Goal and source in South American

Emilia Roosvall

Department of Linguistics Bachelor’s Programme in Linguistics 180 ECTS credits Spring semester 2020 Supervisor: Bernhard Wälchli Swedish title: Mål och källa i sydamerikanska språk Goal and source in South American languages Emilia Roosvall

Abstract

This study primarily investigates the expression of two local roles, goal and source, inSouth American languages. Local roles describe the direction of movement or locatedness in relation to a physical , a ground, in a motion event. While goal expresses motion to or towards and source expresses motion from a ground, these are not always distinguished from one another but sometimes encoded indifferently. A previous cross-linguistic study by Wälchli and Zúñiga (2006) shows that the encoding of goal and source tends to be distinct in Eurasia, North Africa, and Australia, and more diverse in the Americas and New Guinea. However, the sample used in their study is not representative in the Americas. The principal aim of the present study is to determine whether the encoding of goal and source is distinct or indifferent in a representative sample of South American languages, using both reference grammars and parallel texts consisting of Bible translations. The local role path, expressing motion through a ground, is also studied to the extent that this is possible given the data. The findings show that distinct encoding of goal and source is most common in the sample. Indifferent languages are still attested for, yet to a smaller extent than in Wälchli and Zúñiga’s2006 study( ).

Keywords goal, source, South American languages, motion events, linguistic typology

Sammanfattning

Denna studie undersöker främst uttryck av två lokalroller, mål och källa, i sydamerikanska språk. Lokalroller beskriver riktningen av förflyttning eller befintlighet i förhållande tillett fysiskt objekt, en bakgrund, i en händelse som uttrycker rörelse. Även om mål uttrycker rörelse till eller mot och källa uttrycker rörelse från en bakgrund är dessa inte alltid särskilda utan markeras ibland neutralt. En tidigare tvärspråklig studie av Wälchli och Zúñiga (2006) visar att markering av mål och källa tenderar att vara distinktiv i Eurasien, Nordafrika ochAus- tralien, och mer varierad i Amerika. Ehuru är urvalet av språk i deras studie inte representat- ivt i Amerika och Nya Guinea. Det huvudsakliga syftet med den här studien är att avgöra om markeringen av mål och källa är distinktiv eller neutral i ett representativt urval av sydamerik- anska språk, med både grammatikor och parallelltexter bestående av bibelöversättningar som datakällor. Lokalrollen path, som uttrycker rörelse genom en bakgrund, studeras också iden utsräckning detta är möjligt givet datan. Resultaten visar att distinktiv markering avmåloch källa är vanligast i urvalet av språk. Neutrala språk finnes också, men i mindre utsträckning än i Wälchli och Zúñigas studie (2006).

Nyckelord mål, källa, sydamerikanska språk, rörelse, lingvistisk typologi Contents

1 Introduction 3

2 Background 4 2.1 Motion events ...... 4 2.2 Local roles ...... 6 2.2.1 Basic local roles ...... 6 2.2.2 Source and goal ...... 7 2.2.3 Path ...... 10 2.2.4 Syntactic structure ...... 11 2.2.5 Syncretism ...... 14 2.3 Marking with different noun types ...... 17 2.4 The linguistic situation in ...... 18 2.4.1 Genealogy ...... 18 2.4.2 Typology ...... 18 2.4.3 Previous description ...... 20

3 Purpose and research questions 22 3.1 Purpose ...... 22 3.2 Research questions ...... 22

4 Method 23 4.1 Data ...... 23 4.1.1 Methodological background − Parallel texts ...... 25 4.1.2 Methodological background − Bible translations ...... 26 4.2 Sample ...... 27 4.2.1 Methodological background − Sampling methods ...... 27 4.2.2 The sample of this study ...... 28 4.2.2.1 Genealogical stratification ...... 28 4.2.2.2 Areal stratification ...... 29 4.3 Analysis ...... 30

5 Results 32 5.1 The encoding of source and goal ...... 32 5.1.1 Distinct encoding ...... 32 5.1.1.1 Complex markers ...... 32 5.1.1.2 Length of marker ...... 35 5.1.1.3 Analysis from Bible verses vs reference grammars ...... 41 5.1.2 No clear distinct encoding ...... 42 5.1.2.1 Indifferent encoding ...... 42 5.1.2.2 Goal marking with unknown source marking ...... 46 5.1.2.3 Intermediate systems ...... 49 5.1.2.4 Associated motion ...... 50 5.1.2.5 Unresolved cases ...... 50 5.1.3 Areal distribution ...... 50 5.2 The encoding of path ...... 51 5.3 Methodological results ...... 52 6 Discussion 54 6.1 The encoding of goal and source ...... 54 6.1.1 Distinct encoding ...... 54 6.1.2 No clear distinct encoding ...... 55 6.1.3 Areal tendencies ...... 56 6.2 The encoding of path ...... 57 6.3 Methodological results ...... 57 6.4 Additional remarks ...... 58 6.5 Future research ...... 59

7 Conclusions 60

References 62

Appendices 67

A Bible verses 67

B Sample 68

C Suggested linguistic areas of South America 69

D Source and Goal markers in distinctly encoding languages 73

E Encoding in languages with no clear distinction 74

F Syncretism patterns between local roles in the sample languages 75 Abbreviations

1 1st person distr distributive 2 2nd person dsq delayed sequence 3 3rd person evi evidential abl ablative excl exclusive acc accusative f feminine accomp accomplished fin finite act active form formative verbal prefix add additive fsh far from speaker and hearer adv adverbial(izer) fut future advrel relativized adverbial g generic () all allative goal goal andtv andative hab habitual apl applicative impfv imperfective ass assertion incl inclusive aux auxiliary ind indicative caus causative ins instrumental clf classifier ints intensifier comp completive io individual verification, observation orientation. con contrastive locative/temporal lk linker conrel relativized contrastive locative/temporal loc locative conv converb m masculine cts close to speaker mot motative dat dative neg negative/negator decl declarative ni non-indicative def marker of nom nominative dem demonstrative nomcl nominal class desc descriptive obj object det determiner obl oblique dim diminutive past past dir directional pfv perfective dist.past distant past pl

1 place place polint polar interrogative poss possessive ppos postposition prog progressive prs present rec.past recent past refl reflexive rel relativizer sbj sg singular sp specific (classifier) sym sound-symbolic word t/e tense/evidential theme theme top topicalizer vblz verbalizer ventv ventive

2 1 Introduction

Linguistic typology aims at characterizing languages based on functional and structural traits. This study focuses entirely on making such characterizations in regards to the presence or absence of a distinction between certain semantic roles which appear with motion events − that is, events expressing dynamic movement. Two fundamental elements of a motion event are figure and ground. The figure is a physical object whose motion is related to another physical object, namely the ground. Two other cent- ral components of a motion event are motion and direction. Motion, unsurprisingly, conveys the motion of a motion event and direction expresses the spatial relation between the figure and the ground (Talmy 2007). That relation can be either static or dynamic and thus constitute different semantic roles. Static spatial relations express locatedness − to be located e.g. in, at, or on a ground − and have the semantic role place. In , dynamic spatial relations describe movement − to move in relation to a ground. This movement can have different dir- ectional relations to the ground. Two fundamental such relations are motion to or towards, and motion from a ground, represented by the semantic roles goal and source respectively. These are exemplified in (1) below. (1) a. Place: We got engaged in Bremen. (Haspelmath 2019, p. 316) b. Goal: We went to Bamberg. c. Source: She came from Aachen. There are several ways to encode place, goal, and source in different languages, e.g. through case markers, adpositions, verbal markers, or other specialized markers. However, not all lan- guages have a grammatical distinction between goal and source − sometimes they are marked indifferently. A study by Wälchli and Zúñiga (2006), which investigates the encoding of goal and source cross-linguistically, shows that distinct encoding of goal and source is dominant in Eurasia, North Africa, and Australia, while indifferent marking predominantly appears in the Americas, and mixed marking is most frequent in Africa. However, their study does not have representative samples for the Americas or New Guinea. Therefore, this study primarily aims at investigating whether the encoding of goal and source is distinct or indifferent in a representative sample for one of the previously underrepresented areas − South America. South America is an astonishingly diverse linguistic area, both genealogically and typolo- gically. This one continent is home to a quarter of the world’s families andnearly half of the world’s isolates (Campbell 2012a, 2013). There is considerable typological diversity among the South American languages − in both phonology and morphosyntax there are many traits which are over-represented on this continent and several traits which appear to be unique for South American languages. However, South American languages have until re- cently been underrepresented in typological work and generally insufficiently described. Toa certain degree, this remains true, although both the amount of representation and description are increasing. Due to the exceedingly diverse linguistic situation in South America and the history of underrepresentation and -description of its languages, this study focuses entirely on investigating languages of South America, to partially fill this gap. As mentioned above, the main purpose of this study is to investigate whether goal and source are encoded distinctly or indifferently in South American languages. Additionally, the encoding of another spatial relation − path, expressing motion through a ground − is also examined, yet to a more restricted degree due to a lack of data. Moreover, the two data sources used for this study, parallel texts consisting of Bible translations and reference grammars, are compared in regards to reliability and efficiency for typological investigations.

3 2 Background

In this chapter, I firstly discuss motion events and their components. Subsequently, I more pro- foundly review local roles − their semantics, relation to each other, and syntactic structure, as well as syncretism between them. Furthermore, the marking of local roles with different noun types is discussed and, finally, I examine the linguistic situation in South America, re- garding both genealogy and typology, and the many insufficiencies in the previous description of languages from this linguistic macroarea.

2.1 Motion events Talmy (2007, a revised and expanded version of Talmy 1985) uses the term motion event in his discussion of systems expressing motion and location. He defines it as denoting situations either involving dynamic movement or the stationary continuance of a location (2007, p. 70), even though stationary locatedness expressed by motion events may seem counterintuitive. The fundamental components of a motion event are figure and ground. Figure denotes an object located in or moving in relation to another object − the ground. In addition, there are two other basic elements of a motion event − motion and direction. According to Talmy, motion expresses the movement or locatedness in itself of a motion event and direction conveys the direction, or path, of the figure’s movement (a directional motion), alternatively the figure’s static location (a locative motion) (2007, p. 70). The different types of direction, according to Talmy, are illustrated in (2) by different English prepositions.

(2) a. Alex is in/under the car. (locative) (Zwarts 2008, p. 79) b. Alex went into/through the forest. (directional)

Motion events can also be fused with a ‘co-event’ − most commonly expressing manner or cause. Cause, unsurprisingly, expresses the cause of a movement or continued stationary position, while manner describes the nature or style of the motion in a motion event. A few examples of different manners and causes (both in italics) co-expressed with locative ordir- ectional motion are given in (3) below.

(3) English expressions of motion conflated with manner or cause (Talmy 2007, p. 73) a. Locative motion + manner i. The lamp stood/lay/leaned on the table. b. Directional motion + manner i. The rock slid/rolled/bounced down the hill. (non-agentive) ii. I slid/rolled/bounced the keg into the storeroom. (agentive) iii. I ran/limped/jumped/stumbled/rushed/groped my way down the stairs. (self- agentive) c. Directional motion + cause i. The napkin blew off the table. (non-agentive) ii. I pushed/threw/kicked the keg into the storeroom. (agentive)

In a discussion of lexicalization patterns of verbs, Tesnière proposed the distinction between movement and displacement, that is distinguishing between a motion activity (such as ‘walk’ or ‘run’) and a change of location, regardless of the motion, (such as ‘descend’ or ‘enter’) (1959, cited in Wälchli and Zúñiga 2006, p. 298). Wälchli(2001) holds that displacement, in Talmy’s

4 terms, is the co-expression of motion and direction, while movement conflates motion and manner. According to Wälchli (2001), the typologies previously presented on motion events were too centered around verbs and thus insufficient. He argues that all languages express displace- ment, but not solely through verbs. There are three means, or loci, of encoding displacement: verbally (by the verb stem), adnominally (through adpositions or case), or adverbally (by verb particles or verbal affixes). This leads to three main types of languages, based onthelocus which most differentiates displacement: predominantly verbal (V), predominantly adnominal (AN), and predominantly adverbal (AV) languages. Most commonly however, languages use a combination of the three loci to express displacement. Furthermore, all three loci may have more than one syntactic position (or slot), i.e. can be represented multiple times, and none of them are obligatory (Wälchli 2001, p. 301). The three different loci are illustrated in (4) where the verbal locus is underlined, the adnominal locus is in italics, and the adverbal locus is in bold.

(4) English (Indo-European) (Wälchli and Zúñiga 2006, p. 284) The cat ran in from the garden.

In this example cat is the figure, garden is the ground, ran is the conflated expression of motion and manner, and from is the direction. The three loci correspond to Talmy’s distinction between verb-framed and satellite-framed languages − a work that began on the North American language Atsugewi (Hokan) (Talmy 1972). In satellite-framed languages the direction, or relation between direction and ground, is expressed outside of the verb by a satellite − a free word or bound affix, anything but an NPor a PP. Satellite framed languages are a combination of two language types previously identified by Talmy: Type II − expressing manner in the verb and direction in a satellite − and Type III − expressing figure in the verb and direction in a satellite. In verb-framed languages, the direction is encoded in the verb (previously termed Type I). Most satellite-framed languages use solely satellites to encode the directional relation, but many of them express the relation through a satellite alongside an adposition or an adposition alone (Talmy 2000, pp. 221-222). Consequently, verb-framed languages correspond to V type languages and satellite-framed languages correspond to AN and AV types. In certain instances it is difficult to determine which type (V, AN, or AV) a language belongs to, but V type languages appear to be more common than the other two. The adnominal type is ‘always defective’, according to Wälchli, as there is no encoding of displacement without an explicit local nominal expression in these languages (2001, p. 302). Thus AN type languages are also AV types to some degree, to make up for the lack of adnominal marking in the absence of a local nominal expression. The three types are not uniform but show a great deal of internal diversity. Forexample the prevalence of adverbal marking varies immensely between languages as well as within the types but is especially favoured by AN and, unsurprisingly, AV types, while disfavoured by V types (Wälchli 2001, p. 304). Which type of displacement encoding a language belongs to appears to be predictive of the behaviour of certain other aspects of motion verbs (see discussion in Wälchli 2001, pp. 304-310). Conclusively Wälchli states that ‘displacement is a domain that extends across the traditional division of language into grammar and lexicon’ and that it, as a functional domain, belongs to the clause rather than the verb or noun phrase (2001, p. 321).

5 2.2 Local roles A study by Wälchli and Zúñiga (2006) focuses on the semantics of the adnominal locus and what they refer to as local roles. These are described by Fillmore as ‘case-like notions’ needed for a complete account of occurrences of displacement, in contrast to localization which he defines as notions ‘associated with the space semantics of words naming locations and objects’ (e.g. top, bottom, interior, exterior, etc.) (1971/75, p. 26). Wälchli and Zúñiga’s study (2006) investigates the cross-linguistic expression of two local roles − source and goal. Wälchli and Zúñiga (2006), consider source and goal, as well as place and path, as local roles since any of these roles can be assumed by any ground. Localization, on the contrary, depends, at least to some extent, on the form of the ground, e.g. ‘top’ for hills or ‘inside’ for houses (2006, p. 287). This is based on Fillmore (1971/75, p. 26) according to whom local roles encode the basic dynamic and static ‘positional notions’, while localizational notions constitute a dis-joined, more complex system. Most Indo-European languages combine the expression of local role and localization, more or less markedly, but it is far more frequent cross-linguistically to separate their expression (Wälchli and Zúñiga 2006, p. 287). Spatial relation, i.e. the relation between ground and figure, is not seldom expressed through case. Creissels defines case as a system of denoting ‘dependency rela- tions involving NPs’ (2009, p. 609). Spatial cases, without any adpositions, have three main functions. One of these functions is to specify whether an of a motion event is the source, goal, place, or path of the motion, that is which local role the argument functions as. Spatial cases sometimes express relative orientation, or localization, such as (i) ‘in’, (ii) ‘on’, or (iii) ‘at’, in addition to the local roles and are thus bidimensional case systems, yet, as stated above, this is not frequent cross-linguistically in to distinguishing the two. Tridimensional systems also exist, e.g. with an additional dimension of distal as in Tsez (Northeast Caucasian), but are far more scarce (Creissels 2009, p. 615).

2.2.1 Basic local roles Local roles, in language, are needed for a full description of a relation between a figure and a ground, as stated above. That relation can be either static or dynamic. In principle, the relations between figure and ground can be divided into three distinct types:

(5) Types of relations between figure and ground (Kutscher 2010, p. 252) a. Place : A static relation where the figure is located in relation to the ground b. Goal : A dynamic relation where the figure moves or is caused to move towards the ground c. Source : A dynamic relation where the figure moves or is caused to move away from the ground

The three basic local roles are thus place, expressing locatedness; goal, expressing motionto or towards; and source, expressing motion (away) from. There is also a forth local role − path, expressing motion through − discussed in more detail below (see section 2.2.3). Nonetheless, the three local roles in (5) are the most basic ones when expressing spatial relations, however the terminology is less agreed upon (Haspelmath 2019, p. 316). Goal has been called both destination and motion-to, place is often referred to as location, and source has been called motion-from. All three of the roles are also frequently referred to by the case that most of- ten encodes them − allative for goal, ablative for source, and locative for place (Haspelmath 2019, p. 316). Here, the terms place, goal, and source are used (and occasionally, when the

6 roles manifest as case markers, locative, allative, and ablative). For further discussion on the terminology see Haspelmath (2019, pp. 316-317).

2.2.2 Source and goal Zwarts (2008) investigates the relation between direction and ground. He determines that source expressions include the ground in the starting point of a motion (e.g. ‘out of’), while expressions for goal include it in the endpoint of a motion (e.g. ‘into’). He illustrates this with a scale from 0 to 1 where 0 is the starting point of a motion and 1 is its endpoint. If a direction includes the ground in one of the points and not the other the scale is divided into two parts: a positive one, with the ground included, and a negative one, with the ground excluded. Source expressions include the starting point while excluding the endpoint of a motion (as represented in (6a)). Goal expressions include the endpoint while excluding the starting point of a motion (as represented in (6b)). Thus, Zwarts claims that source directions are reversed goal directions (2008, pp. 82-83).

(6) a. Source directions (Zwarts 2008, pp. 82-83) + + + + + −−−−− 0 1 b. Goal directions −−−−− + + + + + 0 1

Wälchli and Zúñiga (2006) investigate whether or not source and goal are, cross-linguistically, encoded distinctively as well as if there is a to some degree grammaticalized encoding of source, and if so of what kind (verbal, adnominal, or adverbal). They also examine how these two features correlate. Their study focuses on source markers due to ‘source being the marked choice of local role in motion events’, i.e. ground elements functioning as source are far less frequent than those functioning as goal since source can often be understood from discourse (Wälchli and Zúñiga 2006, p. 291). This is in accordance with the claim that there is a frequen- tial, extralinguistic, and/or conceptual goal bias in motion events − that goal is the unmarked choice, the naturally preferred reading in ambiguous expressions (see e.g. Stefanowitsch and Rohde 2004). Wälchli and Zúñiga conclude that there are three major types of systems for encoding source and goal: consistent (distinctive encoding of source and goal consistently), mixed (a weakly grammaticalized marker or construction for source), and indifferent (no explicit dis- tinction between the two roles). In addition, there are two minor systems, namely what they call source extended and inanimate. The first refers to systems with consistent marking where the source marker also encodes goal in certain instances and the second to systems with dis- tinctive source and goal markers with non-animate grounds but indifferent marking with an- imate grounds (Wälchli and Zúñiga 2006, p. 295). These are considered subtypes of consistent marking when Wälchli and Zúñiga discuss the areal distribution of their results. They show that distinctive source and goal markers are dominant in Eurasia, North Africa, and Australia, while indifferent encoding is found mainly in the Americas (though itisnot necessarily the dominant type), and mixed marking is most common in Africa. There is most diversity in the Americas and New Guinea, which are also the areas with the least represent- ative samples in the study (Wälchli and Zúñiga 2006, p. 297, 299). Wälchli and Zúñiga also determine that adnominal source markers are significantly more often obligatory, or more grammaticalized, than adverbal and verbal source markers, which tend to be less frequently

7 obligatory, viz. less grammaticalized. Conclusively Wälchli and Zúñiga state that there is a correlation between the type of source marker and source-goal distinguishing, namely lan- guages with adnominal source markers are prone to have distinctive encoding of source and goal while languages without adnominal source markers more often have a non-consistent encoding of source and goal. That the more grammaticalized markers are more obligatory is, according to Wälchli and Zúñiga, precisely what is to be expected from a grammaticalization perspective (2006, pp. 297-298). Wälchli and Zúñiga examine one language, Mapundungun (Araucanian), more closely. Mapundungun does not distinguish source and goal grammatically, but through the lexical meaning of verbs and context. They conclude that grounds exclusively take an oblique ad- nominal marker, mew. Its purpose is simply to mark an NP as oblique, no matter if the NP functions as a source or a goal or even a place or an instrument. In Mapundungun motion verbs have ‘typical role orientations’, that is roles they typically express (Wälchli and Zúñiga 2006, p. 289). This means certain verbs lead to the ground being interpreted as a goal (asin (7b)), while others lead to it being read as a source (as in (7a)).

(7) Mapundungun (Araucanian) (Wälchli and Zúñiga 2006, p. 289) a. Chi narki tripa-y ruka mew. the cat exit-ind house ppos ‘The cat went out of the house.’ b. Puw-i chi kalku taiñ ruka mew. arrive.there-ind the warlock our:pl house ppos ‘The warlock arrived in our house.’

Moreover, Mapundungun does not express clauses with two grounds, but this can be achieved through clause linkage (Wälchli and Zúñiga 2006, p. 290). Wälchli and Zúñiga argue this could suggest that languages that do not distinguish between local roles are inclined to only allow one ground, and therefore one local role, per motion verb and, by extension, that languages with distinctive encoding of the local roles are more likely to allow several grounds per motion verb (2006, p. 290). Wan (Southeastern Mande) is also a language that does not distinguish between local roles, but even so the distinction between goal and source is seldom ambiguous. The lexical mean- ing of the verb, alongside the contextual information, help interpret the local role of a phrase (Nikitina 2009, p. 1137), much like in Mapundungun. Nikitina divides motion verbs in Wan into two subcategories: specialized and non-specialized verbs. Specialized verbs operate as ‘goal verbs’ or ‘source verbs’ and encode the local role of an argument in the argument struc- ture. The non-specialized verbs have a ‘flexible role assignment’, i.e. their arguments canhave different local roles (Nikitina 2009, pp. 1126-1127, 1137). In these instances the local roles are disambiguated by the type of ground, namely whether the ground is a natural goal or source for the type of motion in question, as exemplified in (8).

(8) Wan (Southeastern Mande) (Nikitina 2009, p. 1127) a. è siā tābālí è wā 3sg.sbj fall.past table def under ‘He fell under/⁇from under the table.’ b. è ziā trɔ̄ɔ́ mì 3sg.sbj descend.past ground by ‘He came down to/⁇from the ground.’

8 Nikitina calls this a ‘source/goal asymmetry’ − a difference in properties required of the ground, enabling it to function as either a source or a goal with a certain motion verb (Nikit- ina 2009, p. 1130). In other languages with this system, source/goal asymmetries and flexible role assignment-verbs seem to be roughly the same as in Wan. Furthermore, this pattern of specialized and non-specialized categories can be found for lexical items other than verbs, e.g. in satellite-framed languages (Nikitina 2009, pp. 1135-1136). Nikitina holds that the fact that this pattern applies to more than just verbs reflects a widespread way of expressing meaning compositionally − partially in unambiguous encoding, partially in contextual information − and furthermore that this reflects the complex association between the argument structure of motion verbs and lexical meaning (Nikitina 2009, p. 1138). Schapper (2011) highlights another way to differentiate source and goal without amor- phological distinction, namely through iconicity of sequence. Iconicity refers to language struc- tures which reflect the structure of experience. Iconicity of sequence, consequently, denotes sequences of linguistic forms which reflect sequences of experiences. Schapper shows that two , Kamang and Bunaq (Timor-Alor-Pantar), encode source and goal NPs identically morphologically, however, they are distinguishable by their position in the clause in relation to the motion verb − source precedes the motion verb while goal follows it, thus displaying the start and endpoint of a motion, respectively, iconically. Both Kamang and Bunaq introduce source and goal NPs by a and a locative postposition, however, the languages differ in three regards: whether both source and goal can be expressed inthe same clause; whether both animate and inanimate sources and goals are expressed by iconic sequentiality; and how goals are expressed with transitive displacement verbs. For a full dis- cussion on these differences see Schapper2011 ( ). In (9) below, it is exemplified how Kamang expresses place, source, and goal with one of its locative markers, mi ‘in’, through iconicity of sequence.

(9) Kamang (Timor-Alor-Pantar) (Schapper 2011, pp. 103-104) a. Nal kadii mi. 1sg house in ‘I am in the house.’ b. Nal kadii mi yaangme. 1sg house in go.down ‘I went down from the house.’ c. Nal yaangme il waai mi. 1sg go.down water pond in ‘I went down into the pond.’

Haspelmath (2019) also touches on markers for source and goal in his paper on differential place marking and differential object marking. He argues that goal is more inclined to bezero- marked than source, which thus tends to be more marked (2019, pp. 317, 319). This is primarily based on a study on zero-marking in spatial relations by Stolz et al. (2014). They formulate the frequency of zero-marking in goal, source, and place as is presented in (10a) which can be reformulated to a scale as in (10b):

(10) a. Likeliness of zero-marking (Stolz et al. 2014, p. 281) goal > place > source b. Markedness scale goal < place < source

9 The further to the left a local role is placed on the scales in (10), the more likely itistobe zero-marked or less marked and the further to the right a role is placed, the more likely it is to be more marked and the less likely it is to be zero-marked. Moreover, Stolz et al. conclude that zero-marking of source can predict zero-marking of goal. They confirm that all of their sample languages with zero-marking in source also have zero-marking for at least one of goal and place. There are two languages in their sample with zero-marked sources that cannotbe confirmed to have zero-marked places, Mupun (Chadic) and Itzá (Mayan). Thus zero-marking in source cannot fully predict zero-marking in place, however, all instances of zero-marked sources coincide with zero-marking in goals, presented by Stolz et al. as follows:

(11) unilateral implication in zero-marking (Stolz et al. 2014, p. 281) zero-marked source ⊃ zero-marked goal

Conclusively, source directions, including the ground in the starting point of a motion, are reverse goal directions, which include the ground in the endpoint of a motion. There are three major types of systems for encoding source and goal − consistent, mixed, and indifferent. Languages with indifferent encoding distinguish source and goal by non-grammatical means, such as through the lexical meaning of the verb and context, as in Mapundungun and Wan, or through iconicity of sequence, as in Kamang and Bunaq, and languages with consistent encoding tend to have longer markers for source than goal (which is not seldom zero-marked).

2.2.3 Path One local role that is underinvestigated, and thus rarely discussed, is path1, expressing motion e.g. through or along a ground. Many grammars do not acknowledge path directions at all (Pantcheva 2011, p. 52) and so little is known about it cross-linguistically that Haspelmath excluded analyzing its expression altogether in a recent study (2019, p. 316). If path is expressed through a case of its own, this is often called prolative or perlative, and occasionally translative. However, it is more commonly found that languages encode path by other means than a separate case, e.g. using ablative, or an adposition to encode path, or a spatial case marker alongside an adposition (Creissels 2009, p. 615). A path can be analyzed as either dynamic − a process of traveling along the path − or static − a linear path between two grounds or through a ground (Marciniak and Strube 2005, p. 153). Pantcheva distinguishes between different types of paths depending on how the English prepositions expressing them interact with time-frame adverbials (such as ‘in an hour’) and time-span adverbials (such as ‘for an hour’) (2011, p. 27). Paths expressed by past or along are only compatible with one of these each, while paths expressed by through or across are compatible with both, as seen in (12-14).

(12) a. The boy ran past the tree in one minute. (Pantcheva 2011, p. 28) b. *The boy ran past the tree for one minute. 2

(13) a. *The children walked along the river in an hour. b. The children walked along the river for an hour.

(14) a. The insect crawled through the tube {for two hours / in two hours}. b. Mary limped across the bridge {for ten minutes / in ten minutes}.

1Or route in some works. 2‘This sentence is grammatical under an iterative reading where the boy runs a couple of times pasttheground and the whole macro-event lasts one minute’ (Svenonius, 2010, cited in Pantcheva 2011, p. 28).

10 Pantcheva argues that this shows the difference between transitional, or bounded, paths and non-transitional, or unbounded, paths. Transitional paths have both their endpoints (the be- ginning and the end of the motion) outside of the ground, while non-transitional paths have both their endpoints within the ground. Time-frame adverbials are compatible with the former (as seen in (12)), while time-span adverbials are compatible with the latter (as seen in (13)). Since the prepositions in (14) are compatible with both types of adverbials, they can express both transitional and non-transitional paths (Pantcheva 2011, pp. 28-29). For further discus- sion on the semantics of path see Ganenkov (2002).

2.2.4 Syntactic structure The syntactic structure of source, goal, and place has been studied by Pantcheva (2010). Her intention is to supplement the general that directional expressions include at least two heads: a directional head 3 dominating a place head. The structure can be represented as in (15) and is most clearly displayed by languages with locative markers morphologically incorporated in directional markers as in the Estonian (Finnic) example in (16).

(15) Decomposing directional expressions (Pantcheva 2010)

(16) Estonian (Finnic) (Viitso, 1998, cited in Pantcheva 2010, pp. 1047-1048) a. i. jala-l foot-on ‘on the foot’ ii.

b. i. jala-l-t foot-on-from ‘off the foot’ ii. 3Or path head in Pantcheva (2010).

11 c. i. jala-l-le foot-on-to ‘onto the foot’ ii.

Comparing (16a), expressing adessive (location at), to (16b) and (16c), expressing ablative and allative respectively, clearly discloses that directional expressions can embed locative expres- sions and thus that the directional head dominates the place head (Pantcheva 2010, p. 1048).

Table 1: Examples of ablative markers morphologically containing allative markers

Jingulu (Mirndi) Ingush (Nakh-Daghestanian) locative (place) -mpili -ǧ allative (goal) -ŋka -ga ablative (source) -ŋka-mi -ga-ra

Pantcheva holds that goal and source markers seem to have equal complexity, or simpli- city, in most languages. Nevertheless, there are a few exceptions where the encoding of source contains the goal marker (Pantcheva 2010, p. 1050) − for example in Jingulu (Mirndi) and In- gush (Nakh-Daghestanian) (see Table 1) (data from Blake 1977, and Nichols 1994, respectively, cited in Pantcheva 2010, p. 1051). Both of these unrelated languages exhibit morphologically simple goal markers (-ŋka and -ga) being contained morphologically in complex source mark- ers (-ŋka-mi and -ga-ra). Since Nikitina has never encountered a language showing the oppos- ite hierarchy between the markers, she assumes that this morphological relationship between goal and source indicates a syntactic structure where source markers morphologically contain goal markers. She thus claims that there is a structural difference between source directions and goal directions (Pantcheva 2010, p. 1051). Since it has already been established that direc-

12 tional expressions embed place expressions, one can formulate the syntactic structures of the local roles place, goal, and source with tree diagrams as follows:

(17) a. Place (Pantcheva 2010, p. 1052)

b. Goal

c. Source

In addition to the detailed study of the syntactic structure of source and goal, Pantcheva (2011) has investigated how the syntactic structure of path relates to her previously presented structure (as seen in (17)). She reviews examples from Slovak (Slavic) where path is expressed by adding the preposition po to a prepositional phrase encoding goal (as in (18a)). Without po the prepositional phrase would be interpreted as expressing goal (as in (18b)).

(18) Slovak (Slavic) (Caha, 2010, cited in Pantcheva 2011, pp. 51-52) a. Na Forum Romanum vstupujeme po-pod oblúk-∅ Tita. on Forum Romanum.acc enter.1pl po-under arch-acc of.Tito ‘We entered the Forum Romanum by going under Tito’s arch.’ (lit.: via under) b. Slamu dal pod stôl-∅. hay put.3sg under table-acc ‘He put the hay under the table.’

This suggests that path morphologically contains goal and if, as Pantcheva has previously argued, source also embeds goal this raises the question of what the relation between path and source is.

13 Table 2: Spatial case system in Akhvakh (Daghestanian)

Location Goal Source Path on -g-e -g-a -g-u -g-u-ne at, near -xar-i -lir-a -xar-u -xar-u-ne at -q-e -q-a -q-u -q-u-ne

If one examines local roles in Akhvakh (Daghestanian) it is clear that the expression of path is created through adding the morpheme -ne to the marker -u which creates the complex case marker -u-ne for translative. This pattern has also been observed in Avar (Northeast Caucasian) (see Pantcheva 2011, pp. 52-53). The Akhvakh case system (although slightly reduced) is presented in Table 2 (data from Magomedbekova, 1967, cited in Pantcheva 2011, p. 53). This shows that path does not only embed goal expressions but alsosource expressions, indicating the syntactic structure presented in (19).

(19) Path (Pantcheva 2011, p. 55)

2.2.5 Syncretism Syncretism between cases, not just spatial ones, is frequent cross-linguistically. Creissels con- cludes that the most common syncretism, by far, is the one between allative and dative case, although allative-accusative syncretism is also attested for (2009, pp. 619-620). Furthermore, he discusses syncretism patterns that are common in adpositions but not in case systems. Those are the ones between genitive and ablative or allative, as well as allative and locative. Regarding ergative languages syncretism between ergative and locative or instrumental case is not uncommon. Creissels highlights that the possibility of a syncretism between ablative and ergative case is not mentioned in previous work even though it can be expected to oc- cur since ablative-instrumental is a common syncretism and, as was just mentioned, so is instrumental-ergative (2009, p. 619). Pantcheva concludes that there are five possible syncretism patterns between goal, source, and place. She formulates them as follows:

(20) a. place = goal = source : one single marker (Pantcheva 2010, pp. 1043-1044) b. place = goal ≠ source : one ambiguous marker for place and goal, a separate marker for source c. place ≠ goal ≠ source : a separate marker for each of the three

14 d. place = source ≠ goal : one ambiguous marker for place and source, a separate marker for goal e. place ≠ goal = source : one ambiguous marker for goal and source, a separate marker for place

These are not equally common among languages worldwide. According to Andrews, thefirst three, (20a-c), are quite frequent while the remaining two, (20d-e), are unattested (1985, p. 97, cited in Pantcheva 2010, p. 1044). Pantcheva tested Andrews’s proposed asymmetry between the different syncretism patterns and found confirmation of his generalization. Therewas a clear skew in favour of the patterns presented in (20a-c) and no instances of the patterns in (20d-e) in her sample, both in accordance with Andrews’s suggestion (Pantcheva 2010, p. 1045). However, Pantcheva found that in languages of type (20a), with one and the same marker for the three local roles, the marker functioned as source or goal only in combination with certain verbs − much like in Wan or Mapundungun (see section 2.2.2) − and functioned as place otherwise. Pantcheva argues that these are place markers per default which can be used to express source or goal but are not, in themselves, source or goal markers. Thus the syncretism pattern in (20a), with one marker for all three roles, is in fact non-existent according to Pantcheva (2010, p. 1071). That place is the default interpretation of these markers could perhaps be seen as contradicting the claims by Stolz et al. (2014) and Haspelmath (2019) that place is more marked than goal. Nikitina concludes that there are four strategies for encoding the three local roles goal, source, and place, namely the ones presented in (20a-d) (2009, p. 1116). The system in (20d) − with the same marker for source and place, while goal has a separate one − is unattested according to Pantcheva (2010) and Andrews (1985), however, Nikitina points out that this is not true. She exemplifies the system with data from Dinka (Western Nilotic) and Nivkh (Isolate). In Nivkh a locative/ablative marker is used for both place and source, (21b-c), while a dative/additive marker encodes goal, (21a).

(21) Nivkh (Isolate) (Gruzdeva, 1998, cited in Nikitina 2009, p. 1115) a. N’i-∅ vyn’-∅ d’yr-doχ myr-d’ I-nom copper-nom brim-dat/add rise-fin ‘I rose to the brim of the copper.’ (goal = dat/add) b. T’ivlan̹ čag-ux̹ n̹at’x-∅ vezla-d cold water-loc/abl foot-nom cramp-fin ‘[I] have cramp in [my] foot in the cold water.’ (place = loc/abl) c. Umgu-∅ n’o-x p’u-d’ woman-nom barn-loc/abl come.out-fin ‘A woman came out from the barn.’ (source = loc/abl)

A fifth system of encoding is attested according to Nikitina, although very uncommon, where source and goal have the same marker while place has one of its own (as in (20e) − claimed to be unattested by Pantcheva (2010) and Andrews (1985)) (2009, p. 1116). This pattern can be found in Laz (Kartvelian). In the Ardesen-variety̹ of Laz there is a case called motative which co-expresses goal and source while excluding place (Kutscher 2010, p. 253). The motative marker -sa̹ only conveys that the figure has moved in relation to the motative-marked ground. The ground itselfcan function as either a goal (as in (22a)) or a source (as in (22b)) dependent on the spatial prefix on the verb or context (Kutscher 2010, p. 256). A ground functioning as place, however, is

15 unmarked in this variety of Laz, as is shown in (22c) (Kutscher 2010, p. 261). Other varieties of Laz mark place with a dative marker and source and goal with two separate cases.

(22) Ardesen-variety̹ of Laz (Kartvelian) (Kutscher 2010, pp. 261, 265) a. bere oxori-şa am-ulun child house-mot into-go:3a:sg.prs ‘The child goes into the house.’ b. bere oxori-şa gam-ulun-d child house-mot out-go:3a:sg.prs ‘The child goes out of the house.’ c. sis̹ e̹ masa goo- zunn bottle table on lie:3a:sg.prs ‘The bottle is (lying) on the table.’

The data from the Ardesen-variety̹ of Laz provides clear evidence for the existence of a system previously thought to be unattested, i.e. the one presented in (20e). Nonetheless, itis very uncommon. The typological asymmetry in frequency of syncretism between the different local rolesis represented by Nikitina with a semantic map in the following:

(23) Goal − Place − − Source (Nikitina 2009, p. 1116)

This reflects the likelihood of two roles being encoded by the same marker − the further the distance between them, the less likely they are to be encoded by the same form. By extension, this implies that if two non-adjacent categories in the same semantic space are encoded by the same form, the categories between them are also encoded by that same form (Nikitina 2009, p. 1117). This entails that goal and place are more commonly identically encoded than placeand source, which are, in turn, more commonly expressed by the same form than goal and source. This corresponds to the scales of likeliness of zero-marking and markedness in (10) withplace in the middle and source and goal maximally distant from one another. Syncretism between the local roles could be compared to alignment systems in how they group categories together. If a language differentiates between the core argument of an in- transitive predicate (S), the -like core argument of a transitive predicate (A), and the -like core argument of a transitive predicate (P), this can be done in different ways. Languages can have a tripartite system, with a separate marker for each category, or a neutral system, with one single marker for all categories (much like in (20c) and (20a), respectively, with local roles). It is also possible to encode two of the categories identically and have a separate marker for the third category, as in ergative-absolutive alignment (S, P − A) and nominative-accusative alignment (S, A − P) (much like in (20b) and (20d)). The WALS data- base (Dryer and Haspelmath 2013) features 190 languages in its chapter on alignment systems. 98 of them have neutral encoding, 52 have a nominative-accusative system (with or without marked nominative), 32 have ergative-absolutive alignment and merely 4 of them have a tri- partite system (another 4 have an active-inactive alignment system) (Comrie 2013). From this, one might conclude that it is economical to merge two categories, however, it seems that not just any two categories can be expressed by the same marker. In ergative-absolutive align- ment, the core argument of an intransitive predicate and the patient-like core argument of a transitive predicate are encoded identically. Since one of these appear with transitive predic- ates, and the other with intransitive predicates they will never co-occur and create ambiguity.

16 Therefore, S and P are suitable to be encoded identically. The same applies to nominative- accusative alignment − the identical encoding of S and A is unproblematic since they can never occur with the same predicate. However, the encoding of A and P is always distinct (with the exception of neutral systems, in which there are other means of distinguishing S, A, and P) since they contrast, appearing with the same predicate but expressing different func- tions. Expressing A and P with the same marker while having a separate marker for S would unavoidably lead to ambiguity and, therefore, A and P are generally encoded distinctly. The likeliness of encoding the different core arguments identically can also be represented bya semantic map, much like in the one in (23): (24) A − − S − − P This can, as mentioned, be compared to the encoding of local roles − place can be marked identically with both source and goal (as in (20b) and (20d)) since they never contrast (place being an expression of location, while source and goal are both expressions of direction), how- ever, source and goal are very rarely expressed by the same marker while place has one of its own (as in (20e)) since this leads to ambiguity.

2.3 Marking with different noun types Haspelmath argues that the marker of any local role tends to be shorter with place names, or toponyms, and topo-nouns (nouns commonly functioning as spatial landmarks, e.g. ‘village’, ‘(one’s) house’, or ‘school’) than with common nouns − that is if the marker differs between them at all (Haspelmath 2019, pp. 319, 322, 324). Furthermore, Haspelmath discusses the dif- ference in marking within common nouns. He shows that human, or animate, nouns tend to be more marked than non-human, inanimate nouns. This is in concordance with what Creissels concludes − that geographical names in general are less marked for spatiality, while human nouns generally have special, often longer constructions (2009, p. 611). An example of this is found in Basque (Isolate) where animate nouns require the marker -ga(n) when functioning as a local role, unlike inanimate nouns and toponyms (Creissels & Mounole, 2011, cited in Haspelmath 2019, pp. 319-320) as seen in Table 3.

Table 3: Case marking with different noun types in Basque (Isolate)

‘Irun’ (toponym) ‘the tree’ ‘the boy’ locative Irun-en zuhaitz-ean mutila-ga-n allative Irun-dik zuhaitz-etik mutila-gan-dik ablative Irun-a zuhaitze-ra mutila-gan-a

Table 3 clearly shows that the animate and inanimate nouns have different markers and that the marking of the animate noun ‘the boy’ is longer than that of the inanimate noun ‘the tree’. The markers for the inanimate noun are, in turn, longer than the ones for the toponym ‘Irun’, in line with previous claims. Haspelmath argues that special marking of animate nouns is most likely more common than having special conditions for place names, however this is not confirmed since special marking of animate grounds is not paid much attention inresearch according to Haspelmath (2019, p. 320). He partially links the encoding differences between the noun types to (un-)expectedness and disambiguation, but mostly to coding efficiency (fre- quential differences between the noun types in how often they do or do not function aslocal

17 roles) (see discussion in Haspelmath 2019, pp. 326-330). He illustrates the marking hierarchy of noun types along a scale he calls ‘spatial-reference scale’, as is represented in (25). (25) spatial-reference scale (Haspelmath 2019, p. 323) human noun > common inanimate noun > place name/topo-noun The higher, or further to the left, on the spatial-reference scale a noun type is, the likelieritis to have a marker for local roles and the longer that marker will be. However, Haspelmath does not resolve how toponyms and topo-nouns relate to each other in this regard. He points out that one can argue topo-nouns should be placed to the right of place names since e.g. English cannot omit prepositions with place names, while ‘home’ is grammatical and *‘to home’ is ungrammatical. Still, Haspelmath concludes that the lack of evidence for this is too extensive and that the notion of topo-noun may be too imprecise and thus leaves the relation between toponyms and topo-nouns unresolved (2019, p. 323). The spatial-reference scale is further supported by Stolz etal.(2014). 90% of their sample languages have zero-marking with toponyms, while only 41% of them have zero-marking with common nouns − regardless of (Stolz et al. 2014, p. 287). Of course, it is important to note that their sample is one of convenience in reviewing their results.

2.4 The linguistic situation in South America 2.4.1 Genealogy South America is an exceedingly diverse linguistic continent. There are 108 languages families, 55 of which are isolates4, i.e. languages without any known relatives and therefore language families with only one member, and 53 of which are families with at least two undoubtedly related members. 10 of the 53 families are relatively large, while the remaining 43 are smaller, with 6 or fewer members (Campbell 2012a, p. 59). The 108 language families in South America make up about a quarter of the world’s genealogical diversity since there are roughly 420 languages families, isolates included, in the world. Out of these 420 language families 117 are isolates. This means that about half of the world’s isolates are located in South Americaand thus over-represented on this continent (Campbell 2013, pp. 168-169). Approximately 420 South American languages are spoken today (Campbell 2012a, p. 59) and more than half of these, roughly 240 languages, are located in the Amazon (Rodrigues 2000, cited in Campbell 2012a, p. 59). Some of the languages belonging to South American language families are not spoken on the continent but further north, e.g. in , , , and , as well as on several islands. Moreover, the language families are widely spread. For example, are spoken in over a dozen countries (mainly in the northern half of the continent, but ranging from east to west) and on many Caribbean islands, and Tupían languages are (or were) spoken in nine countries (ranging from the east to west and north to south of the continent).

2.4.2 Typology South American languages are not only genealogically diverse − there are also extensive typo- logical differences, as well as similarities, between the languages. South American languages stand out in several ways in comparison to languages of the rest of the world, both in over- represented traits and unique ones. In this section the typological diversity and specificity of

4However, Campbell notes that these are relative figures and that ‘much depends on how some little-known languages, mostly extinct, are treated’ (2012a, p. 59).

18 South America is illustrated with a few examples regarding phonological and morphosyntactic traits. The presented examples are chosen relatively at random and are not intended tobean exhaustive description of the typological diversity of the continent. Some of the unusual South American traits are phonological. There are several examples of unique speech sounds found in South America. For example, Nivaclé (Matacoan) has a speech sound which, when it was discovered, added a new speech sound to the set of possible sounds in human languages − /k͡l/, a simultaneously released voiceless velar stop and voiced alveolar lateral resonant (Campbell and Grondona 2007, cited in Campbell 2012b, p. 264). This has opposed several suggested universals regarding liquids (see Campbell 2012b, p. 264; Maddieson 1984, p. 88). A second example of a unique speech sound is found in Wari’ and Oro Win (Chapakuran), namely [t͡ʙ̥] − a voiceless laminal dental stop followed by a voiceless bilabial trill, as an allophone of /t/ before /o/ and /u/. However, as was mentioned, these are merely two out of several examples of unique speech sounds found in South America. For a further discussion of phonological traits of South American languages see Campbell (2012b, pp. 263-272), but a few additional, widespread examples are: nasalized (an oral-nasal contrast in vowels); rhinoglottophilia (the of vowels next to /h/); nasal harmony (the spreading of nasality across segments); and small inventories with unusual phonemes. Concerning morphology and syntax, there are several features of interest among South American languages. For instance, every type of basic word order (that is, SOV, SVO, VSO, VOS, OVS, OSV) is represented, which gives the continent more diversity in basic word order than any area of the world (Campbell 2012b, p. 272). See Campbell (2012b, pp. 272-275) for a full discussion on this matter. South American languages also famously favour head-marking − that is, morphological marking of grammatical relations are expressed on the head, i.e. the verb, rather than the dependant(s) (Nichols 1986). Furthermore, Campbell claims that all types of alignment are attested in South American languages (2012b, p. 275). It is unclear whether or not this is true since Campbell then goes on to list only languages with nominative-accusative, ergative-absolutive, and active-stative systems and does not mention e.g. tripartite or direct alignment. Nonetheless, at least these three types of alignment, as well as neutral (Comrie 2013), can be found in South America. The Amazon has even been said to be ‘one of the most ergative areas in the world’ (Martins and Martins, 1999, p. 263, cited in Campbell 2012b). Moreover, South America is the only part of the world where active-stative alignment, i.e. systems dividing the marking of subjects of intransitive verbs between two categories, is a distinctive . These systems encode subjects of all transitive verbs and active intransitive verbs (that is, denoting events or some- thing happening) identically, while having a separate encoding for objects of transitive verbs and subjects of intransitive stative verbs (that is, non-events or states) (Campbell 2012b, pp. 277-278). Another feature of several South American languages is the lack of tense and aspect mark- ing in verbal morphology, which is rather common in the world, but what is interesting is that several South American languages use ‘nominal tense’ instead. This means that the tense of the entire proposition is encoded somewhere in the noun phrase, as seen in (26).

(26) Sirionó (Tupían) (Schmidtke, 2006, cited in Campbell 2012b, p. 286) a. Esi-ke oso ñá ií-ra woman-past go near water-to ‘The woman went near the water’

19 b. Jykv-ke uke-rv tiger-past sleep-pfv ‘The tiger slept’

It is unknown how prevalent nominal tense is in South America, but it is strikingly uncommon among languages worldwide and therefore typologically significant according to Campbell (2012b, p. 285). Finally, an interesting feature of several South American languages, in relation to motion events, is associated motion. Associated motion connects or relates motions to a verb event. For example, in Cavineña (Pano-Tanacan), associated motion suffixes can be added to verbs to express that the action is done while moving, exemplified below (Guillaume 2016).

(27) Cavineña (Pano-Tanacan) (Guillaume 2016, p. 2) ba- ‘see’ ba-ti- ‘go and see’ ba-na- ‘come and see’ ba-aje- ‘see while going’ ba-be- ‘see while coming’ ba-kena- ‘see and go’

This is interesting theoretically since it shows that motion is not inevitably expressed by lexical verbal roots but rather can be expressed grammatically. According to Guillaume, this contrasts what is commonly presumed in most studies of motion (such as those following Talmy, 1985) (2016). Associated motion was first identified in Aboriginal languages of central Australia, spe- cifically in a study by Koch (1984) on Kaytej (Pama-Nyungan). Since then it has been used to describe languages from other parts of the world, in particular . Guil- laume (2016) concludes that associated motion is widespread among languages located in or near the western and south western edge of the Amazon. Some further morphosyntactic traits of South America I will not discuss in detail are: sub- ordinate clauses as nominalizations; frustrative; switch-reference; rich systems of articles and demonstratives (deictics); large systems; inclusive-exclusive contrast in first per- son plural pronouns; noun classifiers; genitive classifiers; and a lack, or very small set,of adjectives (see Campbell 2012b, pp. 272-299). In Wälchli and Zúñiga’s study (2006) there are 14 South American languages, although three of them are Chibchan languages and not all of them are actually spoken on the continent of South America. Of these 14 languages two are considered mixed (i.e. featuring a weakly grammaticalized marker or construction for source), seven are classified as consistent (i.e. with consistent distinctive encoding of source and goal), and five are considered indifferent (i.e. with no explicit distinction between the two roles).

2.4.3 Previous description South American languages have long lacked proper representation in typological works and adequate documentation − before the 1960s, linguistic descriptions of South American indi- genous languages were practically non-existent (Adelaar 2012a, p. 16). In the WALS database (Dryer and Haspelmath 2013) the South American languages are under-represented. Only 76 South American language families are included (i.e. lacking 30% of the genealogical diversity of the continent) but, more importantly, most of the included

20 languages only have a few features registered in the database. This means that much more than 30% of the South American data is missing. Furthermore, the groundbreaking typological work Linguistic Diversity in Space and Time by Nichols (1992) has no more than 17 South American languages in its sample (seemingly from only 15 different families), thus lacking at least 84% of the genealogical diversity ofthe continent. Of course, this is a now a nearly 30 year old book, and the documentation of South American languages has improved since then, but the under-representation of South American languages in such influential, pioneering typological work is still important to highlight. However, South American representation is indeed increasing. For example, Ethnologue (see Eberhard et al. 2020) listed only 63 language families from South America in 2012 (Camp- bell 2012b) − that is, just over half of the continent’s total − but features nearly all of them today, about 94%. Since the 1990s the interest in South American languages, and endangered languages in general, has increased and only a few South American languages remain ‘un- touched by research’ toady (Adelaar 2012a, p. 24). However, the documentation for many languages is still very uneven in regards to coverage and quality (Muysken and O’Connor 2014). Undoubtedly, it is still very important to keep investigating the linguistics of South Amer- ica, not only due to the remaining, although decreasing, lack of documentation but also due to the extreme diversity, both genealogical and typological, of the languages. Typological stud- ies focusing on solely South American languages are useful in several respects, for example they can fill gaps in previous investigations or further attract the attention of typologists to the relevance of studying the languages of this continent.

21 3 Purpose and research questions

3.1 Purpose The purpose of this study is − with a starting point in Wälchli and Zúñiga’s study (2006) − to explore and elaborate on the encoding of source and goal in South American languages and thereby address the previous lack of proper representation of this area. The study aims at determining whether the local roles source and goal are expressed distinctively or indifferently in South American languages as well as how they relate to each other. The expression of path in South American languages is also examined, to the extent that this is possible given the data, since this is a particularly under-investigated local role. In addition, the study includes a strong methodological component in comparing the two data sources used.

3.2 Research questions 1. Are goal and source expressed distinctly or indifferently in South American languages?

2. If they are expressed distinctly, how do they differ from one another?

3. Are there any areal tendencies to the encoding of goal and source?

4. How is path expressed, alternatively what can be deduced about its encoding, in South American languages?

5. How do the two types of data, parallel texts (Bible translations) and reference grammars, differ in terms of efficiency and reliability?

22 4 Method

In this chapter, the data used for the study − both the parallel texts and the reference gram- mars − are described, followed by a methodological background on using parallel texts and Bible translations for typological studies. Subsequently, a brief methodological background on sampling methods is given and the process of creating the sample for this study is accounted for. Lastly, the individual steps of the analysis are explained.

4.1 Data Two data sources are used in this study − parallel texts, consisting of Bible translations, and reference grammars. The parallel texts are raw, unanalyzed translations of the New Testament collected from a large corpus of Bible translations (Mayer and Cysouw 2014). The study focuses entirely on certain Bible verses where either of two place names, ‘Jerusalem’ and ‘Macedonia’, occur in certain semantic roles, namely different local roles. The choice to study the expression of local roles with toponyms in the Bible translations is based on them generally being easy to recognize, no matter the language or orthography (see e.g. Finnveden 2019 for a study using personal names to find case marking in Bible translations). Furthermore, local rolesare not expected to behave substantially different with place names than with other noun types, despite the tendency to have shorter marking with toponyms than with other noun types (see section 2.3, specifically (25)). As Haspelmath (2019) mentions, animate nouns are probably more likely to have special marking of local roles than place names. The specific place names studied were chosen due to previous knowledge of their occurrence in the desired contexts. In total, 140 contexts from the Bible translations are used in this study. All of them are listed in Appendix A according to which local role and toponym they express. They are not from 140 different verses since one and the same verse may express e.g. motion both toandfrom one of the toponyms, or motion to one toponym and from the other. The number of contexts studied varies between languages since the toponym in question occasionally gets left out of a translation and some verses are left blank in some translations (if they are particularly difficult to translate, see de Vries (2007)). In both of these cases, the verses are excluded when discussing frequencies of markers and, naturally, these verses are not used in attempts to identify markers for local roles. Many of the contexts are similar to each other (at least in the English translations). For example, most of the verses expressing goal include the verbs ‘go’ or ‘come’ alongside ‘to’ or ‘up to’, but there are a few exceptions, such as ‘depart for’, ‘arrive in’, or ‘bring/take/carry’ something ‘to’. Depending on the translation, some verses also express ‘toward’. The verses expressing source share many similarities as well. The majority of them feature the verb ‘come’ alongside ‘from’, but a few of them express ‘leave’, ‘get out of’, ‘starting/beginning from’, or ‘send/turn over’ something ‘from’. The verses expressing path are very few − two of them express ‘pass through’ and one of them expresses ‘go back through’ (in some translations they express ‘on one’s way to’ or ‘by way of’). Lastly, the verses expressing place mainly behave similarly as well, expressing locatedness in or at one of the toponyms. A few verses expressing genitive or some other non-local role function, such as ‘the churches of Macedonia’ or ‘the whole of Jerusalem’, are included among the verses expressing place when grouping verses together (explained further in section 4.3. below) since these verses are not analyzed in detail in this study − occasionally the verses expressing place are examined to determine whether a marker expresses more than one local role or to estimate how frequent a marker is, but this is only on very few occasions, most often they are not examined at all since the expression of

23 place is not investigated in this study. Some particular verses, in the English translations, feature the toponyms in contexts which are more difficult to analyze or that, for some other reason, stand out. These verses are,for each sample language, marked with an asterisk to highlight the need to study these verses attentively. Some verses are marked with two asterisks to furthermore attract attention tothe possible difficulty in analyzing the verse. Some examples of such verses are given below.

(28) a. [There] a vision appeared to Paul in the night: a man from Macedonia stood pleading with him and saying, Come over to Macedonia and help us! (Acts, 16:9)5 b. [I wanted] to visit you on my way to Macedonia, and [then] to come again to you [on my return trip] from Macedonia and have you send me forward on my way to Judea. (2 Corinthians, 1:16) c. And behold , that very day two of [the disciples] were going to a village called Emmaus, [which is] about seven miles from Jerusalem. (Luke, 24:13) d. And Barnabas and Saul returned from Jerusalem , having fulfilled the service [en- trusted to them], taking also with them John, surnamed Mark. (Acts, 12:25) e. And after Barnabas and Saul had done the work they were sent to do, theywent back to Jerusalem with John, whose other name was Mark. (Acts, 12:25)6

In (28a) ‘Macedonia’ appears twice, once expressing the hometown of a man, and once expressing goal. This verse is marked with an asterisk to make sure that the right occurrence of ‘Macedonia’ is examined. Roughly the same applies to (28b) where ‘Macedonia’ also ap- pears twice − once expressing goal and once expressing source. However, not all translations feature ‘Macedonia’ twice in this verse, sometimes only the goal expression is included. In (28c) ‘Jerusalem’ does not denote the starting point of a motion, rather than the point from or to which a distance is measured. Therefore this verse is marked with two asterisks, tomake sure that if it is examined this is done very carefully. Of course, the verse could be excluded all together, however I choose not to exclude any data but rather mark problematic verses with asterisks. Lastly, (28d) and (28e) present the same verse from different English translations of the Bible. From these two translations it appears as though ‘Jerusalem’ can function as either a goal or a source in this verse. Due to this inconsistency, this verse is also marked with two asterisks to make sure that it is studied with caution if studied at all. The verses expressing place are, as mentioned, not used to the same extent in this study and are therefore not examined in this amount of detail beforehand. Hence, none of these verses are marked with asterisks even though some of them may also be difficult to analyze or stand out in some way. Regarding reference grammars, at least one, but preferably several, are used for the ana- lysis of each sample language. If a grammar is not available for a given language other de- scriptive sources, such as dictionaries, papers, or in rare instances teaching material, are used. Since most of them are indeed grammars (and most of the dictionaries feature a short gram- mar), henceforth ‘grammar’ (or ‘reference grammar’) is used as an umbrella term for the afore- mentioned sources. Several of the reference grammars used for this study are not written in English, but most commonly in Spanish or Portuguese, and occasionally in French or German. This entails that some detailed information about an encoding may have a higher risk ofbeing

5The translations used here, and throughout the rest of this thesis unless another source is specified, arefrom the Amplified English Bible. The reference includes the book the verse is featured in, followed bythegiven chapter of that book and the number of the verse itself (separated by a colon). 6This translation is from the Contemporary English Bible.

24 missed than if the grammars had all been written in English. However, the risk of missing something is always present, no matter what language a source is written in. The two sources of data fill rather different functions. While the reference grammarsare descriptive, specifying how a given feature functions through description and/or glossed ex- amples, the parallel texts offer more information about the usage of a given feature, e.g.how frequently it occurs in different contexts. This entails that a trait which is not mentioned ina grammar may merely not have been focused on, while a trait not appearing in a parallel text indicates that the trait is not used in this context (at least in the language variety of the given translation). In the following two sections some methodological background is provided by reviewing the advantages and disadvantages of using parallel texts in typological studies and discussing some specific challenges posed by using Bible translations.

4.1.1 Methodological background − Parallel texts Wälchli (2007) discusses the advantages and disadvantages of parallel texts as a source of data for linguistic typological studies both generally and in regards to certain criteria, namely diversity, domains, analysis, perspective, quality, representativity, and comparability. For a full review of these criteria see Wälchli (2007). In the following discussion, a brief summary of Wälchli’s assessment is given and, in that, a few comparisons are made to two other sources of data − questionnaires and reference grammars. Wälchli (2007) argues that linguistic structure is only accessible indirectly, in individual utterances, and that typology therefore always is a typology of texts and only byextension a typology of languages. A big advantage of studying parallel texts is thus that one is more aware of the textual basis. Working with several translations of the same text, rather than ones with different sources, is an another asset of the parallel text method because it entails knowing the meaning of the text as well as the structure of it. Thus, it is possible to identify shorter passages where the investigated construction or concept is most likely to occur and the segmentational analysis and glossing on those parts of the text (Wälchli 2007). Naturally, a disadvantage of working with translations is that they can feature errors, or other peculiarities, which can the analysis of the text in question and by extension the typology it is used for. Nonetheless, when considering frequencies, there can always be in- dividual errors − what is of most importance is whether a feature occurs with its natural frequency in the text or is over- or underrepresented (Wälchli 2007). A clear advantage of using parallel texts which also applies to the questionnaire method is that they illustrate language internal variety and therefore allow the study of the internal behaviour of the chosen domain. However, unlike parallel texts, questionnaires are depend- ent on informants, which limits the amount and diversity of the languages that can be in- cluded. Undoubtedly, reference grammars encompass more diversity, both genealogical and areal, than can be covered by any questionnaires studies and most parallel texts. Nonetheless, there is one set of parallel texts that are available in large numbers of genealogically diverse and geographically widespread languages, namely the gospels (Wälchli 2007). A benefit of questionnaires that does not apply to parallel texts or reference grammars is that they can be designed to exhaust all situations relevant for the construction or concept to be investigated. Parallel texts and reference grammars cannot be influenced by the researcher in such a way. Nonetheless, two different sources of data can to some degree each other according to Wälchli (2007).

25 There are a number of advantages with parallel texts, in comparison to reference grammars, which Wälchli subsumes under the term perspective. Grammars tend to have a system-bias − that is, be biased toward describing systematic structures, or describing structures as sys- tematic, and often pay less attention to exceptions. Furthermore, grammars tend to describe smaller structural units, such as morphemes, rather than constructions. These biases are not found in texts. According to Wälchli, parallel texts present a domain in a certain context from which a form can be deduced, resulting in a function-form, rather than system, orientation (2007). A final advantage of the parallel text method is that concrete examples are directly com- parable between languages. Ideally, the same domain is exemplified in the same context, the same register, and with the same degree of emphasis. These advantages often also apply to questionnaires, although sometimes with the exception of the same degree of emphasis. In studies of parallel texts, comparability can be further increased by selecting a few clauses which are expected to be featured in all translations and instantiate the investigated construc- tion or concept (Wälchli 2007). Ultimately, Wälchli shows that one of the most important advantages of the parallel text method is that it requires being conscious of the difficulties in comparing languages, due toall of its faults. When using parallel texts for typological studies, one must consider several biases such as a bias toward written language, conscious or planned language use, certain registers or varieties, non-native use of language, or translated language (rather than the original use). However, several of these biases also apply to other sources, such as reference grammars and dictionaries. In fact, staggering numbers of grammars and dictionaries are, at least to a certain extent, based on translated texts, and authors of grammars and dictionaries are not seldom also involved in Bible translation. Finally, Wälchli concludes that there is no perfect source of data for typological studies, but that the possible results of studying different sources of material are of such importance that it need be done even if it is done imperfectly (2007).

4.1.2 Methodological background − Bible translations Texts or works translated into many languages spread worldwide are, as shown above, a sig- nificant and intriguing source of information for typological studies. For example, thesetexts can consist of Bible translations. These are often available to the public and can rather eas- ily be used to create large parallel text corpora. Using translations of the Bible for linguistic purposes, however, entails certain specific challenges − partially since every translation is in- fluenced by the community in which it is created, but more importantly since the translations differ regarding which source texts they are based on (the Hebrew or the Greek), whichbooks of the Bible are in- or excluded, which readings are chosen when there are several possibilit- ies, what style level is chosen, and what amount of influence the source language has onthe translation (foreignization or naturalization) (de Vries 2007). Since a translation can never reflect all aspects of the source text, translations always en- tail selectivity and underspecification (or “deficiency” and “exuberancy” (Becker 1995, cited in de Vries 2007)). To elude these problems translators set up criteria arising from the skopos of the translation. Skopos is the Greek word for ‘purpose’ (established as a term in translation studies by Vermeer (2000, cited in de Vries 2007)). Nord (1991, cited in de Vries 2007) defines the purpose, or skopos, of translation as creating a functional text in the target language that retains a relationship with the source text − a relationship that is determined by the target text’s demanded or intended function. Several factors can influence the skopos of a trans- lation, especially regarding translations of the Bible. Translating the Bible involves deciding between textual traditions; which books to include; standpoint regarding the relation between

26 Divine Author and humans authors of the Bible; views on the relation between tradition and Church, the individual believer, and the Bible translation; and translational type − foreign- izing types (high degree of influence from the source language, ‘Holy Inspiration skopos’) or domesticating, neutralization types (low degree of influence from the source language, ‘mis- sionary skopos’). All these factors determine how a single translation functions, its skopos, and make Bible translations a unique and rather intricate source of data for linguistic studies (de Vries 2007, p. 151). For a more detailed review of the skopos determining factors see de Vries (2007). The discussion by de Vries (2007) can be summarized as follows: Bible translations are, just like any translation, influenced by its skopos, however, due to the religious nature of itssko- pos, Bible translations pose certain complications not posed by other translations. There are three main sources of difficulties when using Bible translations in studies with parallel texts. The first is the textual multiplicity. During the course of history, different textual versions of the Bible have been considered authoritative by different communities resulting in vari- ous source texts for Bible translations. Knowing which version a translation is based upon proves very important when comparing them. Secondly, since communities differ in canons, different sets of books are found in different Bibles. Sometimes even the order ornamesof books vary, creating more difficulties in comparing Bible translations. Finally, the third setof complications arises from the different types of translation − whether the translation is more or less literal and, by extension, whether there is more influence from the source or the target language . De Vries concludes that Bibles can be used in linguistic studies, but only if one considers the skopos of the translations, perhaps through consulting Bible translation experts, and what that entails in terms of base text, canon, and translation type (2007, p. 157).

4.2 Sample Sampling is the procedure of choosing an appropriate sample, or representative entities of a population, to establish traits that characterize the population in its entirety, since studying the whole population often demands resources, both in time and money, most researchers do not have. In this section, I firstly discuss sampling methods in general, followed by a description of how the sampling is done in this study to secure a sample with substantial genealogical and areal representation. The final sample, consisting of 37 languages (6 of which are isolates), is found in Appendix B.

4.2.1 Methodological background − Sampling methods In linguistic typology, the composition of the sample is a central methodological question. The type of sample required depends on the research question. In this study, the aim is to have a variety sample − that is, one capturing as much of the linguistic variety as possible − rather than a random sample7 or a probability sample − used for statistical testing. One can achieve such a sample in various ways, e.g. with the Diversity Value method (see Rijkhoff, Bakker et al. 1993, Rijkhoff and Bakker 1998) or the Genus-Macroarea method (see Miestamo 2003, 2005).

7Variety sampling uncovers linguistic variety decidedly better than random sampling (see Miestamo etal. 2016).

27 One can assume that the more connections, be them areal/cultural, genealogical, or oth- erwise, between two languages, the larger the chance that they are similar to each other. Therefore, the more diversity a sample offers in these respects, the likelier it istodisplay more linguistic diversity in the investigated domain. In discussing language sampling, Bell (1978, cited in Miestamo et al. 2016) introduces three terms: universe, frame, and sample. 8 The universe is the category of objects being studied, in linguistic typology that isallnatural languages, the frame is the channel through which one can access the universe, i.e. all the lan- guages one can find data for, and the sample consists of the objects being observed, namely the select languages being studied. The ultimate data for linguistic typology of course includes all possible languages (past, present, or future), however, in practice the goal of sampling is to ‘bridge the gap between the frame and the universe’ (Miestamo et al. 2016, p. 236). To do this, one must minimize the influence of extralinguistic historical factors on the sample. For example, sampling should not be based on the sizes of genealogical groupings since this clearly is a consequence of extralinguistic historical factors. Choosing languages that are independent of each other, i.e. with maximal genealogical and areal distance between them, can be achieved through stratification. When stratifying a sample, one divides the universe into subgroups, for instance genealogical or areal, and chooses a representation from each subgroup at random, rather than choosing at random from the entire universe. Thus, the representation is maximized and the biases are minimized (Miestamo et al. 2016). The subgroups in the stratification can have different bases, e.g.small linguistic areas vs. macroareas, or branches of language families vs. entire language families. The Diversity Value method creates samples based on genealogical stratification, namely the internal diversity of genealogical groupings, and has no areal stratification. The Genus- Macroarea method has both genealogical stratification and areal stratification, unsurprisingly at the genus and macroarea levels respectively, following the definition of genus by Dryer (1989, 1992, 2005, 2013, cited in Miestamo et al. 2016) as a ‘level of genealogical classification intended to be comparable across the world in terms of time depth’ (Miestamo et al. 2016, p. 239).

4.2.2 The sample of this study The sample of languages for this study is chosen so that they, as far as the data allows,are representative of the different families, isolates and areas of South America. As mentioned above, the final sample, with 37 languages (6 of which are isolates), is presented inAppendix B.

4.2.2.1 Genealogical stratification When creating the sample for this study, the genealogical classifications of (Ham- marström et al. 2019) are used due to their detailed and systematic classification procedure9, as well as their transparent account of the sources and evidence used for this classification. Due to the time limit of this study, I stratify genealogically at the level of language famil- ies, rather than using the Diversity Value or Genus-Macroarea method which stratify at lower levels, to reduce the size of the sample. As was mentioned in section 2.4.1, South America is home to exceptional genealogical diversity. There are 108 languages families, 55 of which are isolates, constituting about a quarter of the world’s language families. The aim is to include one randomly chosen language from each family in the sample (consequently including all

8These are originally statistical terms regarding sampling. 9Described in detail at https://glottolog.org/glottolog/glottologinformation.

28 of the isolates) with as much geographical distance between the languages as possible (see section 4.2.3 below). However, since Bible translations are an important source of data for this study, each language included in the sample must have a translation of at least the New Testament. Furthermore, since this is an investigation of South American languages, all lan- guages included in the sample must be spoken on the continent of South America. This means excluding , even though they are listed as languages of South (and North) America on Glottolog (Hammarström et al. 2019), since none of the languages are actually spoken in South America. Other excluded language families are the ones listed as spoken in more than one additional macroarea alongside South America. This means that all families listed as spoken in South America, but also e.g. Eurasia and Africa (such as Austronesian or Indo-European) are excluded, while families listed as spoken in both North and South Amer- ica, and no additional macroareas, (such as Arawakan or Chibchan) are included as long as they have at least one language, with a Bible translation, spoken in South America.

4.2.2.2 Areal stratification Certain features which are spread widely among South American languages suggest similar- ities for the continent as a whole. Yet, it is also clear that languages with certain traits group together in smaller constellations across families and, therefore, several linguistic areas have been proposed within South America. However, it is worth noting that all of these suggested areas and groupings are in need of further investigation. Most of them are not clearly distin- guished from adjoining areas regarding which linguistic traits motivate which area (Campbell 2012b, p. 301) − in most cases the traits overlap. Generally, there seems to be a typological division between eastern and western South America and languages along the and the Andean foothills appear to be especially connected to each other (Muysken, Hammarström et al. 2014). In Appendix C, there is a review of some linguistic traits that cover the entire continent, as well as a discussion of the possible justification, or lack thereof, of a few ofthe suggested linguistic areas (namely the ones reviewed by Campbell (2012b)). For a more de- tailed review of these areas see Campbell (2012b). The discussion in Appendix C can be summarized as follows: numerous linguistic traitsin South America form regional patterns, but most of them are distributed extensively, through- out several suggested linguistic areas. Moreover, many traits that are suggested to imply a linguistic area also fail to include several of the languages in that region. In short, linguistic areas in South America are anything but well-defined. Campbell even calls attempting tode- termine geographically distinguished linguistic areas in South America a ‘futile effort’ (2006, cited in Campbell 2012b, p. 310). Thus, I choose not to take these suggested linguistic areas into consideration when creating the sample for this study. Instead, for every language fam- ily, the languages with the furthest distance to the rest of the sample languages (based on their geographical placement on Glottolog (Hammarström et al. 2019)) are identified and one of them is chosen at random. Since the geographical distribution of the sample is used as a reference for which languages to include when creating the sample, there is a substantial dif- ference in how this affects the choice of language in the early versus late stages of the process. Therefore, I reiterate over all families with 7 or more languages (the limit used by Campbell for a relatively large , see section 2.4.1) to possibly choose a new language with further geographical distance to the rest of the sample as it is then constituted. When reiterating, I begin with the smallest family and end with the largest. The geographical distribution of the sample languages is presented in Figure110 below.

10The same map, although displaying which language is which, is foundat: http://arcg.is/0jOrSW

29 Figure 1: The geographical distribution of the sample languages.

4.3 Analysis The analysis of data in this study is done manually and is divided into two stages. Firstly, the Bible translations are examined. The select number of Bible verses studied are grouped together based on which local role they express − place, goal, source, or path − to more clearly highlight the differences and similarities between them. The aim is to identify possible encodings of source and goal for each sample language, or lack thereof, and, by extension, to attempt to determine whether the encoding of source and goal is distinct or indifferent ina given language. Secondly, the reference grammars are reviewed to try to distinguish the markers for source and goal, and whether or not they are distinct. If a marker of source, goal, or source and goal can be identified, the aim is to determine as much about the marker as possible, i.e. thetypeof marker, how many allomorphs there are, and in what context the marker, or different markers, appear. Furthermore, if the encoding of source and goal is distinct, it is also analyzed how the markers differ form one another, e.g. regarding complexity and length. The measuring ofthe length of markers will only take in to consideration the number of phonemes. Of course, this can be a difficult measurement since one phoneme can be represented by a sequence ofletters orthographically. Thus, certain sequences of letters which presumably represent one phoneme together are analyzed as one when determining of the length of the markers (this is discussed

30 more specifically in section 5.1.1.2). It is also my intention to examine how the reference grammars affect the initial analysis, based on the parallel texts. The possible differences between the results from the two sources of data constitute the basis for answering the methodological research question. When analyzing the encoding of source and goal, as discussed above, the aim is to de- termine whether they are encoded distinctly or indifferently. Thus, this study has a slightly different classification than the study by Wälchli and Zúñiga(2006). Their study differen- tiates between five types − consistent, mixed, indifferent, source extended, and inanimate − while the present study mainly distinguishes two types − clearly distinctive (correspond- ing to Wälchli and Zúñiga’s consistent) and not clearly distinctive. However, the not clearly distinctive languages have several sub-types, such as indifferent or intermediate (see section 5.1.2). The encoding of path is examined after the study of source and goal has been performed. Thus, the source, goal, or source/goal markers will already be known, if they are presentat all in a given language, and simplify the analysis of possible path markers considerably. Con- sequently, the process of determining the encoding of path, or lack thereof, is not included in the methodological analysis. With the exception of the methodological evaluation, the ana- lysis of path markers applies the same methodology, described above, as the analysis of source and goal markers. In this study, source is defined as the start point of a motion (encoded by e.g. prepositions like ‘from’) and goal is defined as the endpoint of a motion (encoded by e.g. prepositions like ‘toward(s)’ and ‘to’) in line with Kutscher (2010) (see section 2.2.2).

31 5 Results

In this chapter, firstly the results regarding the encoding of source and goal are reviewed − the languages with distinct encoding, followed by the languages with no clear distinction and the areal distribution of the languages. Secondly the encoding of path is briefly discussed, and thirdly, the methodological results are evaluated.

5.1 The encoding of source and goal In this section, the distinctly encoding languages are discussed as a group, whereas the lan- guages with no clear distinction are discussed separately since they behave very differently from one another. Thus only one example is given per type, or pattern, regarding distinctly encoding languages (in section 5.1.1), while all of the languages with no clear distinction are discussed individually (in section 5.1.2).

5.1.1 Distinct encoding 23 out of 37 sample languages clearly have distinct encodings of source and goal. A table listing these languages as well as their source and goal markers is found in Appendix D. One of the distinctly encoding languages is Huallaga Huánuco Quechua (Quechuan) where -pita marks source and -man marks goal, as shown in (29a-b) below, however, the endpoint of a motion can also be expressed through the object marker -ta with certain verbs, as in (29c). (29) Huallaga Huánuco Quechua (Quechuan) (Weber 1989, pp. 182, 189, 197) a. …queshpi-r aywa-ku-n wasi-n-pita… flee-adv go-refl-3 house-3.poss-abl ‘…they go fleeing from their house…’ b. Pillku-man aywa-shaq. Pillku-goal go-1.fut ‘I will go to Pillku.’ c. Pillku-ta aywa-shka-: awardenti-man. Pillku-obj go-pfv-1 firewater-goal ‘I went to Pillku to get firewater.’ Both -pita and -man can express meanings beyond the start and endpoint of a motion, re- spectively, such as the material out of which something is made for the source marker, and the endpoint of a metaphorical motion towards e.g. a new state for the goal marker (Weber 1989, pp. 192, 198). Undoubtedly, several, if not most, of the markers presented in this section will have numerous meanings. For the sake of brevity, clarity, and relevance I will henceforth only remark additional meanings if they display a relevant syncretism.

5.1.1.1 Complex markers At least eleven of the 23 languages with a clear distinction have complex markers for local roles − that is, one of the markers for local roles morphologically contains the marker for another local role (as discussed in section 2.2.4) or some other marker. 11 All but one of the complex

11It should be noted that this study is primarily focused on the relation between source and goal markers and, therefore, a place marker which is morphologically contained by another marker may have been missed. However, several instances of this is found as well (as shown below) and, thus, clearly not all of them are missed.

32 markers found in these languages are source markers containing other markers − the place marker, the goal marker or the place and goal marker when they are one and the same. The last complex marker is a goal marker containing a place marker. All of this is in concordance with Pantcheva’s formulated syntactic structure of the local roles − source markers tend to embed goal markers which, in turn, tend to embed place markers (see section 2.2.4, specifically (17)). Table 5, at the end of this section, lists the eleven (or possibly twelve) sample languages with complex markers and which other markers they contain. For some of the languages, however, only one allomorph of a local role is embedded in the complex marker. Below, a few examples of complex markers are given and one more difficult case is discussed. In Camsá, the goal marker is the suffix -oy (or -bi-oy with animate nouns 12), but more importantly, the source marker is the suffix -okan (or -bi-okan with animate nouns). -okan appears to be a combination of the locative marker -ok and -an (which, as far as O’Brien (2018) knows, is not a marker in itself). Thus, the locative marker is morphologically embedded in the ablative marker. Regarding toponyms in Camsá, it is specified that if they end in -ok, as many of them do since it is a locative marker and they are names of places, the first half of the ablative marker is left out and only -an expresses the ablative meaning. Example (30) below illustrates allative, locative, and ablative marking in Camsá (in (30a-c)), as well as ablative marking with toponyms ending in -ok (in (30d)).

(30) Camsá (Isolate) (O’Brien 2018, pp. 81, 83-84) a. së-n-j-a tj-oy 1sg-evi-vblz-go mountain-all ‘I went to the mountain/countryside.’ b. s̈lofts̈ i-nt-s-emn beti-ok bird 3sg-evi-prog-be tree-loc ‘The bird is in the tree.’ c. ats̈ së-n-tap tj-okan 1sg 1sg-evi-come mountain-abl ‘I came from the mountain.’ d. ch benach i-nd-emën chatjok-an asta bastok. det path 3sg-hab-be mocoa-abl until pasto ‘The road goes from Mocao until Pasto.’

In Murui Huitoto (Huitotoan) the source marker, -mona, morphologically contains both the place and goal marker at once since they are one and the same, -mo (glossed as a locative marker in the grammar). The second part of the source marker, -na, may be the topical non-S/A marker -na (Wojtylak 2017, p. 119), however, this is not established.

(31) Murui Huitoto (Huitotoan) (Wojtylak 2017, pp. 300, 302) a. bi-rui-yaɨ-do nofɨko-mo i-tɨ-kue this.cts-clf:day-pl-ins La.Chorrera-loc exist-lk-1sg ‘Nowadays, I live in La Chorrera.’ b. oo moo iye-mo aɨma-jai-d-e 2sg father river-loc fish-andtv-sg-3 ‘Your father is going away to the river to fish.’

12Which corresponds with both Haspelmath’s (2019) and Creissels’s (2009) findings that markers for animate nouns are longer than the marking for inanimate nouns (see section 2.3).

33 c. Ikato-mona duaɨbi-tɨ-kue El-Encanto-abl chew.coca.ventv-lk-1sg ‘I came from El Encanto to chew coca.’ There is one language which is more difficult to analyze that could possibly constitute another instance of a source marker containing a goal marker, namely (Ticuna-Yuri). In Ticuna, the encoding of goal is the suffix -wa and the encoding of source is ne13 (Anderson 1 1 1989, pp. 23, 387, 396). 14 Yet, almost all sentences expressing source, both in the grammar and in the Bible verses, also feature -wa (as in (32b) below). The only instances where -wa does 1 1 not co-occur with ne, according to Anderson, is when there is no explicit ground, but rather a 1 ‘location noun’ (1989, p. 359) such as in (32c). I suspect that this also applies to ngexta, ‘where’, since all examples in the grammar with this interrogative word are marked with ne but not 1 -wa15, however, it is not clear whether ngexta is included in the term ‘location noun’. It should, 1 at this point, be noted that -wa does not simply mark goal, but also meanings like ‘for’, ‘at’, 1 and ‘in’, and is at one point called a ‘location suffix’ by Anderson (1989, p. 359). Therefore, I have interpreted the suffix as a locative marker, to best reflect all of its usages. (32)below presents tentatively glossed and segmented examples of the expression of goal and source in Ticuna with -wa and/or ne. 1 1 (32) Ticuna (Ticuna-Yuri) (Anderson 1989, pp. 12, 15, 23) a. Pauru-pata-wa 3 5 2 3 1 Paul.poss-house-loc ‘to Paul’s house’ b. Yea, cha-pata-wa ne chaxū 1 4 5 2 3 1 1 3 23 there 1sg.poss-house-loc from 1sg.go ‘I’m coming from my house over there.’ c. Dauquena ne naxī 3 1 5 1 2 35 upstream from 3pl.go ‘They come from upstream.’ In the Bible verses the majority of the source expressions are marked with both -wa and 1 ne. There are two instances of -wa appearing without ne, interestingly, and no instances of ne 1 1 1 1 appearing without -wa. See table 4 for a full description of these frequencies. 1

Table 4: Frequency of source markers in the Bible verses for Ticuna (Ticuna-Yuri)

ne -wa ne -wa Other/unclear Total 1 1 1 1 Verse with source 0 18 2 6 26

Whether or not the source marker in Ticuna is complex, containing the goal (or locational) marker, is unclear. It seems, from the grammar, that -wa is always featured in expressions of 1 source unless there is an expression of an inherent place, which is supported by the frequencies

13The grammar does not specify what kind of marker ne is, though it appears to be a postposition. 1 14The numbers below the vowels in the markers indicate . 15With the exception of questions consisting of simply the interrogative word, ¿Ngexta? ‘Where?’.

34 of the different markers in the Bible verses. Possibly, this could indicate that the markingof source in Ticuna does contain other locational markers, but the presence of -wa in sentences 1 expressing source is not commented in the grammar. The source marker -pita in Huallaga Huánuco Quechua (Quechuan) is most likely also complex. It does not contain a marker for another local role, but rather the accusative marker -(k)ta (Adelaar and Muysken 2004, p. 215). Lastly, there is one language in the sample where the goal marker is complex. In Akawaio (Cariban) the goal marker -pona morphologically embeds the place marker -po (Caesar-Fox 2003, pp. 108, 115).

Table 5: Languages with complex markers for local roles

Source contains Source contains Source contains Goal contains Source contains goal place goal/place place other marker Lomeriano-Igna- Camsá Murui Huitoto Akawaio Huallaga Huánuco ciano Chiquitano Quechua Northern Emberá Candoshi-Shapra Aguaruna Palikúr Chipaya Shawi ?Ticuna

As mentioned above, Table 5 lists the eleven or twelve languages with complex markers for local roles and which other markers they contain. Ticuna is included among the languages where the encoding of source embeds the goal/place marker, however it is marked with un- certainty (‘?’).

5.1.1.2 Length of marker As was discussed in section 2.2.2, goal markers are more likely to be zero-marked than source markers, which thus tend to be more marked, i.e. longer, than goal markers. Place markers tend to be more marked than goal, but less marked than source. This relation can be repres- ented as follows (repeated from (10)):

(33) a. Likeliness of zero-marking goal > place > source b. Markedness scale goal < place < source

Of course, this can only apply to languages with distinct encoding, and therefore separate markers, of local roles, hence the discussion of this matter of this section. As was mentioned in section 4.3, the length of markers is measured in number of phon- emes. Since one phoneme can be represented by two or more letters orthographically, certain sequences of letters which are suspected to represent one phoneme are analyzed as onewhen determining the length of markers. More specifically, this applies to qu, sch, ch, and aa in this study. Most instances of complex markers discussed in section 5.1.1.1 display a source marker containing a goal marker, a place marker, or a goal/place marker. These languages concord with the general tendency in the world’s languages to have longer source markers than place and/or goal markers.

35 In Chipaya (Uru-Chipaya) the length of the markers for local roles correspond perfectly to the scale of likeliness of zero-marking presented by Stolz et al. (2014) (also discussed by Haspelmath (2019)) and the reformulation of this into a markedness scale (as presented in (33)). Source in Chipaya is expressed by the ablative marker -kiztan(a) and goal is expressed by the (when this appears with a motion verb) which is zero-marked, -∅. Thus, the goal marker is not only shorter than the source marker, but also than the place marker, -kiz(i) or -kin(a) (Cerrón-Palomino 2006, pp. 125, 128-129). This is exemplified in (34) below.

(34) Chipaya (Uru-Chipaya) (Cerrón-Palomino 2006, pp. 125, 128-129) a. wer-ki we-t qhuya-l-∅ oqh-u-tra 1sg-top 1sg-poss house-1sg-∅ go-1sg-decl ‘I go/am going to my house.’ (‘Yo voy a mi casa.’) b. Orur-kiztan wer thon-u-tra Oruru-abl 1sg come-1sg-decl ‘I come from Oruro.’ (‘Yo vengo de Oruro.’) c. wer-ki Chipay wath-kiz qam-u-tra 1sg-top Chipaya village-loc live-1sg-decl ‘I live in the Chipaya village/village of Chipaya.’ (‘Yo vivo en el pueblo de Chipaya.’)

Interestingly, there are a few languages that oppose the scale presented in (33). One such language is Palikúr (Arawakan). In Palikúr, the source marker contains the goal marker while place is zero-marked. Thus place has a shorter marking than the goal marekr -t (with the allo- morphs -(ri/i/a/u)t and -te) (Launey 2003, p. 139). This opposes the length relation presented in the markedness scale. In Bora (Boran), the place markers, -pañe and -ři-yi, are longer than the source marker, -tu (Thiesen 1975, pp. 17-18), which contradicts the length hierarchy presented in (33) as well. There are also a few languages with longer goal markers than source markers. For example, in Tenetehara (Tupian) the goal marker -kutir is clearly longer than the source marker -wi (Bonfim Duarte 2007, p. 41), exemplified in (35) below.

(35) Tenetehara (Tupian) (Bonfim Duarte 2007, p. 50, 208) a. ihe ɑ-dɨwɨr kwehe ʔɨ ɑpɨr kutɨr 1sg 1sg-come dist.past river headboard towards16 ‘I came towards the source of the river.’ (‘Eu vim em dirçeão à cabeceira do rio.’) 17 b. kon dɑnɨpɑw u-de-ʔok i-di-wi when genipapo 3sg-refl-release 3sg-refl-from18 ‘When the genipapo drops from it (the tree).’ (‘Quando o genipapo se solta dela.’)

However, -kutir barely appears in the Bible verses − it is only present in nine out of the 58 Bible verses expected to express goal, while pe and -ehe, both meaning ‘in’, constitute 47 of the goal expressions together. In the Bible verses expected to express source, on the other hand, the marker presented by the grammar, -wi, is appears in the majority of the verses. See Table 6 for a more detailed presentation of the frequencies.

16Original glossing: eu eu-vir dpass rio cabeceira em dirçeão a 17These, and the following translations from Portuguese to English, have been discussed with and approved by a native speaker of Portuguese with a professional knowledge of English. 18Original glossing: quando genipapo ele-reflex-soltar-se ela-reflex-de

36 Table 6: Frequency of markers for local roles in the Bible verses for Tenetehara (Tupian)

pe -ehe -kutir wi Other/unclear Total Verse with goal 42 5 9 1 1 58 Verse with source 6 1 0 15 0 22

Some Bible verses showing how pe and -ehe seem to be used in the same contexts as -kutir (kutyr in the Bible translation) are given below (although, without glossing or segmentation). (36a-b) show -kutir and -ehe (with the oblique prefix r-) appearing in roughly the same contexts and (36c-d) display the use of -kutir and pe in similar contexts. The Bible verses in (36) are rather long − to clarify which part of the verse is relevant for this illustration a shorter passage is underlined and the marker in question is presented in bold. (36) Tenetehara (Tupian) a. Upytu’u teko uhapukaz re wà. Na’e omono’og Pawru uzeruzar ma’e wamuwà uzepyr wà. Uze’eg wanupe wamurywete kar pà, wamukàg kar pà. A’e re —Zazur rihi, i’i wanupe no. Oho a’e wi wanuwi. Maxeton ywy kutyr ihon kury. ‘And when he had seen the vision, we [including Luke] at once endeavored to go on into Macedonia, confidently inferring that God had called us to proclaim the glad tidings (Gospel) to them.’ (Acts, 20:1) b. Amo ’ar mehe zaiko Ew tawhu pe zane. A’e re ihe zutyka’i aha Maxeton ywy rehe ihe. Heho ’ym mehe we aze’eg nezewe newe. —Epyta xe Ew tawhu pe nehe ty, a’e newe. —Amo awa xe har a’e wà, upurumu’e Tupàn ze’eg ’ym rehe a’e wà, a’e newe. —Ne ty, eze’eg eho wanupe nehe ty, a’e newe. —Pepurumu’e zo kwez ze’eg rehe nehe ty wà, ere wanupe nehe , a’e newe. ‘As I urged you when I was on my way to Macedonia, stay on where you are at Ephesus in order that you may warn and admonish and charge certain individuals not to teach any different doctrine,’ (1 Timothy, 1:3) c. Nezewe uze’eg ire oho Zezuz Zeruzarez tawhu kutyr a’e kury. ‘And after saying these things, Jesus went on ahead of them, going up to Jerusalem.’ (Luke, 19:28) d. Uzexak kar Zezuz Merez taw pe a’e, Zutez ywy rehe. Ero wiko a’e ywy rehe har wanuwihawete romo a’e ’ar mehe a’e. Nan kwehe tete ur amo awa Zeruzarez tawhu pe a’e wà. Zahytata kwaw par romo wanekon wà. ‘NOW WHEN Jesus was born in Bethlehem of Judea in the days of Herod the king, behold, wise men [astrologers] from the east came to Jerusalem, asking,’ (Matthew, 2:1) Furthermore, it should be noted that -kutir is translated as ‘towards’, or the Portuguese em dirçeão a, in (35a) − clearly longer than the translation of -wi ‘from’, or the Portuguese de, in (35b). If -kutir is longer than -wi solely because it denotes ‘towards’ – which is longer than ‘from’ in both English and Portuguese − the length relation between source and goal may not be particularly noteworthy in Tenetehara. Nevertheless, it is not established whether -kutir also marks ‘to’, and since ‘towards’ and ‘to’ are grouped together as goal markers in this study Tenetehara is treated as a language with a longer goal than source marker. A language which may display a longer goal than source marker, but where this is difficult to analyze, is Nadëb (Naduhup). In Nadëb the ablative marker is the postposition bʉ́́ and

37 the marker described as ‘contrastive locative’ (‘locativo contrastivo’) (Weir 1984, p. 96) is the postposition hẽnh. It is unclear exactly what ‘contrastive locative’ refers to, but it appears to denote a location that is different to something as in the locative tii hẽnh ‘there’ (see (37a)). However, there is one example in the grammar where hẽnh seems to function as goal marker, or at least be featured in a goal expression (see (37b)). (37) Nadëb (Naduhup) (Weir 1984, p. 31, 190) a. dooh gʉb tii hẽnh neg cardinal dem con ‘There are no cardinals there.’ (‘Não há cardninais lá.’) b. Manaaw hẽnh ı̃ɨh ha-hɨng con 1sg conrel-descend+ind ‘It is to Manaus I will descend to.’ (‘É para Manaus que vou baixar.’) In the Bible verses, hẽnh is found in 50 out of the 57 verses expected to contain a goal expression, and does not appear to a noteworthy degree in the verses expressing other local roles. A full account of these frequencies is given in Table 7. Clearly hẽnh has a tendency to indicate goal, nonetheless, it cannot be determined that it is a goal marker in Nadëb.

Table 7: Frequency of the marker hẽnh in the Bible verses for Nadëb (Naduhup)

Verse expressing: Goal Source Place Path Verses with hẽnh 50 6 7 1 Total number of verses 57 20 55 3

The postposition bʉ́ is, as mentioned above, glossed as an ablative marker in Nadëb. How- ever, Weir claims that this is simply a ‘convenient term’ for the postposition and that the use of it ‘does not always match up with the generally accepted idea’ of ablative (Weir 1986, p. 301). The wording of that quote indicates that bʉ́ still sometimes expresses the expected meaning of ablative, but this cannot be established without further investigation. Weir does present ex- amples of bʉ́ in a locative meaning, both with stationary location (see (38b)) and with motion (see (38a)), but mainly expressing other meanings, such as in a negative construction with the negation marker dooh (dooh … bʉ́) (as in (38c)): (38) Nadëb (Naduhup) (Weir 1984, pp. 174, 179) a. ı̃ɨh ba-hɨng bʉ́́ 1sg advrel-descend+ind abl ‘when I descended/descend’ (‘quando eu baixei/baixar’) b. Manaaw bʉ́́ Manaus abl ‘in Manaus’ (‘em Manaus’) c. dooh ı̃ɨh a-ód bʉ́́ neg 1sg form-cry+ni abl ‘I am not crying.’ (‘Não estou chorando.’) Weir argues that bʉ́ can express non-contrastive location (‘locativo não-constrastivo’) (Weir 1984, p. 174) as in (38a-b). This, as with hẽnh, of course does not entail that bʉ́ is a source

38 marker in Nadëb, however it does suggest that it stands in contrast to hẽnh. Whether this is a contrast between a source and a goal marker or merely a non-contrastive and a contrastive location marker needs further investigation. Therefore Nadëb is not counted as one ofthe sample languages with a clear distinct encoding of source and goal. Nonetheless, if they are source and goal markers, they indeed oppose the general tendency in languages to have a shorter goal than source marker, and if bʉ́ is not a source marker but a place marker (as in (38b)) Nadëb also opposes this tendency, since goal markers tend to be shorter than place markers as well. Finally, there is one language which has a source marker that seems to be both longer and shorter than the goal marker, namely Aguaruna (Chicham). The locative marker -numa (-(n)ĩ with demonstratives and 19 suffixes) can, in addition to location, express motion both into and towards. (39) Aguaruna (Chicham) (Overall 2007, pp. 222-224) a. wawiku-numa puhu-ha-i Wawik-loc live-1sg-decl ‘I live in Wawik (village)’ b. waã-numa tʃaat akuna-u cave-loc sym caus.enter.pfv-rel ‘he made it go zip! into the cave’ c. aha-numa wɨ-a-ha-i garden-loc go-impfv-1sg-decl ‘I’m going to the garden’ The source marker in Aguaruna is -ia, which can appear directly on locational nouns, but must co-occur with the location marker on regular nouns. (40) Aguaruna (Chicham) (Overall 2007, pp. 43, 226) a. kanusa-ia Santiago.River-abl ‘from the Santiago River’ b. hı̃ɰa-numa-ia house-loc-abl ‘from the house’ When the encoding of source consists of the ablative marker alone, with locational nouns, the source marker is shorter than the goal/place marker in Aguaruna − in contrast to the general tendency in the world’s languages. However, with regular nouns the source marker is longer than the goal/place marker − in accordance with the general tendency. In the Bible verses for Aguaruna, source is marked by -num-ia in the majority of the cases and -ia alone never marks source. In the verses expressing goal, -numa (realized as -num) does appear, but most often the marking is unclear. Table 8 presents the relation in length between the markers for local roles in the 23 lan- guages with distinct encoding. The length of place markers has only been included forthe languages where these happen to be known − place markers have not been searched for spe- cifically since they are not covered by this study. For some languages, not all allomorphs are

19A term coined by Dixon (2010), based on the Latin verb pertinēre ‘belong’, which encodes that the referent belongs to someone or something.

39 Table 8: Relation in length between markers for local roles

Source > goal Goal > source Source > place Goal > Place Place Same length (or goal/place) > goal > source Yagua Central Tunebo Candoshi-Shapra Palikúr Bora Central Aymara Camsá Tenetehara Chipaya Akawaio Páez Cuiba Arabela Cofán Amarakaeri Aguaruna Chácobo ?Nadëb Shawi Murui Huitoto Huallaga Huánuco Quechua Lomeriano-Igna- ciano Chiquitano Northern Emberá Aguaruna ?Ticuna

the same length and therefore the relation in length between the local roles cannot be de- termined definitively. To be able to asses the length relations in all of the distinctly encoding languages, the languages with allomporphs of different lengths will still be classified ashav- ing either longer source or goal markers, although with some difficulty. This difficulty ismost apparent with Northern Emberá (Chocoan) and Arabela (Zaparoan) whose source and goal markers are presented in Table 9 below.

Table 9: Source and goal markers in Northern Emberá (Chocoan) and Arabela (Zaparoan)

Language Source marker Goal marker Type of marker Source Northern -pa, -de-pa, -(e)da, -ma(a), Case suffixes (Mortensen 1999, Emberá -maʉ-pa (-de, -mae) pp. 47, 52-56) Arabela -cuaji, -ji -Vcuara, -(V)cua20, Suffixes (Rich 1999, pp. 74- -jinia(-ra) 75)

Since two source markers in Northern Emberá are longer than all of the goal markers I classify the language as having longer source than goal markers, even though one of the source markers is of the same length, or shorter than, the goal markers. In Arabela, there are two source markers, one long and one short, and three goal markers, two long and one short. The long source marker is shorter than the two long goal markers, and the short sourcemarker is shorter than the short goal marker. Therefore, Arabela is classified as having shorter source markers than goal markers, even though the short goal marker is indeed shorter than the long source marker. Furthermore, Aguaruna is included in two categories in Table 8 since the length of the source marker differs depending on the noun, and Ticuna and Nadëb are included, although tentatively for reasons discussed above.

20V = .

40 5.1.1.3 Analysis from Bible verses vs reference grammars For the vast majority of the languages with distinct encoding it is relatively unproblematic to determine the source and goal markers in the Bible verses. In these instances, the analysis from the Bible translations and the grammars is the same, with the exception of a few spelling differences21 and some additional allomorphs found in the grammars. Since the two sorts of analyses have not differed much for these languages, the discussion above is predominantly based on grammars due to the convenience of presenting already glossed and segmented ex- amples, as well as the assurance in citing language experts. There are three languages with a large number of markers for source and goal whichcom- plicates the analysis greatly, namely Cuiba (Guahiboan), Arabela (Zaparoan), and Northern Emberá (Chocoan). In these cases a few, or even several, of the markers were found in the ini- tial analysis, although among other supposed markers, however, when consulting the gram- mar many more markers were added. Other languages with several (yet fewer) markers for source and/or goal were also slightly more difficult to analyze than the ones with one marker per role, yet not to the same extent as the three languages mentioned above. To specify which languages do or do not have several markers for source and/or goal, and thus clarify how this discussion relates to the different languages, Table 10 shows the languages with onlyone marker per local role and those with several markers for one or both of the local roles source and goal. Ticuna is marked with uncertainty since it has not been determined whether -wa 1 is included in the expression of source − resulting in both ne and -wa ne as possible source 1 1 1 markers – or not − resulting in only one marker per local role.

Table 10: Languages with one or several markers for the local roles source and goal

One marker per role Several markers per role Several markers for one role Central Aymara Palikúr Shawi Bora Aguaruna Chácobo Camsá Northern Emberá ?Ticuna Candoshi-Shapra Cuiba Akawaio Arabela Central Tunebo Lomeriano-Ignaciano Chiquitano Cofán Amarakaeri Murui Huitoto Páez Yagua Huallaga Huánuco Quechua Tenetehara Chipaya

In Tenetehara, as confirmed above, the goal marker is underrepresented and the place markers are over-represented in the verses expressing goal. Therefore the initial analysis did not recognize the goal marker at all, but rather the two place markers. In Chácobo (Pano-Tanacan), source and goal are expressed through directional postposi- tions, however, some of these postpositions differ dependent on the of the verb. The encoding of source with transitive verbs is ka (with the allomorphs ʔoka and -noʔoka) and

21Not counting variations in the use of diacritics, the following have been interpreted as spelling differences in this study: p−b, ʂ−x, b−v, a−aa, j−h, o−u, i−y and i−ee.

41 with intransitive verbs it is ʔaʂ̌ (with the allomorphs ʔoʔaʂ̌ and -noʔoʔaʂ̌) (Prost 1962, p. 110). The encoding of goal, (-)ki, does not vary with the valency of the verb, however, it can occur either as a suffix and a postposition (Prost 1962, p. 111). In the analysis of the Bible verses the goal marker and the intransitive source markers were both easily identified, but the transitive source markers were not identified at all. Most likely, this is due to very limited amountof Bible verses studied and the verbs featured in them. There are a few, approximately five, verses with transitive verbs which express source, however none of these display the marker (-)ki. There are four instances of the marker ca in the position where ʔaʂ̌ otherwise appears and at least two of them are in verses expected to have a transitive verb. Nonetheless, it is unclear to me whether ca and (-)ki are the same marker.

5.1.2 No clear distinct encoding The languages with no clear distinct encoding of source and goal display a wide varietyof systems for encoding local roles. These are described in the sections below. Three tables (Table 13-15) listing the languages with no clear soure-goal distinction and, to the largest extent possible, describing their encoding of local roles is found in Appendix E.

5.1.2.1 Indifferent encoding Seven languages are presented in this section − two with a confirmed morphological indif- ferent encoding of source and goal, one with a suspected morphological indifferent encoding, two with no suspected morphological encoding where the difference is presumed to be en- coded solely by the lexical meaning of the verb, and finally two where the encoding of local roles is more difficult to analyze. Two sample languages have an explicitly ambiguous morphological encoding of source and goal. As was mentioned in section 2.2.2, Mapundungun (Araucanian) encodes source and goal indifferently with the postposition mew (occasionally mo or mu), however, the roles can be distinguished by the lexical meaning of the verb or through context. In (41) below it is exemplified how the source-goal distinction can be made by the verb.

(41) Mapundungun (Araucanian) (Zúñiga 2006, p. 55) a. Amu-ɑ-n wɑriɑ mew. go-fut-1sg city ppos ‘I am going to the city.’ b. Wɑriɑ mew küpɑ-n. city ppos come-1sg ‘I have come from the city.’

The postposition mew appears in the majority of the studied Bible verses, both express- ing source and goal, yet there are also verses without mew. In many of those verses deictic verbal suffixes, like -pa ‘here, hither’ and -pu ‘there, thither’ (Zúñiga 2006, pp. 41-42), appear, however, these do not express a source-goal distinction. The second language with a definite indifferent morphological marker for local rolesis Kakua (Kakua-Nukak). The locative marker =bǔ encodes place, source, and goal, and is some- times accompanied by the object marker =diʔ (Bolaños 2016, p. 206).

42 (42) Kakua (Kakua-Nukak) (Bolaños 2016, pp. 132, 206-207) a. tʃamu-ǐʔ mi=bǔ Samu-poss house=loc ‘(we were) at Samu’s house’ b. ʔIna wâptʃiʔ=bǔ mi=beh-bip=na=ka Ina manioc.field=loc 3sg.f=go-fut=decl=ass ‘Ina will go to the manioc field’ c. tʃena=bǔ kan=ná wât-jùʔ-ɲɨʔ-ɨp two/both=loc 3sg=decl blossom.of.big.flowers-toss-stop-past ‘(it) blossomed from both sides’ d. hiw ʔã=ɲap=na=be nin=bǔ=diʔ=héʔ jaguar 3sg.m=jump=decl=rec.past this=loc=obj=ints ‘The jaguar jumped right here’ The locative marker was easily identified in the Bible verses expressing both goaland source, and the object marker appeared in several verses as well. Tatuyo (Tucanoan) has two locative markers which both appear in the Bible verses for source and goal, albeit whether or not they express direction is not specified in the grammar. When expressing spatial location in Tatuyo, the specifically locative marker is -´pɨ̀, however the marker of circumstance, -rè, also occurs. The two markers can also co-occur as -´pɨ̀-rè, alternatively the spatial location is not marked at all or expressed by a locational noun. The variation in marking is related to the meaning of the verb, the direction of the movement, and properties possessed by the referent according to Gomez-Imbert (1982, p. 68). It is not specified in the grammar just what the relation between the marking and these factors is,but in the one example of a source expression the locative marker -´pɨ̀ is featured (see (43a)), while it only appears in one of two very similar goal expressions (see (43b-c)).

(43) Tatuyo (Tucanoan) (Gomez-Imbert 1982, p. 68) a. ̃wàtí-a-bàà-´pı̀ yı-∅-à(tí)-∅-wɨ̀ forest.spirit-clf-path-loc nomcl.1sg-time.gone.by-come-evi-accomp ‘I came from the Path of Spirits.’ (‘Je suis venu du Chemin des Esprits.’) b. sãtiagu-tıı́ ́ yı-∅-àá-∅-wɨ̀ Santiago-pres nomcl.1sg-time.gone.by-go-evi-accomp ‘I went to Santiago.’ (‘Je suis allé auprès de Santiago.’) c. wàì-yà-´pı̀ yı-∅-àá-∅-wɨ̀ fish-clf(=stream)-loc nomcl.1sg-time.gone.by-go-evi-accomp ‘I went to Piraparaná.’ (‘Je suis allé au Piraparaná.’)

However, it seems that -´pɨ̀ and -rè are not semantically interchangeable as shown by dif- ference between the two unglossed examples in (44a) and (44b) below (Gomez-Imbert 1982, p. 69) − -´pɨ̀, unlike -rè indicates that a given purpose or goal was achieved or reached.

(44) Tatuyo (Tucanoan) (Gomez-Imbert 1982, p. 69) a. ká-èhá-yú-pá’-o, wíì-rè ‘She arrived at the house, they say.’ (‘Elle arriva à la maison, dit-on.’) b. ká-èhá-yú-pá’-o, wíì-´pɨ̀ ‘She reached the house, they say.’ (‘Elle arriva jusqu’à la maison, dit-on.’)

43 It remains undetermined whether the locative markers function as source and goal markers as well, and precisely how different factors affect which marking of spatial relation appears in a given sentence. However, both -´pɨ̀ and -´pɨ̀-rè are heavily featured in the Bible verses expressing source as well as goal. One of the two languages seemingly without morphological encoding of source and goal is Cha’palaa (Barbacoan). Cha’palaa has many markers of spatial relations (Vittadello 1988) yet none of them seem to express the start or endpoint of a motion, rather they predominantly express location and position. Both the Bible data, showing no clear tendencies, and one of the grammars (namely Abrahamson 1962) indicate that there is no source-goal distinction in Cha’palaa, and furthermore, that there is no grammatical encoding of these local roles. As is shown in the translation in (45) below, ‘from’ and ‘to’ are parenthesized − and therefore ‘added to smooth out the English translation’ (Abrahamson 1962, pp. 238-239) − as well as left without indexation − clearly suggesting that source and goal are not overtly expressed in Cha’palaa but rather encoded indifferently. (45) Cha’palaa (Barbacoan) (Abrahamson 1962, p. 238-239) 1huŋga 2hitu 3enu 4himi 5san 6lurenzusha 2Having gone 1there (from) 3here (we) 4went (to) 5San 6Lorenzo. Regarding the other language which seems to lack morphological encoding, Southern Nambikuára, five grammars (Kroeker 2003, 1982; Kroeker 2001, 1996; Lowe 1999) were consul- ted in the attempt to determine the encoding of source and goal. They indicate that thereisno grammatical encoding of these local roles, and that the distinction is made by lexical meaning of the verb (as exemplified below). It is not known for certain whether this analysis is correct, however, ɑ3li3 ‘leave’ and ɑ̃3nxɑi3 ‘go’ appear in a great number of the verses expressing source and goal, respectively. Apart from these verbs, the Bible translations indicate no patterns. (46) Southern Nambikuára (Nambiquaran) (Kroeker 2001, pp. 3, 9) a. Hī1nɑ2su2 hɑ3lo2ɑ2 u2lxi3 ɑ̃ 3nxɑi3-nɑ1-tu1-wɑ2. today in.the.fields far.away go-1sg-fut-impfv ‘Today, I’m going to the distant fields.’ b. Jo3ɑ̃ u2-ɑh3lɑ2 ɑ3li3-∅-nɑ2hẽ3-lɑ2. John-man leave-3sg-t/e.io.past-pfv ‘John left.’ c. Jo3ɑ̃ u2-ɑh3lɑ2 ɑ3li3-∅-nũ2lɑ2 sɑx3we3nɑ2 ɑ̃ 3wih1-∅-nɑ2hẽ3-lɑ2. John-man leave-3sg-dsq forest enter-3sg-t/e.io.past-pfv ‘When John left, he went into the forest.’ Paumari (Arawan) is one of the languages where the encoding of local roles is more difficult to analyze. Paumari has approximately 26 directional suffixes which express meanings like: ‘to go from mid-water towards the shore’, ‘to go from the shore towards the jungle’, ‘to go from open ground into the jungle’, or ‘to bring two vertical objects together’; ‘coming from jungle towards water’ or ‘going from shore to the middle of a lake or river’; ‘parallel to water’s edge’ or ‘from one house to another’; ‘movement within a defined area’; and ‘coming from within a defined area’ (Chapman and Derbyshire 1991, pp. 311-315). These are merely some of the directional markers, not even mentioning the positional and locational suffixes in Paumari. Undoubtedly, there are grammatical encodings of directionality in Paumari, however, whether or not there is a source-goal distinction is more difficult to determine. According to Wälchli (2009), nothing in Paumari functions like the English prepositions ‘to’ and ‘from’. Instead,

44 Paumari uses the motion verb gai (which can express ‘come’ and ‘go’ both) in combination with one of its many directional suffixes. The suffixes do not denote the start or endpointofa motion, but rather the route in its entirety − specifying both the direction and the nature of the ground, e.g the shoreline or the jungle. This is exemplified in (47) where gai and the suffix encoding an up bank direction, -mori, express both source, (47a), and goal, (47b). Both (47a) and (47b) are examples from Bible translations.

(47) Paumari (Arawan) (Wälchli 2009, p. 1) a. …a-vi-gai-mori’a-ha away-pl-go/come-up.bank-comp-theme ‘…and when he was come out of the ship…’ b. …arabo nama-hi-ki-a o-gai-mori-’a-ha mountain above-aux-desc-erg/obl away-go/come-up.bank-comp-theme ‘…and he goes up into a mountain…’

While both situations in (47) are similar in terms of direction, they are also clearly different regarding the relation between the direction and the ground. And while they are expressed by the same means in Paumari, they are not in English. This clearly suggests that the categor- ization systems in Paumari and English are substantially different. These findings suggest that Paumari has indifferent encoding, however, this system differs greatly from the ones discussed above. It seems that the sense of source and goal is absent in Paumari − the focus of the encoding is on the whole route rather than its start or endpoint. Unsurprisingly, the Bible verses for Paumari display a wide array of markers and no clears patterns. Thus, they were not helpful (within the time limit of this study) in establishing the encoding of local roles in Paumari. Lastly, in Maxakalí (Nuclear-Macro-Je) the encoding of source and goal is seemingly also ambiguous, however, it is very difficult to determine how the indifferent markers function from the discussion of them in the grammar. According to Pereira, the notions of direction and origin are close and ‘interdependent’ to some degree (‘estão bem próximas e são de uma certa forma interdependentes’) in Maxakalí (1992, p. 71). It seems that there are three different ‘spatial areas’ (‘àreas espaciais’) in Maxakalí discourse − origin, destination, and an interme- diate area, called ‘nuclear’, ‘marginal’, and ‘transitional’ respectively (Pereira 1992, p. 72). The nuclear marker is tu, the marginal marker is ha, and the transitional marker is ta (tu’, ha’, and ta’ in the grammar, however the glottalstop (-’) is glossed as a third person marker). It is not clear exactly how these expressions of origin and destination relate to source and goal from this short discussion by Pereira. However, in (48) it is shown, with examples from a Maxakalí- English dictionary, how both ha and tu can express goal. The dictionary translates both ha and tu into both ‘to’ and ‘from’ (with no mention of ta expressing any spatial relation) − there are no glossed examples of ha or tu expressing source, but a tentatively glossed, unsegmented example of tu denoting source is found in (48c) below (Popovich and Popovich 2005, pp. 5, 43).

(48) Maxakalí (Nuclear-Macro-Je) (Popovich and Popovich 2005, pp. vi, xiii, 56) a. Ũgmũug mõg nɑ̃ pet hɑ nũy xɑ hɑ̃ pxop ũ pop. 1pl.excl go market to in.order.to 2sg food some buy ‘We (excluding you) are going to the open market to buy you some food.’

45 b. Piya ‘ mõg ‘ɑx ãxa õgnũ yip? komẽn ũm tu-k mõg Belo where 3sg go can (hearsay) 1sg.poss jeep town any to-1sg go Belo Horizonte tu k-mõg ‘ɑx… Horizonte to 1sg-go can ‘’Where could you go,’ they say, ’(if you had) your own jeep?’ I could go to a city. I could go to Belo Horizonte…’ c. Pet te yĩktix tu yĩxoho. Peter subject.indicator mountain from descend ‘Peter came down from the mountain.’ In the Bible verses, tu is considerably more frequent than ha. tu marks goal in about 77% of the verses and source in roughly 52% of the verses, while ha constitutes approximately 18% of the encoding of goal and about 10% of the encoding of source. ta does appear in the Bible verses, although never in connection to either of the toponyms. See Table 11 for a full description of the frequencies.

Table 11: Frequency of markers for local roles in Bible verses in Maxakalí (Nuclear-Macro-Je)

ha tu ta Other/unclear Total Verse with goal 10 43 0 3 56 Verse with source 2 11 0 8 21

The findings from the dictionary and the Bible verses combined suggest that theencoding of goal and source is indifferent in Maxakalí, however, the function of these markers remains unclear. In the Bible verses they appear to modify the verb somehow, occurring especially often with light verbs like mõg ‘go’, nũn ‘come’, xip ‘be’, and yũm ‘sit’, yet they also frequently appear directly after the toponyms, seemingly as postpositions − in line with how Popovich and Popovich (2005) describe them.

5.1.2.2 Goal marking with unknown source marking There are four sample languages with more or less apparent goal markers, but the markingof source, if there is one, is unknown. These languages are reviewed in this section. Two of the languages have an allative marker specified in their respective grammars. In Yanomámi (Yanomamic) the encoding of allative is the enclitic =ɑmɨ (sometimes realized =ɑmë), however, this is also the encoding for locative − =ɑ/ɑmɨ (Perri Ferreira 2011, p. 42). (49) Yanomámi (Yanomamic) (Perri Ferreira 2011, pp. 42, 63) a. huru=ɑmë ɑ=u=huri farm-ala 3sg=go-dir:there ‘He went to the farm.’ (‘Ele foi para a roça.’) b. urihi=ɑmë uru të=pë=ɑ̃ rë=u=ɑ=pɑ=ri=o=wëi forest=loc child clf.gen=pl=dim=go=distr=act=pfv=⁇?=rel të=howe=h[vv]=ɑ=mɑ=rɑ clf.gen=be.good=neg=lk=neg=pres ‘Little boy cannot/should not walk alone in the bush/jungle.’ (‘Menino pequeno não pode andar sozinho no mato.’)

46 Neither the grammar nor the Bible verses indicate a source marker for Yanomámi. How- ever, the goal marker specified in the grammar is featured in many of the Bible verses express- ing goal. The second language with a specified allative marker but lacking information aboutthe encoding of source is Wichí Lhamtés Nocten (Matacoan). The suffix -hu (glossed as applicative) can mark both beneficiaries or receptors and goal (Terraza 2009, p. 220), exemplified in (50a-b) below. However, it seems -hu can denote place as well, as in (50c).

(50) Wichí Lhamtés Nocten (Matacoan) (Terraza 2009, p. 220) a. n-wen-hu pala Miguel 1-have-apl shovel Miguel ‘I give the shovel to Miguel.’ (‘Doy la pala a Miguel.’) b. empleados de INAI to i-hu-hu Oran employees of INAI sub 3-go-apl Orán ‘INAI employees who went to Orán.’ (‘Los empleados del INAI que fueron a Orán.’) c. i-tat-hu močila la-coy 3-direct-apl backpack 3.poss-clothes ‘He/she/it put the clothes in the backpack.’ (‘Mete la ropa en la mochila.’)

Wichí Lhamtés Nocten has directional markers for ‘upwards’, ‘downwards’, and ‘forwards’, as well as locational markers expressing ‘far (away)’, ‘in(side)’, and ‘place’. Nonetheless, there seems to be no source-goal distinction. The two most apparent source expressions in the grammar are encoded by -hi, ‘inside’, and -po, ‘upwards’, respectively, as exemplified below.

(51) Wichí Lhamtés Nocten (Matacoan) (Terraza 2009, pp. 208, 213) a. taa-toy-hi tox -i-hi 3-disappear-loc sub 3-be-loc ‘Disappear from here’ (‘Desaparece de aqui’’) b. n-ne-po to n-toɫ-kye-xwi n-wet 1-leave-dir sub 1-pass-distr-apl 1.poss-house ‘I leave my house (step from the inside to the outside).’ (‘Salgo de mi casa (paso de adentro hacia afuera).’)

The Bible verses for Wichí Lhamtés Nocten do feature the goal marker (among afewother unidentified markers), but the marking of source is unclear. In Mocoví (Guaicuruan) there are several verbal enclitics marking location and direction, expressing 14 different meanings. Three of them encode goal-like notions − =pegeʔ or =peʔ, ‘up to’; =kena, ‘towards here’; and =igi, ‘towards (there?)’ (Grondona 1998, p. 132). While -pegeʔ does occur a few times in the Bible verses expressing goal, it cannot be determined that it is always connected to the toponym in question. =kena also appears sporadically in the verses with goal, yet it cannot be determined that it ever occurs in connection to the toponym in question, while =igi does not appear in the Bible verses at all. Regarding source markers in Mocoví, they are not discussed or exemplified in the grammar and the Bible verses do not indicate any specific markers. Lastly, the fourth language with a suspected goal marker and an unknown source marker is Chamacoco (Zamucoan). The only grammar for Chamacoco which has been available isnot actually a grammar but an ethnographic dictionary. It is very difficult to attempt to determine anything from this, however, the Bible verses expressing goal indicate that (-)tɨ̃r marks goal

47 in Chamacoco − it appears clearly connected to the toponym in 45 out of the 61 verses and in an additional three verses where it may be linked to the toponym. When examining the dictionary (Sušnik 1970), it appears as though this marker denotes ‘towards’ (‘hacia’). In (52) different examples indicating that (-)tɨ̃r (or some similar form) expresses goal are presented. These examples are not segmented or glossed due to the inadequacy of the dictionary. (52) Chamacoco (Zamucoan) (Sušnik 1970, pp. 135-136) a. tīr prāpo ‘where? (towards where?)’ (‘¿hacia donde?’) b. tû tīr pórt ‘moving up/upwards’ (‘en movimiento hacia arriba’) c. toǵâxa tû tī̄rn Bāya ‘I am going to/towards Black Bay.’ (‘Me voy a/hacia Bahia Negra.’) Perhaps (-)tɨ̃r, or (-)tīr, is not a dedicated goal marker, but it is undeniable that it expresses directionality of some kind. The possible goal marker was found accidentally in the dictionary (in a discussion onpo- sitional expressions). There was no such stroke of luck regarding source markers. Nothing has been found about the encoding of source in the dictionary, but the Bible verses expressing source do suggest one marker − otãra. This marker appears in half of the verses clearly con- nected to the place name in question, while the other half of the verses are split rather evenly between other markers or unclear marking. Furthermore otãra does not appear in any of the verses expressing goal, perhaps indicating that it stands in contrast to goal expressions. As with (-)tɨ̃r, otãra may not be a dedicated source marker, nonetheless it seems to indicate source more so than any other marker in Chamacoco. See Table 12 for the frequencies of different markers in verses expressing source and goal in Chamacoco.

Table 12: Frequency of markers, possibly for local roles, in the Bible verses for Chamacoco (Zamucoan)

(-)tɨ̃r otãra -ɨ̃hɨ -chɨ Other/unclear Total Verse with goal 45 0 3 0 13 61 Verse with source 2 10 3 2 3 20

In (53) below, the occurrence of (-)tɨ̃r and otãra in the Bible verses are exemplified. In (53a) it appears as though (-)tɨ̃r expresses goal, and in (53b) otãra seems to express source. Since the verses are long, the marker and toponym in question are presented in bold in each verse. Note that several place names appear as the source of a motion in (53b) and, seemingly, they are all marked with otãra − further supporting the hypothesis that otãra could mark source in Chamacoco. (53) Chamacoco (Zamucoan) a. Eseekite uje ich yuwɨrke ese pehebɨt data uje oshiyokorı̃hɨ kanshɨt ehetɨke hn Pablo shuu otɨbii wɨr Porrosht aabo shuu otakɨsho lawich hn sakɨt õr. Uje ich yuwɨrke hn tatɨm õr ɨm: —Takaha tɨtɨ̃r Macedonia.— ‘AFTER THE uproar had ceased, Paul sent for the disciples and warned and con- soled and urged and encouraged them; then he embraced them and told them farewell and set forth on his journey to Macedonia.’ (Acts, 20:1)

48 b. Uje oso ochunt nos wɨr kuche uje Jesús shiyokõr ich okɨhniya uje otãra wahacha hnɨmich Judea ohno õya ɨre hn yewo otãra Jerusalén hn yewo otãra hnɨmich Idumea hn otãra onoota Jordán pɨt hn otãra dɨt Tiro hn dɨt Sidón. Ohno otı̃r Jesús par uje oteychɨm. ‘And from Jerusalem and Idumea and from beyond the Jordan and from about Tyre and Sidon – a vast multitude, hearing all the many things that He was doing , came to Him.’ (Mark, 3:8)

5.1.2.3 Intermediate systems There are a few languages which cannot neatly be classified as either having orlackinga source-goal distinction. One such language has been discussed above (in section 5.1.1.2), namely Nadëb (Naduhup). Perhaps the contrastive and non-contrastive locative markers, hẽnh and -bʉ́́, indeed pose a source-goal distinction, however, this cannot be determined within the limits of this study. Therefore I classify Nadëb as an intermediate case − not clearly displaying indifferent encoding nor a distinction. Another possibly intermediate system of encoding goal and source is found in Central Tunebo (Chibchan). Central Tunebo is one of the 23 languages classified as having a source- goal distinction, and furthermore a language with a longer goal marker, acam, than source marker, bin. However, with animate nouns the postposition quin marks both source and goal (Headland 1994, p. 31), making Central Tunebo a language which in Wälchli and Zúñiga’s terms (2006) would be classified as inanimate − a language with distinct encoding with inan- imate nouns, and indifferent encoding with animate nouns. In this study, it is included among languages with a clear distinction since there is undoubtedly a distinct encoding of source and goal in this language, however, solely with inanimate nouns. This is somewhat in accordance with Wälchli and Zúñiga (2006) since inanimate is regarded as a subcategory of consistent, i.e. distinct encoding, when discussing areal tendencies. Lastly, another language which is considered to feature distinct encoding of source and goal, Cuiba (Guahiboan), could possibly also be classified as having an intermediate system. In Cuiba, source and goal are encoded by three ‘simple verbs’ (‘verbos sencillos’) each. Thus, Cuiba may at first appear to distinguish source and goal lexically, yet there seems to bemor- phological encoding present as well. The simple verbs can for example be suffixed onto nom- inals (Kerr 1995, p. 125) − suggesting an ongoing grammaticalization process. The unglossed and unsegmented examples below show the simple verbs iya ‘go to/towards the north or south (or outside)’ and dena ‘come from the north or south (or outside)’ mark direction on nouns as suffixes. (54) Cuiba (Guahiboan) (Kerr 1995, p. 128) a. Bɑpon meñɑ (mene iyɑ) ponɑ. ‘He went to/for the river.’ (‘El fue por el río.’) b. Pɑpɑ́tɑme mendenɑ (mene denɑ). ‘They came from the river (north or south).’ (‘(Uds.) llegaron del río (norte osur).’) Furthermore, direction can be marked by complex verbs in Cuiba. These are formed by adding one of the proclitics be=, ‘to, towards’, or we=, ‘from’, to the simple verbs. Since there are six simple verbs twelve complex verbs can be formed, resulting in a total of 18 markers of source and goal in Cuiba. Interestingly, only the complex verbs appear in the Bible verses − no instances of nominal suffixes of simple verbs are found. Since the marking ofsource and goal seems to always feature the proclitics in use, Cuiba is considered to have distinct morphological encoding of source and goal.

49 5.1.2.4 Associated motion As discussed in section 2.4.2, associated motion connects motion to verb events, encoding e.g. an action being performed while moving (Guillaume 2016). There are two sample languages which feature associated motion, yet, curiously, they behave rather differently from one an- other. One of the languages, Yagua (Peba-Yagua), has a clear distinction between source and goal. In Yagua, there are suffixes for ‘bounded movement’ which describe actions done upon arrival at a reference point, upon arrival away from a reference point, upon departure, or en route. Interestingly, the source marker, siy, is the same as the marker for ‘action done upon departure’, -siy/-chiy. However, no such connection is seen between the goal marker, -jù, and the different markers for ‘action done upon arrival’, -nuvɨɨ̜ ̜ and -nuvee (Payne 1985, p. 256). The source and goal markers appear in the Bible verses rather straightforwardly, andthe markers for ‘action done upon arrival’ do not appear at all in the verses. Mosetén-Chimané (Isolate), the second sample language with associated motion, in con- trast shows no indication of a source-goal distinction. There are six markers of associated motion in Mosetén-Chimané: -ki- ‘go away to do an action’, -ti-/-sh- ‘come to do an action’, -kho- ‘to perform on the way there’, -chhi- ‘to perform on the way here’, -ji- ‘perform an ac- tion in various places’, and -min- ‘interruption of the movement to do an action’ (Sakel 2004, p. 272). Both -ki- and -kho- mark movement away from the deictic center (i.e. the speaker, alternatively a central place or person in the story or discourse), while -ti-/-sh- and -chhi- both mark movement to the deictic center. However, it cannot be determined that this corresponds to a source-goal distinction. In the Bible verses for Mosetén-Chimané, however, the associated motion suffixes do not appear to a large extent. Instead, the toponyms are nearly always marked with the inessive22 marker -khan, both in the verses expressing source and goal. In Mosetén-Chimané, toponyms appear with a fixed marker for local relation, no matter which local relation is being expressed (Sakel 2004, p. 77). Hence, it is not surprising to find only one marker in all of the verses.

5.1.2.5 Unresolved cases The encoding of source and goal is left completely unresolved in one language, namely Lengua (Lengua-Mascoy). Four grammars (Lowes 1954; Sušnik 1958, 1977; Unruh and Kalisch 1999) have been used in attempts to determine the encoding of source and goal in Lengua, yet without success.Two of the grammars are old and difficult to navigate, while the other two sources are a dictionary and some kind of teaching material, i.e. not ideal sources for a study of grammatical encodings. Moreover, the Bible verses for Lengua show no clear tendencies. In Wälchli and Zúñiga’s study (2006), Lengua was classified as mixed, i.e. as having a weakly grammaticalized marker or construction for source, due to the verb kyïñeêkteyï ‘come from’. However, their study uses a different Bible translation from this study.

5.1.3 Areal distribution The languages with a clear distinct encoding of source and goal are mainly locatedinthe northwest of South America. At the same time they are relatively mixed with the languages not clearly distinguishing source and goal. The geographical distribution of the languages with

22It should be noted that the markers in Mosetén-Chimané called inessive, addessive, or other names generally given to case markers do not necessarily function like case markers − they do not only appear with nominals but also with e.g. verbs (Sakel 2004).

50 a clear distinct encoding (represented by red dots) and no clear distinct encoding (represented by black dots) is presented in Figure 223.

Figure 2: The geographical distribution of languages with a clear distinct encoding (red dots) and no clear distinct encoding (black dots) of source and goal

5.2 The encoding of path A grammatical encoding of path is specified in the grammar for four of the sample languages. Palikúr (Arawakan) has a separate case for encoding path − the perlative -w, with the variants -iw, -riw, and -ew (Launey 2003, p. 139). Two out of the three Bible verses expected to express path include the allomorph -iw. Akawaio (Cariban) has the same encoding for source and path, the postposition bai (Caesar- Fox 2003, p. 115), however, this marker is not displayed in any of the three Bible verses ex- pected to express path. Instead boro (or poro), ‘via’, appears in all three of the verses. The grammar for Murui Huitoto (Huitotoan) also describes an encoding of path whichdoes not appear in any of the Bible verses − the instrumental case marker -do which can express ‘by, through, on’ (Wojtylak 2017, pp. 306-307), exemplified below.

23The same map, with the possibility to tell the languages apart, is accessible at: https://arcg.is/1ObXOf

51 (55) Murui Huitoto (Huitotoan) (Wojtylak 2017, p. 307) a. beno-mo enɨ-e-do bi-t-e here.clf:sp.place-loc land-clf:g-ins come-lk-3 ‘He came here by land.’ b. bai-e-do bi-tɨ-kaɨ that-fsh-clf:g-ins come-lk-1pl ‘We came through that (e.g. road, path).’

However, the instrumental marker does not appear in any of the verses expected to express path. Instead, the goal marker -mo appears in all of them. In Huallaga Huánuco Quechua (Quechuan), path is encoded by the genitive suffix -pa, ‘through, by way of’ (Weber 1989, p. 210), as illustrated below. This marker appears in one of the three Bible verses.

(56) Huallaga Huánuco Quechua (Quechuan) (Weber 1989, p. 210) a. Noqa heqa-mu-shaq huk punta-pa… 1sg come:over-afar-1.fut a ridge-gen ‘I will come over a ridge (afar off) …’ b. Niykur Llaksa-pa paasa-nchi. then Llaksa-gen pass.through-1pl.incl ‘Then we go through Llaksa (on our way to Wahaq).’

Regarding the remaining 33 sample languages, not much can be said about their encoding of path. Three languages with indifferent encoding feature their source/goal marker inone, two, or all three of the verses expressing path. Eight languages with distinct encoding use their goal marker in two of the three verses. The source marker appears in two of the three verses for two languages with distinct encoding, and in one of the three verses for another two languages (without having the other two verses include a goal marker). 15 languages showed no source, goal, or source/goal marking − featuring e.g. a locative marker or a verb expressing ‘pass through’ instead, alternatively showing no clear tendencies. The remaining three languages show tendencies too weak or unclear to fit in any of these groupings.

5.3 Methodological results As discussed above (see section 5.1.1.3), the analysis of the distinctly encoding languages is roughly the same from the Bible verses and the reference grammars. The grammars present more allomorphs than the Bible translations, although, since only selected verses are examined in this study no conclusion can be drawn regarding how many of a given set of allomorphs the Bible translations generally present. Due to the highly selective Bible passages studied, the context is often similar in many verses. For example, the verses frequently express the same verbs (e.g. ‘come’ and ‘go’) and have a third person singular subject. While the subject marker would be the same between the source and goal verses the verbs would not, resulting in a high risk of selecting the wrong marker. The similarity between the verses also contributes to the lack of allomorphs presented for some languages, such as with Chácobo where only the intransitive and not the transitive source markers could be identified form the Bible verses. In the following, a few general tendencies displayed by the data sources are mentioned. These are further reviewed in the discussion (see section 6.3).

52 The more markers of spatial relation a language has, the more difficult it istodetermine how source and goal are encoded. This applies to both distinctly and indifferently encoding languages. Languages which are easily analyzed with one source of data are easy to analyze also with the other source of data − that is, markers which are clear in the Bible verses, i.e. in use, are also clear in the language descriptions. Several grammars are extremely difficult to work with, using terminology in an unexpec- ted way (as with e.g. Nadëb), or presenting data haphazardly so that one can only hope to accidentally stumble upon what one is looking for (as with e.g. Chamacoco). With several languages, the marker mentioned in the grammar does not appear at all, or at least to a very small extent, in the Bible translations and sometimes another marker appears frequently instead.

53 6 Discussion

In this chapter, firstly the results regarding the encoding of source and goal are discussed − the languages with clear distinct encoding, the languages with no clear distinction, and the areal tendencies in the encoding of source and goal. Secondly, I review the results concerning the encoding of path and, thirdly, the methodological results are analyzed. Finally, some additional remarks and suggestions for future research are given.

6.1 The encoding of goal and source 6.1.1 Distinct encoding Almost two thirds of the sample languages have a clear distinction between source and goal, which partially differs from Wälchli and Zúñiga’s findings (2006). In their study, distinct en- coding is also the most frequent type among South American languages (as mentioned in sec- tion 2.4.2), but the distribution between distinctly and indifferently encoding sample languages differs from this study. Furthermore, it is an unexpected finding that the majority ofthemark- ers identified in the present study are adnominal, i.e. adpositions or case markers, sinceSouth American languages are predominantly head marking and therefore expected to favour verbal marking24. However, Wälchli and Zúñiga (2006) find a strong correlation between distinctive encoding and adnominal markers, which perhaps could explain these somewhat surprising results. Nonetheless, regarding the distribution between distinctly and indifferently encoding languages as well as the distribution between the loci of markers the results of this study differ from previous findings. Regarding complex markers and relation in length between markers the distinctly encod- ing languages are in concordance with the general tendencies. All of the complex markers displayed in the sample follow the model formulated by Pantcheva (2010) which states that source markers are inclined to morphologically contain goal markers which, in turn, are in- clined to contain place markers. However, the findings of this study are not particularly in concordance with Pantcheva’s claim that most languages have equally complex source and goal markers since eleven (or twelve, if Ticuna is included) of 23 distinctly encoding languages have a complex marker for either source or goal, most often source, while the other has not been established as complex, at least in this study. The findings regarding relation in length between markers of local roles primarily agree with previous research. The majority of the languages with a difference in length between the markers for local roles conform to the scale presented by Stolz et al. (2014), which claims that source tends to be more marked than place which, in turn, tends to be more marked than goal (see (10) or (33)). Nonetheless, there are a few languages which display opposite patterns. For example, Palikúr and Akawaio have longer goal than place markers, and Bora features a place marker which is longer than the source marker. Furthermore, in Central Tunebo the goal marker acam is longer than the source marker bin. However, the other languages with a longer goal than source marker are more complicated. In Tenetehara, the goal marker -kutir is clearly longer than the source marker -wi, yet in the Bible verses goal is more frequently encoded by the locative markers pe or -ehe − thus suggesting that goal in Tenetehara may not solely be expressed by -kutir. In Aguarana the goal marker is only longer than the source marker when the latter is not complex − that is, when the source marker occurs with a locational noun.

24I have not investigated precisely what the distribution is between head- and dependent-marking languages in this sample since the loci of markers are not a focus in this study.

54 Arabela has three goal markers and while all three of them are longer than one of the source markers, only the two of them are longer than both of the source markers. Finally, in Nadëb it is undetermined whether the distinction between contrastive and non-contrastive locative corresponds to a source-goal distinction. Nevertheless, if it does, the goal marker in Nadëb would be longer than the source marker. In summary, most languages with a length difference between the markers are in concordance with the markedness scale, and the marking in the languages which may have longer goal than source markers is much more problematic to quantify in this way, with the exception of Central Tunebo. It is important to note, when discussing the encoding of goal, that both markers denoting ‘to’ and ‘towards’ are included in this study. Solely from the English prepositions it is clear that while to is shorter than from, towards is not. It is possible that the languages with a longer goal than source marker have goal markers which denote ‘towards’ rather than ‘to’, as is the case with -kutir in Tenetehara. The length relation would therefore not necessarily bea noteworthy difference. However, since markers denoting ‘to’, ‘towards’, or both are grouped together in this study, I leave it to future research to determine whether the relations described by the markedness scale are affected by the more specific meaning of goal markers. Conceivably, the tendency of source to be both more complex and more marked than goal is no coincidence. It seems highly likely that source markers are longer than goal markers, at least in part, because they are often more complex. This is in line with Pantcheva’s claim (2010) that source and goal are structurally different.

6.1.2 No clear distinct encoding The languages with no clear distinction between source and goal behave very differently from one another − they range between clear indifferent encoding, suspected distinct encoding, intermediate system, and no clear pattern at all. The languages with indifferent encoding seem to both feature and lack grammatical mark- ing of direction. Of course, it is not determined that Cha’palaa and Southern Nambikuára lack grammatical marking, nor that the locative markers in Tatuyo indeed encode direction as well as location, however, both sources of data indicate these analyses. Regarding Maxakalí, it seems fairly evident that the encoding of source and goal is indifferent, yet establishing how the markers for spatial relation function still remains. Another undetermined matter is whether other sample languages than Mapundungun have ‘typical orientation verbs’, or ‘spe- cialized verbs’ as they are called in Wan, however, this is left for future research to determine. Regarding the languages with a suspected or determined goal marker and unknown (if any) source markers, the Bible verses frequently feature the goal markers while not indicating a marking of source, with the exception of Chamacoco. Possibly these languages could affect the length relation between markers described by the markedness scale (see (10) or (33)) − if they indeed feature grammatical marking of goal but not source, the source marker would be shorter than the goal marker and thus not conform to the relation described by the markedness scale. Whether or not there is a source marker in these languages, and whether or not that marker is otãra in Chamacoco, will have to be determined by future research. Furthermore, how these languages could potentially affect the length relations in the markedness scale will also be left for future investigations to determine. In Nadëb it is unclear to me whether the contrastive and non-contrastive locative markers constitute a source-goal distinction − and if not, it is still unresolved how the markers function and more precisely what they denote. Since the two sample languages with associated motion differ regarding source-goal dis- tinctions, and only one of them has a clear connection between one of the markers for local

55 roles and one of the markers for associated motion, it can be concluded that associated motion and the notion of source and goal can relate to each other in very different ways. The issue of determining the possible relations more precisely, both in Mosetén-Chimané and in other languages with associated motion, will be left for future research. Lastly, how source and goal are encoded in Lengua will entirely be left for future studies to determine. Conclusively, languages with no clear distinction between source and goal leave much more to be determined by future research than the languages with a clear distinction. Un- deniably, the methodology in this study was more prepared to cope with distinctly encoding languages since many challenges presented by the languages with no clear distinction are not possible to analyze sufficiently within the limits of this study. It is evident that languages without a clear grammatical source-goal distinction behave astonishingly different from one another. While some languages encode source and goallex- ically (i.e. in the meaning of the verb), others encode them grammatically, e.g. through an oblique marker or a locative marker. Yet, some other languages have much more specific sys- tems, e.g. the contrastive-non-contrastive locative opposition in Nadëb, the encoding of the entire route in Paumari, or the markers in Maxakalí whose function could not be established given the data. Although the majority of the distinctly encoding languages behave similarly, there are some that display more unusual or specific patterns, such as Chácobo with its differ- ent source markers dependent on the valency of the verb, Cuiba which marks source and goal through an extensive set of simple (sometimes suffixed) and complex verbs, and Central Tun- ebo with its distinct encoding with inanimate grounds and indifferent encoding with animate nouns. Furthermore, the two languages with associated motion also display very different systems − while Yagua has a source-goal distinction and the encoding of source is the same as the marker for ‘action done upon departure’, Mosetén-Chimané has no clear source-goal distinction and no clear connection between markers of spatial relations and markers for as- sociated motion. Moreover, the source and goal encoding is even entirely unresolved in one language, Lengua, possibly due to a unique system of encoding. Even though several languages distinguish source from goal very clearly and in similar ways (that is, the distinctly encoding languages), the sample as a whole still displays the wide variety of typological behaviour one would expect from South American languages. Regarding syncretism between the local roles, the sample, again as a whole, display the three most common patterns, presented below (repeated from (20a-c)):

(57) a. place = goal = source : one single marker b. place = goal ≠ source : one ambiguous marker for place and goal, a separate marker for source c. place ≠ goal ≠ source : a separate marker for each of the three

There is one language which partially displays a syncretism between source andplace(as presented in (20d)), namely Amarakaeri (Harakmbut). Amarakaeri appears to have three place markers − one which only encodes place, one which can encode both place and goal, and one which can encode both place and source (Tripp 1995, p. 196). All syncretism patterns found in the sample languages are presented in Table 16 and 17 in Appendix F.

6.1.3 Areal tendencies The areal tendencies in the findings of this study, represented in the map in Figure 2,suggest that distinct encoding is favoured along the west coast of South America and the Andes. This

56 would be in support of previous findings that there is an east-west division in South America, and that languages in the Andes are rather uniform. However, several languages with distinct encoding are also found far to the east and along the northern coast of the continent and languages with no clear distinct encoding also appear along the Andes and in the western part of the continent. I argue that this could be expected given the previous work on possible areality in South America (see section 4.2.2.2 and Appendix C). Earlier literature aimed at distinguishing linguistic areas or areal traits on this continent (see e.g. Campbell 2012b and the many works referenced there) exhibit these issues: almost any given trait is distributed over the entire continent, even if it seems to form a regional pattern; if a trait forms a regional pattern, several, or at least a few, languages of that region will not have this trait; and possible linguistic areas are not well-defined, especially regarding boarders between adjacent areas − they most frequently overlap. One could possibly argue that languages from the southern most part of the continent seem to not clearly distinguish source from goal. Of course, this is an area where rather few of the sample languages are located (due to a lack of languages with a Bible translation), yet, all five sample languages south of do not make this distinction grammatically. However, this suggestion is tentative, considering the amount of languages in this area not considered in this study, and further investigation is needed to determine whether this is indeed an areal tendency.

6.2 The encoding of path The findings, or lack thereof, concerning the encoding of path are, as much as can beregarding a lack of results, in concordance with previous claims. As Pantcheva (2011) concluded, not many grammars mention the encoding of path − only the grammars of four of the 37 sample languages discuss path markers (although, far more than 37 grammars have been used in this study). One of these languages, Palikúr, has a separate case for path, however, the remaining three express path through syncretism of some kind. In Akawaio the encoding of source and path are the same, in Murui Huitoto path is marked by the instrumental case, and in Huallaga Huánuco Quechua it is marked by the . All three of these syncretism patternare in line with Creissels’s claims (2009) that instrumental and ablative case often mark path, and that there often is syncretism between ablative and genitive case (by extension supporting the syncretism between path and genitive shown in Huallaga Huánuco Quechua). The remaining sample languages predominantly display unclear encoding of path. How- ever, if a language does feature one of its markers for local roles in the Bible verses expressing path this is more often the goal marker than the source marker, in contrast to Creissels’s claim (2009) that source and path often are encoded identically.

6.3 Methodological results As is discussed in section 4.1, the reference grammars and the parallel texts function in very different ways, however, this allows them to complement each other to a large extent. While the distinctly encoding languages were mainly unproblematic to analyze with either of the sources, the languages with no clear distinction, in contrast, were much more dependent on both sources of data being used and often nothing could still be determined with certainty. With grammars that were imprecise, the parallel texts tended to offer much needed insight on the usage of different markers, e.g in Nadëb where the Bible verses clearly suggest hẽnh is fea- tured in goal expressions while the grammar described it with the undefined term ‘contrastive

57 locative’, or in Maxakalí where one grammar, in a very short and imprecise discussion on spa- tial expressions, does not shed much light upon how these expression relate to a source-goal distinction, while the Bible data clearly suggest that source and goal are marked indifferently. While the selective set of Bible verses used in this study fail to include many allomorphs of different markers, they still offer an understanding of the markers in use. For example, the marker identified by the grammar may be sparsely, or not at all, used in the Bibleverses, while some other marker appears frequently. This could of course be an error, however, it seems more likely that the Bible translations rather better reflect the actual language use − perhaps some other construction or marker can also encode the function and does so more frequently. Thus, parallel texts present features in use which can reflect language internal variation and favoured or disfavoured markings. However, the presence of different features in parallel texts is decided by the context, while grammars can in- or exclude any given feature. Due to the small set of Bible verses studied, there are often several similarities between them, i.e. not only the markers for local roles are the same between verses but also e.g. verbs and subject markers. This increases the likeliness of selecting an incorrect marker andifone was to study these verses automatically, it seems rather likely that other meanings, such as ‘come’, ‘go’, or ‘he’, would skew the results. However, an automatic or more time-consuming manual study could include more data which could possibly outweigh the potential skew. As mentioned above, languages with no clear source-goal distinction are more difficult to analyze with these data than the languages with a clear distinction and leave more unanswered questions for future studies. This could be expected since a given construction or feature cannot be proven non-existent in a certain language on an empirical basis with any source of data − one always can have missed a rare instance. It is thus impossible to determine the lack of a source-goal distinction in a language, and while the presence of an indifferent encoding can be proven this is often not specified in grammars and very difficult to decidedly determine from parallel texts. This leaves mere indications as the only feasible results to expect fromthis group of languages, which explains the many issues left undetermined. This study features a sample of languages from a rather large number of language famil- ies, yet they have all been analyzed (with the exception of Lengua) within the time limit of the study. This is partly due to the effort invested, but also partly due to suitable data − the encodings of local roles in all sample languages are, in more or less detail, possible to describe which indicates that these are useful sources of data for studying the encoding of source and goal. However, this does not apply to all parts of locational semantics. One has to look no fur- ther than the results regarding the encoding of path in this study to see that these data cannot sufficiently describe all spatial expressions. Arguably, this indicates that a similar procedure, studying motion in a given language, should initially examine source and goal, rather than analyzing e.g. path or associated motion first.

6.4 Additional remarks Using reference grammars for research always leads to a high risk of getting uneven amounts of data, with uneven specificity. Even though, as discussed above, the Bible translations have functioned as an indispensable source of data alongside the grammars in this study, perhaps questionnaires could have offered a more in depth understanding of the languages withno clear source-goal distinction. However within the scope of this study, especially due to the strong emphasis on having a representative sample and the time limit, questionnaires would not have been a sufficient alternative. Perhaps future research could benefit from using ques- tionnaires in determining the encoding of source and goal, as well as the function of the mark-

58 ers, in languages with no clear source-goal distinction. The sample used in this study was rather skewed towards northwestern South America. If there indeed is an east-west typological division among South American languages, this could explain the high number of uniformly behaving languages found on this astonishingly typolo- gically and genealogically diverse continent. Although, the over-representation of languages in western South America is difficult to avoid since there are more individual languages fam- ilies, and in particular more isolates, in this region than others. Perhaps a study with a stricter areal stratification could clarify whether the findings of this study are representative ofSouth American languages or distorted by the geographical distribution of the sample. Furthermore, future studies, with more resources, could also include several languages from each language family to examine possible variation within families. Such family-internal variation could potentially indicate areal tendencies more specifically as well.

6.5 Future research The findings of this study, in particular regarding the languages with no clear distinct encoding, leave many unanswered questions for future research. Although several of these have been mentioned above, a brief summary of them, with a few additions, is given below. Most of the unresolved issues pertain to the functions of different markers and whether certain systems are indeed indifferently or distinctly encoding. Moreover, future studies could examine whether the indifferently encoding languages have ‘specialized’ or ‘typical orienta- tion’ verbs, what the relation is between number of local roles (and thus grounds) allowed per motion verb and style of encoding (as discussed by Wälchli and Zúñiga (2006)), how the notion of source and goal relates to associated motion, and how the markedness scale (as presented in (10) and (33)) is potentially affected by different meanings of goal markers as well asthe potential lack of source markers in a few languages with goal markers. As mentioned above, future research may benefit from a different method in examining specific languages, namely the ones without a clear source-goal distinction. This mayalso apply to future studies of path expressions − since the two data sources used in this study prove rather inefficient in describing or illustrating encodings of path, one might consider using questionnaires for this purpose in the future. Lastly, as discussed in section 6.4, a study with stricter areal stratification could possibly give a better understanding of source and goal encodings in all parts of South America − and by extension potentially present more representative results − and a study with several languages per language family could reveal possible family-internal variation.

59 7 Conclusions

The aim of this study has been to determine whether source and goal are encoded distinctly or indifferently in South American languages, due to the previous lack of proper representation in this area. Furthermore, the study has also aimed to investigate the encoding of path in South American languages, and the differences between the two sources of data used for the investigation, namely parallel texts consisting of Bible translations and reference grammars. In this section, a brief answer is given to each research question, alongside some additional conclusions.

1. Are goal and source expressed distinctly or indifferently in South American languages? While the encoding of source and goal varies, the majority of the sample languages have distinct, adnominal encoding.

2. If they are expressed distinctly, how do they differ from one another? They differ in terms of complexity and length, primarily conforming to previously con- firmed general tendencies in the world’s languages with source being the more complex and the more marked of them.

3. Are there any areal tendencies to the encoding of goal and source? There is a tendency towards distinct encoding along the west coast of South Americaand the Andes. However, as in previous attempts at defining areal traits in South America, many languages with this trait are found outside of the area, and many languages without it are found within the area − further reflecting the difficulty of determining linguistic areasor areal traits in South America.

4. How is path expressed, alternatively what can be deduced about its encoding, in South American languages? The encoding of path is not mentioned in the grammars for the vast majority ofthesample languages. The four languages with a confirmed encoding of path use a case marker,most often through syncretism with another case rather than a case of its own. The languages with distinct encoding without a mentioned path marker use the goal marker more often than the source marker to encode path.

5. How do the two sources of data, parallel texts (Bible translations) and reference grammars, differ in terms of efficiency and reliability? In most cases, the two sources of data complement each other. Regarding most distinctly encoding languages, either the grammar or the Bible translation would suffice in determ- ining the source and goal markers. In languages with no clear distinction both sources of data are most commonly required to establish the marking of local roles, and in several instances more data than there are available is needed to determine markers and/or their functions with certainty. While grammars are, unsurprisingly, more descriptive (often dis- playing more allomorphs or more information about the functions of the markers), the Bible translations give more context, better describing the usage of the markers.

Conclusively, source-goal distinctions seem to be highly frequent in South America, and when they do occur, they are highly regular and uniform, suggesting that the difference

60 between source and goal is a salient one. The languages with no clear distinction present widely different systems, reflecting the diversity of South American languages. In comparison, path is seemingly not as important to distinguish, and thus not as salient. Perhaps it is less important to express, due to the fact that is does not display as clear of an opposition as the one between source and goal, or it is more dependent on specific verbs, unlike source and goal which commonly occur with light verbs. In short, the circumstances for developing a marking of path are not as good as for source or goal.

61 References

Abrahamson, Arne (1962). ‘Cayapa: Grammatical notes and texts’. In: Studies in Ecuadorian Indian languages 1. Ed. by Benjamin F. Elson. Vol. 7. Linguistic Series. Norman: Summer Institute of Linguistics of the University of Oklahoma, pp. 217–247. Adelaar, Willem F.H. (2012a). ‘Historical overview: Descriptive and comparative research on South American Indian languages’. In: The indigenous languages of South America: Acom- prehensive guide. Ed. by Lyle Campbell and Verónica Grondona. Berlin: De Gruyter Mouton, pp. 1–58. — (2012b). ‘Languages of the Middle Andes in areal-typological perspective: Emphasis on Quechuan and Aymaran’. In: The indigenous languages of South America: A comprehensive guide. Ed. by Lyle Campbell and Verónica Grondona. Berlin: De Gruyter Mouton, pp. 575– 624. Adelaar, Willem F.H. and Pieter C. Muysken (2004). The Languages of the Andes. Cambridge: Cambridge University Press. Anderson, Doris (1989). Conversational Ticuna. Yarinacocha, : Instituto Lingüístico de Ver- ano. Bolaños, Katherine (2016). ‘A Grammar of Kakua’. PhD thesis. Universiteit van Amsterdam. Bonfim Duarte, Fabio (2007). Estudos de Morfossintaxe Tenetehára. Belo Horizonte: Faculdade de Letras de Universidade Federal de Minas Gerais. Borman, M. B. (1976). Vocabulario cofán: Cofán-castellano, castellano-cofán. Vol. 19. Serie de vocabularios indígenas ‘Mariano Silva y Aceves’. Quito: Instituto Lingüístico de Verano. Caesar-Fox, Desrey Clementine (2003). ‘Zauro’nödok Agawayo Yau: variants of Akawaio spoken at Waramadong’. PhD thesis. Houston: Houston: Rice University. Campbell, Lyle (2012a). ‘Classification of the indigenous languages of South America’. In: The indigenous languages of South America: A comprehensive guide. Ed. by Lyle Campbell and Verónica Grondona. Berlin: De Gruyter Mouton, pp. 59–166. — (2013). Historical Linguistics: An Introduction. Edinburgh & Cambridge, MA: Edinburgh University Press & MIT Press. — (2012b). ‘Typological characteristics of South American indigenous languages’. In: The in- digenous languages of South America: A comprehensive guide. Ed. by Lyle Campbell and Verónica Grondona. Berlin: De Gruyter Mouton, pp. 259–330. Cerrón-Palomino, Rodolfo (2006). El chipaya o la lengua de los hombres del agua. Lima, Peru: Fondo Editorial de la Pontificia Universidad Católica del Perú. — (2000). Lingüística aimara. Vol. 21. Biblioteca de la Tradición Oral Andina. Lima: CBC: Cusco: Centro de Estudios Regionales Andinos. Chapman, Shirley and Desmond C. Derbyshire (1991). ‘Paumarí’. In: Handbook of Amazonian Languages. Ed. by Desmond C. Derbyshire and Geoffrey K. Pullum. Vol. III. Berlín 1986, 1990, 1991, 1998: Berlin: Mouton de Gruyter, pp. 161–354. Comrie, Bernard (2013). ‘Alignment of Case Marking of Full Noun Phrases’. In: The World Atlas of Language Structures Online. Ed. by Matthew S. Dryer and Martin Haspelmath. Leipzig: Max Planck Institute for Evolutionary Anthropology. Available online at http: //wals.info/chapter/98, Accessed on 2020-03-13. Creissels, Denis (2009). ‘Spatial cases’. In: The Oxford handbook of case. Ed. by Andrej Mal- chukov and Andrew Spencer. Oxford: Oxford University Press, pp. 609–625. Curieux, Rojas and Tulio Enrique (1998). La lengua Paez. : Ministerio de Cultura.

62 de Vries, Lourens (2007). ‘Some remarks on the use of Bible translations as parallel texts in linguistic research’. In: STUF - Language Typology and Universals 60.2, pp. 148–157. doi: https://doi.org/10.1524/stuf.2007.60.2.148. Dixon, Robert M W (2010). Basic linguistic theory, Vol. 2, Grammatical topics. Oxford: Oxford University Press. Dryer, Matthew S. (2013). ‘Order of Adjective and Noun’. In: The World Atlas of Language Struc- tures Online. Ed. by Matthew S. Dryer and Martin Haspelmath. Leipzig: Max Planck Insti- tute for Evolutionary Anthropology. Available online at https://wals.info/chapter/ 87, Accessed on 2020-03-22. Dryer, Matthew S. and Martin Haspelmath, eds. (2013). WALS Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. url: https://wals.info/. Eberhard, David M., Gary F. Simons and Charles D. Fennig, eds. (2020). Ethnologue: Languages of the World. 23rd ed. Dallas, Texas: SIL International. Online version: http : / / www . ethnologue.com. Fillmore, Charles (1971/75). Santa Cruz lectures on Deixis. Reproduced by the Indiana Univer- sity Linguistics Club 1975. Bloomington, Indiana. Finnveden, Gustav (2019). ‘Finding case through personal names in parallel texts’. B.S. Thesis. Stockholm University. Ganenkov, Dmitrii S. (2002). ‘Tipologija padežnix značenij: semantičeskaja zona prolativa’. In: Grammatikalizacija prostranstvennyx značenij. Issledovanija po teorii grammatiki 2. Ed. by Vladimir A. Plungian. Moscow: Russkie slovari, pp. 35–55. Gomez-Imbert, Elsa (1982). ‘De la forme et du sens dans la classification nominale en Tatuyo (Langue Tukano Orientale d’Amazonie Colombienne)’. PhD thesis. Université de la Sor- bonne (Paris IV). Grondona, Verónica M. (1998). ‘A Grammar of Mocovi’. PhD thesis. Ann Arbor: University of Pittsburgh. Guillaume, Antoine (2016). ‘Associated motion in South America: Typological and areal per- spectives’. In: Linguistic Typology 20.1, pp. 81–177. doi: https://doi.org/10.1515/ lingty-2016-0003. Hammarström, Harald, Robert Forkel and Martin Haspelmath (2019). Glottolog 4.1. Jena: Max Planck Institute for the Science of Human History. url: https://glottolog.org/. Hardman, Martha J. (2001). Aymara. Vol. 35. LINCOM Studies in Native American Linguistics. München: Lincom. Hart, Helen (1988). Diccionario chayahuita-castellano. Serie lingüística peruana, 29. Yariha- cocha and Pucallpa: SIL. Haspelmath, Martin (2019). ‘Differential place marking and differential object marking’. In: STUF – Language Typology and Universals 72.3, pp. 313–334. doi: https://doi.org/10. 1515/stuf-2019-0013. Headland, Edna R. (1994). ‘Diccionario bilingüe tunebo-español con una breve gramática’. MA thesis. Arlington: University of Texas. Kerr, Isabel J. (1995). Gramática pedagógica del cuiba-wámonae. Santafé de Bogot: Asociación Instituto Lingüístico de Verano. Koch, Harold (1984). ‘The category of “associated motion” in Kaytej’. In: Language in Central Australia 1, pp. 23–34. Krüsi, Dorothee and Martin Krüsi (1975). ‘Phonology of Chiquitano’. In: Work Papers of the Summer Institute of Linguistics Riberalta, Beni, Bolivia, 1972-1976. Ed. by Ursula Wiesemann. Riberalta: Summer Institute of Linguistics, pp. 53–93.

63 Kroeker, Barbara (2003). Aspectos da Língua Nambikuara. 2nd ed. Cuiabá: Sociedade Inter- nacional de Lingüística. — (1982). Aspectos da língua nambikuara. Cuiabá: Sociedade Internacional de Lingüística-SIL. Kroeker, Menno (2001). Gramática Descritiva da Língua Nambikuara. Cuiabá: Sociedade Inter- nacional de Lingüística. — (1996). Txa²wã¹wãn³txa² Kwa³jan³txa² Wan³txa² Hau³hau³kon³nha²jau³su²: Nambikuara-português ; português-nambikuara. Porto Velho, RO: SIL Brasil. Kutscher, Silvia (2010). ‘When ‘towards’ means ‘away from’: the case of directional-ablative syncretism in the Ardeşen variety of Laz (South-Caucasian)’. In: STUF - Language Typology and Universals 63.3, pp. 252–271. doi: https://doi.org/10.1524/stuf.2010.0021. Launey, Michel (2003). Awna Parikwaki: Introduction à la langue Palikur de Guyane et de l’Amapá. Didactiques. Paris: IRD. Lowe, Ivan (1999). ‘Nambiquara’. In: The Amazonian Languages. Ed. by R. M. W. Dixon and . Cambridge Language Surveys. Cambridge: Cambridge University Press, pp. 269–292. Lowes, R. H. G. (1954). ‘Alphabetical list of Lengua Indian words with English equivalents’. In: Journal de la Société des Américanistes de Paris 62, pp. 85–109. Maddieson, Ian, ed. (1984). Patterns of sounds. Cambridge: Cambridge University Press. Marciniak, Tomasz and Michael Strube (2005). ‘Modeling and annotating the semantics of route directions’. In: Proceedings of the 6th International Workshop on Computational Se- mantics (IWCS-6). Tilburg, The Netherlands, pp. 151–162. Mayer, Thomas and Michael Cysouw (2014). ‘Creating a massively parallel Bible corpus’. In: Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14). Reykjavik, Iceland: European Language Resources Association (ELRA), pp. 3158– 3163. url: http://www.lrec-conf.org/proceedings/lrec2014/pdf/220_Paper. pdf. Miestamo, Matti (2003). ‘Clausal negation: A typological study’. PhD thesis. Helsinki: Univer- sity of Helsinki. — (2005). Standard negation: The negation of declarative verbal main clauses in a typological perspective. Berlin: Mouton de Gruyter. Miestamo, Matti, Dik Bakker and Antti Arppe (2016). ‘Sampling for variety’. In: Linguistic Typology 20.2, pp. 233–296. doi: https://doi.org/10.1515/lingty-2016-0006. Mortensen, Charles A. (1999). A Reference Grammar of Northern Embera Languages. Vol. 7. Studies in the Languages of . Also SILPL 134. Dallas: SIL: SIL International and University of Texas at Arlington. Muysken, Pieter, Harald Hammarström, Joshua Birchall, Swintha Danielsen, Love Eriksen, Ana Vilacy Galucio, Rik van Gijn, Simon van de Kerke, Vishnupraya Kolipakam, Olga Krasnoukhova, Neele Müller and Loretta O’Connor (2014). ‘The languages of South Amer- ica: deep families, areal relationships, and language contact’. In: The Native Languages of South America: Origins, Development, Typology. Ed. by Loretta O’Connor and Pieter Muysken. New York, New York: Cambridge University Press, pp. 299–322. Muysken, Pieter and Loretta O’Connor (2014). ‘Introduction: South American indigenous lan- guages; genealogy, typology, contacts’. In: The Native Languages of South America: Origins, Development, Typology. Ed. by Loretta O’Connor and Pieter Muysken. New York, New York: Cambridge University Press, pp. 1–26. Nichols, Johanna (1986). ‘Head-Marking and Dependent-Marking Grammar’. In: Language 66, pp. 56–119.

64 Nichols, Johanna (1992). Linguistic diversity in space and time. Chicago & London: The Univer- sity of Chicago Press. Nikitina, Tatiana (2009). ‘Subcategorization pattern and lexical meaning of motion verbs: a study of the source/goal ambiguity’. In: Linguistics 47.5, pp. 1113–1141. doi: https:// doi.org/10.1515/LING.2009.039. O’Brien, Colleen (2018). ‘A grammatical description of Kamsá: A of Colombia’. PhD thesis. University of Hawai’i at Mānoa. Overall, Simon (2007). ‘A Grammar of Aguaruna’. PhD thesis. Bundoora, Victoria: LaTrobe University. Pantcheva, Marina (2011). ‘Decomposing Path: The Nanosyntax of Directional Expressions’. PhD thesis. University of Tromsø. — (2010). ‘The syntactic structure of Locations, Goals, and Sources’. In: Linguistics 48.5, pp. 1043– 1081. doi: https://doi.org/10.1515/ling.2010.034. Payne, Doris (1985). ‘Aspects of the Grammar of Yagua: A Typological Approach’. PhD thesis. Ann Arbor: University of California, Los Angeles. Pereira, Deuscreide Gonçalves (1992). ‘Alguns aspectos gramaticais da língua Maxakali’. MA thesis. Belo Horizonte: Universidade Federal de Minas Gerais. Perri Ferreira, Helder (2011). Esboço gramatical do Ỹaroamë: Língua Yanomami falada do Serra do Pacu/RR. Rio de Janeiro: Museo do Indio. Popovich, Harold A. and Frances B. Popovich (2005). Maxakalí-English Dictionary: English Maxakalí Glossary. Cuiabá: Sociedade Internacional de Lingüística. Prost, Gilbert (1962). ‘Signaling of Transitive and Intransitive in Chacobo (Pano)’. In: Interna- tional Journal of American Linguistics 28.2, pp. 108–118. doi: 10.1086/464678. Rich, Rolland (1999). Diccionario Arabela-Castellano. Vol. 49. Serie Lingüística Peruana 49. Lima: Instituto Lingüístico de Verano. Rijkhoff, Jan and Dik Bakker (1998). ‘Language sampling’. In: Linguistic Typology 2, pp. 263– 314. Rijkhoff, Jan, Dik Bakker, Kees Hengeveld and Peter Kahrel (1993). ‘A method of language sam- pling’. In: Studies in Language 17, pp. 169–203. Sakel, Jeanette (2004). A Grammar of Mosetén. Vol. 33. Mouton Grammar Library. Berlin: Mouton de Gruyter. Schapper, Antoinette (2011). ‘Iconicity of sequence in the coding of source and goal intwo Papuan languages of south-east Indonesia’. In: Linguistics in the Netherlands 28, pp. 99– 111. doi: https://doi.org/10.1075/avt.28.09sch. Stefanowitsch, Anatol and Ada Rohde (2004). ‘The goal bias in the encoding of motion events’. In: Motivation in Grammar. Ed. by Klaus-Uwe Panther and Günter Radde. Berlin & New Yor: Mouton de Gruyte, pp. 249–268. Stolz, Thomas, Sander Lestrade and Christel Stolz (2014). The crosslinguistics of zero-marking of spatial relations. Berlin & Boston: De Gruyter Mouton. Sušnik, Branislava (1970). Chamacocos II: Diccionario Etnográfico. Asunción del : Muséo Etnográfico ‘Andrés Barbero’. — (1958). ‘Eenlet Appaiwa. Lengua-Maskoy: estructura gramatical’. In: Boletín de la Sociedad Científica del Paraguay 2, pp. 1–92. — (1977). Lengua Maskoy: Su hablar- su pensar- su vivencia. Vol. VI. Lenguas Chaqueñas. Asunción del Paraguay: Muséo Etnográfico ‘Andrés Barbero’. Talmy, Leonard (2007). ‘Lexical typologies’. In: Language typology and syntactic description III: Grammatical categories and the lexicon. Ed. by Timothy Shopen. 2nd ed. Cambridge: Cambridge University Press, pp. 66–168.

65 Talmy, Leonard (1972). ‘Semantic structures in English and Atsugewi’. PhD thesis. University of California at Berkeley. — (2000). Toward a cognitive semantics, Vol. II: Typology and Process in Concept Structuring. Cambridge, MA: MIT Press. Terraza, Jimena (2009). ‘Grammaire du Wichí: Phonologie et Morphosyntaxe’. PhD thesis. Uni- versité du Québec à Montréal. Tesnière, Lucien (1959). Éléments de syntaxe structurale. Paris: Klincksieck. Thiesen, Wesley (1975). ‘Un informe breve de la morfología bora’. In: Datos Etno-Lingüísticos 18. Tripp, Robert (1995). Diccionario Amarakaeri-Castellano. 1st ed. Vol. 34. Serie Lingüística Peru- ana. Yarinacocha: Ministerio de Educación and Instituto Lingüístico de Verano. Tuggy, John (1966). Vocabulario candoshi de Loreto. Serie Lingüística Peruana 2. Yarinacocha, Peru: Instituto Lingüístico de Verano. Unruh, Ernesto and Hannes Kalisch (1999). Vana Peema. Nentenyey’a Nelhpaqmeesma Nem- peema. Guía del Maestro Maskoy para el Aprendizaje del Idioma Guaná. Lecciones 1 a 16. Ya’alve-Saanga: Ya’alve-Saanga: Nengvaanemquescama Nempayvaam Enlhet. Vittadello, P. Alberto (1988). Cha’palaachi: El idioma Cayapa. Vol. 102. Miscelánea Antro- pológica Ecuatoriana, Serie Monográfica 5. Guayaquil: Pontificia Universidad Católica del , Sede de Esmeraldas, Museo del Banco Central del Ecuador. Weber, David (1989). A Grammar of Huallaga (Huánuco) Quechua. Berkeley: University of Cali- fornia Press. Weir, E. M. Helen (1984). ‘A Negação e outros Tópicos da Gramática Nadëb’. MA thesis. Uni- versidade Estadual de Campinas. — (1986). ‘Footprints of yesterday’s syntax: diachronic development of certain verb prefixes in an OSV language (Nadëb)’. In: Lingua 68, pp. 291–316. Wälchli, Bernhard (2001). ‘A typology of displacement (with special reference to Latvian)’. In: STUF – Language Typology and Universals 54.3, pp. 298–323. doi: https://doi.org/10. 1524/stuf.2001.54.3.298. — (2007). ‘Advantages and disadvantages of using parallel texts in typological investigations’. In: STUF - Language Typology and Universals 60.2, pp. 118–134. doi: https://doi.org/ 10.1524/stuf.2007.60.2.118. — (2009). Motion events in parallel texts: A study in primary-data typology. Bern: University of Bern Habilitationsschrift. Wälchli, Bernhard and Fernando Zúñiga (2006). ‘Source-Goal (in)difference and the typology of motion events in the clause’. In: STUF – Language Typology and Universals 59.3, pp. 284– 303. doi: https://doi.org/10.1524/stuf.2006.59.3.284. Wojtylak, Katarzyna (2017). ‘A grammar of Murui (Bue): A witotoan language of Northwest Amazonia.’ PhD thesis. Cairns, Australia: James Cook University. Zúñiga, Fernando (2006). Mapudungun. Vol. 376. Languages of the World/Materials. München: Lincom. Zwarts, Joost (2008). ‘Aspects of a Typology of Direction’. In: Theoretical and Crosslinguistic Approaches to the Semantics of Aspect. Ed. by Susan Rothstein. Amsterdam & Philadelphia: John Benjamins Publishing Company, pp. 79–106.

66 Appendix A Bible verses

The Bible verses used for this study are presented here as they are represented in thecorpus of Bible translations used (see Mayer and Cysouw 2014). In each verse number, the first two digits refer to which book the verse is from, the following three digits represent the chapter of said book, and the last three digits refer to the number of the verse itself. The books are numbered with the two digit codes described in Mayer and Cysouw (2014, p. 3162). The codes for the books used in this study are: 40 − Matthew, 41 − Mark, 42 − Luke, 43 − John, 44 − Acts, 45 − Romans, 46 − 1 Corinthians, 47 − 2 Corinthians, 48 − Galatians, 50 − Philippians, 52 − 1 Thessalonians, and 54 − 1 Timothy. Some of the verses are marked with asterisks. They are used to indicate that a verse isdif- ficult to analyze or stands out somehow and are meant to highlight the need for attentiveness when examining these verses (see section 4 for further discussion on this matter).

Verses expressing goal Jerusalem: 40002001, 40020017, 40020018, 41010032, 41010033, 41011001, 42004009, 42018031, 43012012, 44011002, 44019021, 44020022, 44021017, 44025003, 45015025, 41011015, 41011027, 41015041, 42009051, 42009053, 42019028, 43002013, 43005001, 44009002, 44013013, 44021012, 44021015, 46016003, 44013031, 40016021, 40021001, 41011011, 42024033, 44022017, 44025020, 42002041, 44001012, 43011055, 44015002, 44025001, 44022005, 42002022, 42002045, 44024011, 44025009, 48001017, 48001018, **44012025, 42024052, 48002001, 42017011, 44008025, 42013022, 44009026, 44015004 Macedonia: 44016010, *44016009, 44019022, 44020001, 54001003, *47001016, 47002013, 47007005 Verses expressing Jerusalem: 40015001, 42005017, **42024013, 44001004, 44022018, source 42024047, 41003008, 45015019, 41007001, 44028017, 43001019, 44025007, 42010030, 44008026, 44011027, 41003022, 42006017, 40004025, **44012025 Macedonia: 44018005, 47011009, 50004015, *47001016 Verses expressing place Jerusalem: 44001008, 44001019, 44002005, 44002014, 44004005, (and occasionally genit- 44004016, 44008001, 44009013, 44009028, 44010039, 44023011, ive or some other func- 44026004, 44026010, 42002043, 42009031, 43010022, 44016004, tion) 44021004, 45015026, 42013004, 42002025, 44011022, 44013027, 44025024, 44026020, 43002023, 43004045, 44021013, 44008014, 42023007, 43004020, 45015031, 43005002, 44009021, 44020016, 44025015, 44006007, 43004021, 44021011, 40021010, 43011018, 42019011, 42024018, 44005016, 40005035, 44021031, 40002003 Macedonia: 45015026, 52001007, 52001008, 44016012, 47008001, 52004010 Verses expressing path Jerusalem: Macedonia: 46016005, 44019021, 44020003

67 Appendix B Sample

Family Language ISO 639-3 code Araucanian Mapundungun arn Arawakan Palikúr plu Arawan Paumari pad Aymara Central Aymara ayr Barbacoan Cha’palaa cbi Boran Bora boa Cahuapanan Shawi cbt Camsá (Isolate) Camsá kbh Candoshi-Shapra (Isolate) Candoshi-Shapra cbu Cariban Akawaio ake Chibchan Central Tunebo tuf Chicham Aguaruna agr Lomeriano-Ignaciano Lomeriano-Ignaciano cax Chiquitano (Isolate) Chiquitano Chocoan Northern Emberá emp Cofán (Isolate) Cofán con Guahiboan Cuiba cui Guaicuruan Mocoví moc Harakmbut Amarakaeri amr Huitotoan Murui Huitoto huu Kakua-Nukak Kakua cbv Lengua-Mascoy Lengua leg Matacoan Wichí Lhamtés Nocten mtp Mosetén-Chimané (Isolate) Mosetén-Chimané cas Naduhup Nadëb mbj Nambiquaran Southern Nambikuára nab Nuclear-Macro-Je Maxakalí mbl Páez (Isolate) Páez pbb Pano-Tanacan Chácobo cao Peba-Yagua Yagua yad Quechuan Huallaga Huánuco Quechua qub Ticuna-Yuri Ticuna tca Tucanoan Tatuyo tav Tupian Tenetehara tqb Uru-Chipaya Chipaya cap Yanomamic Yanomámi wca Zamucoan Chamacoco ceg Zaparoan Arabela arl

68 Appendix C Suggested linguistic areas of South America

In this appendix, some continent-wide trais of South American languages are presented, fol- lowed by a review of a few different suggested linguistic areas of South America. As mention in section 4.2.2.2, none of these areas are well defined and all of them need further investigation. The discussion of each area is followed by a short comment on how the sample languages of this study relate to area in question. This is mainly based on the characteristic language families for each area presented in Campbell (2012b) and does not aim to be a complete rep- resentation of this relation nor include all sample languages.

C.1 Continent-wide traits There are a few widespread, rather specific features among South American languages from all regions, noted by David Payne (1990, cited in Campbell 2012b), that should be mentioned before discussing possible linguistic areas of the continent. These widespread features are: a negative morpheme roughly shaped ma; a causative affix roughly shaped mV (here V = vowel); a causative verbal prefix, most commonly a single ; a directional or locative suffix roughly shaped pV or Vp; an auxiliary expressing ‘to have’, ‘to do’, or ‘to be’, often coinciding with the lexical verbs ‘to say’ or ‘to work’, usually containing ka, and often with a valency- changing verbal affix of the same or similar shape; and complex verbal morphology orahigh degree of polysynthesis (this is also found in many languages of Central and North America) (Campbell 2012b, pp. 300-301).

C.2 Amazon(ian) linguistic area This area has been treated under several names, referring to slightly different geographical spaces, but most commonly it is referred to as the ‘Amazonian area’. Some suggested lin- guistic features of this area are: noun classifiers or gender systems (Dixon and Aikhenvald 1999, cited in Campbell 2012b); regularity in order of phrase constituents − namely, post- positions, genitive-noun, and noun-adjective − virtually regardless of the main clause’s word order; tendency towards ergative alignment (albeit, this needs more detailed investigation) (Derbyshire 1986, cited in Campbell 2012b); a very small set of lexical numbers; and shared discourse particles across families (Beier et al. 2002, cited in Campbell 2012b). The last two features may, according to Campbell, be ‘true features’ of the Amazon, although he also notes that the discourse particles may just as well be a cultural feature as a structural one (2012b, pp. 303-304). Some constructions that are suggested to be characteristic of this area, which are also com- mon elsewhere in South America are: head marked possessor constructions (and head mark- ing in general) − common in both the Americas ; only one core argument cross-referenced on verbs − a feature of Chaco languages as well; mainly prefixes and few suffixes − a widespread feature in South America (Dixon and Aikhenvald 1999, cited in Campbell 2012b); subordinate clauses involving nominalized verbs − commonly found in South American languages out- side of Amazonia; adverbs and adpositions that can be incorporated into the verb − also this a characteristic beyond Amazonia, e.g. in the Chaco and Andean regions; lack of an agentive passive construction − passive constructions often lack an overt agent in languages gener- ally; complex verbal morphology − found in all regions of South America; evidential markers − also found in many South American languages beyond Amazonia (Derbyshire 1986, cited in Campbell 2012b); and few oblique cases − another feature that is not just common in the

69 languages of Amazonia (Dixon and Aikhenvald 1999, cited in Campbell 2012b). Campbell (2012b) does not specify any language families as located in this area. The sample languages roughly located in the Amazon rainforest are Kakua, Tatuyo, Yanomámi, Paumari, Nadëb, Ticuna, Yagua, Bora, Murui Huitoto, Arabela, Chácobo, and Southern Nambikuára. The encoding of source and goal is very varied among these languages.

C.3 Lowland South American linguistic area This area is not well distinguished and it overlaps with the proposed Amazonian areatoa significant extent. Doris Payne (1990, cited in Campbell 2012b) divided the area into two sub- areas − a western and an eastern one. The western group of languages tend to have ahigh degree of polysynthesis; directionals expressed in the verb (which sometimes also have tense- aspect-modality functions); noun classifiers; verb-initial basic word order; and postpositions. Languages of the eastern part of the proposed area are more isolating; have few or no direc- tionals in the verbal morphology; and lack noun classifiers. The western group could compose a larger diffusion area by including the languages of the Chaco region since they share most of the features of the western group. The sample languages from the present study which would be included in the western group are Chácobo, Arabela, and Yagua, and those which would be included in the eastern group are Akawaio and Tenetehara − all distinctly encoding languages.

C.4 Andean linguistic area This proposed area is mainly defined on phonological features and the apparent similarities between Quechuan and Aymaran languages. Some of the shared phonological traits are:sys- tems of five to six vowels; contrastive vowel length (rather rare among other South American languages); a distinction between velar and uvular stops; glottalized obstruents; and series of plain, glottalized, and aspirated obstruents. Quechuan and Aymaran are also saidtobe characteristically agglutinative (Adelaar 2012b). For further discussion and detail see Ade- laar (2012b, pp. 602-607). Some other, more structural traits these languages share are: SOV basic word order; mainly suffixes; similar morphological structures; adjective-noun order; in- terrogative words clause-initially; accusative case; genitive case; a passive construction; and distinctly similar lexical items (seemingly loans) (Campbell 2012b). However, no languages of the suggested area except Quechuan and Aymaran languages are suffixing and agglutinative in a straightforward way (Adelaar 2012b, p. 594). In this study, the Andean region would be represented by Central Aymara, Huallaga Huá- nuco Quechua, and Chipaya, and possibly Cofán and Chácobo as well. All of these languages have distinct encoding of source and goal.

C.5 Colombian-Central American linguistic area The traits shared by languages in this region are: a voicing contrast in obstruents; exclus- ively SOV basic word order (as in the Andean region, see above); exclusively postpositions; mainly genitive-noun order; predominantly noun-adjective order (the most common order in all of South America (Dryer 2013)); noun-number order; primarily negation through suffixes or postpositioned particles; lack of gender in pronouns and inflectional morphology; lack of accusative marking; and a lack of an alienable-inalienable contrast (in contrast to most other languages of South America) (Umaña 1991, cited in Campbell 2012b). Several of

70 these features are expected to co-occur with SOV basic word order, such as the postpositions, genitive-noun order and noun-adjective order, and may therefore be insignificant to list among traits that could motivate a linguistic area. Furthermore, the lack of accusative marking may also be insignificant since it seems to be characteristic of most South American languages outside of the Andes according to Campbell (2012b, p. 306). No languages families are specified for this region in Campbell (2012b), but the sample languages from northern and western Colombia (which are likely to be included in this region) are Northern Emberá, Central Tunebo, Cuiba, Páez, and Camsá − all languages with a source- goal distinction.

C.6 -Amazon linguistic area This suggested linguistic area is based on a cultural area in the northern Amazon identified by Migliazza (1985, cited in Campbell 2012b). The proposed traits of this area are: ergative alignment; SOV or OVS basic word order (that is, O-before-V); absence of an active-passive distinction; relative clauses constructed by apposition and nominalization; and a suggested east to west spreading of nasalization, aspiration, and glottalization. All of these traits, except for the phonological ones, are also proposed to be characteristic of the Amazonian linguistic area (Campbell 2012b). The sample languages from this region are Akawaio and Yanomámi − one distinctly en- coding language, and one with a confirmed goal marker but an unknown source marker.

C.7 Venezuelan-Antillean linguistic area Some proposed features of Venezuelan-Antillean area are: the lack of a voicing contrast in obstruents; exclusively VO basic word order (and therefore lack of SOV order); numeral-noun order; noun-genitive order; and prepositions (Umaña 1991, cited in Campbell 2012b). All of these word order traits are expected to co-occur with VO basic word order and a lack of voicing contrast in obstruents is fairly common among languages, both in and outside of South Amer- ica (Campbell 2012b, p. 307). All of these suggested traits may therefore be unremarkable. The only language in the present study which may be from this region isAkawaio − a distinctly encoding language.

C.8 Vaupés(-Içana Basin) linguistic area This is a small region centered around the Vapués river on the boarder between Colombiaand . According to Campbell, this area is mainly suggested due to the Arawakan language Tariana having absorbed several traits from the in the area. The features shared in this area are: a tonal contrast (or pitch accent); classifiers (used in possessive con- structions as well as with demonstratives and numerals); two futures; topic-advancing verbal derivation; cross-referencing clitics; switch reference; serial verbs; nominative-accusative align- ment (with one case for topical non-subjects); evidential systems with four to five evidentials; a system of core case marking (cf. Zúñiga 2007); [dy] or [ʤ] as a word-initial variant of /y/; nasalization as a prosodic feature (Aikhenvald 2002, 2006a, cited in Campbell 2012b); and as- sociative plural (a suffix typically on a person’s name expressing ‘those associated withthe named person’) (Aikhenvald 2006b, cited in Campbell 2012b). Campbell notes that many of the traits are not unique to this area, but that the arguments may be substantial due to the amount and kind of shared features among the languages of the area (2012b, p. 308).

71 In the present study this region is represented by Tatuyo and Kakua − two indifferently encoding languages.

C.9 The Chaco There are two features that are ‘area-wide’ in the suggested Chaco area, namely and an unspecified possessor marker for possessed nouns (Campbell 2012b, p. 308). Some further traits that are only shared by most of the languages in the Chaco region are: SOV basic word order; active-stative alignment, complex sets of directionals (in verbal mor- phology); absence of verbal morphology expressing tense, which is instead expressed through nominal tense; complex demonstrative systems (with several semantic parameters, often more than just a contrast between visible and non-visible); polar negative adjectives; and genitive classifiers (and few or no classifiers of other kinds) (Campbell 2012b). A few phonological traits have been proposed, such as palatalization (Messineo 2003, cited in Campbell 2012b), contrast- ive nasalization, vowel harmony, and simple consonant clusters (Adelaar and Muysken 2004, cited in Campbell 2012b), although these are not widespread among Chaco languages. There are three languages from this region in the sample of this study. Two of themhave suspected goal markers but unknown source markers − Mocoví and Chamacoco − and the third language is Lengua − an unresolved case.

C.10 The Southern Cone This area, suggested by Klein (1992, cited in Campbell 2012b), overlaps to some extent with the Chaco region. The proposed characteristics of this area are: palatalization (as is also sug- gested for Chaco languages); more back than front ; morphological expression of semantic notions of position; several ways to express the location of a noun subject or object in relation to the speaker; tense, aspect, and number signalled morphologically in location, direction, and motion; and SVO basic word order. The last trait is perhaps mistaken since the languages of the Southern Cone display several different basic word orders (see Campbell 2012b, p. 309). All of these traits can be found in other suggested areas, apart from having more back consonants than front ones − although, this can probably also be found among Andean and Chaco languages according to Campbell (2012b, p. 309). This region is represented by Mocoví and Mapundungun in the present study − a language with a suspected goal marker but unknown source marking and an indifferent language.

C.11 Fuegian languages The Fuegian languages, an areal-typological grouping suggested by Adelaar and Muysken (2004, cited in Campbell 2012b), overlap partly with languages of the Southern Cone. Some of the shared traits among these languages are: low frequency of retroflex articulations, low frequency of voiced and glottalized consonants; mainly suffixing (although there aresome prefixing languages); mainly enclitics (although there are some languages with proclitics); compounding; reduplication; and mainly OV basic word order. Many of these traits are also distinctive features of Chaco and Andean languages. Thus, Campbell, as well as Adelaar and Muysken (2004, cited in Campbell 2012b), argues that this cannot be accepted as a linguistic area and that both the Fuegian and the Southern Cone languages need further investigation and distinction between them (2012b, p. 309). Campbell (2012b) presents a few languages located in this area, however, none of these are featured in sample of the present study.

72 Appendix D Source and Goal markers in distinctly encoding languages

Language Source marker Goal marker Type of marker Source Palikúr (plu) -tak, -(ri/i/a/u)tak -t, -(ri/i/a/u)t, -te Case suffixes (Launey 2003, pp. 139, 159) Central Aymara (ayr) -tʰa, -ta,(-t) -ru,(-r) Case suffixes (Cerrón-Palomino 2000, pp. 208-209; Hardman 2001, pp. 148-152) Bora (boa) -tu -bu Suffixes (Thiesen 1975, p. 17) Shawi (cbt) inanquën, ina quëran,(quëran) -quë Prepositions, (Hart 1988, pp. 354, 358, 477) Nominal suffix Camsá (kbh) -okan (-biokan) -oy (-bioy) Case suffixes (O’Brien 2018, pp. 82-84) Candoshi-Shapra (cbu) -’schochi -’pi Suffixes (Tuggy 1966, pp. 242-243) Akawaio (ake) abai,((a)bai) pona Postpositions (Caesar-Fox 2003, p. 108, 115, 136) Central Tunebo (tuf)a bin acam Prepositions (Headland 1994, pp. 58, 75) Aguaruna (agr) -ia, -numa-ia,(nmaya, numia) -(n)ĩ, -numa,(-num) Case suffixes (Overall 2007, pp. 222-223, 226-227) Lomeriano-Ignaciano auqui au Prepositions (Krüsi and Krüsi 1975, pp. 90, 92) Chiquitano (cax) Northern Emberá (emp) -pa, -de-pa, -maʉ-pa -(e)da, -ma(a),(-de, -mae) Case suffixes (Mortensen 1999, pp. 47, 52-56) Cofán (con) -ne -ni Nominal suffixes (Borman 1976, p. 117) 73 Cuiba (cui) (-)dena, (-)tsica, (-)tsina; we=, (-)iya, (-)icha, (-)deca; be= Simple verbs or nominal suffixes; (Kerr 1995, pp. 125, 128, 159) Proclitics to form complex verbs with the simple verbs Amarakaeri (amr) -ya’ -yo Case suffixes (Tripp 1995, p. 196) Murui Huitoto (huu) -mona -mo Case suffixes (Wojtylak 2017, pp. 300, 304) Páez (pbb) -hu -na Case suffixes (Curieux and Enrique 1998, p. 117) Chácobo (cao) ʔaʂ̌ (ʔoʔaʂ̌, -noʔoʔaʂ̌), (-)ki Postpositions, (Prost 1962, p. 110-111) -ka (ʔoka, -noʔoka) (occasionally suffixes) Yagua (yad) -siy -jù Postpositions, can be suffixed (Payne 1985, p. 161) Huallaga Huánuco -pita -man Case suffixes (Weber 1989, pp. 189, 196) Quechua (qub) Ticuna (tca) ne,(-wa ne) -wa Perhaps a postposition, (Anderson 1989, pp. 23, 387, 396) 1 1 1 1 suffix Tenetehara (tqb) -wi -kutir Suffixes (Bonfim Duarte 2007, p. 41) Chipaya (cap) -kiztan(a) -ø Case suffixes (Cerrón-Palomino 2006, pp. 125, 129) Arabela (arl) -cuaji, -ji -Vcuara, -(V)cuab, -jinia(-ra) Suffixes (Rich 1999, pp. 74-75)

aWith animate nouns the postposition quin marks both source and goal (Headland 1994, p. 31). bV = vowel. Appendix E Encoding in languages with no clear distinction

Table 13: Indifferently encoding languages

Language Marking Source Mapundungun (arn) mew,(mo/mu) − Postposition (Zúñiga 2006, p. 55) Kakua (cbv) =bǔ − Enclitic (Bolaños 2016, p. 206) Tatuyo (tav) Possibly -´pɨ̀,(-rè, -´pɨ̀-rè) − Locative suffix, (Gomez-Imbert 1982, p. 68) Possibly the lexical meaning of the verb Cha’palaa (cbi) Possibly the lexical meaning of the verb (Abrahamson 1962, pp. 238- 239) Southern Nambikuára Possibly the lexical meaning of the verb (Kroeker 2001, pp. 3, 9) (nab) Paumari (pad) The miotion verb gai in combination with a (Wälchli 2009, p. 1) directional suffix − Encodes the entire route Maxakalí (mbl) ha, tu, and possibly ta − Possibly adverbal (Popovich and Popovich markers, possibly postpositions 2005, pp. 5, 43; Pereira 1992, pp. 71-72)

Table 14: Languages with goal markers but unknown source markers

Language Goal marker Source Yanomámi (wca) =ɑmɨ (=ɑmë) − Enclitic (Perri Ferreira 2011, p. 42) Wichí Lhamtés -hu − Case suffix (Terraza 2009, p. 220) Nocten (mtp) Mocoví (moc) Possibly =pegeʔ/=peʔ, =kena, =igi − Enclitics (Grondona 1998, p. 132) Chamacoco (ceg) Possibly (-)tɨ̃r (otãra possibly marks source) (Sušnik 1970, pp. 135-136) − Unknown marker type(s)

Table 15: The remaining languages with no clear distinction

Language Description of marking Source Nadëb (mbj) Intermediate system with contrastive and (Weir 1984, pp. 96, 174) non-constrictive locative markers

Mosetén-Chimané Toponyms appear with a fixed marker for (Sakel 2004, p. 77) (cas) local relation. Associated motion can express movement to or from the deictic center

Lengua (leg) Nothing has been established. Wälchli and (Grondona 1998, p. 132) Zúñiga (2006) classified it as mixed, i.e. as having a weakly grammaticalized marker or construction for source

74 Appendix F Syncretism patterns between local roles in the sample languages

Table 16: Syncretism patterns in the sample languages

place = goal = source place = goal ≠ source place ≠ goal ≠ source No grammatical Unknown or other marking type of system Mapundungun Murui Huitoto Candoshi-Shapra Cha’palaa Paumari Kakua Lomeriano-Ignaciano Chipaya Southern Nadëb Chiquitano Nambikuára Tatuyo Shawi Palikúr Mosetén-Chimané Aguaruna Central Aymara Lengua Cofán Bora Chamacoco Camsá Maxakalí Akawaio Central Tunebo Cuiba Páez Chácobo Yagua Huallaga Huánuco Quechua Ticuna Tenetehara

Table 17: Languages displaying more complicated syncretism patterns

Language Description of syncretism pattern Northern Emberá Several markers for source, goal, and place − place = goal ≠ source and place ≠ goal ≠ source

Amarakaeri Several place markers − place ≠ goal ≠ source, place = goal ≠ source, and place = source ≠ goal

Arabela Several markers for source, goal, and place − place = goal ≠ source and place ≠ goal ≠ source

Yanomámi Unknown source marking − place = goal, source = ?

Wichí Lhamtés Nocten Unknown source marking − place ≠ goal, source = ?

Mocoví Unknown source marking, several place and goal markers − palce = goal and place ≠ goal, source = ?

75 Stockholm University SE-106 91 Stockholm, Sweden Telephone +46 (0)8 16 20 00 https://www.su.se/