switch reference in western south americaffffffff-be1c...1 switch reference in western south america...

50
1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference systems occur in a number of languages spoken in a contiguous area in western South America, across language families, and even across macro culture areas (Andes and Amazon). At first sight, this is suggestive of contact-induced diffusion, but the different systems show rather a lot of variation. This paper gives an overview of the switch-reference structures found in Western South America, taking a multi-variate approach to the phenomenon. In addition, it discusses the likelihood of horizontal transmission hypotheses to explain the current distribution of switch reference in South America. 1. Introduction The following examples come from genealogically unrelated languages spoken in the adjacent Upper Amazon and Andean regions of western South America (1) Cavineña [TACANAN], Guillaume 2008: 702, 725 a. udya =tu-ke [imeta-tsu] mare-kware then =3SG-FM (=1SG.ERG) point.at-IDENT shoot.at-REM.PAST ‘Then I pointed my rifle at it (a peccary) and shot it.’ b. [tu-ra mare-wa=ju] =tu pakaka-wa 3SG-ERG shoot.at-PERF=NONID =3SG(-FM) fall-PERF ‘He (Lucio) shot at it (the porcupine), and it fell down.’ (2) Yurakaré [ISOLATE], Van Gijn 2011: 181 a. ti-bëjta-ø-ja ti-la-mala-ø samu 1SG-see-3-IDENT 1SG-MAL-go.SG-3 jaguar ‘When the jaguar saw me, it ran away from me.’ b. bëjta-y-ti ti-la-mala-ø samu 1SG.PRN see-1SG.S-NONID 1SG-MAL-go.SG-3 jaguar ‘When I saw the jaguar, it ran away from me.’ (3) Cofán [ISOLATE], Fischer & Van Lier 2011: 237

Upload: phungnhi

Post on 26-Jan-2019

220 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

1

Switch reference in western South America

DRAFT - DO NOT CITE WITHOUT PERMISSION

Rik van Gijn

University of Zürich

Switch-reference systems occur in a number of languages spoken in a contiguous area in western

South America, across language families, and even across macro culture areas (Andes and Amazon).

At first sight, this is suggestive of contact-induced diffusion, but the different systems show rather a

lot of variation. This paper gives an overview of the switch-reference structures found in Western

South America, taking a multi-variate approach to the phenomenon. In addition, it discusses the

likelihood of horizontal transmission hypotheses to explain the current distribution of switch reference

in South America.

1. Introduction

The following examples come from genealogically unrelated languages spoken in the adjacent Upper

Amazon and Andean regions of western South America

(1) Cavineña [TACANAN], Guillaume 2008: 702, 725

a. udya =tu-ke =Ø [imeta-tsu] mare-kware

then =3SG-FM (=1SG.ERG) point.at-IDENT shoot.at-REM.PAST

‘Then I pointed my rifle at it (a peccary) and shot it.’

b. [tu-ra mare-wa=ju] =tu pakaka-wa

3SG-ERG shoot.at-PERF=NONID =3SG(-FM) fall-PERF

‘He (Lucio) shot at it (the porcupine), and it fell down.’

(2) Yurakaré [ISOLATE], Van Gijn 2011: 181

a. ti-bëjta-ø-ja ti-la-mala-ø samu

1SG-see-3-IDENT 1SG-MAL-go.SG-3 jaguar

‘When the jaguar saw me, it ran away from me.’

b. së bëjta-y-ti ti-la-mala-ø samu

1SG.PRN see-1SG.S-NONID 1SG-MAL-go.SG-3 jaguar

‘When I saw the jaguar, it ran away from me.’

(3) Cofán [ISOLATE], Fischer & Van Lier 2011: 237

Page 2: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

2

[khasheye=ndekhu=ja ñoña]=si te [matachi=ja tsa=ma ondikhu]=pa

old.man=CLF(PL)=DEF make=NONID REP clown=DEF that=ACC wear=IDENT

tsa='ka=en=ja ko'fe=‘ya

that=CMPR=ADV=DEF play=MIR

‘Reportedly, after the elders made (the clothes), the matachi clown wore them and played like

that.’

(4) Cuzco Quechua [QUECHUAN], Lefebvre & Muysken 1988: 117-8

a. wawa-kuna runasimi-ta rima-spa-nku allin ka-sha-nku

child-PL quechua-ACC speak-IDENT-3PL good be-PROG-3PL

‘When the children speak Quechua they are/feel well.’

b. wawa-kuna runasimi-ta rima-qti-nku mana vali-n-man-chu

child-PL quechua-ACC speak-NONID-3PL not worth-3-POT-NEG

‘If the children spoke Quechua it would not be good.’

These examples look rather similar at first sight: they all involve what seem to be classic switch-

reference (henceforth SR) systems: a combination of two clauses, where one of the clauses is marked

for whether the subject of the next clause is the same or different from the subject of the dependent

clause.1 There are many more examples of languages, from different linguistic families, with similar

constructions in western South America.

This raises the question how this situation came about. One appealing possibility is to say that SR

spread through contact between the different groups, which has been suggested before (e.g. Dixon &

Aikhenvald 1999, Aikhenvald 2002, Crevels & Van der Voort 2008, van Gijn 2012). However, in

order to substantiate this claim, we need to look in more detail at the similarities and differences

between the different SR systems in the languages of western South America, and then evaluate

whether there is a contact scenario that would fit the results.

The purpose of this paper is twofold. In the first place, it gives a general overview of SR systems in

western South America. Whereas overviews of SR in North America (Jacobsen 1983, McKenzie,

forthc.), Australia (Austin 1980, 1981) and Papua New Guinea (Roberts 1997) are available, no such

overview exists for South America. This paper intends to fill that gap. In the second place the paper

evaluates - to the extent that this is possible - the likelihood of contact-induced diffusion as a possible

explanation for the current distribution of SR systems in western South America. The paper is

structured as follows: Section 2 briefly introduces the geographical area under consideration, and the

sample of languages considered in this study. Section 3 focuses on the delimitation and

parameterization of switch reference as it is approached in this paper. Section 4 discusses the patterns

1 The opposing choices in the SR systems discussed in this paper are glossed as ident (identity) vs. nonid (non-

identity) throughout the paper for the sake of correspondence.

Page 3: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

3

found for the different language families and language isolates in the family. Section 5 is a discussion

of the result in the light of potential contact versus inheritance scenarios. Section 6, finally, concludes

the discussion by summing up the major findings.

2. Western South America and the sample

Western South America is one of the most diverse areas in the world in terms of genealogy, containing

many isolate languages and smaller language families, but also in terms of structural diversity (see e.g.

Dahl 2008). Within this area, a distinction is often made between highland languages of the Andes,

and lowland languages of the upper Amazon. Both areas are generally regarded as diffusion areas.

Different scholars have proposed lists of shared features for the languages of the Andes (e.g. Büttner

1983, Torero 2002) as well as for the Amazon (e.g. Derbyshire & Pullum 1986, Payne 1990, Dixon &

Aikenvald 1999). Many of these features are contrastive (see Torero 2002, Dixon & Aikhenvald 1999,

Van Gijn 2014a), suggesting that the two macro areas form distinct cultural-linguistic blocks. On the

other hand, archaeological and ethnological evidence suggests that the ‘separation’ of these two

culture areas is a rather recent phenomenon, and that contacts between lowland and highland cultures

were common in the past (Renard-Casevitz et al. 1988, Taylor 1999). This latter view is corroborated

by more recent linguistic studies, which take a wider perspective: several continent-wide typologies

(Michael 2010 on phoneme inventories, Krasnoukhova 2012 on NP structure, Birchall 2014 on

argument encoding) suggest a broad east versus west split which does not coincide with the traditional

border between the Andes and Amazon. Similarly, Van Gijn (2014a) shows that the Upper Amazon,

stretching from the eastern slopes of the Andes well into the Amazon conforms neither to the

Amazonian nor to the Andean profile, calling into question the reality of the split.

As mentioned above, there are many unrelated families in western South America whose members

seem to have a comparable system of marking some of their complex sentences for SR. Families with

SR systems include Quechuan, Aymaran, Tacanan, Panoan, Jivaroan, Tucanoan, Barbacoan, Uru-

Chipaya as well as a few isolate languages such as Kwazá, Yurakaré, Paez, Cofán, and Urarina. In the

sample to be discussed below, I have tried to have each family known to have a SR system represented

by at least two, preferably three members, plus all of the isolates. I briefly describe the families and

isolates with SR systems here.

Quechuan: language family with a large extension, ranging from northern Chile and Argentina to

southern Colombia, mostly in the Andean regions, although some lowland languages also exist. A

main subdivision is made between Quechua I and Quechua II languages, the former referring to the

languages spoken in central Peru, the latter to the languages spoken to the north and south of that area.

The spread of Quechua II is associated with the Inca Empire; Quechua I spread is older and more

Page 4: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

4

compact, dated around the first half of the first millennium by Torero (1984, cited in Adelaar &

Muysken 2004: 181), but there is no clear consensus.

Tacanan: small language family of limited geographical distribution in north Bolivia, with some spill-

over into Peru. Tacanan is linked to Panoan in deeper time (Key 1968), but there is no absolute

consensus about this link (see Hammarström et al. 20142). For that reason Tacanan and Panoan are

considered as separate families in this paper.

Panoan: one of the major language families of the Peruvian lowlands, with some extensions into Brazil

and Bolivia. The languages are generally thought of as relatively homogeneous, arguably reflecting a

relatively recent diversification. Adelaar & Muysken (2004: 419) mention that “all Panoan groups

presumably originate from the Cumancaya culture (1000 AD)”.

Jivaroan: small language family consisting of four closely related languages, spoken in north-western

Peru and southern Ecuador. There are intricate links with Andean societies, mostly belligerent. The

Jivaroan languages have split very recently, perhaps as recent as 500 years ago (Maurizio Gnerre,

Martin Kohlberger, Simon Overall p.c.).

Tucanoan: languages belonging to this language family are spoken in two major, non-contiguous

zones in north Peru/Ecuador on the one hand, and eastern Colombia/Brazil on the other. The main

division is threefold, between western, eastern, and central Tucanoan languages. Especially westrn

Tucanoan languages are underdescribed and as a result also underrepresented in this study. Eastern

Tucano languages are spoken in the complex Vaupés area, involving a high degree of interaction,

especially with Arawakan Tariana.

Barbacoan: small language family spread along the Ecuadorian coastal area and southern Colombia.

There is a division between northern (including Awa Pit) and southern (including Tsáfiki) languages,

the latter being more closely related than the former (Curnow & Liddicoat 1998). Barbacoan may have

had a wider distribution in pre-colonial times, possibly including the important language Cara,

although there is no conclusive evidence that this was indeed a Barbacoan language (see Adelaar &

Muysken 2004: 142, Bruil 2008: 10-12).

Aymaran: today, the language family consists of three languages. The biggest language, Aymara, is

mainly found in the southern tip of Peru around Lake Titicaca, and in the adjacent area in the north-

western Bolivian Andes. The other two extant languages, Cauqui and Jaqaru, are spoken in west-

central Peru, just south of Lima. There is a long-term and intimate interaction with Quechuan.

2 http://glottolog.org/resource/languoid/id/taca1255, accessed 26-09-2014.

Page 5: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

5

Although Aymaran languages surviving today have reduced or lost their SR systems, comparative

research suggests a more elaborate system in the past.

The isolates and odd-ones-out: Several (near) isolates also have (remnants of) a SR system. From

north to south: Páez (South-western Colombia), Cofán (northern Ecuador), Urarina (northeastern

Peru), Cholón (central Peru), Uchumataqu (western Bolivia), Kwazá (western Brazil, close to the

Bolivian border), and Yurakaré (central Bolivia). Apart from these isolates, there are a few languages

that belong to families that do not have a SR system generally, but which have nevertheless developed

one. They are the the Arawakan language Tariana (western Brazil, close to Colombia), the

Nambikwaran language Mamaindê (Rondônia), and the Chibchan language Tunebo. Further removed,

the Chibchan language Panare also seems to have a SR system.

Two further languages are not part of the sample for various reasons.

Tupian: the vast Tupian stock consists of over 40 extant languages that are spread almost over the

entire northern half of the continent. A number of Tupian languages have a coreferential third

argument marker as well as a non-coreferential one. The system in most of these languages, however,

is different from what is considered to be SR here, since the use of the coreferential marker extends to

the noun phrase (possession) and often has taken on other functions (e.g. it marks intransitive third

person in Mekens - Galucio 2014). Moreover, in most languages it does not extend beyond third

persons - although some languages, Like Kayabí, have coreferential markers for all persons (Dobson

2005).

Macro-Jê: Rodrigues (1999) mentions that some Macro-Jê languages have what looks like a switch

reference system in coordinate clauses, notably Kipea, Canela (Timbira) and Maxakali, all three

languages spoken in the east of Brazil. Although the existence of these SR systems in eastern South

America is highly interesting, the area falls out of the scope of this paper.

The sample used for the present study is given in Map 1; the numbers in the map refer to the languages

in Table 1, the gray dots represent languages that are discussed to some extent, but do not form part of

the sample for which distance measures were calculated, due to lack of sufficient data.

Page 6: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

6

MAP 1: Sample used in this study

Table 1: Languages of the sample and their affiliations

Nr Iso code Name Family

1 tuf Tunebo Chibchan

2 pbb Páez Isolate

3 gum Guambiano Barbacoan

4 cub Cubeo Tucanoan

5 kwi Awa Pit Barbacoan

6 gvc Guanano Tucanoan

7 cbi Cha'palaachi Barbacoan

8 des Desano Tucanoan

9 tae Tariana Arawakan

10 tue Tuyuca Tucanoan

11 qvi Imbabura Quechua Quechuan

12 snn Siona Tucanoan

13 con Cofán Isolate

14 cof Tsafiki Barbacoan

15 qug Chimborazo Quichua Quechuan

16 acu Achuar Jivaroan

17 jiv Shuar Jivaroan

1

23

456

7

8910

11

12

1314

15

1617

181920

2122

2324

2526

27 2829

30 313233 34

353637

38 39

40 41

42

4344

Page 7: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

7

18 mpq Matís Panoan

19 ura Urarina Isolate

20 agr Aguaruna Jivaroan

21 mcf Matsés Panoan

22 quk Chachapoyas Quechua Quechuan

23 qvs San Martín Quechua Quechuan

24 shp Shipibo Panoan

25 swo Shanenawa Panoan

26 cbr Kashibo-Kakataibo Panoan

27 qxo Ancash Quechua Quechuan

28 qub Huallaga Quechua Quechuan

29 amc Amawaka Panoan

30 qvp Pacaraos Quechua Quechuan

31 qvn North Junin Quechua Quechuan

32 xwa Kwaza Isolate

33 qxw Jauja Wanca Quechua Quechuan

34 ese Ese Ejja Tacanan

35 aro Araona Tacanan

36 wmd Mamaindê Nambikwaran

37 cav Cavineña Tacanan

38 quy Ayacucho Quechua Quechuan

39 quz Cusco Quechua Quechuan

40 ure Uchumataqu Uru-Chipaya

41 yuz Yurakaré Isolate

42 cap Chipaya Uru-Chipaya

43 ayr Central Aymaran Aymaran

44 quh Bolivian Quechua Quechuan

3. Switch reference

In this section I will first give the outlines of the type of constructions I look at in this paper, and then

discuss a multivariate approach to the phenomenon under study.

3.1. Delimiting SR

Switch reference has been the object of a considerable amount of studies, and yet it is still elusive

as a phenomenon, in part because SR is not a single-variable phenomenon, in part also because it has

aspects in common with a host of related phenomena (see Van Gijn & Hammond, this volume). The

major reference work scholars still refer to for a canonical definition of switch reference is the edited

volume by Haiman & Munro (1983), who define the phenomenon as “an inflectional category of the

verb, which indicates whether or not its subject is identical with the subject of some other verb” (p.

ix). The ongoing discussions especially since Haiman & Munro (1983) have made it clear that switch-

Page 8: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

8

reference is a multi-layered, multi-factorial phenomenon with many connections to other types of

reference tracking and discourse-structuring systems. Attempting to isolate a group of structures

within this realm and call them switch-reference systems is a serious challenge, and will most

probably always involve some arbitrary decisions. Nevertheless, on the basis of the parameters given

above, I will attempt to come to a description of the kinds of structures that I will consider for this

paper, which enables me to compare languages to each other.

In the first place, I will look at switch-reference systems. That is, most if not all languages may

have complex sentence constructions that can only be used for same-subject relations, or perhaps

(though probably less commonly) only for different-subject relations, but the hallmark of switch-

reference systems is that they form opposing, and mutually exclusive pairs of encoding patterns used

in the same type of construction, where one encoding is characterized as indicating identity, and the

other as non-identity. So the first limitation we pose onto switch-reference systems is that a) there is a

- minimally binary - opposition between two encodings that denote similar interclausal semantic

relation types, b) the opposition is principally determined by referential (non-)identity between two

participants in different events - such that both or all construction types are highly restricted in their

referential interpretation.3

Second, I compare only those systems that can be said to form a morphological paradigm, i.e.

where all, or all but one of the possible values in the opposition are marked by some overt morpheme

or morphological operation as compared to independent clauses. This includes (bound) pronouns that

are formally different from their counterparts in independent clauses (for at least one of the opposite

SR values). Not included are those oppositions characterized by the presence or absence of pivotal

arguments only.

Third, I look at those constructions that involve interpretation domains for which the interpretative

resolution lies outside the clause or predication proper, but inside the next interpretative domain,

whether a sentence or a clause chain. This criterion rules out long-distance reflexives that can also

have their antecedent within the clausal domain. Interpretative resolution involves at least two

arguments, a referentially dependent or relative argument and an anchor or controller argument.

Fourth, I only look at those systems that apply regardless of person value. Some reference-tracking

systems, like logophoric systems or so-called fourth-person systems as found in e.g. Central Alaskan

Yup’ik (see e.g. Payne 1980, Woodbury 1983) and similar systems in a number of Tupian languages

(see e.g. Jensen 1999), tend to apply only when third persons are involved. This means that the system

as it exists in most present-day versions of Tupian languages (which are spread throughout the

continent, including some parts of western South America) are not taken into consideration, as they

have special pronominal forms for coreferent third persons only, even though a system has been

reconstructed for Tupian which involved all persons (Jensen 1990, 1998). Furthermore, there are some

3 This does not mean that I do not consider pragmatic factors that can override the referential basis, but I regard them as

conditions, thus highlighting the role of referential identity, which seems to play a role in all of these systems, even if they

are heavily influenced by pragmatics of inter-event coherence (Stirling 1993).

Page 9: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

9

other differences between the Tupian system and the SR systems as found in western South America,

such as that the coreferent bound pronoun in Tupian is also used for the possessor, and the fact that it

is the coreferent rather than the non-coreferent forms that are marked.

3.2. A multivariate approach to SR

Even within the confines of the definition above, switch-reference systems can differ from each other

on a number of aspects. It is precisely these variable aspects that may give us clearer insights into

possible connections between SR systems in different languages. The approach taken in this paper,

therefore, is to regard these aspects as variables, which in turn form the basis for a distance measure

between the different SR systems presented in Section 5 below. Such a multivariate approach to

comparative linguistics has proven to be useful in typology (Bickel 2010a, b, 2012, Corbett 2005,

2012, Van Gijn & Zúñiga 2014) but also in family-internal comparison (Van Gijn et al. 2015) and

contact situations (Van Gijn & Hammarström in prep).

The variables considered here are based on the typological literature on SR (notably Haiman &

Munro 1983, Foley & Van Valin 1984, Stirling 1993, Huang 2000, Bickel 2010b, Matić et al. 2014,

and on Van Gijn & Hammond, this volume). A relational database was set up that allows for the

systematic storage and comparison of the relevant information of the values for these variables. The

database is sorted by marker: each marker forms a separate entry in the database; all the variables and

values that follow are assessed with respect to each individual marker.

Figure 1 (Van Gijn & Hammond, this volume) shows the several meaningful parts of a SR

construction.

Figure 1: The parts of a SR construction

The marked clause necessarily contains a SR marker, which refers to a reference clause participant

(the controller participant). The pivotal argument of the marked clause is the relativ participant. There

is furthermore a relationship between the clauses as well as between the pivotal arguments, and one or

domain

Marked clause Reference clause

Marker

relative participant

controller participant

context

relation

relation

Page 10: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

10

more markers encoding at least the latter relation. Wider aspects of SR concern the domain within

which it operates (typically a multi-clausal unit) and the wider context (often a narrative context).

Each of these component parts of a SR construction gives rise to one or more variables.

Unfortunately the available information on the languages in the sample, though generally of high

quality, did not allow for the consideration of the full range of variables, but most key variables could

be assessed to the extent that makes a comparison possible.

In what follows I discuss the variables that are kept track of in the database in three thematic

groups: the markers, the marked clause, the inter-clausal relations, and the categories of the relative

and controller participants. Finally, I briefly discuss contextual conditions that may interfere with the

basic system.

Since at least N-1 of the total number of constructions that are in opposition to each other is

required to have some formal marker, we can keep track of the markers themselves. In the first place,

marker-related variables involve the phonetic form of the marker, but also - more abstractly - its

morphological properties, like the type of morpheme, what host it attaches to, and the position with

respect to that host. The database contains the variables and their values displayed in Table 2.4

Table 2: SR markers - formal variables and their values

variable values

Morpheme form (open value)

Fusion isolating, concatenative, non-linear, zero

Host verb, verb phrase, clause, none

Position pre, post, in, free

The phonetic form variable obviously has an open value list, since the variation is in principle not

restricted. This variable is mainly important to be able to determine whether or not phonetic substance

is transferred from one language to another, as well as the family-internal stability of the actual forms

of the markers.

The fusion values relate to the phonological independence of the SR marker(s), combinations are

obviously possible. The host values give information about the selection restrictions of the SR

markers, and the position distinguishes markers that precede a host from those that follow, or those

that are expressed simultaneously with their host (in), or those that are not positioned with respect to a

host (free). The list of values is a reflection of what is encountered in the languages of the sample and

by no means complete.

Apart from marking identity or non-identity, a SR marker may encode further information, which

leads to en expansion of the number of SR markers beyond two. The additional values also form an

4 Most of these values are built up from the data provided by the languages in the database, so that the values for each

variable are not exhaustive. Since I look at oppositions, “zero” morphology is also taken into consideration, though without

making any claims about the “reality” of zero morphemes. The sets of questions (apart from the ones relating to interclausal

semantics) are asked separately with respect to identity and non-identity markers.

Page 11: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

11

open-ended list, which is built up on the basis of the empirical facts presented by the languages of the

sample, leading to languages whose SR markers also encode TMA values, in particular mood

distinctions (realis versus irrealis), those that mark referential aspects of participants (person, number,

gender), and those that additionally mark the type of interclausal semantics of the construction. The

additional functions of SR markers in the data set are summarized in Table 3.

Table 3: SR markers - additional functions of SR markers

variable values

Additional inflectional categories TMA, referential values

A switch-reference construction typically consists of one or more marked clauses and a controller

clause. One of the possible differences between SR systems between languages is whether or not the

presence of a SR marker has any consequences for person inflection in the marked clause. The

variable and its values pertaining to the marked clause are summarized in Table 4.

Table 4: Marked clause variables and their values.

variable values

person marking in marked clause as in independent clause (present), as in independent clause

(absent), different from independent clause (absent), different

from independent clause (present - i.e. different person inflection)

Another criterion of potential cross-linguistic variation that relates to the construction as a whole is

the order of clauses. Is SR marking anaphoric or cataphoric, i.e. is the order marked clause - reference

clause, or vice versa, or perhaps both options are possible. If both options are available, but there is a

clearly stated preference for one over the other, this is considered as a separate situation type (see

Table 5).5

Table 5: The variable “clause order”and its values

variable values

clause order marked-reference; reference-marked; no preference; preference

for marked-reference; preference for reference-marked

Languages with an SR system may show variation in terms of the semantic or syntactic functions

involved in SR. Potential controllers and relative arguments of SR constructions may be describable in

5 Admittedly, this allows for some degree of subjectivity. In practice, however, positional preference seems to be

clearly distinguishable.

Page 12: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

12

terms of syntactic roles (e.g. subject), in terms of semantic roles (e.g. actor), or pragmatic roles (e.g.

topic). Since these categories may or may not be equivalent for controller and relative arguments, they

need to be coded separately for identity and non-identity relations. Table 6 gives the values

Table 6: The categorial restrictions variables and their values

variable values

Participant roled involved in SR S/A (nominative), S/P (absolutive), S (intransitive

participant), A (ergative), P (accusative), Agent, Topic

Languages (or SR markers within languages) may furthermore differ with respect to the types of

semantic contexts in which they occur. It is rather difficult to assess the semantic range of SR markers

sometimes, as it may depend on context. I have tried, on the basis of explicit statements by the authors

as well as examples given in the grammatical descriptions, to be as complete as possible. The values

have been generated in a bottom-up fashion and defined semantically, based on the idea that contact-

induced language change is rooted in semantics (and pragmatics) - see e.g. Heine & Kuteva (2005),

Matras & Sakel (2007). Nevertheless, it is important to note that the semantic types per marker are

open to modifications in the light of more detailed data analyses in some cases. The values are given

in Table 7.

Table 7: Interclausal semantics.

variable values (bottom up)

semantics

clause linkage

temporal (simultaneous, overlapping, after, before), condition (standard, counterfactual),

purpose, reason, concession, desire complement, perception complements, knowledge

complements, fear complements, modifying relations (relativization), coordination

3.3. Additional remarks

Stirling (1993) argues for a much broader functionality of switch reference. In her perspective, switch

reference is about congruence between eventualities, of which referential continuity forms a sub-

function. One common aspect to the systems surveyed by Stirling is nevertheless that referential (non-

)identity (whether syntactic, semantic, or pragmatic) always seems to form part of the pivot, even if it

may be overridden by other, non-referential factors, whereas the other pivot types may be absent. For

this reason, I have chosen to take referential (dis)continuity as the basis for coding. Nevertheless, a

number of languages in the sample show deviations from the referential (dis)continuity basis, and in a

number of cases these deviations occur to satisfy another, discourse related principles, in line with

Stirling’s work. I have kept track of these discourse principles as well, but they were not taken into

account for the distance calculations, because they were too disparate. I come back to the types of

Page 13: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

13

discourse-related principles at work in the languages of the sample throughout section 4 and in section

5.

The variables discussed in the previous subsection form the basis for the comparison between the

South American languages in this study. However, further variables are imaginable that have not been

coded here for lack of sufficient useful data. I will briefly mention a few additional constraints that

may provide fruitful lines of inquiry for future research.

Typologists have been concerned with what constitutes same and different in terms of

referentiality, beyond the specific role that is being tracked. SR systems may differ in whether or not

they allow for inclusion of the tracked participant in one of the clauses into the group of participants in

the other clause to be interpreted as an identity relation. Perhaps another way to state this point is

whether an SR system is sensitive to person value, or to person + number value. In principle, this

would yield three types of systems: one where the pivot may include the controller under an identity

reading, one where the controller may include the pivot, and one where strict identity is required for an

identity reading. Unfortunately the languages surveyed in this study show too many gaps with respect

to this parameter for it to be useful from a comparative perspective.

Another potential variable is whether, in a situation of a chain of marked clauses and one controller

clause, marked clauses always refer to the next clause (whether a marked clause or a controller clause)

or whether all marked clauses refer to the one controller clause. This could lead to a typology of local

scope systems versus global scope systems, and perhaps a third, flexible, system. Some grammars of

the languages coded in the database do actually discuss this aspect, and I will briefly mention some

aspects relating to this variable, but there is not enough material to warrant a useful comparison

between all languages in the sample.

Finally, languages may differ in terms of whether they employ their SR system for tail-head

linkage (de Vries 2005). Again there is some information on this for a few languages, which I will

briefly discuss below, but not enough material for a full comparison.

4. SR patterns in the Andes and Upper Amazon

This section describes the patterns found for the languages in the sample, discussing the facts for each

family (4.1 - 4.8), the isolates (4.9) and the odd-ones-out (4.10).

4.1. Quechuan

To my knowledge, all Quechuan languages have a switch-reference system, and there is quite a bit of

common ground between the systems in the different languages. However, the Ecuadorian Quechuan

languages show deviant patterns. Figure 2 indicates how the Quechuan languages in the sample relate

to each other, based on the classification in Hammarström et al. (2014).

Page 14: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

14

Figure 2: The Quechuan languages in the sample

An overview of SR markers and their functions is shown in Table 8. The first column displays the

languages in the sample, the second the subclassification into Quechua I (central Peru), Quechua IIB

(Northern Peru, Ecuador), and Quechua IIC (Southern Peru, Bolivia); the status of what is often called

Quechua IIA (though Hammarström et al. 2014 give no additional letter to this group) is controversial

because it does not form a unity in itself but rather a group that falls in between Quechua I and II

(Adelaar with Muysken 2004: 186). The third and fourth columns show the identity markers for

temporal/conditional clauses (column 3) and for purpose clauses (column 4). The non-identity markers

for the same two contexts are shown in columns 5 and 6. All markers are suffixes that attach to the

verb.

Table 8: Quechuan SR markers

IDENTITY MARKERS NON-IDENTITY MARKERS

TEMPORAL/CONDITIONAL PURPOSE TEMP/COND PURPOSE

Jauja Wanca I l pti

Tarma I r pti

Huallaga I ʃpa r pti

Ancash I ʃpa r(nin) pti

Pacaraos II ʃpa pti

San Martin IIB ʃpa pti

Chachapoyas IIB ʃ(pa) ti

Imbabura IIB ʃpa ngapah hpi ʧun

QUECHUAN

Pacaraos

San Martin Chachapoyas Imbabura Chimborazo

Cuzco Ayacucho Bolivian

QUECHUA I

QUECHUA II IIB

IIC

Huallaga Jauja Wanca North Junín Ancash

Page 15: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

15

Chimborazo IIB ʃpa ngapak kpi tun

Ayacucho IIC spa stin pti

Cuzco IIC spa qti

Bolivian Quechua IIC spa qti

As can be seen in Table 8, there is a fair amount of overlap in the forms, which moreover suggest a

diachronic development within the family. The markers shared by most languages are -ʃpa/-spa for

temporal/conditional same-subject clauses and -pti with several (related) variations for different-

subject clauses. The Quechuan subgroups can be recognized to some extent on the basis of the

markers: the same-subject marker -r/-l only occurs in Quechua I languages, and the purpose markers

are an innovation in Quechua IIB. In fact, San Martín and Chachapoyas Quechua are geographically

intermediate between the Quechua I area and the other languages of Quechua IIB. Imbabura and

Chimborazo are furthest removed from what seems to be a core Quechua system. Not only do they

have separate markers for purpose clauses, they also are the only two languages where the DS marker

for temporal/conditional clauses without a /t/. In the IIC group Ayacucho differs from Cuzco and

Bolivian Quechua. The latter two are often regarded as part of the same dialect continuum (Adelaar

with Muysken 2004: 187).

Huallaga Quechua has both -ʃpa and -r. According to Weber (1989: 299) there is no functional

difference between the two markers, they are used in parallel contexts. However, the fact that -ʃpa

requires an inflectional subject marker and -r does not allow for any overt inflectional person marker

makes the former the preferred option in some cases. Cole (1983), for Ancash, mentions that the

difference between -r and -ʃpa in Ancash is determined by relatedness of events: -r is used when the

events are related (e.g when the occurrence of one event depends on the other), -ʃpa is used when this

is not the case. The form -rnin in Ancash is in free variation with -r. The element -nin is homophonous

with the third person subject agreement marker, but it does not have that function in the form -rnin

(Cole 1983: 14).

In terms of same-subject marking, Pacaraos (Adelaar 1987) sides with the Quechua IIC

languages, which all have only the marker -ʃpa for same-subject temporal clauses.6 As mentioned, the

northernost Quechua IIC languages Imbabura (Cole 1982, 1983) and Chimborazo (Beukema 1975)

have acquired SR markers for purpose clauses. The opposition -chun (non-identity) versus -ngapaj

(identity) found in Imbabura Quechua can be found in purpose clauses, but it extends to desire

complements. Imbabura Quechua, unlike most of its relatives, developed a SR opposition for purpose

relations, of which the identity marker -ngapaj shows a potential affinity with Awa Pit’s (non-identity)

purpose marker -napa7 and whose non-identity marker -chun may be related to e.g. Tsáfiki’s purpose

6 There is an alternation in Chachapoyas Quechua (Weber 1975) between - ʃpa, which is mainly used in conditional contexts,

and -ʃ which is used in other contexts. 7 The source of this marker, however, is probably Quechuan, from a combination of irrealis nominalizer -na and benefactive

marker -paq.

Page 16: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

16

marker (non-SR sensitive) -chu.8 Quechua was introduced only late, after the Spanish arrival, in an

area formerly dominated linguistically by an extinct, probably Barbacoan language Cara (Bruil 2008),

so mutual influence is probable, also in the form of (Barbacoan?) substrate in Imbabura Quechua. In

Imbabura Quechua, the “subjunctive” SR system has expanded to mark complements of desire verbs

as well, possibly under Spanish influence (Bruil 2008).

Ayacucho Quechua (Hartmann, ed. 1994) has two same-subject temporal markers, -spa and -stin.

The latter is used when the events occur simultaneously. The marker -ʃtin as an adverbial clause

marker is found in North Junín, or Tarma Quechua, but not with SR functionality, and regionally

restricted (Adelaar 1977: 101). The opposed different-subject/simultaneous construction is marked

with the durative marker -ʧka and the different subject marker -pti.

In terms of their semantics, the group of temporal/conditional markers usually has a rather broad

range of possible interpretations, including temporal overlap, succession, simultaneity, condition

(standard and concessive), and often also reason. In San Martín Quechua, SR extends to relatives as

well (Howkins 1977). Certain interpretations are the result of the combination of the SR marker with

some other marker. For instance, in Ayacucho Quechua, different-subject reason interpretations result

from the combination of the SR markers with the assertive enclitic =mi, and a concessive reading in

combination with the additive enclitic =pas.

Another type difference between the markers is the effect they have on person marking on the

dependent predicate. Many of the Quechuan SR markers have a nominalizing effect in that they

require the subjects of the dependent verbs to be marked by possessor suffixes.

(5) Huallaga Quechua [QUECHUAN], Weber 1983: 281

chaya-pti-nchi qoyku-shaq

arrive-NONID-1INCL.POSS give-1FUT

‘When we (incl) arrive, I will give it to him.’

Some of the markers are incompatible with person inflection.

(6) Bolivian Quechua [QUECHUAN] , van de Kerke 1996: 8

p’acha-ta t’aqsa-y ni-spa ni-wa-rqa

cloth-ACC wash-IMP say-IDENT say-1OBJ-3SG.PST

‘“Wash the clothes” saying, she said to me.’

There is some variation between the Quechuan languages with respect to this parameter, although the

tendency is clearly towards nominal person inflection for non-identity marked clauses, and no person

8 Here there is a possible, but less likely Quechua source -chun for third person imperative (Bruil 2008).

Page 17: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

17

inflection for identity marked clauses, as is shown in Table 9.9 In Pacaraos Quechua, different subject

clauses are combined with person inflection, which is identical for possessive and subject (Adelaar

1987) do strictly speaking it is impossible to tell whether there is verbal or nominal inflection.

Table 9: subject person inflection in marked clauses in Quechuan languages

NO PERSON INFLECTION NOMINAL PERSON INFLECTION

Huallaga identity -r identity -shpa non-identity -pti

Jauja Wanca identity -l non-identity -pti

Tarma identity -r identity -r, non-identity -pti

Ancash identity -r(nin) non-identity -pti

Pacaraos identity -shpa non-identity -pti

San Martin identity -shpa non-identity -pti

Chachapoyas identity -sh(pa) non-identity -ti

Imbabura all

Chimborazo all

Ayacucho identity -stin identity -spa, non-identity -pti

Cuzco identity -spa non-identity -qti

Bolivian Quechua identity -spa non-identity -qti

Imbabura Quechua controllers and pivots can to a large extent be defined syntactically, involving

the S/A argument in both cases, but Cole (1983) indicates the following exceptions to this pattern:10

(7) Imbabura Quechua [QUECHUAN], Cole 1983: 6-7

a. ali-mi Ø/ñuka/kan/*pay Juzi-wan parla-ngapaj

good-EVID one/I/you/*he José-COM speak-IDENT

‘It is good that one/I/you/*he speaks with José.’

b. ali-mi Ø/*ñuka/*kan/pay Juzi-wan parla-chun

good-EVID one/I/you/*he José-COM speak-NONID

‘It is good that *one/*I/*you/he speaks with José.’

Impersonal main clauses in Imbabura Quechua control the SR suffixes in an unusual way, the

identity marker -ngapaj forces either an impersonal reading or a SAP reading, the non-identity marker

forces a personal non-SAP reading. It is not clear to what extent ‘deviations’ from the subject pattern

9 In Pacaraos Quechua, different subject clauses are combined with person inflection, which is identical for possessive and

subject (Adelaar 1987) do strictly speaking it is impossible to tell whether there is verbal or nominal inflection. They are

nevertheless grouped with the nominal inflection group. In Tarma Quechua, subject inflection for identical subject clauses is

optional, therefore it appears in both coleumns. However, an r-clause with person inflection is rare (Adelaar 2011). 10 Curnow (1997: 280) reports for the Barbacoan language Awa Pit that similar clauses to the ones in (7) are always marked

with the same-subject (infinitive) marker, even though the subjects are different. Since in these contexts there is no different-

subject counterpart, they are not considered in this study.

Page 18: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

18

occur in other Quechuan languages, and whether they show any systematicity. Howkins (1977)

mentions that there are instances where constructions with identical subjects are nevertheless marked

with the non-identity marker -pti in San Martín Quechua. Unfortunately no generalizations are made,

and Howkins calls it a marginal phenomenon. Ebina (2011: 34) claims for Cuzco Quechua that “the

use of -spa or -qti seems to depend on whether the two actions are considered to comprise one (in case

of -spa) or two different events (in case of -qti)”, unfortunately without examples. Bruil (2008) shows

that in Imbabura Quechua, the purposive complement of certain verbs (e.g. munana ‘want’ and

cachaca ‘send’) can get identity markers on them in cases of different subjects. It seems likely that

more ‘non-standard’ usages will come to light on closer inspection.

All in all, the Quechuan languages show a picture of a clear genealogical core, which must have

been in place before the diffusion of the different languages. However, it also shows signs of contact-

related phenomena, particularly in the Ecuadorian languages, which have developed a SR system for

purpose clauses which extended further, possibly under Spanish influence, into complements of desire

verbs at least in Imbabura Quechua.

4.2. Tacanan

The small Tacanan language family consists of 7 (Hammarström et al. 2014) languages spoken in

northern lowland Bolivia. The patterns of SR are much less homogeneous than they are for Quechuan

languages. The three Tacanan languages in the sample are Cavineña, Ese Ejja, and Araona. There is

also information available for Reyesano, or Maropa (Guillaume 2012) but that language does not seem

to have a SR system. The position of the sample languages in the Tacanan tree are given in Figure 3.

Figure 3: The Tacanan languages of the sample

Of the three Tacanan languages in the sample, Cavineña seems to have the simplest system, with two

opposing markers, -tsu (identity) and =ju (non-identity), shown in (8), a repetition of (1).

(8) Cavineña [TACANAN], Guillaume 2008: 702 & 725

a. udya=tu-ke =Ø [imeta-tsu] mare-kware

then=3SG-FM (=1SG.ERG) point.at-IDENT shoot.at-REM.PAST

TACANAN Ese Ejja

Araona

TAKANIK-

CHAMIC

ARAONA-

TROROMONO

Cavineña

Page 19: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

19

‘Then I pointed (my rifle) at it (a peccary) and shot it.’

b. [tu-ra mare-wa=ju] =tu pakaka-wa

3SG-ERG shoot.at-PERF=NONID =3SG(-FM) fall-PERF

‘He (Lucio) shot at it (the porcupine), and it fell down.’

The markers for same subject (-tsu) and different subject (=ju) are not equivalent in that the former is a

suffix that attaches to verbs, and the latter an enclitic that attaches to the last element of the

subordinate clause. The opposition between the markers is found for temporal clauses of succession or

overlap. Like is the case in all subordinate clauses in Cavineña, there is no person marking in the

marked clause.

The basic system of Araona in the description of Pitman (1980) is given in Table 10.

Table 10: The SR system of Araona as described in Pitman (1980)

RELATION TYPE IDENTITY NON-IDENTITY

sequence -tso -hao

purpose/reason ezae poho

concession bime powe

Pitman (1980) analyzes the top two markers as suffixes and the rest as free forms. The system is

slightly different in Emkow (2006) who analyzes =tso as a clitic that attaches either to the verb or to

the auxiliary if there is one. She has no examples of clauses marked with -hao in her corpus. Emkow

furthermore mentions that, while =tso requires the subject of the marking and reference clause to be

the same in most circumstances; this is not the case when either the marked or the reference clause is a

copula clause.

(9) Araona [TACANAN], (Emkow, 2006: 682)

[hana ti a=tso] zotokana a-bei pi=po=rna

food eat AUX.TR-SEQ jaguar COP-aggressive NEG=AUX.COP=NEG

'After (the Araona) had given food to the jaguar the jaguar was not aggressive,'

The concessive opposition is not discussed in Emkow, the purposive identity versus non-identity

construction is recognized; the opposition for purpose clauses pa= ... =pojo versus pa=... ezae,

whereby SR marking effectively takes place with the opposition =pojo /poho/ (a verbal clitic with the

same distribution as =tso) and a phonologically free form ezae /ɛzaɛ/.

The system in Ese Ejja is quite different. Table 11 displays the SR markers of Ese Ejja (Vuillermet

2012, 2014). Vuillermet analyzes the SR markers of Ese Ejja (Tacanan) as enclitics, because they are

not part of the phonological word, although they are positionally and selectionally restrictive.

Page 20: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

20

Table 11: The SR markers of Ese Ejja (Tacanan)

IDENTITY

MARKER RELATION TYPE CONTROLLER DEPENDENT

xehoho Reason Subject (S) Absolutive (S/P)

axehoho Reason Ergative (A) Subject (S)

xemo Conditional (standard) Subject (S) Absolutive (S/P)

axemo Conditional (standard) Ergative (A) Subject (S)

ximawa Temporal When Subject (S) Absolutive (S/P)

ximawaa Temporal When Ergative (A) Subject (S)

axe Temporal When Subject (S/A) Object (P)

maxe Temporal When Subject (S/A) Subject (S/A)

NON-IDENTITY

hoho Reason Subject (S/A) Subject (S/A)

hoxemo Conditional (standard) Subject (S/A) Subject (S/A)

ximawaho Temporal When Subject (S/A) Subject (S/A)

aho Temporal When Subject (S/A) Subject (S/A)

The Ese Ejja SR system has in common with Araona that it has different sets of markers for different

semantic relations between clauses, but the commonalities end there. In fact, as we will see in the next

section, Ese Ejja SR is rather like Panoan SR. Like Panoan languages, apart from having markers

specialized for the type of relation between the marked clause and the final clause, the language has

different markers for different types of combinations of categories of controlling and dependent

participants, including a split between A and S and a subject to object coreference marker. There are

also some differences (like an absolutive pivot, an S to A pivot, a marker for reason clauses). In terms

of the phonetic form of the markers, moreover, there are many differences with the Panoan ones, and a

possible overlap with Cavineña and Araona only in the temporal (when) non-identity marker.

Vuillermet (2014) shows that most of the markers are in fact taken from other parts of the grammar, in

particular case markers. Another salient aspect of the phonetic form of the markers is the recurrence of

phonetic strings across the paradigm, suggesting a past where perhaps a number of the feature values

had a more separative exponence. This may hint at the fact that the SR part of the system could in fact

have been much simpler. Whether or not person inflection appears on the verb of the marked clause is

dependent on other criteria, and in fact orthogonal to SR marking (see Vuillermet 2012, 2014 for more

details).

The Tacanan languages, in conclusion, show a rather disparate pattern, with each language

displaying a rather different system. Whereas the system in Cavineña is not unlike the Quechua style

SR, with two opposed markers applying in temporal contexts, the system of Ese Ejja is rather more

Page 21: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

21

like the Panoan languages. This makes for an interesting case because Tacanan, the family to which

Cavineña and Ese Ejja belong, is sometimes claimed to be related to Panoan, although this has not

been convincingly proven.11

If Ese Ejja indeed is more conservative than the other Tacanan languages

in terms of its SR system, perhaps Araona, and especially Tacanan, have lost many of the features of

their original system due to contact with Inca factions (see Renard-Casevitz et al. 1988), making the

system rather more like Quechuan.

4.3. Panoan

The relevant categories of controllers and relative arguments for most languages in the sample are

readily describable in terms of subject (S/A). Deviations from this generalization are found in the

Panoan languages, which have a complex system interacting with transitivity, as well as with

interclausal semantics. At the core of Panoan systems is a system with different markers for events

that precede, are simultaneous with, or follow the event in the reference clause in time. In addition,

identity markers in all Panoan languages are sensitive to the transitivity of the controller event, leading

to a minimum of 6 identity markers, as indicated in Table 12:

Table 12: Minimum ingredients of Panoan SR systems

Relative time dependent event relative argument controller argument

anterior S/A S

S/A A

simultaneous S/A S

S/A A

posterior S/A S

S/A A

Non-identity markers generally exist for the three relative time situations, but do not interact with the

transitvity value of the reference clause. The systems of the individual languages are more complex

than this core system. The Panoan languages of the sample (see Figure 4) all have these basic

oppositions, but also make additional distinctions.

11 Interestingly, Loos (p. 227) mentions the forms of the SR markers as a possible point of overlap between the Panoan and

Tacanan languages. This is not corroborated by the present study.

Page 22: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

22

Figure 4: The Panoan languages of the sample

The systems of Shipibo-Konibo (Valenzuela 2003) and Kashibo-Kakataibo (Zariquiey 2012)

overlap to a considerable extent, including the form of the markers, clearly suggesting a common

source. Table 13 displays the identity markers in both languages, for the different relative time types,

organized by form of the marker.12

Table 13: identity markers of Shipibo and Kashibo

Marker Relative time Controller Relative

Shipibo Kashibo Shipibo Kashibo Shipibo Kashibo Shipibo Kashibo

-ax -ax After After/While* S S S/A S/A

-xon -xun After After/While* A A S/A S/A

-taanan -tanan After While S/A S/A S/A S/A

-a -ia After While S/A S/A/P P P

-i -i While While S S S/A S/A

-kin -kin While While A A S/A S/A

-anan -anan While While S/A S/A** S/A S/A*

-nox -nux Prp Prp S S S/A S/A

-no(n)xon -nuxun Prp Prp A A S/A S/A

-tankëx After S S/A

-tankëxun After A S/A

-këtian After P S/A/P

-këx After S P

-këxun After A P

12 * = Also used for conditionals; ** =Only if objects are different

PANOAN Kashibo-

Kakataibo

Shipibo-

Konibo MAINLINE PANO-NAWA

Matsés Matis

MAYORUNA

KASHIBO

HEADWATERS Amawaka Shanenawa

Page 23: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

23

It is clear from the table that, even though there are a number of differences between both systems,

they share a good part, both in terms of organization of the system and in the forms of the markers13

,

although there have apparently been some semantic shifts, and Kashibo-Kakataibo has a more

elaborate system, in particular with respect to successive relations.

Matís and Matses belong to a different branch than Kashibo Kakataibo and Shipibo, and they are

also geographically rather far away. Nevertheless, their systems of SR are also unmistakably Panoan,

including the forms of the markers. The systems of Matses and Matis allow for a systemic

organization of Table 14. Data from Matsés come from Fleck (2003), for Matis from Ferreira (2005).

Table 14: Identity markers in Matsés and Matis

Relative time Relative participant Controller Matsés Matis

Anterior S/A A ash ash

S/A S shun shun

S/A A tanec

S/A S tanquin

S/A A anec

S/A S anquin

S/A P ac ak

Simultaneous S/A S ec ek

S/A A quin/en kin

S/A P sho sho

Posterior S/A S nush nush

S/A A nun nun

S/A S nuec

S/A A nuen

The system of Matses, with a number of extra markers for anterior events, shows some similarities

with the system of Kashibo-Kakataibo, suggesting that the original Panoan system was rather more

complex than the core system indicated in Table 12. The several markers for succession relations in

Matsés have subtle meaning differences, forming three pairs, on the basis of how much time may pass

between the events.14

(10) Matsés [Panoan], Fleck 2003: 1093

a. -ash/shun adjacent or with intervening time periods

b. -anec/anquin two parts of the same episode (first is a locomotion verb)

13

The SR markers in Panoan languages, which in most cases seem tob e complex, have similar correspondences

to the case systems in the different languages oft he family, a feature also found in for instance in Tacanan and

Jivaroan languages. This merits a study in ist own right, but goes beyond the scope of this paper. 14

Similar factors play a role in Shipibo-Konibo fort he markers -tian (temporally adjacent versus -n (mostly

temporally non-adjacent).

Page 24: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

24

c. -t-anec/t-anquin adjacent sequential episodes

They moreover differ in the additional meanings they can take on: -anec and -anquin have a single

interpretation ‘after’, the element -t preceding these markers adds further possible interpretations as

reason clauses. The suffixes -ash and -shun can have additional interpretations condition, reason, and

concession (Fleck 2003: 1083). The suffix -ec in Matsés can also be used for purpose relations.

Finally, they differ in terms of a default object co-reference, which is strongest for -anquin, weakest

for -shun (Ibid. 1093).

With regard to non-identity markers, there is slightly more variation, as can be seen in Table 15,

which also includes data from Amawaka (Sparing-Chavez 2007) and Shanenawa (Vieira-Cândido

2004).

Table 15: Non-identity markers in the Panoan languages of the sample

shipibo cashibo amawaka shanenawa matses matis

anterior ketian an cun kɨn an an

ken bon bo

simultaneous nontian këbë(tan) hain aj na

ain mainun

aintian

posterior non nun non nun teno

Apart from temporal adjacency in at least Matsés and Shipibo-Konibo, there seems to be little

influence of discourse cohesion on the SR systems of Panoan languages. Nevertheless, in Kashibo-

Kakataibo there is a special construction involving two non-identical subjects, which may trigger

identity morphology.

(11) Kashibo-Kakataibo [PANOAN]. Zariquiey 2011: 592

a. uni pakët-këbë kaisa xanu kwan-akë-x-ín

man.ABS fall.down-NONID:S/A/P NAR.REP.3 woman.ABS go-REM.PST-3-PROX

‘It is said that, when the man fell down, the woman went.’

b. uni pakët-ia o-i kaisa xanu

man.ABS fall.down-IDENT.S/A/P>P FACT-S/A>S NAR.REP.3 woman.ABS

kwan-akë-x-ín

go-REM.PST-3-PROX

‘It is said that, when the man fell down, the woman went (but she saw him or was in

some way interested in or compromised with the event, because he was her husband,

her enemy or something like this).’

Page 25: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

25

In (11a) the dependent clause carries a different subject marker indicating no overlap at all between

the participants of both clauses. In (11b), however, the same two events are encoded by a different

construction, involving a factitive element o- with an identity marker (S/A to S). Moreover, the

dependent verb form pakët-ia carries an identity marker core to P. According to Zariquiey (Ibid, 593),

“the subject of the main clause tends to be interpreted (1) as a perceiver and/or (2) as someone

emotionally concerned with the event”. It seems therefore that the factitive element syntactically

functions as a transitive verb, with the same subject as the main clause, and the same object as the

subject of the dependent.15

Amawaka has different succession identity markers depending on whether the shared participant

has been identified before (-cun for succession, -hain for simultaneity) or not (-havan for succession, -

haivaun for simultaneity).

All in all it seems clear that the systems in the different Panoan languages come from a common

ancestor. So if SR has diffused into Panoan languages through contact, it must have happened before

the languages started to fission. The differences between the systems may be attributable to contact,

but may also be due to internal processes of change. There is no clear independent evidence to

substantiate either claim.

4.4. Jivaroan

The Jivaroan family is rather young, and is therefore expected not to display very much variation

between its members. Nevertheless, the SR system, as shown in the contributions to this volume by

Overall and Gnerre, is rather different from those of the surrounding languages.

Figure 5: the Jivaroan languages considered in this study

The main criterion that sets the Jivaroan SR systems apart from most other SR systems in South

America is the fact that their SR system is intertwined not only with interclausal relation markers, but

also with the person-number agreement system.

The basic system of Aguaruna identity markers is given in Table 16 (Overall 2007: 390).

15

A similar type of incorporation of indirect participation exists in Yurakaré, for directives (see Van Gijn 2011

for details).

JIVAROAN Shuar

Achuar

SHUARIC

Aguaruna

Page 26: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

26

Table 16: Aguaruna identity markers

SINGULAR PLURAL

First person -nu suppression of apocope / -i

Second person -mɨ

Third person nasalization of stem-final vowel

As can be seen in Table 16, the identity markers change according to the number of the subject, and -

in the singular - according to the person value of the subject. It can also be observed that some of the

person/identity markers are non-linear operations, another uncommon element in South American SR

systems.

The non-identity markers are slightly more complex, as they interact directly with interclausal

relation type and stem type as well, as can be seen in Table 17 (Overall 2007: 395)16

Table 17: Aguaruna non-identity markers.

RELATION PERSON/NUMBER

2SG 2PL 1/3

non-temporal - -taĩ

sequential -mɨ-nĩ -humɨ-nĩ

-mataĩ

simultaneous -ĩ

Table 17 shows that marking of non-identity relations, apart from person marking, interacts with

interclausal semantics. For the second person forms a different-subject element -nĩ can be

distinguished. The different interclausal semantics are in a number of cases overtly signalled: non-

temporal identity clauses carry an additional marker -sa (not present in non-temporal non-identity

clauses) which precedes the person/SR marker. Simultaneity clauses are additionally marked wih -ku

in both identity and non-identity clauses. A further conspicuous feature of non-identity markers in

Aguaruna is the first person - third person syncretism or collapse.

Finally, Aguaruna has a non-canonical subsystem of switch-reference with two markers, indicated

in Table 18.

Table 18: Non-canonical SR marking in Aguaruna

Suffix Role in marked clause Role in controlling clause

-ma non-subject subject

-tatamana subject object

16 The overview is slightly adapted to the needs of this paper, and is in fact more complex - see Overall 2007 and this volume

for details.

Page 27: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

27

A number of characteristics and markers fom the Aguaruna system are also found in the other two

Jivaroan languages. In Shuar (Gnerre 1999, Saad 2014), the identity markers can be described by the

same template, as shown in Table 19 (Saad 2014: 121). The main differences are found in the plural

identity markers.

Table 19: Identity markers in Shuar

SINGULAR PLURAL

First person -n -r(i)

Second person -mɨ -rum:

Third person nasalization of stem-final vowel -war/-r:

Similarly, the non-identity markers of Shuar look very similar to those in Aguaruna, following the

same template, with some slight differences between the markers. In Shuar, like in Aguaruna, an

element can be isolated for the second person forms, which contributes the different-subject meaning:

-n, shown in Table 20 (based on Saad 2014: 115).

Table 20: Non-identity markers in Shuar

RELATION PERSON/NUMBER

2SG 2PL 1/3

non-temporal - -taĩ

sequential -mi-n -rumi-n

-mataĩ

simultaneous -inia (...) -ĩ

Some of the interclausal semantics are overtly marked in the same way as in Aguaruna: -sa for non-

temporal identity clauses, -ku for simultaneity clauses. Ku- appears between the elements -inia and -ĩ

of the non-second person form.

Like Aguaruna, Shuar also has a non-canonical part to its SR system. The suffix -ma marks that the

object of the marked clause is identical to the subject of the reference clause, as can be seen in

example (12) where the relevant connection between the clauses triggering the marker -ma is the first

person participant.

(12) Shuar [JIVAROAN]. Saad 2014: 142

suma-ru-ka-ma nuat-ka-mia-ha

propose.to-1SG.OBJ-INTS.PFV-IDENT.OBJ>SUBJ marry.woman-INTS.PFV-REM.PST-1SG

‘After she proposed to me, I married her.’

The characteristics of the system in Achuar are less clear. The examples given in Fast (1981) give rise

to Table 21

Page 28: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

28

Table 21: SR markers in Achuar temporal clauses

1SG 2SG 3SG 1PL 2PL 3PL

SS kun kum k kur krum iniak

DS -when mtai k(u)min mtai krin krumin inai-mtai

DS-while jai amin jai arin armin uinai

There are some clear formal correspondences to Shuar forms as well as to Aguaruina forms - though to

a lesser extent, and in terms of the system, Achuar displays a 1-3rd

person syncretism in non-identity

relations (although the third person plural forms are slightly deviant, interaction with interclausal

semantics in non-identity clauses, and a recurring element -ku which may reanalysable as the

dependency-marking suffix also present in the other two Jivaroan languages. Achuar is currently

studied by Martin Kohlberger, which will undoubtedly lead to a more complete picture of the SR of

the language.

In sum, although the languages (or the analyses) seem to differ on some points, there is a clear

commonality to all systems, hallmarks of which are the forms of the markers, the intertwining with

person marking, a 1-3rd

person syncretism for the non-identity forms, and interactions between no-

identity marking and interclausal semantics.

4.5. Barbacoan

The Barbacoan language family consists of 7 (Hammarström et al. 2014) languages, mainly spoken in

western Ecuador, and south-western Columbia. The four Barbacoan languages of the sample (see

Figure 6) are Tsafiki (Dickinson 2002), Cha’paalachi (Vittadello 1988, Floyd 2010), Guambiano

(Branks 1980), and Awa Pit (Curnow 1997).

Figure 6: The Barbacoan languages of the sample

Table 22, from Floyd & Norcliffe (this volume) shows the basic structures of the SR systems in the

four languages.17

17 Vittadello does not regard -nu/-sa as SR opposites, and has some DS examples for -nu.

Tsafiki Cha’palaachi

Guambiano BARBACOAN COCONUCAN

CAYAPA-

COLORADO

UNCLASSIFIED

BARBACOAN Awa Pit

Page 29: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

29

Table 22: SR markers in Barbacoan languages

TEMPORAL PURPOSE

IDENTITY NON-IDENTITY IDENTITY NON-IDENTITY

Cha’paala tu / ʔ ɲu / ɲa nu sa

Tsafiki to sa

Guambiano a en

Awa Pit na napa

A much more detailed analysis of the Barbacoan languages is found in the contribution by Floyd and

Norcliffe to this volume, I confine myself to the main variables of this study. All markers are suffixes

that attach to verbs, but there are quite some differences between the forms of the SR markers,

although there seems to be some overlap between closely related Cha’paala and Tsafiki, whose

sequential identity markers seem to be related, as well as the non-identity purpose marker of Cha’paala

and the non-identity sequential marker of Tsafiki. Tsafiki also has a marker -nan that indicates disjoint

reference in reason clauses (though no same reference counterpart), and -ñu /-ña in Cha’paala seems to

extend to reason clauses.

(13) Cha’paala [BARBACOAN] (Floyd 2010: 118)

Ke-mu de-e-ñu'=mityaa jeen uyala ti-la-a-ka

do-AG.NMLZ PL-become-NONID=RES wild foreigner say-COL-FOC-DUB

‘It is because they do that that they call them “wild (forest)” foreigners.’

The temporal SR markers in Guambiano are unspecified for either sequentiality or simultaneity. There

are alternative construction types to specify (non-)completion but hey are much less common, and not

SR sensitive (Branks 1980: 14). Branks (Ibid.) also mentions the opposition -endu vs. -gocha for

counterexpectational events, but it is unclear whether the opposition refers to an identity versus non-

identity distinction.

Tsafiki hardly has any person/agreement marking on the verb; there is only a non-obligatory

subject plural marker -la. It is unclear whether that marker can appear in the marked clauses. In Awa

Pit, marked clauses, like nearly all subordinate clauses, are non-finite and lack person marking. In

Guambiano verbs marked with a SR marker do not carry any further person inflection. The switch-

reference markers may also occur on the conjunction marker inch- (inchen versus incha) to form

when/while clauses.

One of the salient aspects of Tsafiki SR is that it seems to encode event continuity rather than strict

subject (non-)coreference, although subject (non-)coreference is one of the factors that can trigger

either disjoint or same reference markers. Example (14) shows that referential continuity can be

Page 30: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

30

overridden by temporal continuity, in this case causing the disjoint reference marker to appear in spite

of referential continuity of the subject participants.

(14) Tsafiki [BARBACOAN] Dickinson 2002:138

Junni manjanasa, watate aman chide laribi manjimanti' .

junni man=ja-na-sa wata=te aman chide

then again=come-PROG-NONID year=LOC now bone

la-ri-bi man=ji-man-ti-e

come.out-CAUS-PURP again=go-SIT-REP-DCL

‘They say then coming back, after one year he went to take out the bones (of his dead wife).’

Event (dis)continuity can lead both to the appearance of identity markers in the absence of referential

identity, and - as in (14) to non-identity marking in the presence of referential identity (Dickinson,

p.c.)

The Barbacoan languages seem to show less coherence than some of the other families reviewed in

this paper, and it may well be that these languages were influenced by other languages. However,

many elements remain unclear, especially for Guambiano and Cha’paala. If one should posit an

original system for the Barbacoan languages, however, one would be inclined to include both the

temporal and purpose clauses, since the other languages in the area either do not have purpose clauses,

or if they do (like e.g. Imbabura and Chimborazo Quechua) they are deviations from the genealogical

pattern.

4.6. Tucanoan

The Tucanoan languages discussed in this study are given in Figure 7. Unfortunately, western

Tucanoan languages are seriously underdescribed and therefore underrepresented here.

Figure 7: The Tucanoan languages of the sample.

The basic pattern for eastern Tucanoan languages is slightly different from the previous cases

discussed in this section, with a single marker for non-identity, and a paradigm of nominalizers with

TUCANOAN

Cubeo Desano

Guanano Tuyuca

WESTERN

EASTERN

EAST

Siona

WEST

Page 31: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

31

information on gender and number for identity relations.18

Table 23 displays the set of SR markers for

Tuyuca, used in conditional clauses (see Barnes 1975, 1990, Galeano & Barnes 1977).

Table 23: SR markers of Tuyuca

identity non-identity

masculine singular gɨ

ri feminine singular go

plural ra

All of the other eastern Tucanoan languages in the sample, Guanano or Kotiria (Waltz 1981, Stenzel

2013, this volume), Desano (Miller 1999), Cubeo (Maxwell & Morse 1999) follow similar basic

patterns, although there do seem to be some differences in the categories recognized in the same-

subject categories: the Guanano system includes a special marker for non-SAP participants, whereas

for Cubeo, the relevant distinction is between animate and non-animate. The four Eastern Tucanoan

languages are compared in Table 24.

Table 24: Basic SR markers in four Eastern Tucanoan languages

IDENTITY NON-ID.

M.SG F.SG 3SG INAN.SG PL PL.INAN 3PL

Tuyuca gɨ go ra ri

Desano gɨ go rã kɨ

Guanano kɨ / i ko ro ro (g)a ʧɨ

Cubeo (ka)kɨ (ka)ko karõ kawɨ ke ereka

The Cubeo system seems to be the most elaborate, extending beyond temporal to conditional clauses

tand purpose clauses as well, and it includes a host of different markers. The Cubeo markers are given

in Table 25 (Maxwell & Morse 1999). 19

Table 25: The Cubeo SR system

IDENTITY NON-ID

M.SG F.SG NEUT PL.NONNEUT PL.NEUT

SIM/COND (ka)kɨ (ka)ko karõ kawɨ ke ereka

WHEN rĩ ere / rõre

PURP ø pe

18 I ignore the so-called implicit chain here described by Longacre for Guanano, because it does not meet the defining criteria

of SR (morphological marking). Moreover, as is discussed in Stenzel (this volume) it is questionable to what extent the DS

interpretation of those chains is a 19 The conditional uses the same bound forms as the simultaneous construction, but in addition is combined with the element

Page 32: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

32

CONC wakari wareka

The system of Siona is slightly different, although some similar building blocks recur. In Siona (Bruil

2014) temporal and (semantically) coordinate clauses, but has two systems for different temporal

situations. In the (relative) present tense, the dependent verbs are inflected for person, though in a

different way than would be the case in independent clauses. A three-way distinction is made between

feminine singular (-ko) masculine singular (-kɨ) and plural (-hɨ) in present tense contexts. In addition,

different-subject clauses are marked with the SR suffix -na. Effectively, this is a zero versus -na

system of marking SR. In the relative past tense, there is an opposition -ni (SS) versus -na (DS), and in

addition, the non-identity cases are marked for number/gender with a slightly different set of markers

(see Table 26).

Table 26: The Siona SR system

TENSE/ASPECT PERSON/NUMBER IDENTITY NON-IDENTITY

NUM/GEN SR NUM/GEN SR

PRESENT M.SG -ko -ko

F.SG -kɨ Ø -kɨ na

PL -hɨ -hɨ

PAST M.SG -o

F.SG Ø ni -ɨ na

PL -de

Although the system in Siona, the only western Tucanoan language in the sample, is slightly different

from the other systems, the basic contours in that there is an interaction with the number-gender

markers, are also in place, albeit in a different way.

For the Tucanoan languages, we can also conclude that the SR system shows clear similarities

across the family, which suggests that the SR system was established before the dispersal of the

languages. The fact that Siona, as a western Tucanoan language, has a slightly different system

suggests post-dispersal change, as does the more elaborate system of Cubeo.

4.7. Aymaran

Most Aymaran languages or dialects have reduced or lost their SR system, but comparative evidence

suggests that the family may have had a more elaborate SR system that interacted with person

marking. Central Aymaran dialects that have preserved an identity vs. non-identity opposition have a

system as in Table 27 (adapted from Cerrón-Palomino 2000, 2008).

Table 27: Aymaran SR system

Page 33: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

33

identity non-identity

simultaneous -sa -ipana

successive -sina

At least the non-identity marking can be expanded to mark conditional relations by adding the marker

-xa. A parallel construction exists in Quechuan languages (Ibid: 149).

There is evidence that the SR system was more elaborate in proto-Aymara. The Morocomarca and

Huancané dialects preserved an entire DS paradigm (Briggs 1993, cited in Cerrón-Palomino 2000:

244), shown in Table 28.

Table 28: The SR system of Morocomarca Aymara

Person value different-subject marker

1st person -inyana

2nd

person -imana

3rd

person -ipana

4th

person -isana

The paradigm, perhaps under the influence of Quechuan, broke down, and only the third person non-

identity form was preserved and generalized as a different-subject marker for all person values. As is

discussed by Cerrón-Palomino (2008) the extant SR systems of Aymara to a large extent have become

functionally equivalent to Quechuan SR systems.

The Aymaran languages, in spite of their historical and current importance, are underdescribed, and

many aspects of the SR systems in the languages of this family remain unclear. Therefore they are not

taken into account in the distance measures.

4.8. Uru-Chipaya

The small Uru-Chipaya family consists of two languages, Uru, or Uchumataqu, and Chipaya. Chipaya

has an active system of SR, whereas the now extinct languages Uchumataqu (Hanns 2008) seems to

have only remnants of a SR system. The SR system in Uchumataqu is strictly speaking not a SR

system in terms of the definition given above, but it is included here for the sake of correspondence (it

is not included in the database and the graph derived from it). It marks a combination of relative tense

and coreferentiality as indicates in Table 29 (Hannss 2011: 288):

Table 29: The SR markers of Uchumataqu

+coreferential/+simultaneous -ku

-coreferential/-simultaneous -na

Page 34: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

34

Hannss (Ibid. 289) stresses that the scanty data of this extinct language do not allow for any definite

conclusions about the SR system of Uchumataqu. The Uchumataqu system does seem rather different

from the one found in its sister language Chipaya, which has an infinitival marker -z which marks

same-subject complements of verbs of desire and knowledge. Different-subject counterparts for these

types of complements interact with tense/mood and aspect. Irrealis complements are marked with the

suffix -ñi, realis complements of knowledge verbs are marked with -chi or -ta depending on the

aspectual value, the former adding a resultative focus. The non-identity markers -chi and -ta have

nominalizing qualities, and all of the above markers have wider distributions where they do not seem

to mark referential discontinuity. In addition, Chipaya has a SR system for its purpose clauses, marked

by the suffixes -japa (identity)20

and -jo (non-identity), although there are exceptions to the (non-

)coreferentiality requirements (Cerrón-Palomino 2006: 262). Finally, Chipaya has a SR system for

temporal clauses, consisting of the markers in Table 30 (Ibid: 152).

Table 30: The temporal SR markers of Chipaya

identity non-identity

simultaneous -kan(a) -an(a)

consecutive -zhku; -a -tan(a)

The difference between -zhku and -a is that the former expresses a necessary condition or prerequisite

for the main event to take place, whereas the latter does not.

The surviving SR markers of Uchumataqu are probably related to the simultaneity markers of

Chipaya, suggesting an older SR system which broke down in Uchumataqu, retracting to mark

temporal clauses only, perhaps under the pressure of Quechuan and Aymaran languages, as well as

Spanish (Hannss 2008: 8-10). Language contact and language loss seems to have had a destructive

effect on the SR system in the case of Uchumataqu.21

4.9 Isolates

It is difficult to say anything about the diachrony of the SR systems in the isolates of the area. There

are SR systems in Yurakaré, Kwazá, Paez, Cofán, and Urarina. The first two languages are spoken in

the south of the area under consideration in this paper (Bolivia/east Brazil), the latter three in the north

(north Peru, Ecuador, south Colombia).

A review in terms of formal overlap, given in Table 31 does not reveal obvious correspondences.

Potential formal correspondences, e.g. between the non-identity markers of Kwazá and Cofán, or

between the identity markers of Kwazá and Páez are excluded to be the result of contact, given the

20

The identity marker requires the presence of mediopassive -z(i). 21

See also Hannss (2011) for a more complete comparison between Uchumataqu and Chipaya subordination

strategies, which tend to be more analytical in Uchumataqu and more synthetic and concatenating in Chipaya.

Page 35: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

35

enormous geographical distance between the languages. Even Páez and Urarina, whose non-identity

markers have the element -na are spoken too far apart for them to have been of direct influence to each

other.

Table 31: SR markers of the isolates in the sample

Identity Non-identity

Yurakaré (Van Gijn 2006) -ja [ha]

-ya [ja]

-ti [ti]

Kwazá (Van der Voort 2004) -t(j)a [ta], [tja] -si; -dyta [si] [dɨta]

Paez (Jung 2008) -tsja [ ts

ja]

-i: [i:], [j]

-mej [mej]

-te [te]

-i:na [i:na] [jna]

-mejna

Cofán (Fischer 2007) -pa [pa] [mba] -si [si]

Urarina (Olawsky 2006) -ne [ne] -na [na]

The Yurakaré SR system partly looks like (some) Quechuan languages22

in that SR is in particular

found in temporal and conditional clauses, and in that it has both finite (person-marked) and non-finite

constructions, but is different in that the finite forms are not nominalizations with a possessor subject.

In terms of semantic relation types Yurakaré goes further than most other languages in that it includes

relative clauses and perception complements into the SR system. Like Quechuan languages, chains are

usually short, and the language makes extensive use of tail-head linkage (Van Gijn 2014b) - as does

Cavineña (Guillaume 2011). Interestingly, the SR markers are also found elsewhere in the grammar,

with different functions. The marker =ja is a theme marker, especially used on subject NPs, the irrealis

identity marker =ya is homophonous with the reportative from the finite verbal paradigm, which in

turn probably derives from the verb -ya- ‘to answer’. The non-identity marker -ti possibly derives from

the demonstrative ati. There may be potential for a contact-induced influence, possibly from Quechua,

or perhaps from Tacanan.

The SR system in Kwazá (as well as in the nearby isolate language Aikanã) is characterized by

longer chains. There is furthermore a slightly different situation for the third person as compared to the

SAPs: whereas the latter function on a -ta versus -si opposition, the former have an opposition of -t(j)a

versus -dyta. The SR system may be used for a number of different semantic relations, like coordinate

relations, purpose, manipulation (to order), reason, consecutive, and overlapping temporal relations, a

cosubordinate clause “procures its meaning from the context” (Van der Voort 2004: 654). Although it

is unclear what - if any - the diachronic precedents of the Kwazá system are, there is a potential case of

convergence with another isolate of the area, Aikanã (Hein van der Voort, p.c.).

The three northern isolates, Paéz, Cofán, and Urarina are also quite different. The Paéz markerd

opposition between -tsja and -te can be used for (at least) temporal ‘when’ relation, hypothetical

22 Among them the closest variety: Bolivian Quechua (see Lastra 1968).

Page 36: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

36

conditionals, and reason relations, and the addition of the suffix -pa to each of the SR markers encodes

concessive relations. There are additional oppositions: -i: (or -j after vowels) versus -i:na (or -jna after

vowels) for simultaneity relations, and -mej versus -mejna for ‘without’-clauses. These seem to be

reanalyzable into one opposition of zero (identity) versus -na.

Cofán markers -pa (/mba/ after a nasal vowel) and -si mark sequential temporal relations (and by

extension also reason clauses). SR-marked clauses often appear in chains, and seem to display

different degrees of integration with the main clause. Positionally they are relatively free, appearing

before, inside and after main clauses (Fischer 2007).

In Urarina, the only construction that falls within the definition of SR given above are complements

of desire. Same-subject want complements are marked with the infinitive marker -na and do not carry

any person inflection; different-subject want complements are marked with the suffix -ne and do carry

person inflection.

(15) Urarina [ISOLATE], Olawsky 2006: 762, 767

a. kʉ kunʉhe-ʉrʉ hãʉ ɲãe atiĩ be-na

there insist-PL-3 because already nevertheless tell-IDENT

here-ne

want-NEG:3

‘As they insisted, she still did not want to tell.’

b. tʉrʉ-s=ne heri-ji

arrive-3=NONID want-NEG:3

‘He(i) does not want him(j) to come.’

For the northern isolates, it makes sense to look at Barbacoan, Tucanoan, and Jivaroan languages for

possible contact influences. Siona has a different subject marker -na for temporal clauses, Aguaruna

and Shuar have a different-subject temporal clause marker -n and Cha’palaa has a different-subject

temporal marker -ñu or -ña. Imbabura and Chimborazo Quechua, as well as Awa Pit have similar

markers for non-identity purpose clauses, containing the element -na. The correspondences between

these markers and with the non-identity markers in Páez and Urarina may certainly be coincidence,

given the rather common and short form /na/ (and especially given the fact that the far-away Uru-

Chipaya family has similar non-identity markers) but they may also be the result of contact. It remains

a conspicuous fact that the area contains languages from several different families that have SR

systems that particularly mark temporal clauses (Urarina excepted).

4.10. Odd-ones-out

Page 37: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

37

A few languages in the area that belong to families that, as far as I am aware do not display SR

systems, have nevertheless developed a SR system. They are Tariana (Arawak), Mamaindê

(Nambikwaran), and Tunebo (Chibchan).

Chibchan does not seem to have SR marking among its features23

but Tunebo seems to have

developed some sort of SR system. The Tunebo system (ignoring the unmarked chain suggested by

Headland & Levinsohn 1977 because it does not meet the defining criteria discussed in Section 2)

consists of two markers which are not entirely equivalent semantically. The suffix -r marks same-

subject for sequential clauses, and the suffix yat/yet marks different subjects for (partially) overlapping

events.

(16) Tunebo [CHIBCHAN], Headland 1994: 42-43

a. carretera cú-ro-ri. cu-r carro cuit bi-ro-ri.

road arrive-DCL-CONTR arrive-SEQ.IDENT car INS go-DCL-CONTR

‘I arrived at the road. Having arrived, I went in a car.’

b. cuat biht-in yar-yat esi cu-ca-ro.

but adult-GRP die-WHEN.NONID five blow-PRES-DCL

‘But when an adult dies, they blow for five nights.’

There is a third marker, yiror, which encodes sequential events where the subject of the first clause

leaves the scene, while the scene remains the same.

(17) Tunebo [CHIBCHAN], Headland 1994: 43

as bi yiror canorí tani ya-wi.

I go after work well do-IMP

‘After I leave, do the work well.’

Headland & Levinsohn (1977) discuss an unmarked chain which has apparent commonalities with the

system in Tucanoan Guanano as described by Longacre (1983) - though see Stenzel this volume for a

critique of Longacre’s study - suggesting the Tucanoan languages as a possible source for the SR

constuctions in Tunebo, though the marked constructions discussed here do not show an affinity with

Tucanoan structures.

Mamaindê is a Nambikwaran language spoken in Rondônia, Brazil. For a few of its interclausal

relations, it makes a difference between same-subject and different-subject constructions, as is shown

in Table 32 (after Eberhard 2009):

23

Quesada (2007) does discuss a system that comes relatively close. In Boruca, the topic continuity marker ang

(a former focus marker) can sometimes be affixed to the verb (p. 214), and in some contexts it can be opposed to

the marker ki (which is not suffixed to verbs) which marks new (focal) subjects. Ang is also used as a relativizer

of subjects (versus ki the relativizer of objects).

Page 38: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

38

Table 32: The SR markers of Mamaindê

identity non-identity

Sequential kʰatoʔ hĩʔ

sĩʔ

Purposive teʔ

teʔn(a)taʔ

siʔ

sihta(ku)ʔ

If this is a contact-induced system, the most likely source from a geographical proximity perspective,

is Kwazá. The most conspicuous observation from the comparison between Kwazá and Mamaindê is

the existence of similar non-identity forms: Kwazá -si versus Mamaindê -sĩʔ and - sihtaʔ as well as

elements related to the form [ta] in both languages.24

Like Kwazá, Mamaindê has clause chains, and

although Kwazá is more implicit about the interclausal semantics than Mamaindê, SR chains in Kwazá

include temporal and purpose relations as well. Contact-induced change seems to be at least a possible

explanation.

SR is not a feature of Arawakan languages. Nevertheless, the Arawakan language Tariana has

developed a switch-reference system, under the influence of intense contact with speakers of

Tucanoan languages. Tariana is spoken in the Vaupés area around the Brazilian-Colombian border is a

well-studied linguistic area where speakers of different languages, especially Arawakan and Tucanoan

are in contact with each other (Aikhenvald 2002). Tariana and a number of Eastern Tucanoan societies

(Desano, Wanano, Piratapuya, Tucano, and Bará)25

practice institutionalized exogamy, and a very

strong sense of identity, of which their respective languages (acquired through patrilineal descent) are

important manifestations. There is a general prohibition to borrow words from the other languages,

though there is widespread bilingualism. Tucano acquired a special status during the time of the

Catholic missions, for which it became the lingua franca, which changed the status of the Tucano, and

gave the unilateral linguistic diffusion from Tucano to Tariana an important impetus. Strict

segregation of languages was also relaxed more recently.

Aikhenvald (2002: 158-161) discusses diffusion of clause-linking strategies from Tucano to

Tariana, including a SR system. Tucano, as most Eastern Tucanoan languages, marks identity clauses

with a paradigm of nominalizations that also convey information on person, gender (both in the

singular only), and number. The animate oppositions of Tucano for simultaneity/reason relations are

given in Table 33 (from Ramírez 1997, and Aikhenvald 2002)

Table 33: Tucano SR markers

24

The element -taʔ in the Mamaindê system seems to mark purpose rather than SR per se. Nevertheless, the

same element is found on same-subject simultaneity clauses (which have no DS counterpart, and are therefore

not shown in Table 32). 25 Not all combinations are possible. Tha Tariana, for instance cannot marry Desano (Aikhenvald 2002: 22).

Page 39: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

39

identity non-identity

singular non-feminine -gɨ

-kã singular feminine -go

plural -rã

Tariana uses a number of enclitics for identity and non-identity clauses (Aikhenvald 2003: 516), with a

basic opposition between sequential/causal and simultaneous clauses, given in Table 34.26

Table 34: Tariana SR markers

-hyume ‘after, because’ -kayami ‘after’

-kapua ‘because’

-kakari ‘while’

-nikhe ‘while’

-kariku ‘while’

-nisawa ‘while’

The Tariana enclitics do not have any gender or number information on them, but person marking is

obligatory on these SR-marked predicates. Some of these markers have nominalizing qualities to the

extent that they allow for case markers. In terms of semantic relation types there is also overlap

between the two systems. Aikhenvald furthermore mentions that Tariana has an implicit-chain-like

structure where there is no dependency marking on the verbs, and implied combinations of

sequentiality/simultaneity and same or different subject. Finally, Tariana has adopted other kinds of

structures to do with complex clauses from Tucano like nominalizations as subordination strategies for

forming complement clauses and relative clauses, as well as other parallel structures.

In the case of Tariana, then, the SR system adopted differs in a number of ways from the one found

in Tucano, and would not necessarily strike one as the result of diffusion through contact, were it not

for the many other parallels with Tucano structures and the ethnological evidence of the contact

relations between the speakers of Tucano and Tariana. Moreover, Tariana differs from its sister

languages in that it has SR. This may serve as a warning against discarding contact-induced accounts

of superficially similar phenomena too soon.

5. Discussion

On the basis of the variables discussed in section 2 I have created a questionnaire on the basis of

which distances between the languages can be calculated. These distances are visualized in the

Neighbor-Net network (Bryant & Moulton 2003) in Figure 8, which will serve as a basis for

discussion in the remainder of this section.

26 The oppositions are not entirely equivalent, there is a meaning difference between -kariku (overlapping) and -nisawa

(simultaneous). -kakari and nikhe differ in that only the former allows for case marking. There are furthermore other, non-

productive markers that are or were SR sensitive.

Page 40: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

40

Figure 8: Neighbor-net of distances between languages of the sample.

Comparing the languages of western South America in broad terms, there are a number of clear

similarities between the SR systems. In terms of SR markers there is a very clear preference for

suffixes that attach to the dependent verb, noth for identity and for non-identity relations. Deviations

from this pattern seem to be genealogically determined. Some of the Tacanan languages have enclitic

markers that attach to the clause (Cavineña) or to the verb+auxiliary group (Araona)27

and Araona also

has phonologically independent SR markers. Non-linear morphology to mark SR is found in the

Jivaroan languages. Zero markers for identity relations are found in Siona, and possibly also in Páez.

Although languages differ in terms of the interclausal semantics of SR clauses, there is a very clear

and almost implicational preference for marking SR in temporal clauses. Many languages also make a

distinction between simultaneity and succession clauses. Other common interclausal semantics are

reason, condition, purpose, and concession.

27

Vuillermet (2013) also analyzes the SR markers of Ese Ejja as enclitics because they behave differently from

affixes in phonological terms, though not in terms of selection restrictions, which was the criterion I followed.

Quechuan Tucanoan

Barbacoan

Jivaroan

Panoan

Tacanan

Page 41: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

41

Some more differences are found with respect to the orthogonal categories that are marked by the

SR markers. A number of languages have SR markers that simultaneously mark referential features of

the subject, like the identity markers of te Eastern Tucanoan languages and the SR markers of Jivaroan

languages and the non-identity markers of the full Aymaran paradigm. Interactions with mood are

found in Chipaya, Yurakaré, and Imbabura Quechua, and simultaneous expression of SR and

interclausal semantics are found in Tacanan (except Cavineña), Panoan28

, Jivaroan languages, as well

as some individual languages like Cubeo, Páez, and Chipaya.

Most languages have SR systems that function on the basis of a subject (S/A) pivot, although clear

exceptions are the Panoan languages, Tacanan Ese Ejja, and - to a lesser extent - Jivaroan languages

Aguaruna and Shuar. For a number of languages, although the S/A category forms the basis for the SR

system, other factors interfere, like topicality-related ones such as referentiality in Imbabura Quechua,

person (SAP versus non-SAP) in Kwazá, Guanano and to some extent in Jivaroan languages, animacy

in Cubeo. Whether or not the participant has been identified is of importance for some of the identity

markers in Amawaka. Other types of pragmatic co-determination of the SR system include inter-event

cohesion in Tsafiki, Ancash Quechua, and possibly Cuzco Quechua, as discussed above, and temporal

coherence for some of the succession markers in Shipibo-Konibo and Matsés. Issues relating to

indirect participantion are found in Yurakaré, Imbabura Quechua and Kashibo-Kakataibo.

Although there were no clear patterns in the data on the additional factors playing a role in SR

systems in South America, it is likely that a closer study of “exceptions” to the referential (non-

)identity patterns in the different languages will yield more data, and will perhaps reveal some

interesting areal and/or genealogical patterns. At present it seems premature to draw any conclusions

about this aspect of SR in South American languages.

For most families, an family-internal pattern can clearly be discerned. This is true for Quechuan,

Tucanoan, Panoan, and Jivaroan, which show up as relatively clear clusters in Figure 8, as well as for

Aymaran and Uru-Chipaya (which cannot be seen in Figure 8 because they are not represented in it,

except for Chipaya) and it seems to be true to a somewhat lesser extent for Barbacoan and Tacanan

families. This suggest pre-dispersion presence of SR in the area.

In addition to the commonalities, some family-internal differences are also clear. The northern

Quechuan languages are markedly different from the central and southern ones, in the sense that they

have developed SR for purpose clauses and in the sense that they have lost all person marking in SR

clauses. In Figure 8 it can be seen that the Ecuadorian varieties Imbabura and Chimborazo are clearly

separated from the Quechuan group, although they do not end up close to Barbacoan Awa Pit and

Tsafiki, presumably because those systems have undergone some developments not shared by these

northern Quechua varieties. There are also evident differences between the different Panoan systems

in terms of how elaborate they are, though their commonalities are more impressive, resulting in the

Panoan cluster in Figure 8. Compared to Chipaya, Uchumataqu has a much poorer system, which is

28

Alternatively the Panoan system can be analyzed as interacting with (relative) tense.

Page 42: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

42

also true for Cavineña and Araona in comparison to their Tacanan sister Ese Ejja, although

commonalities between Araona and Ese Ejja are still observable, so that they end up relatively close

together in Figure 8 (though Ese Ejja is in fact closer to Shipibo-Konibo than to Araona). Cavineña is

positioned relatively close to the Quechuan languages, which may have influenced the SR system of

the Tacanan language. As mentioned, most Aymaran systems have eroded considerably or even

disappeared. The Barbacoan languages, only represented by Awa Pit and Tsafiki in Figure 8, suggest a

common underlying system, but also important differences between the languages. There seems to be

a small northern group in Figure 8 close to the Quechuan languages, consisting of Barbacoan Awa Pit

and Tsafiki, Cofán, Tunebo, and Urarina.

Contact-induced language change is the most likely cause of these family internal differences. The

most likely scenario for the rise of SR constructions for purpose clauses in Ecuadorian Quechua

varieties is contact with Barbacoan languages. The presence of Barbacoan languages in Ecuador was

probably more dominant in the past (Bruil 2008) and it is also possible that Barbacoan to Quechuan

language shift took place. In this scenario, the original Barbacoan SR system included SR marking for

temporal and purpose clauses which diffused to the northern Quechuan varieties. The SR marking

system for purpose clauses in Imbabura Quechua seems to be in a process of re-interpretation as a

system of marking subjunctive, expanding to mark complements of verbs of desire and allowing for

exceptions to the same versus different subject pattern under the influence of Spanish (Bruil 2008).

Cerrón-Palomino (2008) shows that the SR structures of Aymaran and Quechuan languages display

isomorphic tendencies, gravitating towards the Quechuan system. For Aymaran this means reduced

complexity of the system, preserving only one of the non-identity forms. This simplification may have

set the stage for loss in a number of Aymaran languages. Likewise, the erosion of the Uchumataqu

system is related to language contact with Quechuan and Aymaran languages, as well as with Spanish

later on, and concomitant language loss starting in the Inca era (and likely also to data availability).

Contact with Quechuan languages may also have been of influence on the Tacanan languages, where

Cavineña, and has acquired a more Quechuan-like SR system, applying to temporal clauses only (to a

lesser extent this is may be true for Araona as well). Archaeological and ethnohistorical evidence

suggest Inca presence in the Beni delta (Renard-Casevitz et al. 1988) which may have led to Quechua-

Tacana language contact.

6. Conclusion

The patterns found for SR marking in western south America suggest that SR has been present in the

area for a relatively long time, dating back to before the fissioning of language families like Quechuan,

Panoan, and Tucanoan, based on the recurring patterns found throughout these families. At the same

time, the data suggest a sensitivity to contact-induced diffusion, evidenced not only by the fact that

languages belonging non-SR families have developed SR systems under the influence of neighbouring

Page 43: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

43

languages, but also by the fact that some members of SR families have become more like their

neighbours. This is in particular true for Ecuadorian Quechua, Aymaran, and perhaps also Tacanan.

Furthermore, there are some tentative traces of contact in the systems of some of the isolates in the

area.

The latest developments in these areal diachronic processes point at erosion and loss of SR

systems, most notably in Aymaran, Uchumataqu, and perhaps Tacanan languages, and more recently

in Kakataibo (Zariquiey, this volume) and Imbabura Quechua (Bruil 2008), under the influence of

Spanish.

Abbreviations

A transitive subject

AG agent

ABS absolutive

ADV adverbializer

ACC accusative

AUX auxiliary

CAUS causative

CLF classifier

CMPR comparative

COL collective

COM comitative

CONTR contrastive

COP copula

DCL declarative

DEF definite

DUB dubitative

ERG ergative

EVID evidential

F feminine

FM formative

FOC focus

FUT future

GEN gender

GRP group membership

IDENT identity

IMP imperative

INAN inanimate

INCL includive

INS instrument

INTS intensive

LOC locative

M masculine

MAL malefactive

MIR mirative

NAR narrative

NEG negation

NEUT neuter

NMLZ nominalizer

NONID non-identity

NUM number

OBJ object

P transitive object

PST past

PERF perfect

PFV perfective

PL plural

POSS possessive

POT potential

PRES present tense

PRN pronoun

PROG progressive

PROX proximal

PURP purposive

REM remote

REP reportative

RES resultative

S intransitive participant

SAP speech act participant

SEQ sequential

SIT situational

SG singular

SR switch reference

SUBJ subject

TR transitive

Page 44: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

44

References

Adelaar, Willem. 1977. Tarma Quechua: grammar, texts, dictionary. Peter de Ridder Press.

Adelaar, Willem. 1987. Morfología del Quechua de Pacaraos. Lima: Universidad Nacional de San

Marcos.

Adelaar, Willem. 2011. Participial clauses in Tarma Quechua. In: R. van Gijn, K. Haude & P.

Muysken (eds.) Subordination in native South American languages. Amsterdam/Philadelphia: John

Benjamins, pp. 267-280.

Adelaar, Willem, with Pieter Muysken. 2004. The languages of the Andes. Cambridge: Cambridge

University Press.

Aikhenvald, Alexandra Y. 2002. Language contact in Amazonia. Oxford: Oxford University Press.

Aikhenvald, Alexandra. 2003. A grammar of Tariana, from northwest Amazonia. Cambridge:

Cambridge University Press.

Austin, Peter. 1980. Switch-reference in Australian languages. In: Pamela Munro (ed.). Studies of

switch-reference. Los Angeles: University of California, pp. 7-47.

Austin, Peter. 1981. Switch-reference in Australia. Language 57 (2): 309-334.

Barnes, Janet. 1975. Notes on Tuyuca discourse, paragraph and sentence. Lomalinda: Instituto

Lingüístico de Verano and Ministerio de Gobierno.

Barnes, Janet. 1990. Classifiers in Tuyuca. In D. Payne (ed.), Amazonian Linguistics: Studies in

Lowland South American Languages. Austin: University of Texas Press, pp. 273-292.

Beukema, Ronald William. 1975. A Grammatical Sketch of Chimborazo Quichua. Yale University.

Ann Arbor: UMI.

Bickel, Balthasar 2010a Capturing particulars and universals in clause linkage: a multivariate analysis.

In: Isabelle Bril (ed.) Clause linking and clause hierarchy. Amsterdam: John Benjamins, pp. 51-

101.

Bickel, Balthasar. 2010b. Towards a multivariate typology of reference tracking. Presentation given at

the meeting of the Research Group 742 “Grammar and Processing of Verbal Arguments”, Leipzig,

April 21. Available at: http://www.uni-leipzig.de/~va/Dokumente/SR_Bickel.pdf.

Birchall, Joshua. 2014. Argument marking patterns in South American languages. PhD Thesis

Radboud University Nijmegen. Utrecht: LOT Publications.

Branks, Thomas. 1980. La cohesion y prominencia en las narrativas del guambiano. Lomalinda,

Colombia: Instituto Lingüístico de Verano.

Briggs, Lucy. 1993. El idioma Aymara: variantes regionales y sociales. La Paz: Instituto de Lengue y

Cultura Aymara.

Bruil, Martine. 2008. Innovations in the Ecuadorian Converb Systems: grammatical change in

language contact situations. MA thesis Leiden University.

Page 45: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

45

Bruil, Martine. 2014. Clause-typing and evidentiality in Ecuadorian Siona. PhD Thesis Universiteit

Leiden. Utrecht: LOT Publications.

Bryant, David & Vincent Moulton : Neighbor-net, an agglomerative method for the construction of

phylogenetic networks. Molecular Biology and Evolution 21(2): 255-265,

Büttner, Thomas Th. 1983. Las lenguas de los Andes centrales: estudios sobre la clasificación

genetica, areal y tipológica. Madrid: Instituto de cooperación iberoamericana.

Cerrón-Palomino, Rodolfo. 2000. Lingüística aimara. Cusco: Centro de Estudios Regionales Andinos.

Cerrón-Palomino, Rodolfo. 2006. El Chipaya o Lengua de los Hombres del Agua. Lima: Fondo

Editorial.

Cerrón-Palomino, Rodolfo. 2008. Quechumara: estructuras paralelas de las lenguas quechua y

aymara. La Paz: Plural.

Cole, Peter. 1982. Imbabura Quechua. Amsterdam: North-Holland.

Cole, Peter 1983. Switch-reference in two Quechua languages. In: John Haiman & Pamela Munro

(eds.) Switch reference and Universal Grammar. Amsterdam & Philadelphia: John Benjamins, pp.

1-16.

Corbett, Greville. 2005. The canonical approach in typology. In: Zygmunt Frajzyngier, Adam Hodges

& David S. Rood (eds) Linguistic Diversity and Language Theories. Amsterdam : John Benjamins,

25-49.

Crevels, Mily and Hein van der Voort. 2008. The Guaporé-Mamoré region as a linguistic area. In

Pieter Muysken (ed.) From Linguistic Areas to Areal Linguistics, 151-179. Amsterdam and

Philadelphia: John Benjamins.

Curnow, Timothy. 1997. A grammar of Awa Pit (Cuaiquier): An indigenous language of south-

western Colombia. PhD Thesis Australian National University, Canbera.

Curnow, Timothy and Anthony Liddicoat. 1998. The Barbacoan languages of Colombia and Ecuador.

Anthropological Linguistics 40, 3, pp. 384–408.

Dahl, Östen. 2008. . An Exercise in a posteriori Language Sampling. Sprachtypologie und

Universalienforschung 61 (3): 208–220.

Derbyshire, Desmond C. & Geoffrey K. Pullum. 1986. Introduction. In: Desmond C. Derbyshire &

Geoffrey K. Pullum (eds.) Handbook of Amazonian Languages, Volume 1, 1-28. Berlin: Mouton de

Gruyter.

Dickinson, Connie. 2002. Complex Predicates in Tsafiki. PhD dissertation University of Oregon,

Portland.

Dixon, R.M.W. & Alexandra Y. Aikhenvald. 1999. Introduction. In: R.M.W. Dixon & Alexandra Y.

Aikhenvald (eds.) The Amazonian languages. Cambridge: Cambridge University Press.

Dobson, Rose M. 2005. Aspectos da língua Kayabi. Cuiabá: Sociedade Internacional de Lingüística

[Série Lingüística, 12].

Page 46: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

46

Eberhard, David M. 2009. Mamaindê Grammar: A Northern Nambikwara language and its cultural

context. PhD dissertation Vrije Universiteit Amsterdam. Utrecht: LOT Publications.

Ebina, Daisuke. 2011. Cusco Quechua. In Yamakoshi, Yasuhiro (ed.), Grammatical Sketches from the

Field, 1-39. Tokyo: Research Institute for Languages and Cultures of Asia and Africa (ILCAA),

Tokyo University of Foreign Studies.

Emkow, Carola. 2006. A grammar of Araona. Unpublished PhD Thesis LaTrobe University.

Ferreira, Rogério. 2005. Língua Matis (Pano): Uma Descrição Grammatical. Universidade Estadual de

Campinas.

Fischer, Rafael. 2007. Clause linkage in Cofán (A'ingae). In: L. Wetzels (ed.): Language

Endangerment and Endangered Languages: Linguistic and Anhropological Studies with Special

Emphasis on the Languages and Cultures of the Andean-Amazonian Border Area. Leiden:

Research School of Asian, African and Amerindian Studies (CNWS), pp. 381-399.

Fischer, Rafael & Eva van Lier. 2011. Cofán subordinate clauses in a typology of subordination. In: R.

van Gijn, K. Haude and P. Muysken (eds.) Subordination in native South American languages.

Amsterdam/Philadelphia: John Benjamins, pp. 221-250.

Fleck, David William. 2003. A grammar of Matsés. University of Houston. Ann Arbor: UMI.

Floyd, Simeon. 2010. Discourse Forms and Social Categorization in Cha'palaa. PhD dissertation The

University of Texas at Austin.

Foley, William & Robert van Valin. 1984. Functional syntax and Universal Grammar. Cambridge:

Cambridge University Press.

Galeano, Luis & Janet Barnes. 1977. Relaciones entre las proposiciones en la lengua tuyuca. In

Heinze, Carol (ed.), Estudios tucanos 2. Bogota: Ministerio de Gobierno, pp. 99-127.

Galucio, Ana Vilacy. 2001. The morphosyntax of Mekens (Tupí). PhD thesis University of Chicago.

Gijn, Rik van. 2006. A grammar of Yurakaré. PhD dissertation Radboud University Nijmegen.

Gijn, Rik van. 2011. Grammatical and semantic integration in Yurakaré subordination. In: R. van Gijn,

K. Haude and P. Muysken (eds.) Subordination in native South American languages.

Amsterdam/Philadelphia: John Benjamins, pp. 169-192.

Gijn, Rik van. 2012 Switch attention (aka switch reference) in South American temporal clauses:

facilitating oral transmission. Linguistic Discovery 10 (1): 112-127.

Gijn, Rik van. 2014a. The Andean foothills and the adjacent Amazonian fringe. To appear in: L.

O'Connor and P. Muysken (eds.) The native languages of South America: origins, development,

typology. Cambridge University Press.

Gijn, Rik van. 2014b. Repeated dependent clauses in Yurakaré. In: R. van Gijn, D. Matic, J.

Hammond, S. van Putten and A. Galucio (eds.) Information structure and reference tracking in

complex sentences. John Benjamins, pp. 291-308.

Page 47: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

47

Gijn, Rik van & Fernando Zúñiga. 2014. Word and the Americanist perspective. Morphology 24(3)

[Special issue ed. by Rik van Gijn and Fernando Zúñiga: Wordhood: Theory and typology from an

Americanist perspective].

Gijn, Rik van, Fernanda Nogueira & Ana Vilacy Galucio. 2015. Subordination strategies in Tupi

languages. Comparative Tupi and Tupi-Guarani linguistics: Genetic and Typological Aspects

[Special issue of the Boletim do MPEG, Belém, edited by Wolf Dietrich and Sebastian Drude].

Gijn, Rik van & Harald Hammarström. A construction‐based approach to measuring distances

between languages: subordination strategies in South America.

Gnerre, Maurizio. 1999. Perfil Descriptivo e historico-comparativo de una lengua Amazonica: El

Shuar (Jíbaro). Nápoles: Istituto Universitario Orientale.

Guillaume, Antoine. 2008. A grammar of Cavineña. Berlin/New York: Mouton de Gruyter.

Guillaume, Antoine. 2012. Maropa (Reyesano). In: M. Crevels & P. Muysken (eds.) Las lenguas de

Bolivia, vol II: Amazonía. La Paz: Plural, pp. 191-229.

Hammarström, Harald & Forkel, Robert & Haspelmath, Martin & Nordhoff, Sebastian. 2014.

Glottolog 2.3. Leipzig: Max Planck Institute for Evolutionary Anthropology.

(Available online at http://glottolog.org)

Hannss, Katja. 2008. Uchumataqu: The lost language of the Urus of Bolivia. A grammatical

description of the language as documented between 1894 and 1952. Leiden: Research School of

Asian, African and Amerindian Studies (CNWS).

Hannss, Katja. 2011. Complex sentences in Uchumataqu in a comparative perspective with Chipaya.

In: R. van Gijn, K. Haude and P. Muysken (eds.) Subordination in native South American

languages. Amsterdam/Philadelphia: John Benjamins, pp.281-306

Haiman, John & Pamela Munro. 1983. Introduction. In: John Haiman & Pamela Munro (eds.). 1983.

Switch-reference and Universal Grammar. Amsterdam/Philadelpha: John Benjamins, ix-xv.

Hartmann, Roswith, ed. 1987. Rimaykullayki. Unterrichtsmaterialien zum Quechua Ayacuchano -

Peru. Berlin: Dietrich Reimer Verlag.

Headland, Edna. 1994. Diccionario bilingüe Tunebo-Español, Español-Tunebo con una breve

gramática Tuneba. University of Texas at Arlington. Ann Arbor: UMI.

Headland, Paul & Stephen Levinsohn. 1977. Prominence and cohesion in Tunebo discourse. In: R.

Longacre & F. Woods (eds.): Discourse grammar Studies in the languages of Colombia, Panama,

and Ecuador, volume 2. Norman: Summer Institute of Linguistics, pp. 133-157.

Heine, Bernd & Tania Kuteva. 2005. Language Contact and Grammatical Change. Cambridge:

Cambridge University Press.

Howkins, Angela Margaret. 1977. Syntactic relations in San Martin Quechua. PhD thesis University

of St. Andrews.

Huang, Yan. 2000. Anaphora. Oxford: Oxford University Press.

Page 48: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

48

Jacobsen, William H. 1983. Typological and genetic notes on switch-reference systems in North

American Indian languages. In: John Haiman & Pamela Munro (eds.) Switch reference and

Universal Grammar. Amsterdam & Philadelphia: John Benjamins, pp. 151-184.

Jensen, Cheryl. 1999. Tupí-Guaraní. Dixon & Aikhenvald (eds.), 125-164.

Jung, Ingrid. 2008. Gramática del páez o nasa yuwe: descripción de una lengua indígena de

Colombia. München: Lincom Europe.

Kerke, Simon van de. 1996. Affix order and interpretation in Bolivian Quechua. PhD Thesis

University of Amsterdam.

Key, Mary Ritchie. 1968. Comparative Tacanan Phonology. The Hague, Paris: Mouton

Krasnoukhova, Olga. 2012. The noun phrase in the languages of South America. PhD Thesis Radboud

University Nijmegen. Utrecht: LOT Publications.

Lastra, Yolanda. 1968. Cochabamba Quechua Syntax. The Hague: Mouton.

Lefebvre, Claire & Pieter Muysken. 1988. Mixed categories: nominalizations in Quechua. Dordrecht:

Kluwer.

Longacre, Robert E. 1983. Switch-reference systems in two distinct linguistic areas: Wojokeso (Papua

New Guinea) and Guanano (Northern South America). In: John Haiman & Pamela Munro (eds.)

Switch reference and Universal Grammar. Amsterdam & Philadelphia: John Benjamins, pp. 185-

208.

Matic, Dejan, Rik van Gijn & Robert van Valin. 2014. Information structure and reference tracking in

complex sentences: an overview. In: Rik van Gijn, Jeremy Hammond, Dejan Matic, Saskia van

Putten & Ana Vilacy Galucio (eds.) Information structure and reference tracking in complex

sentences. Amsterdam/Philadelphia: John Benjamins, pp. 1-42.

Matras, Yaron & Jeanette Sakel. 2007. Investigating the mechanisms of pattern replication in language

convergence. Studies in Language 31 (4): 829-865.

Maxwell, Michael & Nancy Morse. 1999. Gramática del Cubeo. Bogotá: Editorial Alberto Lleras

Camargo.

McKenzie, Andrew. in prep. A survey of switch-reference in North America. Manuscript.

Miller, Marion. 1999. Desano grammar. Dallas: Summer Institute of Linguistics and the University of

Texas at Arlington.

Olawsky, Knut. 2006. A Grammar of Urarina. Berlin: Mouton de Gruyter.

Overall, Simon. 2007. A Grammar of Aguaruna. PhD dissertation LaTrobe University.

Payne, Doris. 1990. Morphological characteristics of lowland South American Languages. In: Doris L.

Payne (ed.) Amazonian Linguistics: Studies in Lowland South American Languages, 213-242.

Austin, TX: University of Texas Press.

Payne, Thomas E. 1980. Who’s kissing who: the fourth person in Yup’ik Eskimo. In: Pamela Munro

(ed.). Studies of switch-reference. Los Angeles: University of California, pp. 65-88.

Page 49: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

49

Pitman, Donald. 1980. Bosquejo de la gramática araona. (Notas Lingüísticas, 9.) Riberalta, Bolivia:

Instituto Lingüística del Verano.

Quesada, Juan Diego. 2007. The Chibchan languages. Cartago: Editorial Tecnológica de Costa Rica.

Ramirez, Henri. 1997. A Fala Tukano dos Ye'pâ-Masa. Manaus: Inspetoria Salesiana Missionária da

Amazônia, CEDEM.

Renard-Casevitz, F., T. Saignes & A. Taylor. 1988. Al este de los Andes: Relaciones entre las

sociedades amazónicas entre los siglos XV y XVII. Lima: Instituto Francés de Estudios Andinos.

Roberts, John R. 1997. ‘Switch-reference in Papua New Guinea: A preliminary survey.’ In A. Pawley,

ed.: Papers in Papuan linguistics, No. 3, 101–241.Pacific Linguistics, A-87.

Rodrigues, Aryon. 1999a. Macro-Jê. Dixon & Aikhenvald (eds.), 165-206.

Saad, George. 2014. A grammar sketch of Shuar. MA thesis Radboud University Nijmegen.

Sparing-Chávez, Margarethe W. 2007. Aspects of Grammar: Amahuaca: An Endangered Language of

the Amazon Basin. Lima: Summer Institute of Linguistics.

Stirling, Lesley. 1993. Switch reference and discourse representation. Cambridge: Cambridge

University Press.

Stenzel, Kristine. 2013. A Reference Grammar of Kotiria (Wanano). University of Nebraska Press.

Taylor. A. 1999. The western margins of Amazonia from the early sixteenth to the early nineteenth

century. In: F. Salomon & S. Schwartz (eds): The Cambridge history of the native peoples of the

Americas, volume 3: South America, pp. 188-256. Cambridge: Cambridge University Press.

Torero, Alfredo. 1984. El comercio lejano y la difusión del quechua: el caso del Ecuador. Revista

Andina 2 (2): 367–402.

Torero, Alfredo. 2002. Idiomas de los Andes: lingüística e historia. Lima: Instituto Francés de

Estudios Andinos, Editorial Horizonte.

Valenzuela, Pilar. 2003. Transitivity in Shipibo-Konibo grammar: a typologically oriented study.

Ph.D. dissertation, University of Oregon.

Viera-Cândido, Gláucia. 2004. Descrição morfossintática da língua Shanenawa (Pano). PhD

dissertation UNICAMP Campinas.

Vittadello, Alberto. 1988. Cha'palaachi = El idioma cayapa. Esmeraldas, Ecuador: Pontificia

Universidad Católica del Ecuador.

Voort, Hein van der. 2004. A grammar of Kwaza. Berlin: Mouton de Gruyter.

Vries, Lourens de. 2005. “Towards a typology of tail-head linkage in Papuan languages”. Studies in

Language 29.2:363-384.

Vuillermet, Marine. 2012. A Grammar of Ese Ejja, a Takanan language of the Bolivian Amazon. PhD

Thesis Université Lumière Lyon.

Vuillermet, Marine. 2014. The multiple co-reference systems in the Ese Ejja subordinate clauses.

In : In: R. van Gijn, D. Matic, J. Hammond, S. van Putten and A. Galucio (eds.) Information

structure and reference tracking in complex sentences. John Benjamins, pp. 341-372.

Page 50: Switch reference in western South Americaffffffff-be1c...1 Switch reference in western South America DRAFT - DO NOT CITE WITHOUT PERMISSION Rik van Gijn University of Zürich Switch-reference

50

Weber, David. 1975. Apuntes sobre el quechua de Lamud. Lima: Instituto Lingüístico de Verano.

Weber, David. 1989. A grammar of Huallaga (Huánuco) Quechua. Berkeley and Los Angeles:

University of California Press.

Woodbury, Anthony C. 1983. Switch-reference, syntactic organization, and rhetorical structure in

Central Yup’ik Eskimo. In: John Haiman & Pamela Munro (eds.) Switch reference and Universal

Grammar. Amsterdam & Philadelphia: John Benjamins, pp. 291-316.