Phasal Polarity Systems in East Bantu

Althea Löfgren

Department of Linguistics Independent Project for the Degree of Bachelor 15 HEC General Linguistics Spring 2018 Supervisor: Ljuba Veselinova Examiner: Francesca Di Garbo Phasal Polarity Systems in East Bantu

Abstract

This study explores a category of expressions akinto not yet, already, still and no longer, called Phasal Polarity (PhP) expressions. They encode the domains of phasal values, polarity and speaker expectations and have previously been described in European languages (van der Auwera 1998) and in a small, genealogically diverse sample (van Baar 1997). Using refer- ence grammars as the primary source of information, the aim of this crosslinguistic study is to describe PhP expressions in a sample of East . It is found that the distribution and behaviour of PhP expressions in East Bantu differ from both European languages and the genetically diverse sample of van Baar. The markers are found to be morphologically diverse and of varied crosslinguistic frequency. Furthermore, the verbal morphotax indicates that the markers are, or are in the process of, being incorporated into the tense-aspect systems of their respective language.

Keywords aspect, Bantu languages, phasal polarity, tense

Sammanfattning

Denna studie utforskar en kategori av utryck besläktade med inte än, redan, fortfarande och inte längre, som kallas Phasal Polarity (PhP) expressions. De uttrycker fas, polaritet och ta- larörväntningar och har tidigare beskrivits i en studie av europeiska språk (van der Auwera 1998) och i en småskalig, genetiskt mångfaldig studie (van Baar 1997). Med referensgram- matikor som primärkälla ämnar denna studie undersöka PhP expressions i ett urval av språk i underfamiljen Östbantu. Resultaten visar att PhP expressions i Östbantu har annorlunda distribution och användning jämört med europeiska språk och van Baars urval. Markör- erna varierar avseende morfologi och tvärspråklig frekvens. Verbens morfotax indikerar att markörerna har, eller håller på att, inkorporeras i respektive språks -system.

Nyckelord aspekt, Bantuspråk, phasal polarity, tempus Table of Contents 1 Introduction ...... 1 2 Background ...... 1 2.1 Phasal Polarity expressions ...... 1 2.1.1 PhP expressions and phasal values ...... 2 2.1.2 PhP expressions and polarity ...... 2 2.1.3 PhP expressions and speaker expectations ...... 4 2.2 Previous crosslinguistic studies ...... 5 2.2.1 Van der Auwera’s European study ...... 5 2.2.2 Van Baar’s crosslinguistic study ...... 5 2.2.3 Kramer’s PhP parameters ...... 5 2.3 Bantu languages ...... 6 2.3.1 Classification and characteristics ...... 6 2.3.2 Tense and aspect in Bantu languages ...... 7 2.3.3 Negation in Bantu languages ...... 8 2.3.4 Grammaticalization in Bantu languages ...... 9 2.4 Aims and research questions ...... 10 2.5 Summary ...... 10 3 Method ...... 10 3.1 Sample ...... 10 3.2 Working definitions ...... 12 3.3 Data collection ...... 14 3.4 Analysis ...... 14 4 Results ...... 15 4.1 Frequency of occurrence and form of PhP expressions in Bantu ...... 15 4.2 Morphological status ...... 21 4.3 Semantics ...... 23 4.4 Summary ...... 25 5 Further correlations and discussion ...... 26 5.1 Crosslinguistic frequency and universals ...... 26 5.2 Co-occurrence of PhP expressions and more general markers ...... 26 5.2.1 Co-occurrence with standard negation ...... 26 5.2.2 Co-occurrence of  and continuous aspect ...... 30 5.2.3 Co-occurrence of  and perfect aspect ...... 31 5.3 A small scale historical comparative analysis ...... 32 6 Conclusions ...... 35 References ...... 38 A Appendix 1 ...... 42 B Appendix 2 ...... 44 C Appendix 3 ...... 46 D Appendix 4 ...... 48 E Appendix 5 ...... 50 F Appendix 6 ...... 52 List of abbreviations

1 1st person  object marker 2 2nd person  optative mood 3 3rd person  persistive 9 prefix for class 9  polite register 15 prefix for class 15  present imperfective  article  perfect  augmentative  pronoun  auxiliary  progressive aspect  counterexpectational  present tense  conditional mood  past tense  consecutive tense  participle  direct evidential  recent past  future  singular  final vowel  subject marker  infinitive  stative  imperfective  subjunctive lgs languages  tense aspect mood  negative . . . Tumbuka-Sena-Nyanja . . North Mozambique  verb sentence marker  nominative case = clitic boundary  non-completive ≠ verbal root ? morpheme unknown

Presentation conventions Metalanguage with small capitals will be used to talk about the PhP expressions on an abstract level when no specific language is discussed (i.e. ,   etc.). Many Bantu languages have grammatical tone. This feature is not consistently indicated in the reference grammars. This is therefore only illustrated when tone is relevant to distinguish one PhP expression from another marker. Otherwise, tone is left out of the discussion. All examples are given as found in the original source. An unglossed morpheme of un- known meaning is glossed with a question mark. The ISO-code of each language is given in square brackets in all examples. 1 Introduction

Languages have different ways of categorising a very abstract part of the human existence – time. This central part of cognition can be expressed in different ways depending onhow one categorises the world. If the point of time of an action is important, the language in question can express this in different temporal distinctions, such as past, present and future tense. If the action is viewed as a whole, it can be expressed as aspectual distinctions. The start, middle and end of an event can also be encoded, mainly through adverbials expressing phasal values. Some languages have grammaticalized phasal values, in addition to tense and aspect. Whether the event is taking place at the time of the utterance or not, is also important and expressed through polarity, i.e. positive and negative values, e.g. the existence or non- existence of a particular situation. All of the categories above can be expressed either lexically or grammatically. There is a category of expressions that combine phasal and polarity values, briefly outlined above: a semantic domain typically referred to as Phasal Polarity, and the ways of its encoding, namely Phasal Polarity expressions (henceforth PhP). PhP items are morphologically diverse markers that facilitate the encoding of primarily three domains: phasal values, polarity and speaker expectations. Phasal expressions received some attention in the late 1990s, but not much has been done since then (van der Auwera 1998; van Baar 1997). Thus the domain remains fairly unexplored. Two ongoing projects are currently investigating this phenomenon. One of them is Phasal Polarity in African languages at Hamburg University, led by Raija Kramer. The other one is Expectations shaping grammar: searching for the link, between tense - aspect and negation at the Department of Linguistics at Stockholm University, led by Ljuba Veselinova. My study contributes to both of these projects by collecting and analyzing data from grammars. While this method is insufficient for a detailed study of this phenomenon, the information found in grammars is an important part of any crosslinguistic study. This study serves as a stepping stone for future research in an otherwise little explored area.

2 Background

In this chapter an overview of the theoretical background will be provided. First, a basic description of PhP expressions will be given (2.1). Second, the findings of earlier studies will be presented (2.2) and third, some relevant features of Bantu languages will be described (2.3).

2.1 Phasal Polarity expressions PhP items are morphologically diverse markers that facilitate the encoding of primarily three domains: phasal values, polarity and speaker expectations. In European languages, they are often adverbials as in the example of English PhP items presented below.

(1) PhP expressions in English (van Baar 1997: 1) a. Peter is already in London. b. Peter is still in London. c. Peter is no longer in London. d. Peter is not yet in London.

1 According to van Baar (1997), the four adverbials in the example are inherently telic/atelic and therefore possess phasal properties. PhP expressions are also affected by the existence or non-existence of a situation at the moment of speech. Therefore they also have polarity values. Many other expressions may appear to be PhP items, but at a closer look they lack either the phasal values or polarity values that are essential to the objects of this study. Consider the examples below.

(2) Expressions similar to PhP items (van Baar 1997: 1) [emphasis added] a. John will not visit me again. b. John will not be discussing this matter any further.

Van Baar explains that the above examples have iterative or additive meaning but are not phasal in the same way as PhP expressions are. To be able to classify an expression as a PhP item, it is essential that they are related by temporal phases as well as polarity.

2.1.1 PhP expressions and phasal values Phasal values are described by Plungian (1999) as being either inchoative, continuative or terminative, i.e. they represent the start, continuation and end of an event. While Plungian agrees that phases are similar to aspect, he argues that they are interconnected but not the same. Aspect describes the inner structure of an event by referring to points in time (phases), and therefore uses phasal values, to some degree, to express aspectual values (Plungian 1999: 314). However, phasal values form a separate functional domain, different from the domains of tense and aspect. The interrelated phasal properties of PhP expressions are best described using the metaphor of a timeline starting with  , followed by ,  and ending with  . If Plungian’s three phases are applied to PhP expressions, we end up with three categories: the inchoative , the continuative  and, in Plungian’s view, also  , and the terminative  . These items are described by van Baar as phasal and sequential, since they encode the three different phases of an event that occur after one another.

2.1.2 PhP expressions and polarity In addition to phasal values, polarity is an equally important part of PhP expressions. Consider a situation x. If x holds, a speaker would use  or  and if x does not hold they would use   or   . In addition to the fact that half of the PhP paradigm is negative and the other half positive, all four items are related to each other through negation. This systematic relationship is called the Duality HypothesisLöbner ( 1989). The Duality Hypothesis is based on the idea that all PhP expressions are related toeach other by external and internal negation Löbner (1989). External negation is where the element falls within the scope of negation (NOT (x)), whereas in internal negation, the element has the negation in it’s scope (x NOT). The English phasal adverbials create the system seen in figure 1. Figure 1 should be interpreted as a situation that either has negation in its scope or falls within the scope of negation. Using   as an example consider a situation p. If this event has yet to come, one could express it as [  p]. According to van Baar (1997) the internal negation of this situation would be [not [ p]] and the external negation would be [ [not p]]. This is true because of the fact that if an event has not already happened, it is true that is is not yet happening. Also, if it is still not happening, it is true that it has not

2 Figure 1: Duality in English (van Baar 1997: 20)

yet happened. This relationship works the same way even with the expressions that are not negated, i.e. the equivalent of  would be [not [  p)] and [  [not p]]. If a situation has not not yet happened, it has started and is already happening. If a situation is no longer not happening, it has also started and is now already happening. Languages make use of different combinations of internal and external negation to create their PhP systems. The examples below show how Spanish uses internal negation to create  and  .

(3) Example of  in Spanish [spa] (Garrido 1992: 361): El niño duerme todavía  child sleep..3  ‘’The child is still sleeping’

(4) Example of   in Spanish [spa] (Garrido 1992: 361): El niño no duerme todavía  child  sleep..3  ‘’The child is not sleeping yet’

The Spanish PhP system consists of an affirmative and anegative ya () and todavía () thus making use of internal negation to construct both   and   (van Baar 1997: 22). The following examples show how Classical Nahuatl uses external negation to create  and  .

(5) Example of  in Classical Nahuatl [nci] (Andrews 2003: 76, 92, 175): oc yōliyah still they.were.living ‘’they are still alive’

(6) Example of   in Classical Nahuatl [nci] (Andrews 2003: 76, 92, 175): ayoc āc .still he/she.is.present ‘’he is no longer here’

In this system external negation is used to create the negative PhP expressions, negating  to create an expression for  , ayoc. This means that   is equal to [not [ p]].

3 2.1.3 PhP expressions and speaker expectations Languages make use of different strategies to express if a situation conforms or deviates from the shared norms in a context. Heine, Claudi and Hünnemeyer (1991) describe these expres- sions as counterexpectation (CE) markers, like too and only illustrated in the example below.

(7) Counterexpectation markers in English (Heine et al. 1991: 192) a. Your house is too small, even if you are only two.

The main function of a CE marker is, according to Heine et al. , to compare a statement to what is being presupposed or assumed. In the case of PhP expressions, all of them are to be viewed as CE markers, but behave differently depending on the context as illustrated in van der Auwera (1998). Van der Auwera proposes a theory where he views PhP expressions from the perspective of speaker expectations instead of negation, called the Double Alternative Hypothesis. Van der Auwera explains that both the surrounding circumstances and the expectations of the speakers involved will affect which PhP expressions are being used. There are three sce- narios that influence the use of PhP items. The first scenario is neutral, where the expectations of the speaker is met (Scenario 1). The second scenario is an event realized earlier thanthe speaker expected (Scenario 2). The third scenario is an event realized later than the speaker expected (Scenario 3). The examples below are inspired by van Baar (1997: 27). To illustrate the different scenarios the examples share the same background:

• 7 a.m. Anna and Mary are meeting at the bus stop to exchange notes. • 8 a.m. Mary will leave for school. • 9 a.m. Mary will arrive at school.

In the neutral scenario neither of the girls are surprised with the outcome of the meeting because they are aware of the background information. Even if Anna oversleeps and arrives at the bus stop at 9 a.m., she would be aware that Mary would have already left and say: ’(I know) Mary is already at school.’ (Scenario 1). On the other hand, if Anna arrived on time, but Mary remembered that her morning class starts at 8 a.m. and not 9 a.m. and therefore left at the agreed meeting time, Anna’s response would be one of surprise: ’(Oh no!) Mary is already at school.’ (Scenario 2). Likewise, if Mary was late for the meeting and thus was late for school, Anna’s statement would be: ’Mary is finally at school.’ (Scenario 3). The second and third scenarios are therefore opposites, in the first one Mary arrives at school surprisingly early and in the second she is surprisingly late. In language specific cases all PhP expressions will not be compatible with all three scenar- ios. According to van Baar (1997: 29) an English example would show that  is only compatible with scenario one and two while scenario three demands a different expression: finally. Telicity also affects the compatibility with some scenarios, as is the case with  and  . Van Baar argues that their lack of a clear finishing point make them incompatible with the third scenario since their inherent lateness presupposes an event that has already taken place.

4 2.2 Previous crosslinguistic studies As already mentioned, the field of Phasal Polarity is understudied. A brief interest was shown in the area during the 1990’s, that gave rise to the studies described below and in 2.1.2 and 2.1.3, but since then the interest has subsided. It is not until now that linguists have started to show interest in the field with two large, ongoing projects that started recently (see chapter 1).

2.2.1 Van der Auwera’s European study Van der Auwera’s (1997) study of European languages gave rise to what he calls the Euroversal accessibility hierarchy. Based on a sample of 45 European languages, the study shows that it is most common for a language that has one PhP expression to have only  . Conversely, if a language is missing one PhP expression it is most commonly . Ordered from most common to least common the following hierarchy is presented by van der Auwera:   > /  > . There seems to be some opposition in the expressibility of the atelic half of the PhP paradigm with   and  on the opposite ends of the hierarchy.

2.2.2 Van Baar’s crosslinguistic study Van Baar’s (1997) findings dispute van der Auweras claims. Based on a crosslinguistic sample of 40 languages, his results show that the PhP expression most frequently missing is either  or   or both. Languages across the world seemingly favour the continuative half of the PhP paradigm, while European languages deviate in the sense of favouring the terminative   above all other PhP items. Van Baar also proposes two coverage universals that explain patterns in the PhP systems found in his sample (van Baar 1997: 166). The fist universal states that if a language has the same expression for  and  , there is no distinct  . This should be understood as it is more common for   to be derived from , and not the other way around. The second universal states that if a language has the same expression for  and  , it also has an expression meaning  and  .

2.2.3 Kramer’s PhP parameters Kramer (n.d.), influenced by the studies above, has developed six parameters under which PhP expressions can be discussed. The parameters are: coverage, pragmaticity, telicity, wordhood, expressibility and paradigmaticity. The first three parameters are used to discuss the semantics of PhP expressions, while the last three are concerned with their structure. The coverage parameter is concerned with how many PhP expressions a language usesper PhP concept. Kramer defines a PhP system with one expression per concept as rigid and a system that has more than one as flexible. The parameter discussing pragmaticity is based on van der Auwera’s three scenarios de- scribed in 2.1.3. Here, Kramer narrows it down to two scenarios: one neutral and one counter- factual. The counterfactual scenario is described as containing two simultaneous phases that are either actual or expected. This explanation leaves out both the unexpectedly early and late outcomes in van der Auwera’s theory, and instead focuses on the opposition between the expectation of the speaker and the reality of the situation.

5 Telicity in PhP expressions includes, as discussed in 2.1, both telic and atelic aspects. Kramer adds to this the concept of earliness/lateness from The Double Alternative Hypothesis, but also states that the telic PhP items can be insensitive to this concept and, thus, be general. The parameter wordhood is concerned with the independence, or grammaticalization, of PhP expressions. Kramer describes PhP items as diverse, ranging from independent auxiliaries with verbal properties to bound affixes. Since PhP expressions are so varied in their form, they cannot be identified based on the wordhood parameter alone, but need to be viewed as structurally diverse, specialised markers that form a paradigm in a language. The concept of expressibility is discussed in the previous section in the form of universals based on earlier findings. Kramer discusses this parameter more generally and states that PhP systems may or may not contain specialised items for all four expressions, no specialised items at all or any variation in between. The paradigmaticity parameter focuses on the symmetrical properties of the PhP system as a whole. Kramer describes this parameter as containing both an internal and external per- spective. From an internal point of view, paradigmaticity is concerned with whether a PhP expression has an alternative item with the opposite polarity. From an external perspective this parameter focuses on whether a system has a one-to-one relationship between PhP items and non-PhP items.

2.3 Bantu languages This study investigates the phenomenon known as Phasal Polarity in a sample of EastBantu languages. The Bantu family was chosen for two reasons: a) it is absent in all studies mentioned above and b) PhP expressions are frequently brought up in the Bantu literature, though not necessarily under this name. In this section relevant features of this family, such as tense, aspect and negation, will be described.

2.3.1 Classification and aracteristics Bantu languages are spoken in Sub-Saharan Africa. They are part of the Niger Congo phylum (Nurse & Philipson 2014: 1). The debate regarding the classification of Bantu languages is ongoing, mainly due to the complex migration patterns of the Bantu people, as well asthe difficulties of distinguishing language from dialect. The issue of classification aside, recent studies claim a number between 500 and 600 languages (Grimes 2000; Maho 2003; Grollemund et al. 2015). The focus of this study is the East Bantu sub-family, where Phasal Polarity hasnot been studied previously. This branch includes 252 languages and is found in the south eastern part of Africa. The Bantu verb is famously complex and is often described using a template toillustrate the morphotactic structure. Table 1 illustrates in what order Bantu verbs allow affixes. Due to its size Table 1 is locates on the next page. In total, there are seven slots devoted to affixes that make up what in English would be a whole sentence, leaving one slot for the verbal stem. Both the subject and the object are marked on the verb, together with , polarity and clause type. The word order most common in Bantu isNurse SVO( & Philipson 2014: 125).

6 Table 1: Verbal template (Güldeman 1999: 546)

Verbal template

SLOT FUNCTION preinitial /polarity initial subject postinitial /polarity preradical object radical verbal root prefinal derivation/ final  postfinal clause type/object/polarity

2.3.2 Tense and aspect in Bantu languages The preinitial, postinitial och final slots are used to express . According to Nurse (2003)  are encoded in Bantu languages using three main components: verbal inflection, tone and auxiliaries. Combining these three strategies, Nurse explains that Bantu languages can have several different past, present and future tenses with different degrees of distance to the point of reference or speech. Adding to this, they also encode many different kinds of aspect, the most common being: perfective, imperfective and progressive. Examples (8)–(11) illustrate the four future tenses of Logooli. From example (8) to (10) the future becomes more and more remote with the fourth future being uncertain.

(8) Example of near future tense in Logooli [rag] (Nurse 2003: 130) [own glossing]: kʊ-ra≠gʊr-a 1.-buy- ‘’We will buy’ [Near future]

The near future in Logooli is expressed with the marker ra after the subject marker.

(9) Example of middle future tense in Logooli [rag] (Nurse 2003: 130) [own glossing]: na-kʊ≠gʊr-i .1-buy-. ‘’We will buy’ [Middle future]

The middle future in the example is found in the subjunctive, but whether this is obligatory or not is uncertain. It is, however, expressed with the marker na before the subject marker.

7 (10) Example of far future tense in Logooli [rag] (Nurse 2003: 130) [own glossing]: kʊ-rika≠gʊr-a 1.-buy- ‘’We will buy’ [Far future]

The far future tense is marked by rika, and like the near future marker, this marker occurs after the subject.

(11) Example of uncertain future tense in Logooli [rag] (Nurse 2003: 130) [own glossing]: kʊ-ri≠gʊr-a 1.-buy- ‘’We will buy’ [Uncertain future]

The uncertain future tense, illustrated in example (11), is constructed with themarker ri. The same pattern is found in the past tenses, the only difference being that Bantu languages usually have a higher number of past tenses than future tenses, according to Nurse (2003). Nurse claims an average of one to three future tenses and one to four past tenses in the Bantu- family. In example (12) and (13) the imperfect and progressive aspect in Haya are illustrated.

(12) Example of imperfect aspect in Haya [hay] (Nurse 2003: 121): tu-Ø≠gúr-a/tu-Ø≠gur-á we-zero-buy-a ‘’We buy’ (IPFV)

(13) Example of progressive aspect in Haya [hay] (Nurse 2003: 121): ni-tu-Ø≠gúr-a -we-zero-buy-a ‘’We are buying’

The imperfect aspect is marked by high tone on either the final vowel a or the verbal stem gur. The progressive aspect on the other hand has amarker ni that occurs before the subject marker giving the verb the aspectual meaning.

2.3.3 Negation in Bantu languages According to Güldeman (1999), negation in Bantu languages is mainly accomplished using two different strategies. A negative utterance can be constructed both using a) a periphrastic con- struction and b) grammaticalized affixes, particles or clitics. Together with these two strate- gies, Bantu languages negate the verb using three of the slots above, shown in Table 1: the preinitial, the postinitial and the postfinal. Negation using the preinitial slot is by far the most common and is thought to be the unmarked negation in many Bantu languages (Güldeman 1999: 555-556). The examples below show preinitial and periphrastic negation in two Bantu languages.

(14) Example of preinitial negation in Standard Swahili [swh] (Güldeman 1999: 551): ha-tu-tak-i 1-want- ‘’Wir wollen nicht’ [we don’t want]

8 Example (14) illustrates preinitial negation in Standard Swahili. The preinitial slot contains the negative marker ha, followed by the subject marker, the verbal stem and lastly the final vowel marking the present tense.

(15) Example of periphrastic negation in Hungan [hum] (Güldeman 1999: 550): beet tu-khoon-aak ku-mon ni beet tu-biing we 1-fail- -see that we 1-win ‘’We don’t think that we’ll win’

In contrast, Hungan uses periphrastic negation illustrated in example (15). The auxiliary verb ’fail’ is marked with a subject marker and a progressive marker and has the main verb as an infinitive complement.

2.3.4 Grammaticalization in Bantu languages To further discuss some of the diachronic implications of this study, an introduction to gram- maticalization is necessary. Broadly speaking, grammaticalizaiton is a diachronic process where items with a more referential meaning are gradually incorporated into the grammar of a language. Typical traits of this process are semantic change, phonological erosion and loss of wordhood (Givón 1971: 12).1 Givón proposes the cycle of grammaticalization to ex- plain how the synchronic morphology and the diachronic syntactic structure are related.

(16) Givón’s cycle of grammaticalization discourse > syntax > morphology > morphophonemics > Ø

With this cycle, Givón wants to illustrate that the syntax of a language is indicative of possible morphological patterns. For example, if an adverb becomes grammaticalized the word order in a language will affect whether it becomes a prefix or a suffix. If adverbs occur before the verb, they will become prefixes and if they occur after the verb they will become suffixes. Traugott (2011) proposes three basic mechanisms of grammaticalization: Reanalysis, ex- tension and repetition. Reanalysis is change in the structure of an utterance that does not have any apparent effect on the surface structure. For example changes in word order or morpheme boundaries. Extension is to be understood as a broader term that can include ex- tension of a wide range of domains, such as semantic extension which expands the meaning of an utterance. The syntax and pragmatics of an utterance can also undergo extension. The repetition of an utterance also contributes to the grammaticalization process. Repetition can lead to phonetic reduction and loss of referential content. A pattern of grammaticalization can be found in African languages using the mechanisms proposed by Traugott. Heine (2011) describes how the African languages develop their com- plex -systems by a constant addition of new forms of tense and aspect-markers. He pro- poses the grammaticalization chain in example (17) to describe this process.

(17) The grammaticalization of auxiliary constructions (auxiliation)Heine ( 2011: 21) a. Main verb (V1) - (non-finite verb) complement b. Auxiliary - Main verb (V2)

The example above can be described in the three steps.

1The literature on this subject is extensive. I only reference the work which is of immediate relevance tothis study.

9 1. V1 takes a non-finite verb as its complement (a). 2. V1 loses some of its semantic and morphosyntactic properties, and gains some gram- matical properties (a). 3. V1 is now an auxiliary and V2 is the main verb (b).

The result of this grammaticalization chain is an auxiliary with inflection typically associ- ated with a main verb and a main verb that is non-finite. In extreme cases V1 will become even further grammaticalized into a clitic or an affix. These types of constructions are commonly found in this study, as will become apparent in chapter 4.

2.4 Aims and research questions The purpose of this study is to describe Phasal Polarity expressions in East Bantu languages, while also contributing to the current discussion on Phasal Polarity. The research questions are as follows:

1. What form do the PhP expressions have? 2. Which PhPs are available and in what languages? 3. What is the morphological status of the markers? 4. What function do the PhP expressions have? 5. Do the markers occur under any restrictions?

2.5 Summary PhP expressions are morphologically diverse markers with opposite phasal values and polarity that encode speaker expectations and together form a system within a language. Earlier studies have found that a sample of European languages differs from a crosslinguistic sample in what the most common PhP items are. Van der Auwera’s European study finds that   is the most common PhP expression, while van Baar claims that  is the most frequent. The ambition of this study is to add to the theoretical framework of PhP expressions by describing Phasal Polarity in a sample of East Bantu languages. The Bantu family was chosen because there is evidence of PhP expressions in the Bantu literature and the family is absent in all previous studies.

3 Method

The sample (3.1), working definitions (3.2), data collection (3.3) and methods of analysis (3.4) are presented in this chapter.

3.1 Sample The 46 languages2 selected for this study are presented in alphabetical order in Table 2. The sample languages are mainly spoken in , Zimbabwe, Tanzania, Mozambique and South Africa and belong to 15 different branches of the East Bantu sub-family. In Table 2 these different genealogical groups are called clusters. The clusters are presented in appendix 6.

2A map illustrating the distribution of the sample can be found here: http://www.arcgis.com/home/ webmap/viewer.html?webmap=8daad53ce7d349159b13f1aee5d1b116

10 East Bantu was chosen because a) it was deemed a good place to start investigating this phe- nomenon, since no previous studies on Phasal Polarity has been conducted in this area and b) it has the best descriptive literature of the African languages. This gives the study a bibliog- raphy bias, but could not be avoided since this study relies solely on descriptive grammars for data collection.

Table 2: Sample

e sample Language ISO code Country Cluster G. zone Bemba bem Zambia Sabi M42 Bena (Tanzania) bez Tanzania S. T. Highlands Bantu G63 Chopi cce Mozambique S. Bantu-Makua S61 Chuwabu chw Mozambique N. Mozambique Bantu P34 Gitonga toh Mozambique S. Bantu-Makua S62 Ila ilb Zimbabwe Botatwe M63 Kalanga kck Botswana Shona S16 Kaonde kqn Zambia Central Western Bantu L41 Kgalagadi xkv Botswana Sotho-Tswana S 31d Koti eko Mozambique N. Mozambique Bantu P311 Lala-Bisa leb Zambia Sabi M51 Lamba lam Zambia Sabi M54 Lenje leh Zambia Botatwe M61 Makhuwa vmw Mozambique N. Mozambique Bantu P31 Makonde kde Tanzania Rufiji-Ruvuma P23 Makwe ymk Tanzania Rufiji-Ruvuma G402 Manyika mxc Zimbabwe Shona S13a Matumbi mgw Tanzania Rufiji-Ruvuma P13 Mwani wmw Mozambique Coastal NEC Bantu G403 Ndali ndh Malawi Corridor Bantu M301 Ndengereko ndg Tanzania Rufiji-Ruvuma P11 Ngoni ngo Mozambique S. T. Highlands Bantu N12 Nsenga nse Mozambique Sabi N41 Nyakyusa-Ngonde nyy Malawi Corridor Bantu M31 Nyamwezi nym Tanzania Sukuma-Nyamwezi F22 Nyanja nya Mozambique Tumbuka-Sena-Nyanja N31a Pedi nso South Africa Sotho-Tswana S32a Sena (Malawi) swk Malawi Tumbuka-Sena-Nyanja N441 Shona sna Zimbabwe Shona S10

11 Continuation of Table 2 Language ISO code Country Language family G. zone Soli sby Zambia Botatwe M62 Southern Sotho sot South Africa Sotho-Tswana S33 Sumayela Ndebele nbl South Africa Nguni-Tsonga S408 Swati ssw Swaziland Nguni-Tsonga S43 Tawara twl Mozambique Shona S11 Tewe twx Mozambique Shona S13b Tonga (Zambia) toi Zambia Botatwe M64 Totela ttl Zambia Botatwe K41 Tsonga tso Mozambique Nguni-Tsonga S53 Tswa tsc Mozambique Nguni-Tsonga S51 Tswana tsn Botswana Sotho-Tswana S31a Tumbuka tum Malawi Tumbuka-Sena-Nyanja N21 Venda ven Zimbabwe Sotho-Makua-Venda S21 Xhosa xho South Africa Nguni-Tsonga S42 Yao yao Mozambique Rufiji-Ruvuma P21 Zimbabwean Ndebele nde Zimbabwe Nguni-Tsonga S44 Zulu zul South Africa Nguni-Tsonga S42 Total: 46

As mentioned in 2.3, Bantu languages are notoriously hard to classify; they are also un- derdescribed. This leads to the main limitation of the sample. To be able to capture asmuch diversity as possible, ideally both the genealogical and geographical distribution would have been considered when constructing the sample (Croft 2002: 21). This has, however, proven difficult due to the number of languages that lack descriptive grammars and the sample has thus been limited to a convenience sample. Nonetheless, great care was taken in selecting languages with as even a distribution as possible. Several languages belonging to genealogical clusters where also chosen to enable small scale historical comparative analysis.

3.2 Working definitions The following section defines the characteristics and functions of all four PhP expressions based on the works of van der Auwera (1998), van Baar (1997) and Kramer (n.d.). These pa- rameters are also used in Veselinovas ongoing project.

1.  : This is a negative continuative expression. It represents an ongoing period of time before an event where the action has not yet taken place, but is expected to do so.

2. : This is a positive, inchoative expression which embodies a point of change when an event starts taking place. This PhP expressions occur in three scenarios.

• Scenario 1: A neutral scenario where no speaker expectations exist.

12 • Scenario 2: A surprisingly early scenario where the speaker expected the event to occur later. • Scenario 3: A surprisingly late scenario where the speaker expected the event to occur sooner.

3. : This is a positive continuative expression. It represents a period of time whenan event is taking place.

4.  : This is a negative terminative expression. It represents a negative pointof change when the event ends.

Consider the examples below.

(18) Example of   in Hausa [hau] (van Baar 1997: 84): datra úwa nóron man doctor  still he.is ‘’The doctor isn’t here yet’

In example (18) the negative continuativity of   is apparent because of the fact that there is a period of time, ongoing at the time of the utterance, where the doctor is not there, but that the speaker expects the doctor to come.

(19) Example of  in French [fra] (van Baar 1997: 75) [own glossing]: L’arbre fleurit déjà the=tree blossom..3 already ‘’The tree is blossoming already (it didn’t bloom before)’

In example (19) it is apparent that  is an inchoative expression since it marks the point where the tree starts blooming, which is now and not before. This example is neutral and belongs to scenario 1.

(20) Example of  in Georgian [kat] (van Baar 1997: 77): me k’idev xel-i unda mo-va-c’er-o I still hand- must 1..-write- ‘’I still have to sign’

Example (20) shows that  is continuative since the speaker had to sign before, and still has to sign at the moment of the utterance. Therefore the situation is ongoing.

(21) Example of   in Burmese [mya] (van Baar 1997: 79): Pita ma-maùn-dáw-ba-bù Peter -drive-anymore-() ‘’Peter is no longer driving’

Example (21) shows that   is terminative since it marks the point when Peter has stopped driving.

13 3.3 Data collection The PhP expressions where collected according to the working definitions described in3.2and are presented in appendices 1–4. Since interaction and correlation with tense, aspect and other grammatical categories have been discussed in earlier studies, information about the perfect and continuous tenses and standard negation were also collected to enrich the discussion. A study that relies as heavily on reference grammars as this one is bound to encounter problems associated with reliability. The ideal grammar is a both current and detailed descrip- tion of a language from which the relevant data can be extracted. However, such a grammar is not always possible to find, especially in the case of Bantu languages, as many grammars are superficial or outdated. There are two main concerns when using an outdated or superficial grammar. The problem most relevant to this study is the fact that an old grammar mightbe missing features simply due to the lack of knowledge of a certain domain at that time. As Phasal Polarity is a relatively new field, there is a considerable risk of it being overlooked in a grammar from the early 1900’s. The second problem is that the presentation conventions might also have changed over the last century, as well as the languages themselves, making both the data and analysis unreliable. The above-mentioned problems are difficult to solve be- cause of the lack of secondary sources. While it is important to mention the insufficiency of a study exclusively based on grammars, it is equally important to note that it is also a necessary stepping stone for further research.

3.4 Analysis Glossed examples are a prerequisite for any type of linguistic analysis. As straightforward as this might sound, it proved problematic early on in the study, as many grammars did not contain glossed examples. Below follows a typical example.

(22) Example of   in Matumbi [mgw] (Odden 1996: 66) nįnakalaanga lį ’I haven’t yet fried’

It is apparent from the example that no morpheme boundaries can be identified and that the meanings of said morphemes are unclear. This problem was solved by glossing the examples containing the relevant markers by comparing them to the chapters in the grammar describing the verb structure, or verbal template, and the relevant affixes. The examples were glossed according to the Leipzig glossing rules, with one exception (Comrie et al. 2008). Since the exact meaning of the verbal stem could not always be established, it was left in the glossing as . The lack of glossed examples posed yet another problem. When working with theagglu- tinative Bantu language family, bound markers are expected to be found. However, because of the lack of glossing, unless the bondedness was mentioned explicitly in the text, the or- thography of the examples was used as the main source when deciding whether a morpheme was bound or free. This makes the conclusions drawn regarding the bondedness of thePhP expressions unreliable. Below follows a typical example.

(23) Example of   in Tsonga [tso] (Ouwehand 1978: 108) a. A mi si heta xana? — have you not yet finished? b. A ku si hlakuriwa la — it has not yet been weeded here

14 Example (23) possibly shows morpheme boundaries by inserting spaces between each of them, making all parts of this structure seem like unbound markers. It might be the case that they are actually bound, but that the presentation in the grammar make them look like free forms. This is not discussed or mentioned anywhere in the grammar, forcing the bondedness criteria to be based solely on the orthography. Kramer’s parameters are mainly applicable to the semantic content of PhP expressions. Again, due to the limited data in the sources, information about the semantic content has been difficult to come by. Therefore Kramer’s parameters are not used when discussing the findings in this study. In his study van Baar (1997) used questionnaires to investigate PhP expressions. This method, of course, provides more in depth information than a study using reference gram- mars as the only source. However, within the scope of this study, I have not been able to use questionnaires. It is important to note that under the the present circumstances, the results can demon- strate tendencies within the sample, but should not be used to make any wider claims.

4 Results

In this chapter the results of the study are presented following the order of the research ques- tions.

4.1 Frequency of occurrence and form of PhP expressions in Bantu The fist two research questions are addressed in this section:

1. What form do the PhP expressions have? 2. Which PhPs are available and in what languages?

PhP expressions are very diverse in regards to their form, but some genealogical tendencies can be discerned (see. 5.3). For a detailed description of all PhP expressions, see appendices 1-4. Table 3 shows the crosslinguistic frequency of PhP expressions among the 46 sample languages. This includes all expressions that correspond to the definitions in3.2.

Table 3: Crosslinguistic frequency of PhP expressions

Crosslinguistic frequency of PhP expressions

PhP       Distribution 36 (0.78) 16 (0.35) 27 (0.59) 12 (0.26)

The most common PhP item is  , which occurs in 78 percent of the languages. The second most common marker is  with a 59 percent occurrence rate.  and   are much less common, as they only occur in 35 and 26 percent of the sample respectively. Ordering the markers from most common to least common the following East Bantuversal hierarchy (inspired by van der Auwera’s Euroversals) can be proposed:

15 (24) The East Bantuversal accessibility hierarchy for PhP   >  >  >   For a more detailed description of the distribution of PhP expressions in each language, see appendices 1-4. The PhP systems found in the sample are illustrated in Table4.

Table 4: PhP systems

PhP systems PhPs per language Function No. of lgs ISO Four    6 [chw][sot]    [ssw][tsn] [ven][zul]    5 [bem][cce] Three  [ilb][xkv] [xho]     5 [kck][nso]  [twx][ttl] [tso]   1 [nbl]      8 [lam][leh] Two [wmw][nyy] [nya][twl] [tsc][yao]    2 [eko][nde]   1 [kqn]   11 [toh][vmw] One [kde][ymk] [mxc][mgw] [ngo][swk] [sna][sby] [toi]  1 [bez]  1 [ndg] None - 5 [leb][ndh] [nse][nym] [tum]

Altogether, ten types of systems could be found, although many of these only have one or two occurrences. As shown in Table 4, six languages have all four PhP expressions and

16 five have none at all. In total, eleven languages have three PhP items each, but in different constellations. Five of these languages have excluded  , five have excluded , and one language has excluded  . The two most common PhP item to exclude ina system with three expressions are  and  . For a more detailed overview of the PhP systems, see Appendix 5. Eleven languages were found to have systems with two PhP expressions. Eight of these systems contain  and  . Two languages have systems with  and  , and one language has a system with  and . The most common two-item system in the sample is, by far, one with  and  . Systems with only one PhP expression were found in thirteen languages. This is the most common PhP system. Eleven of these have  , one has  and one has . The languages with only one item seem to favour   over all other PhP expressions, since this is the most common PhP system found. Makhuwa is one of the languages with only  , illustrated in figure 2. The language specific PhP system is illustrated with plus signs next to the existing markers.

Figure 2: PhP expressions in Makhuwa

(25) Example of   in Makhuwa [vmw] (Van der Wal 2009: 106): mí nki-ná-n-koh-é . .1-1-ask- ‘’I haven’t asked him yet’ The PhP system in Makhuwa only has the PhP expression  , which is expressed with a prefix, ná, affixed to the negated verb. The second most common arrangement, with eight occurrences in the sample, isatwo- marker system containing   and . Figure 3 illustrates this kind of system found in Nyakyusa-Ngonde.

Figure 3: PhP expressions in Nyakyusa-Ngonde

Examples (26) and (27) show the PhP expressions  and   in Nyakyusa-Ngonde.

17 (26) Example of  in Nyakyusa-Ngonde [nyy] (Persohn 2017: 186): tʊ-kaalɪ tʊ-kʊ-job-a 1 1-speak- ‘’we still speak / we are still speaking’

Nyakyusa-Ngonde has a PhP system with two PhP items,  and  . These are both expressed with an auxiliary, kaalɪ, that takes two different complements. This auxiliary is used to express that a situation still holds from a later point in time. When it is marked by a subject marker, forming a periphrastic construction with an inflected verb as its complement, it takes on the meaning of .

(27) Example of   in Nyakyusa-Ngonde [nyy] (Persohn 2017: 187): jɪ-kaalɪ ʊ-kʊ-py-a 9 15()-be(come).burnt- ‘’It’s not yet done.’

When it takes an infinitive complement, like in example (27), the meaning changes to that of  . Therefore, unlike in the example for Spanish in 2.1.2, it is the complement that decides the meaning of kaalɪ, not negation. Six languages have all PhP expressions. An example of a complete PhP system is illustrated by data from Southern Sotho in Figure 4.

Figure 4: PhP expressions in Southern Sotho

The PhP system in Southern Sotho uses three different markers to express all fourPhP items. One marker is used to express  , another is used exclusively for , and the third marker is used both in constructing  and  . Examples (28)–(31) il- lustrate the PhP items in Southern Sotho. No language with a complete system possesses a grammar with glossed examples, therefore the glossing of each example was completed ac- cording to 3.3.

(28) Example of   in Southern Sotho [sot] (Doke & Mofokeng 1957: 213) [own gloss- ing]: ha-kē-es’o-rek-ē 1.-buy-. ‘’I do not yet buy; I have not yet bought’

  is expressed with the marker es’o used in conjunction with standard negation in a bound construction.

18 (29) Example of  in Southern Sotho [sot] (Paroz 1946: 96) [own glossing]: kē sa rut-a 1  teach- ‘’I still teach’

Unlike es’o,  is expressed with an auxiliary, sa, which is used together with the main verb.

(30) Example of  in Southern Sotho [sot] (Paroz 1946: 95) [own glossing]: kē-se kē-rut-a 1 1-teach- ‘’I already teach’

(31) Example of   in Southern Sotho [sot] (Paroz 1946: 95) [own glossing]: kē-se kē-sa-rut-e 1 1-teach-. ‘’I teach no more’

What is described as another auxiliary, se, is used to express both  and  . In this construction the subject marker is found on both the auxiliary and the main verb. In the case of  , there is a negation marker sa, that is described as a dependent negation marker. There is also a group of ten languages of which half have excluded  and half have excluded  . The PhP system in Chopi has all PhP expressions except  , in a system with both periphrastic and bound constructions. Examples (32)–(34) show the PhP expressions in Chopi.

Figure 5: PhP expressions in Chopi

(32) Example of   in Chopi [cce] (Smyth 1902: 95) [own glossing]: ni-sanga-von-a 1.-see- ‘’I have not yet seen’

  is expressed with the marker sanga, directly before the stem of the verb.

(33) Example of  in Chopi [cce] (Smyth 1902: 38) [own glossing]: se-ni-von-ile 1-see- ‘’I have already seen’

19  is expressed with a prefix, se, and the perfect tense. (34) Example of  in Chopi [cce] (Smyth 1902: 95) [own glossing]: ni-nga-di ni-ci-von-a 1 1-see- ‘’I still see’ The construction meaning  is more complex and contains markers that are not ex- plained. In Chopi,  is expressed using a periphrastic construction with an auxiliary, di, marked by both the subject marker and a marker of unknown meaning, nga, which is described as a ’tense marker’. The main verb is inflected with both a subject marker and a consecutive tense marker. Except for the fact that this utterance means , exactly how the consecutive tense marker and the ”tense marker” nga affect the meaning is unknown. In Totela three PhP items could be found. This system uses two markers to express the three different PhP items, illustrated in examples (35)–(37).

Figure 6: PhP expressions in Totela

(35) Example of   in Totela [ttl]Crane ( 2011: 337): ta-ndi-ini ku-kalis-a 1-have- -begin- ‘’I haven’t started (yet)’   is expressed with the negated auxiliary ini together with an infinitive complement (Crane 2011: 337). (36) Example of  in Totela [ttl]Crane ( 2011: 325): ndì-chì-hùpúl-à 1-think- ‘’I’m still thinking’ (37) Example of   in Totela [ttl]Crane ( 2011: 327): ta-tu-chi-izi 1-know. ‘’We no longer know’  and   are expressed with the marker chi after the subject marker, which, in the case of  , is negated using standard negation. To answer research question number two: The PhP systems found in the sample are quite diverse. Six languages have a system containing all PhP expressions and five have none. Eleven languages have systems containing different combinations with three items, another eleven languages have systems with two items and thirteen languages only have one PhP item. As shown in Table 3,   and  seem to be the most favoured expressions.

20 4.2 Morphological status Research question number three is answered in this section:

3. What is the morphological status of the markers?

Table 5 illustrates the morphological status of all PhP items. The largest category in Table 5 is labeled ’Unclear’. This means that the morphological status is not mentioned explicitly in the reference grammar. Because of the problems regarding the orthography and bondedness described in 3.3 the results presented here should be considered tentative at best.

Table 5: Morphological status

Morphological status of PhP expressions

      Verb 2 (0.05) 0 0 0 Adverb 0 3 (0.19) 1 (0.04) 0 Particle 2 (0.05) 1 (0.05) 2 (0.07) 0 Auxiliary 6 (0.17) 3 (0.19) 6 (0.22) 1 (0.08) Prefix 6 (0.17) 3 (0.19) 6 (0.22) 4 (0.3) Infix 1 (0.03) 0 1 (0.04) 0 Suffix 1 (0.03) 0 0 0 Unclear 18 (0.5) 6 (0.38) 11 (0.4) 7 (0.58) Total: 36 16 27 12

  occurs as a prefix six times and an auxiliary six times. The other occurrences are few, for example there are two examples of particles and one suffix.  occurs as an adverb, auxiliary and prefix three times each, and only once as a particle. There are six examples each of  being an auxiliary or a prefix, and two of it being a particle. The other categories are only found once.   occurs four times as a prefix and once as an auxiliary. Below, the most common types of the PhP expressions are given in examples (38)–(44). (38) Example of   in Matumbi [mgw] (Odden 1996: 66) [own glossing]: ni-na-kalaanga̧ li̧ 1.-fry  ‘’I haven’t yet fried’ As shown in Table 5,   most commonly occurs as either a prefix or an auxiliary. The prefix na in example (38) is described as a negative persistive marker with no known positive counterpart. This prefix occurs with the negative marker li.̧

21 (39) Example of   in Makwe [ymk] (Devos 2008: 341): ináamba kuúlya :1-be:yet. 15.eat. ‘’I have not eaten yet’

The auxiliary námba in example (39) is a polarity-item described as meaning ’be (not) yet’. It always co-occurs with negation and takes an infinitive as complement.

(40) Example of  in Zimbabwean Ndebele [nde] (Bowern et al. 2002: 186): se-si-dl-ile 1-eat- ‘’We’ve already eaten’

Just as in the case with  , the most common types of  markers are prefixes and auxiliaries, two categories that tie with the adverb. Example (40) shows  as a prefix in Zimbabwean Ndebele. It occurs word-initially with the verb in the perfect tense.

(41) Example of  in Koti [eko] (Schadeberg & Mucanheia 2000: 147) [own glossing]: ka-áz-’ó-o-sííl-a 1-hear- ‘’I have already heard it (before)’

The auxiliary in example (41) is described as an affirmative counterexpectational that, when no other event is mentioned, means . It is preceded by the subject marker and has an infinitive as a complement.

(42) Example of  in Bena [bez] (Morrison 2011: 269) [own glossing]: ndi-pí-gon-a 1-walk- ‘’I’m still walking’

 is most commonly found as a prefix or an auxiliary. In Bena, example (42), it is described as a persistive prefix, pi, which implies that the subject is ’still V-ing’. It is found after the subject marker and before the verbal stem.

(43) Example of  in Mwani [wmw] (Floor 2010: 15) [own glossing]: n-ingarí ku-fyóm-a 1 -read- ‘’I am still reading’

In Mwani,  is an auxiliary in the form of ingarí, and is inflected with a subject marker and takes an infinitive as a complement.   is most often described as a prefix, as in example (44), where the PhP expression consists of the prefix sá together with standard negation.

(44) Example of   in Sumayela Ndebele [nbl] (Ziervogel 1959: 88) [own glossing]: a-tí-sá-dl-i 3-eat-. ‘’They no longer eat’

22 Table 6: Bondedness

Bondedness of PhP expressions

      Bound 18 (0.5) 4 (0.25) 17 (0.63) 7 (0.58) Periphrastic 14 (0.39) 3 (0.19) 4 (0.15) 3 (0.25) Free forms 4 (0.11) 9 (0.56) 6 (0.22) 2 (0.17) Total: 36 16 27 12

The table above describes the bondedness of the PhP expressions. Half ofthe  - markers are bound, 39 percent are used in a periphrastic construction and eleven percent are unbound morphemes. The marker  is also mainly bound, at 63 percent, followed by a 15 percent section of periphrastic constructions and 22 percent free forms.   follows the same structure. Of the examples found, 58 percent are bound, 25 percent are used in a periphrastic construction and 17 percent are unbound. The one marker that differs from this pattern is , which has 56 percent free forms, 19 percent periphrastic markers and 25 percent bound items. According to section 2.4.4, which explains the grammaticalization process, the bonded- ness of the PhP expressions indicates the level of grammaticalization.  is therefore the most grammaticalized marker, followed by   and  . The high percentage of periphrastic constructions for  and   indicate that they are currently undergoing grammaticalization (see example (16)). In short, the answer to research question number three is that the PhP expressions are of very varied morphological status, however, some tendencies can be discerned. The most common categories for  ,  and   is auxiliary and prefix.  is commonly found to be a prefix, an auxiliary or an adverb. ,   and   all have a high tendency to be grammaticalized, while  does not.

4.3 Semantics Research questions number four and five are treated in this section:

4. What function do the PhP expressions have? 5. Do the markers occur under any restrictions?

The function of the PhP expressions were investigated with great difficulty. Thelackof detail in many grammars often provides the reader with no more than an English translation of the marker, without discussing its usage and meaning further. In three cases   was described as a CE-marker (Guerois 2015; Schadeberg & Mucanheia 2000; Van der Wal 2009), or to have ’expectant implications’ (Stevick 1960; Canonici 1996). In one case  was described as an affirmative CE-marker, and in another it was described as inceptive (inchoa- tive) (Schadeberg & Mucanheia 2000; Canonici 1996).  is described on four occasions as marking an action which has previously been going on and still continues at the time of

23 speech (Gadelii n.d.; Ziervogel 1952; Kiso 2012; Persohn 2017).   is not described as anything other than a negated form of . Even though the function is not mentioned explicitly in the grammars, the markers po- sition in the verbal template can be used to speculate about its status in the grammar of the relevant language. Table 8 shows in what slot in the verbal template (2.4.1) the PhP expressions occur.

Table 7: Position in the verbal template

Position in the verbal template

      Initial 0 3 0 0 Postinitial 14 1 16 7 Postfinal 1 0 1 0 Unclear 1 1 1 0 n/a 20 11 9 5 Total: 36 16 27 12

The unbound or periphrastic constructions are not applicable, hence the last category. The majority of the bound  -markers are found in the postinitial slot, directly preceded by the subject marker. The postinitial slot is described by Güldeman (1999) as being reserved for -marker. The bound -markers show the same tendency, with 16 examples found in the postinitial slot.   has seven markers in the postinitial slot.  is, again, deviant with only one marker in the postinitial slot, and three in the prenitial slot directly preceding the subject marker. The preinitial slot is described by Güldeman (1999) as being reserved for  and polarity. There are only two cases where the PhP expression occurs in the postfinal slot, which is reserved for clause type/object/polarity. Lack of detail in many grammars also led to difficulties in answering question number five. What can be said about the restrictions put upon PhP expressions by other categories is not much. It is, however, clear that they often occur alone, without any other tense or aspect marker. Although not restricting them, there are a few exceptions where there is a co- occurrence of both a PhP-marker and a -marker. Table 7 illustrates which PhP expressions occur with which -markers in each language. Table 8 shows the co-occurrence of PhP expressions and . Due to its size Table 8 is located on the next page. Seven out of 36  -markers occur together with a -marker. Five out of these are past tense markers. Six out of 16 -markers occur with the perfect tense.  occurs four times in all tenses and once only with the continuous tense.   shows similar tendencies with two occurrences in all tenses and one in the perfect tense.

24 Table 8: Phasal Polarity and 

e co-occurrence of PhP expressions and 

      Past 5 (0.14) 0 0 0 Perfect 0 6 (0.4) 0 1 (0.08) Continuous 1 (0.03) 0 1 (0.04) 0 Habitual 1 (0.03) 0 0 0 All tenses 0 0 4 (0.15) 2 (0.17) Total: 7 (0.19) 6 (0.4) 5 (0.19) 3 (0.25)

The answer to question number four is that the majority of the bound PhP expressions seem to function like -markers since they occur in the postinitial slot otherwise reserved for -markers. It is also very rare for all PhP expressions to occur with another -marker. This shows that they are fully incorporated inthe -system of the language. To answer research question number five: Most often, PhP expressions can not occur to- gether with other -markers, but the few exceptions show a tendency towards a specific correlation between a PhP item and a -category.

4.4 Summary In section 4.1, the East Bantuversal accessibility hierarchy for PhP was proposed to illustrate what PhP expressions are the most common in East Bantu. The most common expression is  , followed by , with  and   being far less common. The PhP systems in the samples proved to be diverse but the two most common systems contained either only   or   and . The morphological status of the markers proved to be diverse as well. The majority of the  ,  and  -markers are either prefixes or auxiliaries.  devi- ates in the sense that it was found to be an adverb as well as a prefix and an auxiliary. The bondedness of the markers indicate that the majority of  ,  and   are grammaticalized in the languages.  is found to be mostly unbound which indicates it is not grammaticalized. The majority of the bound markers occur in the postinitial slot, otherwise reserved for /polarity, in the verbal template proposed in 2.4.1. The majority of the markers also do not co-occur with any other -markers. This indicates that they are incorporated in the - system, and function in a way similar to other -markers.

25 5 Further correlations and discussion

This section is organized as follows. The results presented in the previous section andthe results of previous studies are discussed (5.1). The co-occurrence of PhP expressions and stan- dard negation (5.2.1), continuous aspect (5.2.2), perfect aspect (5.2.3) is discussed, together with other relevant findings made when comparing the data.

5.1 Crosslinguistic frequency and universals The findings in this study differ from van der Auweras Euroversal hierarchy. Bantu languages differ from European languages in respect to the crosslinguistic frequency of PhP items.   is the least common marker and not the most common as van der Auwera claims for European languages. However, concerning  there seems to be some similarities.   is the least common marker, but  is the marker in the sample that is the most deviant.  is the least bound of all four PhP items and occurs as an adverb, a category not found in any of the other expressions, except one occurrence in . In van der Auwera’s hierarchy,  and   are said to have the same crosslinguistic frequency. The results of this study differ, since   occurs in 78 percent of the sample. The number for  is much lower, at 59 percent, making   the most common PhP expression in Bantu languages by 19 percent. The crosslinguistic frequency in van Baar’s study, with  and   as the most common and  and   as the least common, is confirmed by the findings of this study. The first of van Baar’s coverage universals, mentioned in 2.2.2, is also endorsed by the results. He claims that   is commonly derived from  which is a result also found here. More on this in section 5.2.1 The second coverage universal states that ifa language has a shared expression for  and  , that languages also has the other two PhP items. There is only one instance when   is derived from , making it impossible to prove or disprove this universal.

5.2 Co-occurrence of PhP expressions and more general markers In this section the co-occurrence of the two negative PhP expressions and standard negation is investigated. Similarly, the co-occurrence of  and the continuous tense and the co- occurrence of  and the perfect is examined.

5.2.1 Co-occurrence with standard negation The co-occurence of PhP items and standerd negation is illustrated in Table 9. Due to itssizethe table is located on the next page. All cases of   co-occurs with negation. In fact,   is most often the negated form of . This is not true for  , however, since there are three cases of   that do not co-occur with negation, illustrated in example (45)-(47). Example (45) gives the structure of   in Yao.

(45) Example of   in Yao [yao] (Sanderson 1922: 187) [own glossing]: ni-gini-ni-ich-e 1.1-come- ‘’I had not yet come’

26 Table 9: Co-occurrence of negative PhP items and standard negation

Co-occurence of negative PhPs and SN

PhP Co-occurrence with SN   33   12

Yao constructs the expression with the   marker using gini. This prefix occurs between two subject markers. The interesting deviation in this construction is the final vowel. In other words, both negative and positive tenses, Sanderson describes the final vowel as a. The other two cases where the final vowel is e is in the immediate future tense and in the subjunctive mood. The future tense includes an additional prefix, but the subjunctive moodis only characterised by the change of the final vowel from a to e. This raises the question whether the subjunctive has anything to do with the PhPexpres- sion. That would mean that the counterexpectational part of   links the construction to the subjunctive, that has similar semantic content. Hypothetically, it is imaginable that this expression is derived from the subjunctive like this: ’I may come later´ meaning ’I have not yet come (but I expect to later)’. Example (46) gives the structure of   in Chopi. (46) Example of   in Chopi [cce] (Smyth 1902: 95) [own glossing]: ni-sanga-von-a 1.-see- ‘’I have not yet seen’ The prefix sanga is described briefly by Smyth and Matthews (1902: 95) as meaning  , but no further explanation is given. Standard negation in Chopi consists of a prefix, a, and a final vowel i, which does not occur in the example. In this case,   seems to be completely grammaticalized into a single prefix. However, upon examining the section in the grammar describing the perfect, a negative prefix was found: nga. This form is also used to negate the subjunctive, in both cases together with the negative prefix a. If this negative prefix is part of the marker described as meaning  , there is a number of interesting implications. It is possible that   in Chopi is derived from either the perfect or the subjunctive, with the marker sa originally meaning , a -marker, being negated by the subjunctive or perfective negative prefix nga. Again, it is possible that   has been derived from a construction originally in the subjunctive mood. Example (47) gives the structure of   in Nyakyusa-Ngonde. (47) Example of   in Nyakyusa-Ngonde [nyy] (Persohn 2017: 187): jɪ-kaalɪ ʊ-kʊ-py-a 9 15()-be(come).burnt- ‘’It’s not yet done.’ Persohn (2017) describes the marker kaalɪ as a persistive marker that can mean either  or   depending on the complement. Possible reasons why the infinitive complement

27 changes the meaning of the marker to   are not discussed. This type of construction is, however, common as noted by both Bernander (2017: 263) and Veselinova (pc). Table 10 gives the bases of derivation of  .

Table 10: Derivation of  

Bases for derivation of  

 2  0 Not derived 34

Although   usually occurs with standard negation it does not seem to be derived from any other PhP expressions, as shown in Table 10. There are only two examples of   being derived from , and no examples of it being derived from . Table 11 gives the bases of derivation of  .   does not behave like   and is for the most part derived from .

Table 11: Derivation of  

Bases for derivation of  

 9  1 Not derived 2

Comparing  and  , it soon became evident that   was most often derived from . This was the case in nine out of twelve languages. Two languages showed no similarities between the two PhP expressions, and one was derived from . Example (48)-(49) showcase the most common way this was done. (48) Example of  in Pedi [nso] (Ziervogel 1954: 70) [own glossing]: ki-sa-khov-a 1-sit- ‘’I still sit, I am still sitting’ (49) Example of   in Pedi [nso] (Ziervogel 1954: 70) [own glossing]: a-ki-sa-khov-i 1-sit-. ‘’I no longer sit’ Comparing  and   in Pedi is rather straightforward. The positive construc- tion contains the subject marker, the prefix sa and the stem. The negated version starts with the negative prefix a, and ends with a negative final vowel i. This is how all expressions of   are derived, except in Kalanga, that has a more complex structure. Examples (50)–(52) illustrate the way Kalanga expresses  and  .

28 (50) Example of  in Kalanga [kck] (Chebanne & Schmidt 2010: 101) [own glossing]: ndi-tja-lim-a 1-plough- ‘’I’m still ploughing’ There are clear similarities between example (50) and (48). Both contain the subject marker, followed by the PhP item and the verbal stem. However, unlike all other languages with a   marker, the construction in Kalanga is limited to the past and future progressive tenses. It is also a periphrastic construction, a characteristic only shared with Tswana, whose   marker is not derived from any other PhP expression. (51) Example of   in Kalanga [kck] (Chebanne & Schmidt 2010: 101) [own gloss- ing]: a-ndi-towo-be ndi-tjowo-be ndi-lim-a 1. 1. 1-plough- ‘’I shall no longer be ploughing’ [future tense] Example (51) illustrates the structure of   in the future progressive tense. This construction consists of two inflected auxiliaries and a main verb. The auxiliaries, be (to be), are both marked with the subject marker, and the first is also negated with both the negative prefix a and the future negative marker towo. The second auxiliary has the marker, tjowo, which described as a contraction of tja () and nowo (). The main verb is, again, marked for person, but with no other inflection. (52) Example of   in Kalanga [kck] (Chebanne & Schmidt 2010: 101) [own gloss- ing]: nda-ka-be ndi-si-nga-tja-lim-a 1 1-plough- ‘’I shall no longer be ploughing’ [past tense] Example (52) illustrates the structure of   in the past progressive tense. This construction differs a lot from the future progressive, as it is not negated and only has one auxiliary. The auxiliary, be, is inflected for person and marked with the past tense marker ka. The main verb repeats the subject marker, in direct conjunction to what is described as a participial marker, si. This also changes the final vowel from a, to e, the same as in the participial mood. The next marker is nga, which is described as a conditional, or eventuality marker. This marker is used to express the conditional, potential, permissive, and the optative (Chebanne & Schmidt 2010: 128). Unlike example (51), there is no negation marker in this construction. Exactly in what way the participial and conditional markers contribute to changing  to   is not explained. There is, however, another example of a negative PhP expression with no explicit negation. Compare example (52) to example (26) of   in Nyakyusa-Ngonde. This is also a periphrastic construction where the main verb is an infinitive. Although not abundant, these examples show that negation is not always needed to construct a negative PhP expression. Olsson (2013) investigates perfective grammatical markers, iamitives, in Southeast asian languges. These markers can also be used as PhP expressions. He concludes that the relation- ship between negation and iamitives is that a negated  in his data becomes  . Furthermore, when  is negated Olsson claims it becomes  . It would seem that the Southeast Asian languages use the the continuative  to derive the continuative  , and the inchoative  to derive the terminative  . As illustrated above, Bantu languages exhibit a different pattern:   is derived from  more often than from , and   is, for the most part, not derived from another PhP item. Southeast Asia and Southeast Africa seem to have different lexicalization patterns.

29 5.2.2 Co-occurrence of  and continuous aspect Table 12 shows that  coexists with the continuous aspect in 23 languages. 16 languages only have a continuous tense and four languages have  without a continuous tense. Three languages showed no evidence of either.

Table 12:  and the continuous aspect

Co-occurrence of  and continuous aspect

Both 23 Only  4 Only  16 Neither 3

Example (53) and (54) illustrate a language with both continuous aspect and grammatical- ized , namely Totela.

(53) Example of the continuous tense in Totela [ttl]Crane ( 2011: 317): ndi-li-kweesi ndi-la-yend-a 1.. 1-walk- ‘’I am walking’

Totela uses a periphrastic construction to express the continuous tense, with the auxiliary kweesi inflected for person and a present tense marker, li. The complement is marked with the subject marker and the non-completive marker la. Another, infinitive complement, can also be used to express continuous aspect.

(54) Example of  in Totela [ttl]Crane ( 2011: 325): ndì-chì-hùpúl-à 1-think- ‘’I’m still thinking’

What is interesting in Totela is that  and the continuous tense show different levels of grammaticalization. The bound marker chi () is more grammaticalized than the construc- tion used to express the continuous tense. This may indicate that the addition of a continuous tense is a more recent evolution in the language. Table 13 illustrates the bondedness of  compared to the continuous tense.  shows 12 bound constructions, six unbound and three periphrastic constructions. Two constructions are mixed and can be expressed with both an auxiliary and a bound morpheme. The contin- uous tense has a few more bound constructions, with 17, and five periphrastic constructions. Only one is unbound. Over all,  seems to be slightly less bound with its six unbound morphemes, but overall the difference is small.

30 Table 13: Bondedness of  and the continuous tense

Bondedness of  and the continuous tense

 Cont. Bound 14 17 Periphrastic 5 5 Unbound 6 1 Mixed 2 0

5.2.3 Co-occurrence of  and perfect aspect Table 14 shows the co-occurrence of  and the perfect tense. As  is one of the least common of the PhP expressions it is no surprise that the largest group of languages are the ones with only the perfect tense, with 24 occurrences. The majority of the languages that have  also have the perfect tense, in fact, twelve out of sixteen, or 75 percent of the languages show this tendency. There are also four languages that only have , and six that have neither  nor the perfect tense.

Table 14: Co-occurrence of  and perfect aspect

Co-occurrence of  and perfect aspect

Both 12 Only  4 Only  24 Neither 6 Total: 46

Upon closer examination, the languages with both an -marker and the perfect tense show that  is currently undergoing grammaticalization as described in section 2.3.4. In these languages there is a division into three categories based on the bondedness of the marker. First, there are the free forms and auxiliaries, then there are the bound markers that co-occur with the perfect tense, and finally there are the bound markers which occur on their own. The examples below show  in different stages of grammaticalization. From these examples it is clear how  is becoming more grammaticalized in some languages. (55) Example of  in Tswana [tsn] (Cole 1962: 130) [own glossing]: kè-sétsè ké-kwàl-á 1 1-write- ‘’I am already writing’

31 Example (55) illustrates  as an auxiliary inflected for person, together with the main verb that also has a subject marker.

(56) Example of  in Zimbabwean Ndebele [nde] (Bowern et al. 2002: 51): se-si-dl-ile 1-eat- ‘’We’ve already eaten’

Example (56) is particularly interesting because of the following: First,  occurs together with the perfect in a bound construction and does not seem to occur in any other tenses. This is fully possible since  and the perfect have similar semantic properties, causing an overlap (Dahl & Wälchli 2016). Second, Bowern et al. (2002) explain that  was originally preceded by another subject marker, making the total of subject markers two as in example (55). The initial subject marker has now been deleted.

(57) Example of  in Zulu [zul] (Canonici 1996: 103) [own glossing and translation]: se-ngi-thand-a 1-love- ‘’I already love’

Example (57) also has an initial -marker, just like the two previous examples. Canonici (1996) explains that this marker was originally an auxiliary that had its own sub- ject marker, but is now a prefix. He also mentions that it can co-occur with the perfect, but that it is not obligatory. Based on these examples the following grammaticalization pattern can be hypothesised:

1.  is an unbound marker. 2.  becomes an auxiliary (example (55)). 3.  becomes bound, retaining both subject markers from the time it was an auxil- iary. 4.  loses the first subject marker, as it is now redundant (example (56)). 5. An intermediary stage in the semantic development could cause  to co-occur with the perfect (example (56)). 6. The co-occurence of  and the perfect becomes optional (example (57)). 7.  becomes a bound marker that occurs as an initial prefix without any other -marker.

It becomes apparent from this hypothesis that the languages in the sample, and especially those in the example, are in different stages of this development. The example languages even share a similar -marker: se or sétsè. Other auxiliaries with the same meaning is the Sumayela Ndebele sele and se in Southern Sotho (example (23)). It seems possible that an increased usage and/or a development from an auxiliary to a bound form could reduce sétsè and sele to se, as it is the more common form, meaning this could have already happened in the other languages.

5.3 A small scale historical comparative analysis This section discusses the similarities found in the clusters. First, some general observations are made and later on the two most common markers,   and , are discussed. Only

32 Table 15: Clusters

Distribution of PhP expressions in the clusters

Cluster No. of lgs       No. of PhPs Botatwe 5 5 1 3 1 10 Sabi 4 2 1 2 0 4 Shona 5 5 0 3 1 9 Sotho-Tswana 4 4 3 4 3 14 Rufiji-Ruvuma 5 4 1 1 0 6 Nguni-Tsonga 7 6 5 6 4 21 N. M. Bantu 3 3 2 1 1 7 T-S-N 3 1 0 1 0 2 Total: 36 30 13 21 10

the clusters with three or more languages were used in the analysis. This means that 10 lan- guages were excluded. Table 15 gives an overview of the PhP items found in the clusters. The clusters are: Bo- tatwe, Sabi, Shona, Sotho-Tswana, Rufiji-Ruvuma, Nguni-Tsonga, North Mozambique Bantu and Tumbuka-Sena-Nyanja. Nguni-Tsonga and Sotho-Tswana are the two clusters with the highest number of complete systems, and the overall highest concentration of PhP items with 21 and 14 items each. It seems like the southern part of the sample, below central Zimbabwe, is where PhP expressions are most common. This is especially true if Chopi and Venda are added, since they also have complete PhP systems and are geographically located close to the two other branches. The more northern part of the sample (Botatwe, Sabi, Shona) and the more north-eastern part (Rufiji-Ruvuma, North Mozambique Bantu) all have a lower frequency of PhP, but still have a high frequency of  . It seems like   is favoured in these parts of the sample. Table 16 illustrates the three clusters where  has a similar form. A total of 16 out of 27 expressions share these similarities, adding up to 59 percent of the sample. The com- mon denominator seems to be a , usually s, followed by the vowel a. Two of the branches are more closely related, raising the possibility of this being a genealogical feature. Both Nguni-Tsonga and Sotho-Tswana are part of the Southern Bantu-Makua sub-family, but Shona branches of earlier. Shona, however, is geographically located just north of the other two clusters, so there is a possibility of contact-related spread of this form of . Table 17 illustrates the three branches in which   share a similar form. The com- mon denominator here seems to be na. In total, 11 of 36 items, or 31 percent, share a common form; a much lower number than for . Two of these clusters are not as closely related as Sotho-Tswana and Nguni-Tsonga either. Botatwe and Rufji-Ruvuma branch off directly under

33 Table 16: Proto-

Proto-

Cluster Language  Nguni-Tsonga Sumayela Ndebele sa Nguni-Tsonga Swati sa Nguni-Tsonga Tsonga ha Nguni-Tsonga Tswa ha Nguni-Tsonga Xhosa sa Nguni-Tsonga Zimbabwean Ndebele - Nguni-Tsonga Zulu sa Shona Kalanga tja Shona Manyika - Shona Shona - Shona Tawara sa Shona Tewe cha Sotho-Tswana Kgalagadi sha Sotho-Tswana Pedi sa Sotho-Tswana Southern Sotho sa Sotho-Tswana Tswana sa Total: 16

the East Bantu sub-family, while North Mozambique Bantu is part of the Southern Bantu- Makua branch. Even though they are much less closely related, they still share this common marker. This could be because they are fairly close together geograpically. The Rufiji-Ruvuma and North Mozambique Bantu clusters are located along the coast of Tanzania and Mozam- bique. Botatwe is mainly spoken in Zambia, very far away from the two other clusters, and is therefore the odd one out. In both Botatwe and North Mozambique Bantu   is mainly bound, while in Rufiji-Ruvuma the expression is bound in half of the languages and used in a periphrastic construction in half. This suggests that the form na is newer in the Rufiji-Ruvuma cluster than in the other clusters. A historical comparative analysis relies mainly on sound changes. Without more diachronic data, any historical comparative analysis will be difficult. Nonetheless, looking at the fren- quencies of the markers it is possible to hypothesise a proto-form of both  and  , although not on the Proto-Bantu level. The most frequent form of  in the Nguni-Tsonga, Sotho-Tswana and Shona branches is sa. The most frequent form of   in the Botatwe,

34 Table 17: Proto- 

Proto- 

Cluster Language   Botatwe Ila na Botatwe Lenje na Botatwe Soli na Botatwe Tonga na/ninga Botatwe Totela ini N. Mozambique Bantu Chuwabu na N. Mozambique Bantu Koti na N. Mozambique Bantu Makhuwa na Rufiji-Ruvuma Makonde ka-naa/ka-naamba Rufiji-Ruvuma Makwe naamba Rufiji-Ruvuma Matumbi na Rufiji-Ruvuma Ndengereko - Rufiji-Ruvuma Yao kana Total: 13

North Mozambique Bantu and Rufiji-Ruvuma branches is na. Therefore the proto-form of  could be sa, and for   it could be na or naamba. As stated in 2.3.4. the gram- maticalization process can result in phonetic reduction. Therefore the longest form, naamba, might also be the oldest. But without futher documentaion, it is difficult to draw any certain conclusions. In any case, both  and   appear to be inherited rather than borrowed features in these clusters.

6 Conclusions

A summary of the answers to the research questions, the characteristics of PhP expressions and Bantu-specific tendencies, as well as recommendations for further research is presented in this section.

The answers to the research questions are:

1. What form do the PhP expressions have?

35 The forms of the PhP expressions vary. For a description of each PhP item, see appendices 1–4.

2. Which PhPs are available and in what languages?

The PhP systems found in the sample are diverse. Six languages have a systemcon- taining all PhP expressions and five have none. Eleven languages have systems contain- ing different combinations of three items. The most common three-item systems are  // and  // . Another eleven languages have systems with two items. The most common two item system is  /. Thirteen languages only have one PhP item, of which twelve are  .

3. What is the morphological status of the markers?

The morphological status of the markers was not mentioned in most grammars. However, in the cases where this was mentioned,  ,  and   were most commonly found to be prefixes or auxiliaries.  was found to be a prefix or an auxiliary as often as an adverb.

4. What function do the PhP expressions have?

The bound PhP items show tendencies of being incorporated inthe -system of their re- spective languages, hence functioning like other -markers. No other functions were found.

5. Do the markers occur under any restrictions?

The majority of the PhP expressions did not occur with any other -markers, indicating that they might be incompatible. No other restrictions were found.

  was the most common marker found in the sample. It is most often a prefix or an auxiliary and was found to be a bound marker in half of the occurrences. The bound  -markers occurred in the postinitial slot in the verbal template.   was not found to be derived from another PhP expression.  was the second most common marker found in the sample. It is most often a prefix or an auxiliary and was found to be a bound marker in more than half of the occurrences. The bound -markers occurred, like  , in the postinitial slot in the verbal template.  was not commonly found in the sample. This marker was found to have varied morphological status and was often a free form. Therefore the verbal template was inapplicable in the majority of cases. A third of the -markers co-occurred with the perfect.   was not commonly found in the sample. The majority of the  - markers were found to be derived from . Therefore, they share the same characteristics. Three tendencies indicate that PhP expressions are a stable category in the East-Bantu family. First, the languages in the sample show a very strong tendency to incorporate   and to a certain degree  in the -system. Second, there is also a probability that   and  are diachronically stable in specific clusters. Third, the languages show a tendency to grammaticalize PhP expressions. To better describe the semantics and usage of PhP expressions, future studies would ben- efit from using a questionnaire like van Baar (1997). This method would avoid two problems

36 associated with grammars. It would increase the chance of finding complete systems, as the questionnaire would be focused solely on PhP expressions, unlike grammars. This method would also provide data that reflects the current usage of PhP items in the target language, as opposed to the many grammars used in this study that are out-dated.

37 References

Andrews, J.R. 2003. Introduction to Classical Nahuatl. University of Oklahoma Press.

Bernander, R. 2017. Grammar and Grammaticalization in Manda: An Analysis of the Wider TAM Domain in a Tanzanian Bantu Language. Department of Languages and Literatures, University of Gothenburg.

Botne, R.D.O., & Schafer, L. 2008. e Chindali Language of Malawi: A grammatical sketch of Cindali, Malawian variety. Lightning Rod Press. American Philosophical Society.

Bowern, C., Lotridge, V., & Alling, E. 2002. Ndebele. Languages of the world: Materials. Lincom Europa.

Canonici, N.N. 1996. Zulu Grammatical Structure. and Literature, University of Natal.

Carter, H., & Kahari, G.P. 1979. Kuverenga Chishóna: An Introductory Shona Reader with Gram- matical Sketch. School of Oriental and African Studies, University of London.

Carter, H., & Kahari, G.P. 1986. Course. Shona Language Course, nos. v. 1–3. African Studies Program, University of Wisconsin.

Carter, H., & Kashoki, M.E. 2002. An outline of Chitonga grammar. Bookworld Publishers.

Chebanne, A.M., & Schmidt, D. 2010. Kalanga: Summary Grammar. CASAS book series. Centre for Advanced Studies of African Society (CASAS).

Cole, T., Mpho Mokaila D. 1962. A course in Tswana.

Comrie, Bernard, Haspelmath, Martin, & Bickel, Balthasar. 2008. e Leipzig glossing rules: Conventions for interlinear morpheme-by-morpheme glosses.

Crane, T.M. 2011. Beyond Time: Temporal and Extra-temporal Functions of Tense and Aspect Marking in Totela, a Bantu Language of Zambia. Ph.D. thesis, University of California.

Croft, W. 2002. Typology and Universals. Cambridge Textbooks in Linguistics. Cambridge University Press.

Dahl, Östen, & Wälchli, Bernhard. 2016. Perfects and iamitives : two gram types in one gram- matical space. Letras de Hoje, 51(3), 325–348.

Dembetembe, N.C. 1987. A Linguistic Study of the Verb in Korekore. Supplement to Zambezia. University of Zimbabwe, Publications Office.

Devos, M. 2008. A grammar of Makwe: (Palma; Mozambique). LINCOM studies in African linguistics. Lincom Europa.

Doke, C.M. 1938. Text Book of Lamba Grammar. Witwatersrand University Press.

Doke, C.M., & Mofokeng, S.M. 1957. Textbook of Southern Sotho grammar. SOLINET/ASERL Cooperative Microfilming Project. Longmans, Green.

Floor, S. 2010. Mwani grammatical sketch. SIL Moçambique.

38 Gadelii, K.E. A grammatical sketch of Xitshwa. August, 1998.

Garrido, J. 1992. Expectations in spanish and german adverbs of change. 26(01), 357–402.

Givón, T. 1971. Historical Syntax and Synchronic Morphology: An Archeologist’s Field Trip. Vol. 7. The Chicago Linguistic Society. Pages 394–415.

Grimes, B.F. (ed). 2000. Languages of the world: ethnologue. SIL International.

Grollemund, Rebecca, Branford, Simon, Bostoen, Koen, Meade, Andrew, Venditti, Chris, & Pagel, Mark. 2015. Bantu expansion shows that habitat alters the route and pace of human dispersals. Proceedings of the National Academy of Sciences.

Guerois, Rozenn. 2015. A grammar of Cuwabo (Bantu P34, Mozambique). Ph.D. thesis. 2015LYO20032.

Güldeman, Tom. 1999. The Gensis of Verbal Negation in Bantu and its Dependency onFunc- tional Features of Clause Types. Pages 545–587 of: Hombert, & Hyman (eds), Bantu Historical Linguistics. Calif: CSLI Publications.

Heine, Bernd. 2011. e Oxford Handbook of Grammaticalization. Oxford University Press. Chap. Grammaticalization in African languages.

Heine, Bernd, Claudi, Ulrike, & Hünnemeyer, Friederike. 1991. Grammaticalization : a concep- tual framework. Chicago: University of Chicago Press.

Hoch, E. n.d.. Bemba grammar notes for beginners. n.p.

Kirsch, B., Skorge, S., & Magona, S. 1999. Xhosa. Teach yourself books. NTC Publishing Group.

Kiso, A. 2012. Tense and Aspect in Chichewa, Citumbuka and Cisena: A Description and Com- parison of the Tense-aspect Systems in ree Southeastern Bantu Languages. Ph.D. thesis, Department of Linguistics, Stockholm University.

Kraal, P.J. 2005. A Grammar of Makonde (Chinnima, Tanzania). Universiteit Leiden.

Kramer, Raija. Position paper on Phasal Polarity expressions. https://www.aai.uni- hamburg.de/afrika/php2018/medien/position-paper-on-php.pdf.

Lanham, L.W. 1955. A study of giTonga of Inhambane. Bantu linguistic studies. Witwatersrand University Press.

Löbner, Sebastian. 1989. ’Schon – erst – noch’: An integrated analysis. Linguistics and Philos- ophy, 12(2), 167–212.

Lukusa, S.T.M., & Monaka, K.C. 2008. Shekgalagari grammar: a descriptive analysis of the language and its vocabulary. CASAS book series. Centre for Advanced Studies of African Society (CASAS).

Madan, A.C. 1906. Wisa Handbook: A Short Introduction to the Wisa Dialect of North-East Rhodesia. Clarendon Press.

Madan, A.C. 1908. Lenje Handbook: A Short Introduction to the Lenje Dialect Spoken in North- west Rhodesia. Clarendon Press.

39 Maganga, C., & Schadeberg, T.C. 1992. Kinyamwezi: Grammar, Texts, Vocabulary. East African languages and dialects. R. Köppe.

Maho, Jouni. 2003. A classification of the Bantu languages: an update of Guthrie’s referential system. In: Nurse, & Philipson (eds), e Bantu languages. Routledge.

Miti, L.M. 2001. A Linguistic Analysis of Cinsenga: A Bantu Language Spoken in Zambia and Malawi. CASAS book series. Centre for Advanced Studies of African Society (CASAS).

Morrison, M. E. 2011. A Reference Grammar of Bena. Ph.D. thesis, Rice University.

Ngonyani, D.S. 2003. A Grammar of Chingoni. Languages of the world / Materials: Materials. Lincom.

Nurse, D. 2003. Aspect and tense in Bantu languages. In: Nurse, D., Philippson G. (ed), e Bantu Languages. Routledge.

Nurse, Derek, & Philipson, Gérard. 2014. e Bantu languages. Routledge.

Odden, D.A. 1996. e Phonology and Morphology of Kimatuumbi. Phonology of the world’s languages. Clarendon Press.

Olsson, Bruno. 2013. Iamitives : Perfects in Southeast Asia and beyond. M.Phil. thesis, Stockholm University, General Linguistics.

Ouwehand, M. 1978. Everyday Tsonga. Sasavona publishers.

Paroz, R.A. 1946. Elements of Southern Sotho. Basutoland, Morija Sesuto Book Depot.

Persohn, B. 2017. e verb in Nyakyusa: A focus on tense, aspect and modality. Contemporary African Linguistics. Language Science Press.

Plungian, Vladimir A. 1999. A typology of phasal meanings. Pages 311–321 of: Abraham, & Kulikov (eds), Tense-aspect, transitivity and causativity: Essays in honour of Vladimir 24 Nedjalkov. John Benjamins Publishing.

Poulos, G. 1990. A Linguistic Analysis of Venda. Via Afrika Limited.

Sanderson, M. 1922. A Yao Grammar. Society for promoting Christian knowledge.

Schadeberg, T.C., & Mucanheia, F.U. 2000. Ekoti: e Maka Or of Angoche. East African languages and dialects. Köppe.

Smith, E. W. 1907. A handbook of the . Oxford University Press.

Smyth, B., Matthews J. 1902. A vocabulary with a short grammar of Xilenge. Society For Promoting Christian Knowledge.

Stevick, W. 1960. Manyika Step-by Step. Central Mission Press.

Ström, Eva-Marie. 2013. e Ndengeleko language of Tanzania. Ph.D. thesis, University of Gothenburg.

Traugott, Elisabeth Closs. 2011. e Oxford Handbook of Grammaticalization. Oxford Univer- sity Press. Chap. Grammaticalization and the mechanisms of change.

40 van Baar, Theodorus Martinus. 1997. Eurotyp: 3: Phasal Polarity. Amsterdam: IFOTT. van der Auwera, J. 1998. Phasal Adverbials in the Languages of Europe. Mouton de Gruyter. Pages 25–145. van der Merwe, D. F., Schapera I. 1943. A comparative study of Kgalagadi, Kwena, and other Sotho dialects. University of Cape Town.

Van der Wal, J. 2009. Word order and information structure in Makhuwa-Enahara. LOT inter- national series. LOT. van Eeden, B.J.C. 1936. Notes on the Grammar of Soli. Nasionale Pers, Beperk.

Wright, J.L. 2007. An Outline of Kikaonde Grammar. Language in Zambia. Bookworld Pub- lishers and UNZA Press.

Ziervogel, D. 1952. A Grammar of Swazi (siSwati):. Bantu grammatical archives, 3. Witwater- srand University Press.

Ziervogel, D. 1954. e Eastern Sotho: a tribal, historical, and linguistic survey (with ethno- graphic notes) of the Pai, Kutswe, and Pulana Bantu tribes in the Pilgrim’s Rest District of the Transvaal Province, Union of South Africa. SOLINET/ASERL Cooperative Microfilming Project. J.L. Van Schaik.

Ziervogel, D. 1959. A Grammar of Northern Transvaal Ndebele, Etc. Pretoria.

41 A Appendix 1

  Language ISO-code Form of   Source Bemba bem ta-?-la-?-a Hoch (n.d.: 21) Bena (Tanzania) bez Morrison (2011) Chopi cce SM-sanga-STEM-FV Smyth (1902: 39) Chuwabu chw NEG.SM-ná-STEM Guerois (2015: 383) Gitonga toh NEG-SM-si-INF-STEM-NEG Lanham (1955: 173) Ila ilb NEG.SM na INF STEM-FV Smith (1907: 165) Kalanga kck a-SM-thu + verb in the past Chebanne & Schmidt consecutive (2010: 107) Kaonde kqn Wright (2007) Kgalagadi xkv NEG ntse SM STEM-NEG van der Merwe (1943: 71) Koti eko NEG-SM-ná-STEM-e Schadeberg & Mucan- heia (2000: 122) Lala-Bisa leb Madan (1906) Lamba lam NEG.SM-nga-STEM-FV Doke (1938: 284) NEG-SM-nga-STEM-FV Lenje leh NEG-SM-ná-STEM Madan (1908: 38) Makhuwa vmw NEG.SM-ná-STEM-e Van der Wal (2009: 106) Makonde kde SM-ka-nàa(va) INF-STEM- Kraal (2005: 242) FV SM-ka-nàamba INF- STEM-FV Makwe ymk NEG.SM-náamba INF.cl15- Devos (2008: 341) STEM-FV Manyika mxc NEG-SM-PST.NEG-ti + verb Stevick (1960: 171) in the recent past participle Matumbi mgw SM-ná-STEM-FV NEG Odden (1996: 66) Mwani wmw NEG-námba INF-STEM-FV Floor (2010: 15) Ndali ndh Botne & Schafer (2008) Ndengereko ndg Ström (2013) Ngoni ngo NEG.SM-kona NEG-INF- Ngonyani (2003: 87) STEM Nsenga nse Miti (2001) Nyakyusa-Ngonde nyy SM-kaali AUG-INF-STEM- Persohn (2017: 187) FV

42   continued Language ISO-code Form of   Source Nyamwezi nym Maganga & Schadeberg (1992) Nyanja nya NEG-SM-RECPST-STEM- Kiso (2012: 157) SBJV-be Pedi nso NEG-SM-ki SM-STEM-FV Ziervogel (1954) Sena (Malawi) swk Kiso (2012) Shona sna NEG-SM-satí Carter & Kahari (1986: 37) Soli sby PST-NEG.PST-na-SM-STEM van Eeden (1936: 25) PST-SM-NEG-na-SM-STEM Southern Sotho sot NEG-SM-es’o-STEM-FV Doke & Mofokeng (1957: 213) Sumayela Ndebele nbl Ziervogel (1959) Swati ssw SM-se SM-NEG-STEM-FV Ziervogel (1952: 133) Tawara twl NEG-SM-cha-AUX SM-INF- Dembetembe (1987: STEM-FV 101) Tewe twx NEG-sa-ti + verb in the past Carter & Kahari (1979: participle 41) Tonga toi NEG-SM-níngá-STEM-FV Carter & Kashoki NEG-SM-ná-STEM-FV (2002: 61) Totela ttl NEG-SM-ini INF-STEM-FV Crane (2011: 337) Tsonga tso NEG SM si STEM-FV Ouwehand (1978: 108) Tswa tsc NEG SM.CONT ha STEM- Gadelii (n.d.: 22) NEG Tswana tsn NEG-SM-ísé + verb in the ha- Cole (1962: 130) bitual subjunctive Tumbuka tum Kiso (2012) Venda ven NEG-SM-athu-STEM-FV Poulos (1990: 344) Xhosa xho NEG-SM-ka-STEM-FV Kirsch et al. (1999: 205) Yao yao SM-kana-SM-STEM-e Sanderson (1922: 56) ngana-SM-STEM-FV Zimbabwean Ndebele nde NEG-SM-zake SM-STEM- Bowern et al. (2002: 45) SUBJ Zulu zul NEG-SM-ka-STEM-NEG Canonici (1996: 105)

43 B Appendix 2

 Language ISO-code Form of  Source Bemba bem apo pene Hoch(n.d.: 51) Bena (Tanzania) bez Morrison (2011) Chopi cce se-SM-STEM-PRF Smyth (1902: 38) Chuwabu chw já SM-PRF-STEM Guérios (2015: 363) Gitonga toh Lanham (1955 ) Ila ilb kle Smith (1907: 213 ) Kalanga kck Chebanne & Schmidt (2010) Kaonde kqn kala Wright (2007: 31) Kgalagadi xkv la sala v.d Merwe and Schapera (1943: 84) Koti eko SM-áz INF-STEM-FV Schadeberg (2000: 147) Lala-Bisa leb Lamba lam Doke (1938) Lenje leh Madan (1908) Makhuwa vmw van der Wal (2009) Makonde kde Kraal (2005) Makwe ymk Devos (2008) Manyika mxc Stevick and Machiwana (1960) Matumbi mgw Odden (1996) Mwani wmw Floor (2010) Ndali ndh Botne (2008) Ndengereko ndg pyaa Ström (2013: 251) Ngoni ngo Ngonyani (2003) Nsenga nse Miti (2001) Nyakyusa-Ngonde nyy Persohn (2017) Nyamwezi nym Maganga and Schadeberg (1992) Nyanja nya Kiso (2012) Pedi nso Ziervogel (1954) Sena (Malawi) swk Kiso (2012) Shona sna Carter (1986) Soli sby van Eeden (1936) Southern Sotho sot SM se SM STEM-a Paroz (1946: 95) Sumayela Ndebele nbl sele, biyo Sekhosana (2009: 330)

44  continued Swati ssw (SM)-se SM-STEM-FV Ziervogel (1952: 132 ) Tawara twl Dembetembe (1987) Tewe twx Carter and Ka- hari (1979) Tonga (Zambia) toi Carter (2002) Totela ttl Crane (2011) Tsonga tso Ouwehand (1965) Tswa tsc Gadelii (1999) Tswana tsn SM-sétsè verb in the present participle Cole and Mokaila (1962: 130) Tumbuka tum Kiso (2012: 235- 236) Venda ven SM-PRF-no-STEM-FV Poulos (1990: 342) Xhosa xho sele/sel/se Kirsch and Scorge (1999: 123) Yao yao Sanderson: 1922 Zimbabwean Ndebele nde se-SM-STEM-PERF Bowern et al. (2002 : 51) Zulu zul se-SM-STEM-a Canonici (1995: 95)

45 C Appendix 3

 Language ISO code Form of  Source Bemba bem ci Hoch(n.d.: 29) Bena (Tanzania) bez SM-pí-STEM-FV pele SM-i- Morrison (2011: 271- STEM-FV 272) Chopi cce SM-NEG-di SM-CONS- Smyth (1902: 38) STEM Chuwabu chw naváno Guérios (2015: 323) Gitonga toh Lanham (1955 ) Ila ilb SM chi STEM Smith (1907: 140, 148) Kalanga kck SM-tja-STEM-a Chebanne and Schmidt (2010: 106) Kaonde kqn SM-(ka)-ki-STEM Wright (2007: 32) Kgalagadi xkv SM sha STEM-CONT Lukusa & Monaka (2008: 145) Koti eko Schadeberg (2000 ) Lala-Bisa leb Madan (1906) Lamba lam SM-ci-STEM Doke (1938: 269) Lenje leh SM-chi-STEM SM-achi- Madan (1908: 35) STEM Makhuwa vmw van der Wal (2009) Makonde kde Kraal (2005) Makwe ymk Devos (2008) Manyika mxc Stevick and Machi- wana (1960) Matumbi mgw Odden (1996) Mwani wmw SM-ingari INF-STEM-FV Floor (2010: 15) Ndali ndh Botne (2008) Ndengereko ndg Ström (2013) Ngoni ngo Ngonyani (2003) Nsenga nse Miti (2001) Nyakyusa-Ngonde nyy SM-kaali SM-PRS-STEM-FV Persohn (2017: 184) Nyamwezi nym Maganga and Schade- berg (1992) Nyanja nya SM-PST.PFV-STEM-FV-be Kiso (2012: 150) Pedi nso Ziervogel (1954)

46  continued Language ISO code Form of  Source Sena (Malawi) swk Kiso (2012) Shona sna Carter (1986) Soli sby van Eeden (1936) Southern Sotho sot SM sa STEM-a Paroz (1946: 96) Sumayela Ndebele nbl SM-sá-STEM-FV Ziervogel (1959: 88) Swati ssw SM-sa-STEM-FV Ziervogel (1952: 97) Tawara twl SM-sa-STEM-FV Dembetembe (1987: 133-134) Tewe twx SM-cha-STEM-a Carter and Kahari (1979: 28) Tonga (Zambia) toi Carter (2002) Totela ttl SM-chi-STEM-FV Crane (2011: 325) Tsonga tso SM ha TM STEM Ouwehand (1965: 107- 108) Tswa tsc SM ha STEM Gadelii (1999: 22) Tswana tsn SM-sà-STEM. SM-sántse SM- Cole and Mokaila STEM-PART (1962: 121) Tumbuka tum Kiso (2012) Venda ven SM-kha-di-STEM-FV Poulos (1990: 334) Xhosa xho SM-sa-STEM Kirsch and Scorge (1999: 37-38) Yao yao SM-AUX ?-FUT-STEM- Sanderson (1922: 88- FUT.FV 89) Zimbabwean Ndebele nde Bowern et al. (2002) Zulu zul SM-sa-(OM)-STEM-a Canonici (1995: 101)

47 D Appendix 4

  Language ISO code Form of   Source Bemba bem Hoch(n.d) Bena (Tanzania) bez Morrison (2011) Chopi cce Smyth (1902) Chuwabu chw já NEG-SM-STEM-PRF Guérios (2015: 363) Gitonga toh Lanham (1955) Ila ilb Smith (1907) Kalanga kck 1. NEG-SM-NEG.FUT-AUX Chebanne and Schmidt SM-STILL.FUT-AUX SM- (2010: 111, 120) STEM-FV 2. PST-SM-PRF- PST-AUX SM-PART-COND- STILL-STEM-PART.FV Kaonde kqn Wright (2007) Kgalagadi xkv v.d Merwe and Schapera (1943) Koti eko Schadeberg (2000) Lala-Bisa leb Madan (1906) Lamba lam Doke (1938) Lenje leh Madan (1908) Makhuwa vmw van der Wal (2009) Makonde kde Kraal (2005) Makwe ymk Devos (2008) Manyika mxc Stevick and Machi- wana (1960) Matumbi mgw Odden (1996) Mwani wmw Floor (2010) Ndali ndh Botne (2008) Ndengereko ndg Ström (2013) Ngoni ngo Ngonyani (2003) Nsenga nse Miti (2001) Nyakyusa-Ngonde nyy Persohn (2017) Nyamwezi nym Maganga and Schade- berg (1992) Nyanja nya Kiso (2012) Pedi nso NEG-SM-sa-STEM-NEG.FV Ziervogel (1954: 70)

48   continued Language ISO code Form of   Source Sena (Malawi) swk Kiso (2012) Shona sna Carter (1986) Soli sby van Eeden (1936) Southern Sotho sot NEG SM sa STEM-a Paroz (1946: 9)5 Sumayela Ndebele nbl NEG-SM-sá-STEM-NEG.FV Ziervogel (1959: 88) Swati ssw NEG-SM-sa-STEM-NEG.FV Ziervogel (1952: 102 ) Tawara twl Dembetembe (1987 ) Tewe twx NEG-SM-cha-STEM-FV Carter and Kahari (1979: 28) Tonga (Zambia) toi Carter (2002) Totela ttl NEG-SM-chi-(TM)-STEM- Crane (2011: 327 ) FV Tsonga tso NEG SM ha TM STEM- Ouwehand (1965: 107- NEG.FV 108) Tswa tsc Gadelii (1999) Tswana tsn NEG-SM-tlhòlé SM-STEM- Cole and Mokaila PART (1962: 130) Tumbuka tum Kiso (2012) Venda ven NEG-SM-tsha-STEM-FV Poulos (1990: 338) Xhosa xho sele/sel/se Kirsch and Scorge (1999: 123) Yao yao Sanderson (1922) Zimbabwean Ndebele nde Bowern et al. (2002) Zulu zul NEG-SM-sa-(OM)-STEM- Canonici (1995: 101) NEG.FV

49 E Appendix 5

PhP Systems Language ISO-code       Bemba bem Yes Yes Yes - Bena (Tanzania) bez - - Yes - Chopi cce Yes Yes Yes - Chuwabu chw Yes Yes Yes Yes Gitonga toh - - - Ila ilb Yes Yes Yes - Kalanga kck Yes - Yes Yes Kaonde kqn - Yes Yes - Kgalagadi xkv Yes Yes Yes - Koti eko Yes Yes - - Lala-Bisa leb - - - - Lamba lam Yes - Yes - Lenje leh Yes - Yes - Makhuwa vmw Yes - - - Makonde kde Yes - - - Makwe ymk Yes - - - Manyika mxc Yes - - - Matumbi mgw Yes - - - Mwani wmw Yes - Yes - Ndali ndh - - - - Ndengereko ndg - Yes - - Ngoni ngo Yes - - - Nsenga nse - - - - Nyakyusa-Ngonde nyy Yes - Yes - Nyamwezi nym - - - - Nyanja nya Yes - Yes - Pedi nso Yes - Yes Yes

50 PhP Systems continued Language ISO-code       Sena (Malawi) swk - - - - Shona sna Yes - - - Soli sby Yes - - - Southern Sotho sot Yes Yes Yes Yes Sumayela Ndebele nbl - Yes Yes Yes Swati ssw Yes Yes Yes Yes Tawara twl Yes - Yes - Tewe twx Yes - Yes Yes Tonga (Zambia) toi Yes - - - Totela ttl Yes - Yes Yes Tsonga tso Yes - Yes Yes Tswa tsc Yes - Yes - Tswana tsn Yes Yes Yes Yes Tumbuka tum - - - - Venda ven Yes Yes Yes Yes Xhosa xho Yes Yes Yes - Yao yao Yes - Yes - Zimbabwean Ndebele nde Yes Yes - - Zulu zul Yes Yes Yes Yes

51 F Appendix 6

Clusters Language ISO-code Country Cluster Bemba bem Zambia Sabi Lala-Bisa bez Zambia Sabi Lamba cce Zambia Sabi Nsenga chw Mozambique Sabi Ila toh Zimbabwe Botatwe Lenje ilb Zambia Botatwe Soli kck Zambia Botatwe Tonga (Zambia) kqn Zambia Botatwe Totela xkv Zambia Botatwe Kalanga eko Botswana Shona Manyika leb Zimbabwe Shona Shona lam Zimbabwe Shona Tawara leh Mozambique Shona Tewe vmw Mozambique Shona Kgalagadi kde Botswana Sotho-Tswana Pedi ymk South Africa Sotho-Tswana Southern Sotho mxc South Africa Sotho-Tswana Tswana mgw Botswana Sotho-Tswana Makonde wmw Tanzania/Mozambique Rufiji-Ruvuma Makwe ndh Tanzania Rufiji-Ruvuma Matumbi ndg Tanzania Rufiji-Ruvuma Ndengereko ngo Tanzania Rufiji-Ruvuma Yao nse Mozambique Rufiji-Ruvuma Sumayela Ndebele nyy South Africa Nguni-Tsonga Swati nym Swaziland Nguni-Tsonga Tsonga nya Mozambique Nguni-Tsonga Tswa nso Mozambique Nguni-Tsonga Xhosa swk South Africa Nguni-Tsonga Zimbabwean Ndebele sna Zimbabwe Nguni-Tsonga Zulu sby South Africa Nguni-Tsonga

52 Clusters continued Language ISO-code Country Cluster Chuwabu sot Mozambique North Mozambique Bantu Koti nbl Mozambique North Mozambique Bantu Makhuwa ssw Mozambique North Mozambique Bantu Ndali twl Malawi Corridor Bantu Nyakyusa-Ngonde twx Malawi Corridor Bantu Tumbuka toi Malawi Tumbuka-Sena-Nyanja Nyanja ttl Mozambique Tumbuka-Sena-Nyanja Sena (Malawi) tso Malawi Tumbuka-Sena-Nyanja Chopi tsc Mozambique Southern Bantu-Makua Gitonga tsn Mozambique Southern Bantu-Makua Bena (Tanzania) tum Tanzania Southern Tanzania Highlands Bantu Ngoni ven Mozambique Southern Tanzania Highlands Bantu Kaonde xho Zambia Central Western Bantu, Luban Mwani yao Mozambique Coastal NEC Bntu Nyamwezi nde Tanzania Sukuma-Nyamwezi Venda zul Zimbabwe Sotho-Makua-Venda

53 Stockholms universitet/Stockholm University SE-106 91 Stockholm Telefon 08 - 16 20 00 www.su.se