive symmetry as in (31),
or simply associative symmetry as in (32) where the participants simply grow up together.
Both the reciprocals formed with 'RR' and 'BECOME'conform to this pattern with the
difference between the two being based on the degree to which the event was spontaneous and
expected. In (29a) and (30a) below, the utterance would be appropriate in a situation in which
the reciprocants were family members or friends. (29b) and (30b) would be used in a situation
in which the reciprocants were strangers or known to dislike each other. In the latter case the
reciprocity is unanticipated.
(29) a. ji-ka-a
3P2=PL house=LOC go-INTS-RECIP/REFLX-COL-DO:VCL-DCL
‘They went to/visited each other's houses.’ (expected)
b. ji-ka-i
3P2=PL house=LOC go-INTS-BECOME-COL-BECOME:VCL-DCL
‘They went to/visited each other's houses.’ (unexpected)
(30) a. kira-ka-a
3P2=PL small=PL=ACC see-INTS-RECIP/REFLEX-COL-DO:VCL-
DCL
‘They cared for each other's children.’ (expected)
b. kira-ka-i
3P2=PL small=PL=ACC see-INTS-BECOME-COL-BECOME:VCL-DCL
‘They cared for each other's children.’ (unexpected)
(31)
not.be-INCL then plant DO-ASSOC.Mplant
ji-ka-a
DO-ASSOC.Mgo-INTS-RR-INF-DCL
‘Not having this, (we) would have to go around planting and planting (it) for
each other.’ (09Oct0301S1.1123)
(32)
SMBL=DO:VCL again=INCL 1F=PL one SP:pair-ST
owa-i-ka-i
like grow-BECOME-INTS-BECOME-COL-BECOME:VCL-CNGR-DCL
‘So, on the other hand, (it's) like we grew up together with each other (despite
the fact we are not related by blood).’ (09Jun0501S1.675.735)
1++ 2
The subject of the reciprocal construction can be coded as a simple plural as in
examples (29, 30) above, or each reciprocant can occur separately marked with the enclitic
comitative = A nominal marked with shares its semantic role—actor, undergoer,
locative, dative, etc.—with another participant. In all constructions when two nominals are
both coded with the enclitic the interpretation is that both arguments share the same
semantic role and are equal in terms of prominence(33).
(33) /=be 3=be
now Martin=COM Samuel Loche=COM
speak-INTS-RR-COL-DO:VCL-EV-DCL
‘Now, Martin and Samuel Loche must have argued with each other.’
(27Oct0304.0621.971)
)$$
$*4-
$=be.
(34) Tsachi=la Dobe=be
person=PL Dobe=COM hit-INTS-BECOME-IMPF.P
BE:AUX-ST=PL-SITU-RP-DCL
‘The Tsachila used to fight with the Dobes (another ethnic group).’
(16Nov0302.2467.021)
However the most common pattern in Tsafiki is to simply elide both arguments or just
the most prominent argument and only the participant marked with remains. Clauses in
which both arguments are individually expressed are fairly rare in the texts and because
Tsafiki allows rampant ellipsis the more prominent argument is usually elided. In (35) I have
given a second, more literal translation, awkward in English, but a closer representation of the
Tsafiki construction.
(35) be
woman=COM argue-RR-COL-DO:VCL-EV-RP-DCL
‘(He) must have argued with (his) wife.’/ ‘(He) with his wife must have argued
with each other.’ (26Aug9601S6.093)
4.2 The collective -la and the intensifier -ka
As noted above all the elements of the reciprocal construction can occur in other
constructions without coding symmetry. The collective can occur in its own to indicate group
participation as seen in example (5) above. The intensifier 56also occurs in other
constructions. It commonly occurs followed by a verb class marker as in (36) below. The -
in this construction can function as a quantifier over the temporal or path measures or to
delineate the quantity of the undergoer participant in a transitive construction (36a). Or it can
intensify the manner in which the event is performed.
(36) a. ka-ki ‘He ate a lot/for a long time/wolfed down his food.’
b. ka-i ‘She ran hard/for a long time.’
c. ka-ti ‘He cried hard/loud/for a long time.’
d. ka-jo ‘She is really sad.’
e. ka-ra ‘It is really stretched tight/long/hard.’
The predicate collective suffix cross-references the most prominent argument,
normally the subject (Dickinson 2002). It could be that in the reciprocal construction the
intensifier serves to quantify the event itself or the object as in the intensifying construction.
Either would be plausible in that the reciprocal codes the participation of both, i.e. ‘he hit her
and she hit him= they hit each other’. If the intensifier does quantify over the object it means
that in the reciprocal construction both the subject and the object are quantified. However this
issue needs to be clarified with further research.
4.3 The reflexive nature of -a 'RR'
The ‘RR’ is no longer an extant free verb in the language but it still does have
predicating properties in that like the generic verbs it determines the semantic class of the
predicate. For example, the normal verb class marker for the verb ‘express’ is -
‘EXPRESS:VCL’ (37a). However when the verb is followed by - the verb class marker is
‘DO:VCL’ (37b). It is reasonable to assume that like all other -class verbs, codes an actor
participant.
(37) a. ti-ti ‘They spoke.’
b. -a--ki ‘They spoke to each other.’
The ‘RR’ also occurs in other constructions. However it is no longer a highly
productive element. There is one coverb ‘step’ which commonly combines with the
generic verb ‘PUT’ to depict an event of stepping on something (38). However when
occurs with the generic verb ‘DO’, the suffix occurs after the coverb ‘step’ to depict
‘dance’. It seems to serve a reflexive purpose here such as ‘to step oneself’9 (39). The
reciprocal form of this verb occurs with two instances of the generic verb ‘RR’ (40).
(38) tere po
3P2=FOC water=LOC step PUT-SR=FOC
water BECOME-INF BE:AUX-RP-DCL-RP-DCL
‘(They) say (they) said if she stepped in the water, she would dissolve.’
(39) tere-a-ki 7
thunder step-RR-DO-SITU BE:AUX-INT
‘Does the thunder dance?’ (11Dec0303.0480)
(40) tere-a-ka-a-la-ki
3P2=PL step-RR-INTS-RR-COL-DO:VCL-DCL
‘They danced with each other.’
Another construction in which the 8appears is formed with a bivalent coverb coding
two participants, an actor and an undergoer. In these constructions the bivalent coverb is
suffixed with ‘RR’ and followed by ‘BECOME’ to form a permissive construction. As in
other Tsafiki constructions the final predicating element determines which participant
receives nominative case and nominative case with active verbs always indicates the initiating
point of the action. In (41) ‘peel’ occurs in an ordinary transitive construction. In (42)
‘peel’ occurs in the permissive construction. With an inanimate subject the interpretation
of the permissive construction is that there is a quality inherent in the subject/undergoer that
permits the action to be carried out. This construction can optionally contain an agentive
participant cliticised with the comitative The comitative is used here because even
though the participants do not have the exact same semantic role, they are both agents of the
event. The undergoer has a double role in that it denotes both the entity that instigates the
action and the one that undergoes the change of state. This sense is clearer when the
undergoer is animate (43). The undergoer is in the nominative, (determined by the final
predicating element, in this case ‘BECOME’), and is interpreted as the initiator of the event or
the one permitting the agent to carry out the action (see Dickinson 2002).
(41) *- lachi
(3P2=PL) skin all peel-COL-DO:VCL-DCL
‘(They) peel all the bark off.’ (14Dec0302S2.191)
(42) *be- lachi-a-i
(3P2=PL=COM) fruit quickly peel-RR-BECOME-NEG-RP-DCL
‘(They) say this fruit doesn't peel (easily) (with/for them).’
(09Jun0501S2.0338.377)
(43) / po-re-a-i
Maria Secundina=COM hair cut-CAUSE-RR-BECOME-DCL
‘Maria let/had Secundina cut (her) hair.’
The details of this construction are too complex to be presented here (see Dickinson
2002). What is of primary interest is that in these constructions appears to have a reflexive
reading, indicating that the actor and undergoer of the coverb refer to the same entity.
However because is no longer highly productive nor an extant free verb in the language it
is difficult to ascertain its exact meaning.
4.4 The function of i 'become'
In this section, I will look at the range of uses of the verb ‘become’, from its use as a
simple verb to its most grammaticalised use as the non-congruent marker The verb
'become' is of course, the quintessential change of state verb and when used as a simple
predicate in Tsafiki simply codes the change of state:
(44)
Roberto evil-ST become-DCL
‘Roberto turned into/became a devil.’
As a verb class marker it classifies primarily motion verbs, but also ‘die’,
‘get lost’, ‘sleep’ as well as itself (see Appendix A). In contrast to ‘do’ class verbs,
the subject of all these verbs undergoes a change of state or location. As a generic verb
‘BECOME’ commonly occurs in complex predicate constructions which code situations which
are often called ‘middle voice’. Middle voice has been described semantically in a variety of
ways including: reflecting a reduction in the elaboration of the event (Kemmer 1993); or as
depicting a situation where the development of an action is contained within the personal
sphere of the agent such that he himself is affected by the action (Shibatani 2006). Both
Maldonado (1993) and Shibatani (2006) analyse voice contrasts as concerning the evolving
properties of actions as they are portrayed in discourse. Middle voice focuses on the change of
state phase of an evolving event. In Tsafiki this function of middle voice is somewhat
transparent since the common pattern in Tsafiki is to form it with an class verb, most
commonly the generic verb ‘BECOME’ itself or the -class generic verb ‘GO’. The
corresponding causative construction is formed with a -class verb (see Appendix A). The
subject of a -class verb is always an actor or an agent.
Some of the types of events described with this construction are: inchoative
positionals (45a); spontaneous events (45b); change-of-state (45c); cognitive/emotional (45d);
and translational motion (45e).
(45) coverb + i-class generic verb coverb + ki-class generic verb
a. ‘stand up’ ‘make sm. stand up’
b. ‘get hot’ ‘heat sm. ’
c. ‘snap’ ‘snap sm. ’
d. ‘get sad’ ‘sadden’
e. ‘go running’ ‘send running’
While a detailed description of Tsafiki complex predicates would lead us far astray
here (see Dickinson 2002), it should be noted that in the majority of the middle voice
constructions, the -class generic verb codes a single participant that undergoes a change of
location or state. The coverb also codes a single undergoer participant. The two semantic
participants converge on the same nominative argument when combined in the complex
predicate construction. Nominative case in Tsafiki is determined by the generic verb.
Therefore the event described by the middle voice construction begins with the undergoer
who is associated with the change of state phase of the event. Whether there was a causal
entity or event is left unspecified in the construction itself. The change of state is the
prominent phase of the event here, not the causal phase. In contrast, in the causative
construction the generic -class verb codes a single actor participant which receives
nominative case while the undergoer participant of the coverb receives accusative case. The
description of the event begins with the actor entity that is associated with the causal phase of
the event (Dickinson 2002).
Both Maldonado (1993) and Shibatani (2006) have noted that middle voice
constructions often can be used to indicate that the event was unexpected, surprising or that
the actor is not in control. An event beginning with the causal phase as opposed to one that
begins with the change of state may naturally be less startling. If the causing sequence is
expressed, the consequential change of state may be expected. While the above constructions
do not explicitly code the contrast between expected vs. unexpected events, because the
causal phase of the event is not explicitly portrayed, these constructions can indicate
spontaneous or sudden change which can imply unexpected change.
11+ &
Non-translational motion verbs follow the same general pattern, but with an important
difference. While in (45) above the difference between the -class and -class constructions
often involves an increase in transitivity—intransitive > transitive, or transitive > ditransitive
—with the non-translational motion construction there is no change in transitivity, both
constructions are intransitive. The notions conveyed by these constructions move closer to
that of mirativity distinctions. The 'BECOME' construction carries notions such as
uncontrolled or unexpected motion. The uncontrolled sense is the more likely interpretation in
the participant situation and the unexpected sense the more likely in the non-participant
situation. The ‘DO’ construction simply codes an activity (46).
(46) coverb + i-class generic verb coverb + ki-class generic verb
‘writhe in an unexpected
or uncontrolled manner’
‘writhe/wiggle’
‘move hips in an
unexpected or
uncontrolled manner’
‘move hips’
!!! ‘blink eyes in
uncontrolled manner
‘blink eyes
‘have diarrhea’ ‘defecate’
In these constructions the generic verb determines whether the subject will be
interpreted as an actor or an undergoer. When these coverbs combine with 'become', which
has a single undergoer participant, there are two changes in comparison with the ‘do’
construction: (1) the focus is on a change from a state of not moving to one of moving—it
depicts a change of state rather than an activity; (2) the participant is coded as an undergoer
rather than an actor.
In a non-participant situation the difference between the two constructions concerns
the degree to which the event is anticipated. In (47) below a moth has flown into a flame and
as a consequence is writhing. The writhing is a predictable, expected consequence given the
circumstances. (48) depicts a situation where someone suddenly just falls down and starts
writhing for unknown reasons—perhaps due to an epileptic attack or spirit possession. The
'become' construction is used here. If the construction were used, it would imply there was
a predictable cause as in (47).
(47) soko ke
Fire=in/on go.in-PF.P BE:AUX-ST writhe DO:VCL-DCL
‘(The moth) that entered the flame writhed (of course).’
(15Oct0304.0359.865)
(48) soko
then hit-send-RR-BECOME-PRG-SR writhe
i-
BECOME-PRG-DCL say-DCL
‘Then (they) say falling down, (he) was writhing (suddenly, unexpectedly and
for an unknown reason).’ (17Nov0303.0967.35)
Both forms can occur with the grammaticalised mirative markers in the participant
situation. Here the distinction between the two constructions often has to do with control and
awareness. With a -class generic verb and the congruent suffix the construction indicates
the speaker was deliberately performing the motion (49), i.e. she is something. The
congruent suffix with the -class generic verb indicates the speaker cannot control the
movement, i.e. something is to her, but she is fully conscious and the event makes
sense given the circumstances—it is normal to blink when dust enters the eyes (50).
(49) yo
1Fman=ACC desire BECOME-SR very blink.eye DO-CNGR-DCL
‘Desiring the man, I blinked my eyes (fluttered my eyelashes at him).’
(50) yo
1Fearth speck because blink.eye BECOME-CNGR-DCL
‘Because of the dust, I blinked my eyes.’
The use of the non-congruent mirative suffix with the non-translational constructions
indicates in both cases that the participant (usually the speaker) was unconscious during the
event. Unlike a normal activity, he was not aware of his movement while it was occurring. He
knows he performed the action because someone else informed him or because of some other
kind of evidence, i.e. in (51) perhaps the bedclothes are completely mussed up and he knows
he spent a restless night full of bad dreams. With the -class form (51) he is more likely to
have some idea as what caused the action although he was unaware at the time. The -class
form indicates he was not only unconscious, he also has no idea what caused the motion (52).
As in (48) this would be appropriately said by the victim of an epileptic attack or spirit
possession.
(51) i
nightmare=INSTR=INCL writhe DO-NCNGR-DCL
‘Because of a nightmare, (I) must have writhed.’ (while I was sleeping)
(52) i
writhe BECOME-NCNGR-DCL
‘I must have writhed.’ (for some unknown reason while I was unconscious)
11, &%i
Finally, as noted above the non-congruent mirative marker used in the participant
situation is semantically and diachronically related to 'become'. As noted above this is used
when the source of the information is a participant in the event. The congruent form is
used when the information is congruent with the speaker's expectations; 9 is used when
the information is not. The older10 non-congruent participant marker is still used in the
less common, more conservative dialect of Tsafiki. In this dialect, usually found with older
speakers, the -occurs after the non-congruent -in certain constructions (53). In the more
common dialect, just the - occurs, although the trace of - can still be found in the lowering
of the preceding vowel (54). Tsafiki exhibits some vowel harmony, especially in rapid speech,
in which a preceding vowel can either raise or lower to conform more closely to the
characteristics of the following vowel—a:o; i:e. This would not occur
with a vowel sequence such as but is predictable with the sequence
(53) Conservative dialect
yai7
ya-i
SP:glass crack-SUF DO-COL-DO:VCL-NCNGR-NCNGR-INT
‘Did (we) accidentally break the glass?’ (Lit: ‘Did we become unexpectedly
breaking the glass?’)
(54) More common dialect
ei7
-(ya)-i
SP:glass crack-SUF DO-COL-DO:VCL-(NCNGR)-NCNGR-INT
‘Did (we) accidentally break the glass?’
Note that the - occurs in the position for finite morphology in Tsafiki and in most
constructions it is not followed by a nominaliser or verb class marker. There is however still
one construction where it can be seen that this - is derived from the verb ‘become’. This
construction uses the suffix -‘be in a state’ followed by the appropriate verb class marker for
the generic verb. This is an evidential construction indicating that the speaker is deducing that
the event occurred from his general knowledge. In (55), with a third person subject in a non-
participant situation the generic verb ‘DO’ is followed by the then the expected verb
class marker
(55) kike
kiki
SP:glass crack-SUF DO-ST-DO:VCL-DCL
‘(He) must have broken the glass.’s
However in (56) the generic verb is followed by , then the nominaliser and then the
verb class marker -appropriate for the verb ‘become’ but not ‘do’. While the in this
construction still carries the verbal property of controlling the class of the predicate, unlike the
middle voice and non-translational motion constructions above and like the non-congruent -
suffixit can only indicate a situation in which the speaker was a participant. Out of context,
the default interpretation is a first person subject and not third person (56).
(56) ii
*-i-i
SP:glass crack-SUF DO-NCNGR-NCNGR-ST-BECOME:VCL-DCL
‘(I/*he) must have broken the glass.’
Finally, in the conservative dialect the non-congruent - is not followed by - when it
occurs with the evidential suffix . This suffix indicates the speaker is inferring that the
event occurred from direct, usually physical, evidence (57). In the more common dialect there
are two strategies. Some speakers simply drop the which again leaves a trace in the
lowering of a preceding high vowel (58). However other speakers are now replacing the
with (59).This use of in a context where it did not previously occur provides fairly strong
evidence that speakers are now interpreting it as the non-congruent marker and not the verb
‘become’. All three constructions carry the same meaning and can only be used in participant
situations.
(57) yanu
ya-nu
SP:glass crack-SUF DO-NCNGR-EV-DCL
‘(I) must have broken the glass.’
(58) nu
-(ya)-nu
SP:glass crack-SUF DO-(NCNGR)-EV-DCL
‘(I) must have broken the glass
(59) inu
-(ya)-i-nu
SP:glass crack-SUF DO-NCNGR-EV-DCL
‘(I) must have broken the glass.’
Semantically the use of both the - and the - may have become redundant in that
both can indicate non-congruency. We have already seen that the can carry these notions,
particularly when it occurs as one half of a contrasting set of constructions with the same
transitivity values coding the same event. The construction with -contrasts with those in
which the congruent appears. (can be preceded by the verb ‘become’ but itis never
followed by )So speakers could drop one of the elements without losing the basic meaning
distinction.
5. COMPARISON OF RECIPROCAL CONSTRUCTIONS WITH OTHER
RELATED CONSTRUCTIONS
In the previous section it was shown that the verb ‘become’ ranges in use from a
simple change-of-state verb to a non-congruent suffix used in participant situations. As a
simple verb or when it occurs in middle voice constructions it does not necessarily carry
mirative notions such as unexpectedness, non-control or surprise. It only begins to take on
mirative notions when it occurs as a member of a contrasting set as with the non-translational
motion constructions described above. Returning to the reciprocal constructions it can be seen
that the same general principles hold. In situations in which either form could be used, the
‘BECOME’ reciprocal and the ‘RFLX/RCP’ construction have meaning differences similar to
the non-translational motion constructions. The form often indicates non-control or
unexpectedness. In (60) the two participants suddenly and unexpectedly begin hitting each
other, hence the form is used in the first mention of the fight. Once the fight is established
the form is used. In (61) the arguing is expected given the speaker’s evaluation of the
mental aptitude of the participants. However in (62) the arguing is unexpected in that brothers
should get along (or at least not publicly display their disagreements according to Tsachi
culture).
(60) -ka-i-
SMBL-SAY/EXPRESS:VCL-PRG-IMPF.P=CNTR suddenly hit-INTS-BECOME-SR
ka-a =
hit-INTS-RR-COL-DO:VCL-RP-DCL-RP-DCL Salun jaguar=COM
‘While (they) were talking, suddenly hitting each other, Salun and the jaguar
were fighting each other.’ (26Aug9601.186)
(61)
first=on/at.region=INCL 3D2=SMBL heart-NCL NOT.BE-ST
pata-a-
because.of argue-RR-IMPF.P=PL BE:AUX:DCL
‘From the very beginning (they) were stupid, so (they) argued (as stupid people
will).’
(62) pata-i
brother=COL argue-BECOME-PRG-IMPF.P=PL BE:AUX
‘The brothers are always arguing with each other (being brothers they should
get along).’ (26Nov9601S4.054)
However, this mirative like contrast only occurs when there are two contrasting
constructions. In the simple reciprocal constructions, formed without the intensifier 8 some
of the coverbs only occur in the ‘BECOME’ reciprocal construction. There is no constrastive
set. These coverbs all belong to the class of what could be called ‘natural reciprocals’—
'hug', 'exchange', 'meet' or the Spanish borrowing 'greet'. In these cases,
the ‘BECOME’ simple reciprocal construction does not necessarily carry the mirative notion
of unexpectedness or non-control. In (63) taken from the elicited data there is nothing
surprising about the participants greeting each other as they meet. These constructions, with
this specific class of coverbs, are closer to the middle voice constructions which do not
necessarily carry mirative notions.
(63)
then 3P2=PL=COL greet-BECOME-COL-BECOME:VCL-DCL
‘Then they greet each other.’ (FA #5)
There is an important difference between the mirative contrasts present in the
reciprocal constructions and the more grammaticalised mirative markers. The
grammaticalised mirative markers consist of two sets that distinguish between participant and
general situations. All the reciprocal examples we have seen so far are general situations, but
these constructions can also be used in participant situations. Either the congruent or the
non-congruent - can be used with some subtle differences in meaning. In (64) the -
reciprocal with the non-congruent marker indicates that the speaker, in the dark, mistook
Mayra for his wife and deliberately went up to her and hugged her. His hugging of Mayra,
rather than his wife is accidental. The non-congruent suffix with the reciprocal, on the other
hand, indicates that the speaker has no idea what happened. Stumbling around in the dark he
found himself unexpectedly hugging a woman he thinks, but he’s not sure, was Mayra (65).
(Note: ‘hug’ only occurs in the simple reciprocal construction with ‘become’, but it
can occur with either reciprocal form in the reciprocal with intensifier construction).
(64) /
1M=COM Mayra=COM very dark BE-DR
ai
hug-INTS-RFLX/RCP-COL-DO:VCL-NCNGR-DCL
‘It being really dark, Mayra and I accidentally hugged each other.’ (I thought
she was my wife).
(65) /
1M=COM Mayra=COM very dark BE-DR
ii
hug-INTS-BECOME-COL-BECOME:VCL-NCNGR-DCL
‘It being really dark, Mayra (I think) and I suddenly and unexpectedly found
ourselves hugging each other.’
The congruent suffix indicates that in both reciprocal constructions the speaker
was aware and conscious that he was hugging Mayra and not his wife. The difference is that
in the reciprocal construction the speaker was in control and acted deliberately (66). In the
reciprocal construction, although he is perfectly aware of what he is doing, he is not in
control. He is simply overwhelmed by the urge to hug Mayra (67). (Note that if the speaker
utters either of the following examples within hearing range of his wife, he may have some
explaining to do).
(66) /
1M=COM Mayra=COM very dark BE-DR
ayo
hug-INTS-RFLX/RCP-COL-DO:VCL-CNGR-DCL
‘It being really dark, Mayra and I hugged each other (deliberately).’
(67) /
1M=COM Mayra=COM very dark BE-DR
iyo
hug-INTS-BECOME-COL-BECOME:VCL-CNGR-DCL
‘It being really dark, Mayra and I spontaneously hugged each other.’ (we
couldn’t control ourselves).
So in summary, the reciprocal constructions semantically resemble the non-
translational motion constructions in that the contrast in meaning involves mirative notions
concerning control and expectedness. However a small set in the simple reciprocal
construction pattern like the middle voice constructions. The key appears to be whether there
is a contrastive set both coding the same basic event with no change in transitivity. While the
reciprocal constructions exhibit mirative like notions, unlike the more grammaticalised
mirative markers they can be used in either the participant or the non-participant situation.
6. ELICITED DATA
Data used for this chapter was elicited using the video clips outlined in Chapter Two.
Four speakers, two men and two women in their thirties participated in the study. All are
bilingual in Spanish and Tsafiki, with Tsafiki being their first, dominant and preferred
language. All four speakers watched all sixty-four of the video clips. There was a great deal of
variety between speakers. The men responded with reciprocals slightly over half the time
(56% and 54%). The two women responded with reciprocals about a third of the time (34%
and 28%). Hence in order to say anything conclusive, it would be necessary to elicit with
more speakers.
As previously noted, the data did clearly show a difference between symmetrical
positionals and symmetrical actions. In all the clips concerning positional relationships such
as ‘next to’ or ‘lean against’ or in coding scenes of ‘chasing’ or ‘following’, the speakers used
a positional coverb and/or the collective enclitic regardless of the specific condition
associated with the clip.
Leaving aside the video clips that were portrayed using positionals, the main factor
that conditioned the use of a reciprocal construction appeared to be whether the condition
included adjacent participants, paired participants, strong symmetry, a chain, a melee or
asymmetrical actions (Table 4).
The only clips in which no reciprocal was used by any speaker at any time and where
speakers did not accept reciprocal constructions as appropriate descriptions, all concerned
asymmetrical sequential conditions. The asymmetrical sequential video clips essentially
portrayed situations in which there was no symmetry or reciprocity: a woman sits on a chair
while another searches her head for lice (#51); someone is hugged without hugging back (#4,
#16); someone looks at someone without the other looking back (#25, #31). The single clip in
the asymmetrical sequential class in which a reciprocal construction used was (#1) which
portrayed one man speaking to another. However the speaker used the complex predicate
which means converse rather than ‘say/express’ or ‘speak’. The sense may
have been that this was a reciprocal communicative event even if only one man was speaking.
1.&$$
Condition Reciprocal Reciprocal Total
construction
used
construction not
used
# % # % # %
Adjacent/simultaneous 04 1.00 0 0 04 .02
Pair/simultaneous 15 .94 01 .06 16 .09
Pair/sequential 07 .88 01 .12 08 .05
Strong/simultaneous 22 .81 05 .19 27 .15
Strong/sequential 23 .55 19 .45 42 .24
Chain/sequential 08 .35 15 .65 23 .13
Melee/sequential 04 .27 11 .73 15 .08
Asymmetrical/sequential 01 .02 42 .98 43 .24
Total 84 .47 94 .53 178 1.0
The other conditions in which reciprocal conditions did not tend to occur were the
melee sequential conditions. In these situations not every participant has a direct reciprocal
relationship with each of the other participants. For example, (#44) concerns four people who
stand and hit each other sequentially, but not every person hits every other person. The
speakers tended to describe these in chained sentences describing each participant’s individual
actions or simply describe the situation with a simple collective, ‘(they) are hitting.’
However unlike the asymmetrical sequential condition, I was able to elicit reciprocals with
these constructions after the fact.
There was also a tendency for simultaneous conditions to receive reciprocal coding
over sequential actions. In both the pair and strong conditions the simultaneous condition was
more likely to receive reciprocal coding. This is probably due to the fact that the speakers
were describing the video clips on-line, i.e. describing the event as it unfolded—"First he hits
her. Then she hits him back.” After the fact they were quite happy to accept reciprocals as a
reasonable description for these video clips.
In general, the speakers showed a fair amount of variety in describing the clips. All of
the following examples were elicited from the same video clip (#30) in which three pairs of
people sequentially exchange items. The first speaker seems to focus on the sequential aspect
rather than the reciprocity. He uses two instances of ‘one’ followed by the dative enclitic
to code the sequentiality and a non-reciprocal group verb form. The reciprocity is
inferred (68). The second speaker ignores the sequentiality and uses the collective and a
reciprocal verb form (69). The third speaker uses a reciprocal verb form to code the
reciprocity and then two instances of ‘one’ with the dative enclitic to code the
sequentiality (70). The fourth speaker makes a finer distinction. He uses a reciprocal verb
form, the ideophone ‘back and forth’, and then specifically codes the event as
occurring between three pairs of participants by using the enclitic after
‘three pairs’ (71).
(68) man=chi man=chi
and one=DAT one=DAT give-COL-DO:VCL-PRG-DCL
‘And then (they) give (things) first to one than another.’ (M #30)
(69) =tala a
Then 3P2=PL=COL give-INTS-RR-COL-DO:VCL-DCL
‘Then they give (things) to each other together.’ (F #30)
(70) man=chi man=chi a
one=DAT one=DAT come.across-CAUSE-INTS-RR-COL-DO:VCL-PRG-
DCL
‘(They) pass (things) to each other, to first one than another.’ (V #30)
(71) beko beko a peman-ka pareja=tala
back.and.forth give-INTS-RR-COL-DO:VCL-
DCL
three-NCL SP:pair=COL
‘(They) give (things) back and forth, between the three pairs (of people).’
(MC #30)
When the elicited data is compared with natural data there is a rather interesting
contrast in terms of the type of reciprocal used. Again leaving out positionals and just
comparing the reciprocal constructions coding symmetrical actions, in natural discourse texts
the four different reciprocal constructions are fairly evenly distributed (Table 5). However in
the elicited data, only one of the four possible reciprocal conditions, reciprocal with intensifier
formed with ‘RR’ makes up well over half of the tokens. Ignoring simple reciprocal
constructions where the low number of simple reciprocal constructions is probably due to a
lack of these specific predicates in the elicited data, and just looking at the reciprocal with
intensifier constructions, speakers chose the form 82% of the time (53\65).
<.2
Texts Elicited
Simple
reciproca
l
Reciprocal
with
intensifier
Total Simple
reciprocal
Reciprocal
with
intensifier
Total
# % # % # % # % # % # %
i 'become' 176 .30 151 .26 327 .56 17 .20 12 .14 29 .35
-a 'RR'109 .18 152 .26 261 .44 02 .02 53 .63 55 .65
Total 285 .48 303 .52 588 1.00 19 .23 65 .77 84 100
In addition out of the seventeen elicited examples that used the 'become' simple
reciprocal construction, all involved predicates such as ‘hug’, ‘exchange’,
‘meet’ or the Spanish borrowing ‘greet’ in which there is no other option. Besides the
‘natural’ reciprocal verbs the other predicates found in the reciprocal with intensifier
constructions were: ‘play’, ‘see’, ‘shake hand/get hand’, ‘touch’,
‘get/catch’, and ‘pull’.
The results in the elicited examples could be simply a reflection of the types of events
coded. The data had a lot of incidents of 'hitting' and 'giving' and it is possible the -
construction is preferred in these types of events. However looking at the verb ‘hit’, in both
the natural texts and elicited data, there is a clear difference. In the natural texts ‘hit’ occurs
in the reciprocal with intensifier construction slightly more often with 'BECOME' (118/211 =
56%), than with 'RR' (93/211 = 44%). In contrast, all fourteen examples (100%) of
reciprocal 'hit/do' in the elicited data occur with 'RR' in the reciprocal with intensifier
construction. The same held for the predicate 'give'. In the natural discourse texts
'give' occurred in the reciprocal with intensifier construction 66% (12/18) of the time as
opposed to 33% (6/18) of the time in the reciprocal with intensifier construction. Whereas
in the elicited examples 'give' occurred in the reciprocal with intensifier construction
in all sixteen of the examples (100%).
=.2
$>$>
Reciprocal with
intensifier
constructions
Elicited data Natural texts
# % # %
‘hit’ with ‘BECOME’ 0 0 118 56
with ‘RR’ 14 100 93 44
‘give’
with ‘BECOME’ 0 0 12 66
with ‘RR’ 16 100 6 33
It appears that the elicitation task itself, rather than the type of event depicted biased
the responses in favor of the reciprocal constructions formed with - This makes sense in
light of the fact that in the elicitation task the same kinds of events are repeated again and
again and the speakers were well aware that these were staged events. After seeing a few
instances of reciprocal actions it might be the speakers began to anticipate that something
similar was about to happen and hence used the construction which indicated that the event
was expected.
7. CONCLUSION
In conclusion it has been shown that the different strategies used in Tsafiki to code
symmetrical situations are comprised of elements that already have established uses in other
constructions. There is no single element dedicated to the coding of symmetry. Positional
symmetry is coded with positional coverbs that can code other types of positional
configurations including asymmetrical positions and those involving the configuration in
space of a single entity. The symmetrical coverbs simply comprise a subset of positional
constructions. In essence this symmetry is lexically coded.
The strategies for coding active symmetry arise out of the possibilities and constraints
imposed by the complex predicate constructions and the complex semantics involved in
symmetrical actions. In reciprocal constructions there are of course two participants, each
participant is both an actor and an undergoer. In the complex predicate constructions the
speaker must not only choose from a limited set of generic verbs, but also from a very limited
set of second order semantic predicate classes. Of the five classes, ‘become’, ‘do’,
‘say/express’, ‘be in a position’ and ‘be’, only the ‘become’ and ‘do’ classes are
plausible categories for the coding of the reciprocal construction. If the speaker chooses the
class he is focusing on the undergoer aspects of the participants in that an class verb always
indicates that the nominative argument undergoes a change-of-state or location. In other
words what is to the participants is emphasised. The opposite is true if a -class
predicator such as ‘RR’ is used. Here the nominative argument will be an actor. In other
words what the participants are will be emphasised. With the exception of a small set of
coverbs in the simple reciprocal construction, the Tsafiki speaker has the option of choosing
either the -class or the -class construction to code the same reciprocal event.
In addition, the class verbs are also used to form middle voice constructions. As
pointed out by Shibatani (2006) and Maldonado (1993) middle voice constructions often carry
mirative notions such as unexpectedness. Middle voice focuses on the change of state phase
of the event, leaving out the causal phase which can lead to an interpretation of spontaneous
or sudden change which in turn can lead to an interpretation of unexpected change—a
mirative notion. While these notions may be only incipient in the middle voice constructions,
they come to the forefront in the Tsafiki reciprocal and non-translational motion constructions
where the same event can be coded by either an -class or -class verb without a major
change in transitivity or event structure.
So while cross-linguistically the types of distinctions made by the Tsafiki reciprocal
constructions may appear somewhat unusual, they are quite normal and natural in Tsafiki,
also occurring in other related constructions. The reciprocal constructions can be seen to fit
quite well into a scale where the verb ‘become’ ranges from a simple verb coding a change
of state, to a verbal suffix coding mirative notions.
Given the types of distinctions made by the Tsafiki reciprocal constructions it is not
surprising that they did not conform well to the different criteria concerning reciprocal
situations presented in the video clips. The Tsafiki reciprocal constructions have fairly unique
characteristics due to the nature of the constructions from which they arise and the overall
grammatical structure of Tsafiki. Finally, in comparison with the natural data, the elicited data
was biased in favor of the reciprocal construction indicating expected actions, perhaps due to
the fact that the speakers knew the events were staged and there was a fair amount of
repetition. The type of discrepancy between the elicited and natural data found here is, of
course, not particularly unusual and just serves to emphasise the need to always compare and
contrast elicited data with natural data.
APPENDIX A
Tsafiki Generic Verbs
?.2@$$%
Generic verb Generic verb
i ‘become’-class # of coverbs* ki ‘do/make’-class # of coverbs
‘become’ 257 ‘do/make’ 269
‘go’ 109 ‘cause to become’ 156
‘come’ 48 ‘cause’ 149
‘go out/up’ 45 ‘send’ 50
‘come out/up’ 50 ‘put’ 43
‘arrive there’ 34 ‘get’ 35
‘arrive here’ 33 ‘talk’ 10
‘go down’ 35 ‘see’ 9
‘come down’ 35 ‘give’ 4
‘go across’ 43 ‘eat’ 1
‘come across’ 35 ‘direct cause’ 76
‘go in’ 37 ‘cause motion’ 39
‘come in’ 33 ‘set/place’ 17
‘lose’ 5 ‘remove’ 15
‘die’ 1 ‘direct cause’ 5
‘sleep’ 4 ‘indirect cause’ 5
‘get into position’ ‘associative cause’ open
‘reflexive/reciprocal’ open
ti ‘say’ class ra ‘be in position’ class
‘say/express’ ‘be in a position’ 125
‘go around’ 33
jo
‘be’-
class
‘have’ 11
‘be’ ‘sit’ 7
‘not be’ ‘lie’ 5
* $ * % <AA-
$
ORTHOGRAPHY AND ABBREVIATIONS
The orthography used in this study is that used by the Tsachila community which is
roughly based on the IPA with a few exceptions. The symbol ‘j’ represents a voiceless glottal
fricative /h/ (note this differs from the Spanish voiceless velar fricative /x/). The symbol ‘y’ is
used for the voiced palatal approximant /j/. A single apostrophe indicates a glottal stop.
Tsafiki does not have diphthongs or long vowels. Each vowel in a sequence is individually
pronounced with a glottal stop separating it from a following vowel. (This phonologically
conditioned glottal stop is not marked in the writing system). Finally a syllable final ‘>
indicates nasalization of the preceding vowel, 9B9, = /C9
Glosses used in this study include: 1F‘1st person feminine’, 1M ‘first person
masculine, 2 ‘second person’, 3P1 ‘third person proximate one’, 3P2 ‘third person proximate
two, 3D1 ‘third person distal one, 3D2 ‘third person distal two’, ACC accusative, ADV
‘adverb’, ASP ‘aspect’, ASSOC.M ‘associated motion’, ATT ‘attemptive’, AUX ‘auxiliary’,
CER ‘certain’, COM ‘comitative’, COL ‘collective’, CNGR ‘congruent’, CNTR ‘contrastive’,
DAT ‘dative’, DCL ‘declarative’, DCL:EMPH ‘declarative emphatic’, DESR ‘desirative’, DR
‘different reference’, DUB ‘dubative’, EMPH ‘emphatic’, EV evidential, FOC ‘focus’,
GRAD ‘grade’, IM ‘immediate’, IMP ‘imperative’, IMPF.P ‘imperfective participle’, INCH
‘inchoative’, INCL ‘inclusive’, INF ‘infinitive’, INGR ‘ingressive’, INSTR ‘instrument’,
INT ‘interrogative’, INTJ ‘interjective’, INTS ‘intensive’, IRR ‘irrealis’, LOC ‘locative’, MIR
‘mirative’, MOOD ‘mood’, NCL ‘nominal classifier’, NCNGR ‘non-congruent’, NEG
‘negative’, OBL.NMLZR ‘oblique nominaliser’, PF.P ‘perfective participle’, PL 'plural', POS
‘possessive’, PRG ‘progressive’, PRO ‘proform’, PURP ‘purpose’, QT ‘quotative’, QUAL
‘quality’, RR ‘reciprocal/reflexive’, RP ‘reportative’, SITU ‘situational’, SMBL ‘semblative’,
SP 'Spanish borrowing', SPEC ‘speculative’, SR ‘same referent’, ST ‘stative’, SUF ‘suffix’,
VCL ‘verb class marker’.
ACKNOWLEDGEMENTS
Research for this paper was supported by the National Science Foundation (0618887) and the
Volkswagen Stiftung Foundation. I would like to thank all the members of the Tsachila
organisation PIKITSA for producing the natural discourse materials used in this study and
their invaluable help in the analysis with special thanks to the director, Alfonso Aguavil
Calazacón as well as Jose Jacinto Aguavil Loche and Catalina Calazacón Aguavil.
REFERENCES
Curnow, Timothy Jowan. 1997. A grammar of Awa Pit (Cuaiquer): An indigenous language
of southwestern Colombia. Ph.D. diss. Australian National University.
Curnow, Timothy Jowan. 2000. Why ‘first/non-first person’ is not grammaticalized
mirativity. ! )3,, ,AAA2 )3
DeLancey, Scott. 1997. Mirativity: the grammatical marking of unexpected information.
3 1: 33-52. Berlin: Mouton de Gruyter.
Dickinson, Connie. 2000. Mirativity in Tsafiki. 3 24(2): 379-401.
Dickinson, Connie. 2002. Complex predicates in Tsafiki. Ph.D. diss, University of Oregon.
Evans, Nicholas. To appear. %$.
%,Ekkehard König and Volker Gast(eds.),Berlin: Mouton de Gruyter.
Evans, Nicholas. (2006). Dyad Constructions. D 3 3,
Keith Brown (ed.).
Evans, Nicholas, Alice Gaby & Rachel Nordlinger. In press. Valencey mismatches and the
coding of reciprocity in Australian Languages. 3
Frajzyngier, Zygmunt & Traci S. Curl (eds.). 1999. . E
Amsterdam: John Benjamins.
Grinevald, Colette. 2006. The expression of static location in a typological perspective.
. 3 2$ 2, Mya Hickmann &
Stéphane Robert (eds.),Amsterdam: John Benjamins.
Haviland, J.B. 1992. Seated and settled: Tzotzil verbs of the body. FG!
H45(6): 543-561.
Kemmer, Suzanne. 1993. /IAmsterdam: John Benjamins.
Maldonado, Ricardo. 1993. Dynamic construals in Spanish. ( 3
)XXII(3): 531-566.
McGregor, William B. 2002. I)Berlin: Mouton de
Gruyter.
Moore, Bruce. 1991. !@2*2- Stephen H. Levinson
(ed.). Quito: Instituto Lingüístico Verano.
Schultze-Berndt, Eva. 2000. Simple and complex verbs in Jaminjung: a study of event
categorisation in an Australian language Ph.D. diss. Nijmegen: Katholieke
Universiteit Nijmegen.
Shibatani, Masayoshi. 2006. On the conceptual framework for voice phenomena. 3
44(2): 217-269.
?.2@$$%
1 See Curnow (1997) and (2000) for a description of the Barbacoan language Awa Pit and
Dickinson (2002) for a description of complex predicates in Tsafiki.
2 These suffixes include: ‘collective’, ‘contrastive’, ‘attempt’, ‘stative’,
‘intensive’ and the enclitic .
3 The situation is somewhat complicated in that nouns or nominalized verbs can take the full
set of finite verbal suffixes. However a noun or nominalized verb suffixed with finite verbal
morphology always forms a stative rather than an active predicate. The primary
morphological distinction between nouns and verbs in Tsafiki is that verbs require
nominalizing morphology to function as nouns.
4 The element can occur after both nouns and verbs, however both its meaning and pattern
of distribution varies. With nouns or nominalized verbs occurs as an enclitic, (as do the
case-markers), usually cliticizing to the final element of the noun phrase. Its position is much
more rigid with verbs, where it occurs immediately after the verb stem as a suffix. In addition,
when = occurs with nominals or nominalized verbs referring to humans it is always
interpreted as a simple plural (Moore, 1991). However number is not an obligatory category
in Tsafiki and when occurs with nominals referring to non-humans or on verbs it functions
as a collective, i.e. a verb or non-human nominal suffixed with – may refer to a singular
referent. The – indicates that this singular referent is part of a group. I cannot go into details
here, so in this paper when =occurs with nominals referring to humans it is glossed as PL.
In all other cases it is glossed as COL ‘collective’.
Generic verb Generic verb
i ‘become’-class # of coverbs* ki ‘do/make’-class # of coverbs
‘become’ 257 ‘do/make’ 269
‘go’ 109 ‘cause to become’ 156
‘come’ 48 ‘cause’ 149
‘go out/up’ 45 ‘send’ 50
‘come out/up’ 50 ‘put’ 43
5 In actual fact it is impossible to determine whether it is the generic verb or the verb class
marker that controls case marking in that the case-marking pattern of the generic verb and the
verb class marker, when it occurs as a simple or generic verb is identical. This is not the case
with the auxiliaries. The case-marking pattern of the auxiliary and generic verb may differ.
The case-marking pattern always follows that of the auxiliary, not the generic verb.
6 Moore (1991) was the first to note the degree to which participation is formally coded in
Tsafiki, noting that the verbal suffix indicates non-participation on the part of the speaker.
7 DeLancey uses the terms ‘conjunct’ and ‘disjunct’. These terms correlate respectively with
what I am calling ‘congruent’ and ‘non-congruent’, with an important difference. In Tsafiki
there is a trinary rather than a binary distinction. The category congruent/conjunct is similar,
but the category ‘disjunct’ in Tibeto/Burman is divided into two categories ‘neutral’ and ‘non-
congruent’ in Tsafiki. These terms, conjunct/disjunct are also problematic in that they have
another, completely unrelated use in linguistics. In an attempt to avoid the confusion, I am
using what I hope are the more semantically transparent terms congruent and non-congruent.
8 There are several morphemes that have the same form in Tsafiki. Each can be
distinguished morphosyntactically, phonologically and by function. Some of the forms are
diachronically related, others are not. The intensifier always carries the main stress in the
word. In the standard orthography used by the community, which I use here, accent is not
marked. See Dickinson (2002).
‘arrive there’ 34 ‘get’ 35
‘arrive here’ 33 ‘talk’ 10
‘go down’ 35 ‘see’ 9
‘come down’ 35 ‘give’ 4
‘go across’ 43 ‘eat’ 1
‘come across’ 35 ‘direct cause’ 76
‘go in’ 37 ‘cause motion’ 39
‘come in’ 33 ‘set/place’ 17
‘lose’ 5 ‘remove’ 15
‘die’ 1 ‘direct cause’ 5
‘sleep’ 4 ‘indirect cause’ 5
‘get into position’ ‘associative cause’ open
‘reflexive/reciprocal’ open
ti ‘say’ class ra ‘be in position’ class
‘say/express’ ‘be in a position’ 125
9 Tsafiki has a reflexive construction unrelated to the method represented with The
reflexive construction is a typical transitive construction with a reflexive nominal or
in object position, although without the accusative suffix.
a) tenkachikenan ‘He saw himself.’
b) tenkachi ‘They are killing themselves.’
10 The majority of younger speakers do not use and some do not even recognize it. All the
older speakers I have worked with recognize , and can use it when asked to do so, even if
they do not normally utilize it in their daily speech.
‘go around’ 33
jo
‘be’-
class
‘have’ 11
‘be’ ‘sit’ 7
‘not be’ ‘lie’ 5
* $ * % <AA-
$