Citation
Topics in Sesotho Control Verbs

Material Information

Title:
Topics in Sesotho Control Verbs
Creator:
Smouse, Mantoa
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (125 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Linguistics
Committee Chair:
McLaughlin, Fiona
Committee Members:
Potsdam, Eric H.
Kane, Abdoulaye
Matondo, Masangu
Graduation Date:
8/7/2010

Subjects

Subjects / Keywords:
Applicative theories ( jstor )
Infinitives ( jstor )
Linguistic complements ( jstor )
Linguistics ( jstor )
Morphemes ( jstor )
Nouns ( jstor )
Sentences ( jstor )
Subjunctive mood ( jstor )
Syntactics ( jstor )
Verbs ( jstor )
Linguistics -- Dissertations, Academic -- UF
control, morpho, referential, verb
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Linguistics thesis, Ph.D.

Notes

Abstract:
Topics in Sesotho Control Verbs This study examines the phenomenon of forward complement control in Sesotho using Principles and Parameters Theory of syntax as a theoretical framework. It explores the interaction between syntax, morphology, and semantics through an examination of the role played by argument-structure changing verbal morphology (verbal extensions). This study suggests that this argument-structure changing verbal morphology is central to understanding the different syntactic behaviors observed across and within verb classes and to the study of control in Bantu languages. Forward complement control is a relation of obligatory co-indexation between the subject or object of the matrix clause and the subject of the subordinate clause. The subject of the subordinate clause is the null PRO which is understood as having the same reference as the pronounceable subject or object of the matrix clause. Sentence (1) illustrates control in Sesotho. (1)[Mmei o -kgothaletsa bana j [PRO *i/j ho - bala dibuka]] 1.mother 1.SM - encourage 2.children INF - read 10.books ?Mother encourages the children to read books? Control verbs select noun phrases, subjunctive clauses as well as the infinitive in Sesotho. Only the infinitive participates in control relations. However, an examination of the other complements of control verbs supports the classification of verbs in Sesotho proposed in this dissertation. The infinitive is introduced by ho which is also the prefix of class 15 nouns. This study compares the status of ho as an infinitival morpheme and ho as a noun class prefix in Sesotho. I suggest that the morphology of the control verbs together with the referential relations between the controller in the matrix clause and the subject of the infinitive help determine the differences between nominal infinitive (class 15 nouns) and clausal infinitives.This study also proposes that argument- structure changing verbal morphology is important in the classification of control verbs in Sesotho. This morphology is also responsible for the syntactic properties of control and directly accounts for the types of control observed in Sesotho. Contrasting this analysis with a typology of control patterns from other languages, the lack of partial control or split control is furthermore directly linked to argument-structure changing morphology. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2010.
Local:
Adviser: McLaughlin, Fiona.
Statement of Responsibility:
by Mantoa Smouse.

Record Information

Source Institution:
UFRGP
Rights Management:
Copyright Smouse, Mantoa. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Classification:
LD1780 2010 ( lcc )

Downloads

This item has the following downloads:


Full Text







As can be observed from the data, Sesotho does not allow the arguments of the control

verb to control PRO jointly. In Sesotho, only one argument of the control verb may control

PRO. Again we conclude that the argument structure requirements of the embedded verb have an

influence on referential dependencies.

5.5 Summary

In this chapter I investigated the role played by argument-structure changing verbal

morphology in accounting for the differences in referential dependencies observed across and

within verb classes. I then show how this verbal morphology impacts on a typology of control

attested in Sesotho. I also explored how control relations help distinguish between the nominal

infinitive and the clausal infinitive.

I subsequently examined control types in Sesotho. I discovered that object control verbs

that allow control shift from object to subject are inherently applicative. Those that allow the

shift from object control to arbitrary control are morphologically applicative or have a

combination of causative and applicative morphology. These are the verbs that help to

distinguish between the nominal infinitive and the clausal infinitive. Finally, I established that,

the MDP and the argument -structure of the embedded clause is responsible for the lack of

partial control and split control in Sesotho.









configuration is such that PRO is understood as either the subject or object of the matrix clause.

This is called exhaustive control.

The other control relations are partial control and split control. In partial control, the

subject of the matrix clause is included as one of the referents of PRO. In split control, the

subject of the infinitival clause refers to two distinct arguments. Stiebels 2007 provides the

representation in Table 4-1 to explain the referential properties of PRO. In the configuration in 5-

1, k refers to PRO, i refers to a subject argument of the matrix clause, whereas refers to an

object argument of the matrix clause. These properties in 5-1 are spread between obligatory

control (OC) and non-obligatory control (NOC) (Williams 1987). Split control is a phenomenon

which is restricted to NOC9. The differences between OC and NOC are based on the phenomena

in Table 5-2 (adapted from Hornstein 2007: 114):

Table 5-1. Referential properties of PRO
Type of control Subject Control Object Control
Exhaustive k=i k=j
Partial kDi klj
Split k= i+j


Table 5-2. Properties of PRO in OC and NOC
Obligatory Control (OC)

(a) *It was expected PRO to shave himself.

(b) *John thinks that it was expected PRO to
shave himself.
(c) *John's campaign expects PRO to shave
himself.

(d) John expects PRO to win and Bill does
too.(=Bill win)
(e) *Johni told Maryji PROi+j to wash
themselves/eachother


Non-Obligatory Control (NOC)
(a) It was believed that PRO shaving was
important.
(b) Johni thinks that it is believed that
PROi shaving himself is important.
(c) Clinton'si campaign believes that PROi
keeping his sex life under control is
necessary for electoral success.
(d) John thinks that PRO getting his
resume is crucial and Bill does too.
(e) Johni told Maryj [that [PROi+j washing
themselves/each other] would be fun]l.


9 This position is still debated, see Madigan 2008 on split control in Korean.

95










The examples in (5e) and (6) are central to this study. Although this type of noun

formation is observed elsewhere, unlike other nouns, the final vowel of the verb remains

unchanged. The various suffixes that we see with other deverbal nouns are not possible with

class 15 nouns as indicated by the examples in (5e) and (6a). This pattern reveals that class 15

nouns are different from the other nouns with respect to noun formation although they employ

the same strategy of prefixing the noun class prefix to verbs.


(5) Nouns derived from the verb rata 'like/love'
(a) mo- rat- i /*o
cl.1- love- Pers Suffix
'lover'

(b) le- rat- o/*i
cl.5- love- Imp Suffix
'love'

(c) that- o/*i
cl.9. love- Imp Suffix
'love'

(d) mo- rat- uw- a
cl.1- love- PASS- FV
the beloved one'

(e) ho- rat- a/*i/*o
cl.15- love- FV
'to love/loving'


(6) Nouns derived from deverbal nouns
(a) ho- nk- a seabo
cl.15 take- FV part
'to participate/ participating'


(b) mo- nk- a- seabo
cl.1 take- FV part
'participant'


(c) mo- nk- *i seabo
cl.l take- Pers Suffix part
Intended: 'participant'









Chomsky, Noam and Howard Lasnik. 1995. The theory of principles and parameters. In The
Minimalist Program by Chomsky, N. The MIT Press. Cambridge, Massachusetts, London,
England.

Chomsky, Noam, and Howard Lasnik. 1977. Filters and control. Linguistic Inquiry. 8:425-504.

Comrie .1984. Language Universals and Linguistic Typology. University of Chicago Press.

Creissels, Denis. 1996. Conjunctive and disjunctive verb forms in Setswana. SNiut African
Journal of African Languages; Vol. 16(4):109-116.

Creissels, Denis. 2005. Tswana verb morphology and the Lexical Integrity Principle. Paper
presented at Fifth Mediterranean Morphology Meeting, Freju.

Creissels, Denis and Daniele Godard. 2005. The Tswana infinitive as a mixed category.
Proceedings of the HPSG05 Conference, University of Lisbon, CSLI Publications.

Culicover, Peter W., and Ray Jackendoff. 2001. Control is not movement. Linguistic Inquiry
32:493-512.

Culicover, Peter W., and Ray Jackendoff. 2003. The semantic basis of control. Language
79.3:517-556.

Culicover, Peter W., and Ray Jackendoff. 2005. Simpler Syntax. Oxford: Oxford University
Press.

Culicover, Peter W., and Ray Jackendoff. 2006. Turn over control to the semantics. Syntax
9(2):131-152.

Culicover, Peter W., and Wendy Wilkins. 1984. Locality and linguistic theory. New York:
Academic Press.

Culicover, Peter W., and Wendy Wilkins. 1986. Control, PRO, and the Projection Principle.
Language 62:120-153.

Davies, William D. and Stanley Dubinsky. 2004. The Grammar ofRaising and Control. A
Course in Syntactic Argumentation. Blackwell Publishing.

Davies, William D. and Stanley Dubinsky (Guest eds.). 2006. The place, range, and taxonomy of
control and raising. Syntax 9:2: 111-117.

Demuth, Katherine. 1995. Questions, relatives, and minimal projection. Language Acquisition
4:49-71.

Demuth, Katherine. 1990. Locatives, impersonals and expletives in Sesotho. The Linguistic
Review 7: 233-249.

Demuth, Katherine and Sheila Mmusi. 1997. Presentational focus and thematic structure in
comparative Bantu. Journal of African Languages and Linguistics 18:1-19.


119









CHAPTER 5
SESOTHO COMPLEMENT CONTROL PROFILE

5.1 Introduction

The primary goal of this chapter is to present the complement control patterns observed in

Sesotho. Complement control refers to a relation of obligatory coindexation between the subject

and the object of the matrix clause and the subject of the subordinate complement clause. The

secondary goal of this chapter serves to address the question of the role played by argument-

structure changing verbal morphology in control phenomena. In Chapter 3 I discussed the verbal

morphology associated with control verbs in Sesotho. One of the areas discussed relates to

argument-structure changing verbal morphology. In this chapter I explore the role played by

argument-structure changing verbal morphology in accounting for the differentiated behavior of

verbs that belong to the same class. I also examine how this argument-changing morphology

impacts on the types of control attested in Sesotho. In so doing, I also explore how control

relations help distinguish between the nominal infinitive and the clausal infinitive

Chapter 5 is organized as follows: In the first part of this chapter I define terminology

associated with control providing illustrative examples. I then go on to discuss control types

attested in Sesotho, followed by an examination of referential dependencies. For each of the

sections I try to determine whether class membership is an adequate determinant of referential

properties in Sesotho.

5.2 Properties of Control

In Chapter 1, I defined complement control as a relation of obligatory co-indexation

between the subject or object of the matrix clause and the subject of the subordinate complement

clause. The referential relations of the controlled element (PRO) are determined by those of the

controller. This is called referential dependency. The two major types of obligatory control are









Case is assigned to NPs by certain heads under the structural relationship of Government

stated in (25). In Sesotho for example, Spec I is a position where the subject NP is governed by

and receives nominative Case from I. The object is governed by and receives accusative Case

from V in a verb-complement position2. The Case Filter forces the movement of the subject to

spec,IP in (19), otherwise, the subject would not be assigned Case and the Case Filter would be

violated, and the derivation would not succeed.

(25) Government and m-command
(a) A governs B iff A m-commands B and no Barrier intervenes between A
and B. Maximal Projections are barriers to government. Governors are
heads.
(b) A m-commands B iff
i. A does not dominate B,
ii. B does not dominate A, and
iii. The maximal projection of A dominates B.

(Chomsky 1986b:8)

The various modules of GB interact to ensure well-formedness. Let us now consider

another module that relies on Government as defined in (25). The Binding Theory (BT)

determines the distribution of anaphors, pronouns, and R-expressions in relation to their potential

antecedents (Haegeman 1991:215). The principles regulating the interpretation and distribution

of these NPs as stated in (26-28) are referred to as the Binding Principles. The binding relations

are expressed through co-indexation of two constituents and the binding principle works as a

filter to determine licit and illicit co-indexations. For our purposes, the governing category can

be understood as the minimal clause.

(26) The Binding theory (Chomsky 1981:188)
A. An anaphor is bound in its governing category


2 am ignoring the subject agreement marker SM here for reasons of simplicity.

24









Argument structure "encodes lexical information about the number of arguments, their

syntactic type, and their hierachichal organization necessary for the mapping to the syntactic

structure" (Bresnan 1995:1). The discussions that look at the relationship between the lexicon

and argument structure continue to investigate whether argument structure is not redundant since

it repeats information that is encoded in the lexicon. Bresnan 1995 points out that work in

lexical semantics show that "much of the information about number, obligatory status, and

hierachichal organization of arguments in argument structure is in fact predictable from

semantics" (Bresnan (1995:2), Levin 1993).

(37) Bresnan 1995 interface model

Lexical semantics

1 Lexicon

a-structure

1 Syntactic projection

Initial syntactic structure (D-structure)

1 Syntactic Transformations

Final syntactic structure (S-Structure)



Bresnan 1995 uses data from English and Chichewa to show that the verbal relation

changes associated with 'transformations' are redundant since these are lexical morphological

processes. Bresnan therefore proposes that the initial syntactic phase should be abandoned. In

this study I take Bresnan's proposal as a starting point because I look at the contribution of these

verbal relation transformations to argument structure. I focus on the causative, the applicative

and reciprocal in particular. I propose that these 'lexical morphological processes' indeed









TABLE OF CONTENTS


page

A C K N O W L E D G M E N T S ..............................................................................................................4


L IST O F T ABFIG U RE S .............................................................................................. .....................
L IST O F FIG U R E S ..................................................................................... 9

L IST O F A B B R E V IA T IO N S ............................................................................. .....................10

A B ST R A C T ......... ......................................................................................... ...... 14

1 IN T R O D U C T IO N ................................................................................................................ .. 14

1 .1 In tro d u ctio n ............................................................................................................ 14
1.2 The A rea of Investigation ................................ ............................... ........... .... 16
1.3 T theoretical B background .................................................................. .. ... ............20
1.3.1 The Principles and Param eters Theory .................................... ..................21
1.3.2 The Interaction between Semantics, Morphology and Syntax .........................28
1.4 Organization of the Study ...................................... ........... .............................. 30

2 THE SIGNIFICANCE OF HO IN SESOTHO.....................................................................32

2 .1 In tro d u ctio n ............................................................................................................ 3 2
2 .2 N om inal M orphology .......................................................................... ................... 32
2 .2 .1 N oun C lasses.....................................................34
2.2.2 G general N om inal D erivation......................................... ......................... 37
2.2.3 N ouns D erived from V erbs.......................................... .......................... 39
2.3 Properties of the Infinitive in Sesotho ........................................ ...... ............... 41
2.3.1 Distributional Properties of N ouns ...................................................... 42
2.3.2 Distributional Properties of Verbs ............................................ ...............45
2.4 O their U ses of H o .................. .. ... ..... ................................ .......................51
2.5 PRO as the Subject of the Infinitive in Sesotho............... ............... ............... 54
2.6 Nominal and Clausal Infinitive.............................................. ............... 59
2.7 Sum m ary ................................. ............................ ...... ...... ........ 62

3 ASPECTS OF SESOTHO MORPHO-SYNTAX RELEVANT TO CONTROL ..................64

3 .1 Introdu action ...................................... ................................................. 64
3.2 The Verb in Sesotho .................. ................................. .. ................. 64
3.3 Verbal Extensions ....................................... .......... ............... 65
3 .3 .1 T h e C au sativ e ............................................................................................... 6 6
3.3.2 T he A pplicative.......... .......................................................... ...... ..... 67
3.3.3 The R eciprocal ......................... ............ ............ ......... 69
3.4 Sum m ary ................................................................ ..... ....... ........ 70









I have also looked briefly at the behavior of the infinitive in relation to other verbs in

Sesotho. This examination reveals that while the infinitive shows some form of inflection this is

not the same type of inflection as fully fledged finite clauses. I have associated the lack of SM in

the infinitive to the lack of necessary morphology such as Case assigner due to some form of

defective Tense morphology. Finally I have used tests to show that the subject of the infinitive as

complements of control verbs is different from the subject of other complements such as the

subjunctive. In Chapter 3 I discuss the structure of verbs in Sesotho with special focus on the

role of verbal extensions in determining the controller and control relations.









[anaphor] and [Ipronominal] could be used to characterize the elements regulated by Binding

Theory, yielding four categories. These combinations are given below (Haegeman 1991:223)

(35) NP types
(a) anaphoror, -pronominal] anaphor
(b) [-anaphor, pronominall] pronominal
(c) [-anaphor, -pronominal] R-expression
(d) anaphoror, pronominall] PRO

The combinations in (35a) -(35c) correspond to anaphors, pronouns and referring

expressions. PRO on the other hand has the features anaphoror, pronominall] represented in

(35d). Considering the Binding Principles above, we can see that being anaphoror] PRO would

have to be bound (Principle A) and being pronominalal] it would have to be free (Principle B) at

the same time since anaphors must be bound and pronominals free. In order to resolve the

contradiction caused by the satisfaction of these two principles, PRO must not have a governing

category. In such a case, it will vacuously satisfy both Principles A and B. This can occur if PRO

is ungoverned, yielding the PRO Theorem. Government and Binding principles, therefore

determine the distribution of PRO. Given that Case is assigned under government, it also follows

from the PRO Theorem that PRO will also not be assigned Case.

1.3.2 The Interaction between Semantics, Morphology and Syntax

The theoretical outline presented in the previous section makes the assumption that

control is a purely syntactic process. The role assigned to the meaning is reduced to

subcategorization frames which encodes the information about the arguments associated with

that particular verb or lexical item. This may be represented structurally as follows (36)4.

(36) Lexicon -- argument structure-* syntactic structure -*control type


4 -* means determines









to some other argument. These data in (4 and 5) answer the question raised in Chapter 2 of how

to distinguish between the nominal infinitive and the clausal infinitive. The answer to this

question lies not only in the internal structure of the infinitival complement or its position in a

sentence, but also on the semantics of the control verb, specifically the morphological makeup of

that verb. As we can observe from the English glosses, whenever the infinitive does not have its

subject co-referenced with the subject or object of the matrix clause it takes a nominal form

(gerundive). This use is similar to the use of the nominal infinitive as a subject or a class 15

noun.

What element of the control verb is responsible for the distinction between object to

subject control shift and object to arbitrary control shift? In (6a) the object control verb, qobella

is co-indexed with PRO because it is the closest c-commanding NP. However, when this object

is omitted the control shifts from the object to the subject since it is the closest c-commanding

NP. This is not the configuration we get in the case of arbitrary control. The initial thought

would be that qobella is a manipulative verb whereas kgAIth,'leit, is an experiencer-object verb.

As observed with (4b) thibela 'prohibit' is also a manipulative verb but does not allow control

shift in the same manner that qobella does. In (4d) we do not even get any kind of control shift

with lemosa 'advise.' The question that we need to ask then is why only a subset of manipulative

verbs allows control shift. The answer to this question may lie with the morphology of these

verbs.

Verbal extensions are used to extend the meaning of verbs but also to derive new verbs.

The extensions that are associated with increasing the number of arguments are the applicative

and the causative. In many instances, the introduction of a causative affix changes the verb class

completely. For example, I pointed out in Chapter 4 that achievement verbs such as hopola









for partial and split control in Sesotho. Putting all these proposals together, this chapter

addressed the questions of the role played by argument-structure changing verbal morphology in

control phenomena as well as the question of the status of ho in Sesotho.

In Chapter 4 I approached the classification of control verbs by looking at whether a

correlation exists between certain verb classes and syntactic properties. I did this by mapping the

syntactic properties onto verbal morphology. The main finding is that the variation in referential

dependencies observed within verb classes can be directly accounted for through verbal

morphology. I discovered that the majority of verbs that are "bare" are associated with subject

control and these are also associated with a lack of shift from subject to object. Object control

verbs on the other hand are either causative, applicative or both. As we observed in Chapter 3,

these are responsible for introducing a benefactive and patient role. It is therefore not surprising

that these verbs are object control verbs. Another important distinction within object control verb

classes is that the subdivisions are based on whether or not these verbs participate in control

shift. This is a property also associated with verbal morphology.

I closed Chapter 4 with an evaluation of verbs that are control verbs in other languages but

not in Sesotho. I proposed that the lack of control with interrogative verbs, such as "wonder' is

associated with the lack of wh-movement in Sesotho, a movement which is associated with these

verbs in languages that allow wh-movement. I also proposed that a close examination of these

verbs reveals that they select the subjunctive rather than the infinitive in Sesotho. This further

supports my observation that certain control verbs prefer the subjunctive over the infinitive.

In Chapter 5 I set out to establish a typology of control in Sesotho. I used this analysis of

the types of control attested in Sesotho to further explore the differences between the nominal

infinitive and the clausal infinitive. As noted in Chapter 2, the nominal and clausal infinitive









(36) #Ke mangvar2 eo mme wa hae var a mvar2 rata-ng?
COP- who 1.REL 1.mother 1.POSS- his 1.SM 1.OM like-REL
'Who does his mother like?'

(37) PRO var ho bina ha haevar2 ho a movar3- kgathatsa.
PRO INF sing 1.POSS his 15.SM FOC 1.OM -bother
'His singing bothers him?'

(38) Ke mang var3 eo PRO var ho -bina ha hae var2 ho mo -var3 kgathats-
ang?
COP who 1.REL PRO INF sing 1.POSS -his 15.SM- 1.OM bother.REL
'Who does his singing bother?'


Example (35) shows two bound variables represented as var1 and var2. If we move var2 over

varl in question formation as in (36) the result should be dubious. However due to the OM that

stays put the sentence is strange but acceptable. In (37) one of the variables is PRO. If we move

the questioned variable over PRO we do not get the WCO effect observed in (36). The test

suggests that the PRO NP is different from other NPs such as mme wa hae.

The VP ellipsis test gives us a better picture. Under VP ellipsis we get a sloppy

interpretation when using the infinitive. In VP ellipsis the verb phrase is omitted, the omitted

phrase is usually replaced by word such as too, as well in English. The two examples (39&40)

show that we obtain a sloppy reading when the subject of the embedded clause is PRO and a

strict reading when it is a lexical NP (40).

(39) Sello o kopa ho -sebetsa, le bana ba a kopa...(ho sebetsa)
1.Sello 1.SM ask INF work, and 2.children 2.SM- FOC- ask... (to work)
'Sello asks to work and the children do too... (ask to work)'

(40) Sello o kopa hore a sebetse, le bana ba a kopa... ( a sebetse)
1.Sello 1.SM- ask that 1.SUBJ work, and 2.children 2.SM FOC ask ... (1.SM work)
'Sello asks that he works and the children ask too ( he works)'

In (39) the omitted part is ho sebetsa of the coordinated sentence. The preceding sentence

creates an environment that can help retrieve the omitted information. Note that ho sebetsa if

added to the second sentence takes the subject as bana as opposed to Sello the original subject.









are those that have been proposed for other languages but do not function as control verbs in

Sesotho. These are interrogative, perception, propositional and modal verbs.

Interrogative verbs that induce control are generally associated with wh-movement.

Sesotho does not show wh-movement. In instances of bi-clausal interrogative sentences, the

embedded clause is introduced by a complementizer which selects the subjunctive form of the

verb. As we have seen in Chapter 3, most of the control verbs also select the subjunctive form as

a complement in non-control relations.

Perception predicates such as 'smell', 'hear', 'feel' also do not participate in control in

Sesotho. These verbs act like other ordinary verbs. Proposition or attitude verbs also do not occur

with infinitival complement in Sesotho. Again, like the interrogative verbs, these verbs select the

subjunctive form. There is one verb however, nahana 'think' which allows a control reading.

This is the only verb from this class that allows control which makes its membership

questionable or it suggests a move from the use of the infinitive to the subjunctive form.

Modal verbs are restricted to raising in Sesotho. Although these verbs select an infinitival

clause, they also allow the insertion of an expletive in the matrix clause. However, when an

expletive is inserted in the matrix clause, the embedded clause is a finite clause. The example

below is used to illustrate this behavior.

(25) Ke tshwanela [ho- sebetsa.]
1SG must to -work
'I must work'

(26) Ho tshwanela [hore ke sebets-e]
It necessary that 1SG. work-SUBJ
'It is necessary that I work'

Although other control verbs also select the subjunctive form as a complement, the ability

to insert an expletive in this manner is limited to this class of verbs. Another thing to mention is

that other verbs such as 'can', 'be able' 'may' can also be expressed with an auxiliary verb '-ka'

91









1.3.1 The Principles and Parameters Theory

The Principles and Parameters theory of syntax is based on the notion of Universal

Grammar (UG). Chomsky (1981b:7) states that "universal grammar may be thought of as some

system of principles, common to the species and available to each individual prior to

experience." Within P&P, all languages obey these universal principles. The differences between

languages are accounted for by differences in parameters rather than principles.

In this view of language the grammar has a lexicon (containing lexical entries with

subcategorization information) and computational system (guided by X-bar Theory). A

derivation goes through four levels of representation: Deep Structure, Surface Structure,

Phonological Form (PF) and Logical Form (LF). There are various principles that apply at

different levels of representation. Let us look at a derivation of a simple Sesotho sentence to

illustrate the various aspects of the theory.

(18) Banana ba bapala mampatile.
2.girls 2.SM play hide.and.seek
'The girls play hide and seek'

In order to derive sentence (18), the computational system first selects the verb and its two

arguments and various functional elements. These are combined into the D-structure in (19),

which adopts a version of the VP-internal subject hypothesis (Koopman and Sportiche 1991 and

others) in which the external argument is base-generated in the specifier ofVP.

(19) [ip [I I [vp Banana bapala mmampatile]]]
The girls play hide.and.seek


The subject then moves to the specifier of IP, yielding the S-structure in (20). I ignore PF

and LF in the rest of the dissertation because it is not relevant to the issue of control.

(20) [ip [I Bananai I [vpti bapala mmampatile]]]









TOPICS IN SESOTHO CONTROL VERBS


By

MANTOA ROSE SMOUSE


















A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2010









(23) Subcategorization1

(a) Hopola: V, [ NP/IP] 'remember'
(b) Batla: V, [ NP/IP] 'want'
(c) Hopotsa: V, [ NP ,NP/IP] 'remind

(d) Hopola: Verb
1 2
NP NP/IP
i j
(e) Batla: Verb
1 2
NP NP/IP
i j
(f) Hopotsa: Verb
1 2 3
NP NP NP/IP
i j k


The subcategorization frames in (23a-c) represent the internal arguments selected by each

of the verbs. The theta-grids (d-f) shows the different theta roles represented by letters i, j and k

assigned by the verbs.

There are transformations that take place between the Deep Structure and the Surface

Structure. Within GB, transformations such as passive movement are subsumed under Move-V.

Movement is optional but is constrained by various restrictions.

One of the movements that applied to our derivation in (19) is the movement of the subject

from spec, VP to spec, IP. This occurs because of Case theory, which requires that all NPs be

assigned Case by Surface Structure. This is stated formally in the Case Filter:

(24) Every phonetically realized NP must be assigned (abstract) Case
(Chomsky 1995: 111)





1 The external argument is excluded in this instance


I









c-commands it. Let us look at some examples of referential dependencies and how they interact

with argument-changing verbal morphology.

5.4.1 Exhaustive Control

Exhaustive control is a referential dependency in which the referents of the controller and

PRO overlap completely. This can be represented as "DPi....PROi." In Sesotho, all verb classes

allow exhaustive control as long as the lexical requirement of the embedded verb is met.

Sentences (10) and (11) illustrate EC with the subject and the object.

(9) Badisanaj ba kgetha PROj ho- tsamaya
2.shepherds 2.SM prefer INF- go
'The shepherds prefer to go'

(10) Badisanaj ba qobellabashanyanak PRO*j/k ho-tsamaya
2.shepherds. 2.SM- force 2.boys INF -go
'The shepherds force the boys to go'

Going back to our MDP notation, we note that the nearest c-commanding NP in (10) is

badisana 'shepherds' and all the referents of PRO are included in badisana 'shepherds.' PRO

cannot possibly refer to some other implicit NP. In (11) on the other hand, the nearest c-

commanding NP is bashanyana 'boys.' Again this is the only NP that PRO is co-referenced

with. These two examples are instances of exhaustive control.

5.4.2 Partial Control

Partial control presents a completely different picture. In partial control relations, the

interpretation of PRO properly includes the referent of the controller instead of being identical to

the controller. We represent partial control as NPj ....PROj+ in line with the MDP. Collective

verbs such as 'gather' kopana are normally used as tests for partial and split control. In Sesotho,

kopana 'meet' is inherently reciprocal. As indicated in Chapter 3, the reciprocal requires a plural

subject or conjoined singular subjects. As the examples in (22-24) show that Sesotho does not

allow partial control.









and nominal properties." Visser 1989 made similar observations about the infinitive in Xhosa.

These statements are also supported by Mugane 2003 where he observes that the behavior of the

infinitive and the gerund in Logoli is that of an extra-sequential hybrid construction.

Table 2-4. Verbal properties of the infinitive
Property Finite verb Infinitive
Subject Agreement Yes No
Object Agreement Yes Yes
Verbal Extensions Yes Yes
Verbal Modifiers Yes Yes
Direct Objects Yes Yes
Tense Yes Yes
(Restricted)
Aspect Yes Yes
(restricted)
Mood Yes No
Negation Yes Yes


The conclusion that can be drawn from the discussion and the observations from other

scholars is that the infinitive is more restricted in object position. Most of the instances where the

infinitive fails a morphological or syntactic test are when used as an object. This is not surprising

since verbs select their internal arguments. As we will observe in Chapter 4, certain verbs have

restrictions on the kind of arguments they select. The infinitive is one such argument. The

question that follows from this conclusion relates to the internal structure of the infinitive. I deal

with these answering this question let us look at other uses of ho in Sesotho.

2.4 Other Uses of Ho

The morpheme ho in Sesotho has various other uses. In one of the uses, the morpheme is

prefixed to noun stems where it functions as a noun class marker. While this is not evident in the

orthography system, some nouns do show this property of ho. In this section, I describe the use

of ho as a conjunction coordinating verbs, as a preposition, as a copulative and as an expletive or

impersonal SM.









share the same properties with minor differences. This comparison in Chapter 2 highlighted the

similarities between the two but did not really explain why the two should be treated differently.

Through an exploration of the referential dependencies as well as the morphological composition

of control verbs, I illustrated that the distinction between the nominal and clausal infinitive rests

with referential dependencies associated with PRO as the subject of the infinitive. For example, I

showed that object-control verbs that are applicative or have a combination of causative and

applicative give the interpretation of PROARB which is an equivalent of the nominal infinitive.

Because PROARB has no c-commanding referent, it can occur in subject position, but PRO that

requires an antecedent is restricted to object position.

Another major contribution associated with control verb classes and the role played by

verbal extensions is that these two directly determine types of control. Subject control is

common with bare verbs, object control is associated with the causative and applicative, and the

reciprocal is responsible for determining referential relations but also responsible for determining

potential control relations. Sesotho lacks partial control and split control because the morphology

associated with these types of control is not compatible with Sesotho verbal morphology.

Chapter 5 has directly addressed questions (1) and (3) above.

6.2 Implications for Control Relations

Recent studies on control may be divided into three areas: Control may be viewed

primarily as a syntactic phenomenon (Chomsky and Lasnik 1993, Hornstein 1999, Boeckx and

Hornstein 2003 amongst others). Others view semantics as a central component in understanding

control (Sag and Pollard 1991, Levin 1993 & 2007, Cullicover and Jackendoff 2001 & 2005

amongst others). There is yet another line of thought which supports the idea that a purely

semantic or purely syntactic approach to obligatory control is not adequate, but rather a

combination of morphological agreement and semantic tense is better suited to account for cross-

114









So far we notice that class 15 nouns pass all the distributional properties of nouns in

subject position. In object position there are some restrictions. Class 15 nouns fail to trigger OM

when used in this position. At this point we also need to remember what we noted earlier that

class 15 nouns also fail to pluralize. The failure of class 15 nouns to pluralize and trigger OM

could be attributed to what Creissels & Godard 2005 term semantic hierarchy. Creissels &

Godard propose that in the hierarchy of semantic objects, class 15 nouns denote an abstract

object, which should be viewed as an eventuality. An eventuality is a notion associated with

verbs. This is not surprising because as I pointed out earlier, studies in Bantu languages show

that class 15 nouns show nominal as well as verbal properties. Let us now turn to the verbal

properties of these nouns.

2.3.2 Distributional Properties of Verbs

Bantu languages are well-known for their rich verbal morphology. Verbal characteristics

relevant for the purposes of this study include subject and object agreement, tense, aspect, mood,

negation as well as the ability to take verbal extensions and verbal modifiers. I deal with verbal

extensions in detail in Chapter 3.The first property that we will look at is the verb's ability to

support subject and object agreement. From now onwards we'll refer to these clauses as

infinitival clauses since we are looking at verbal properties associated with the infinitive. The

examples in (11) illustrate some of the properties.

The first difference between the verb halefisa when used with ho is that it does not support

subject agreement (SM) as demonstrated by (1 la-c). The OM is however supported. In the

examples in (1 la&b) the infinitival verb is extended by the causative extension. This is a

common property of verbs in Bantu.









behavior only would fail to capture the similarities shown available through semantic class

belonging.

4.2.3 Achievement Verbs

Achievement verbs are those verbs that express the bringing about of an achievement,

success, triumph or lack thereof. Achievement verbs in Sesotho are subject control verbs with

no option of shifting to object control. Although these verbs allow the derivation with the

causative and the applicative, in this case they are used in their bare, non-derived forms. The

derived forms of these verbs are realized as belonging to the manipulative verb class. Recall that

desiderative verbs are also subject control verbs, and that these also participate in subject control

only. The sentences in (13) exemplify achievement verbs.

(13) Achievement verbs

(a)Banak ba- lebala PROk ho -tsoha hoseng
2.children 2.SM- forget INF wake-up morning
'The children forget to wake up in the morning'

(b)*Banak ba lebala mmej PROk/*j ho- tsoha hoseng
2.children 2.SM forget mother INF- wake-up morning
'*The children forget mother to wake up in the morning'

Note that while the causative form of the achievement verbs occur as manipulative verbs

(hopola- hopotsa 'remind'; lebala- lebatsa 'cause to forget; tlohela- tlohedisa 'stop' etc) the

applicative forms of these verbs do not participate in the same manner (hopotsa-hopolela 'feel

for'; lebala- leballa 'forgive'; tlohela- tlohella leave/ stop/ permit'). Recall that the applicative

form of the desiderative form is realized as manipulative verbs whereas the causative form is

not? Now we observe that the other set of manipulative verbs is formed by the derived forms, in

this case causative form of the achievement verbs. This is another example of how the

morphological property determines class and syntactic properties.









singular then there must be a post-verbal subject conjoined with the other subject (18) otherwise

the sentence becomes ungrammatical (19).

These properties of the reciprocal are crucial for an analysis of control verbs. In Sesotho,

the reciprocal plays a role in determining referential dependencies. When a verb represents a

collective action, this is expressed with a verb that is inherently reciprocal in Sesotho. Collective

verbs are usually used in expressions that show partial control. As we have noted a reciprocal

verb requires a plural or collective argument in Sesotho. Because of this thematic requirement, it

becomes impossible to obtain partial control with collective verbs in Sesotho. This is a critical

role since it determines the differences between Sesotho and other languages that show partial

control.

3.4 Summary

In this chapter I described the internal structure of verbs in Sesotho in relation to

inflectional and derivational affixes. Inflectional affixes such as tense, aspect and agreement are

prefixed to the verb although some aspectual and tense morphemes occupy the position before

and after the verb. Derivational affixes such as verbal extensions occur as suffixes and are

positioned between the verb root and the final vowel. I then represented this information about

the verb in a linear format indicating the various positions associated with SM and OM. Finally I

took a general look at verbal extensions and how their argument selection impacts on the overall

sentence structure. In Chapter 4 I explore Sesotho control verb classes and their interaction with

verbal extensions.









The structure is evaluated with respect to various principles to ensure well-formedness.

Two important principles are the Projection Principle and the Theta Criterion. The Projection

Principle is a requirement which ensures that the thematic structure associated with lexical items

is saturated in the syntax (Haegeman 1991:63) and that the number of projected arguments is the

same at all levels of representation. Simply stated, the Projection Principle ensures that lexical

information is syntactically represented and prevents insertion of structure beyond D-structure.

This principle is stated formally as follows:

(21) Projection Principle
Representations at each syntactic level (i.e. LF and D- Structure and S-Structure)
are projected from the lexicon, in that they observe the subcategorization
properties of lexical items.

Chomsky further proposed that all clauses must have subjects, regardless of their lexical

requirements. This requirement on clauses together with the Projection Principle came to be

known as the Extended Projection Principle (EPP).

The Theta Criterion is responsible for the assignment and tracking of semantic roles such

as agent, theme, goal and so forth. These theta-roles are part of the lexical information contained

in verb's lexical entry. This assignment and distribution of theta roles derives directly from the

theta criterion. The theta-criterion is stated formally as follows:

(22) Theta-Criterion
Each argument bears one and only one theta-role, and each theta-role is assigned to
one and only one argument (Chomsky 1981:36).

The Theta-Criterion together with the Projection Principle must be satisfied for the

structures to be well-formed. The sub-categorization of some of the verbs we have used in our

examples thus far are given in (23).










understanding how the class 15 noun class is different from the other noun classes in Sesotho.

Below are examples that show the productivity of noun formation. I limit my examples to word

formation that involves the noun class prefix.


(2) Nouns derived from adjectives
(a) mo- holo
cl.1- big/old
'elder'

(b) bo- holo
cl.14- big/big
'size'

(c) se- holo
cl.7 old/big
'ways of the elderly'


(3) Nouns derived from adverbs
(a) bo- hole
cl.14 far
'distance'

(b) le- ho-dimo
cl.5-cl. 16. up
'heaven/sky'

(c) di- ka- hare
cl.10- PREP- inside
'contents'


(4) Nouns derived from nouns
(a) bo- tho
cl.14- person
'being human'

(b) se- sadi
cl.7 woman
'woman-like'

(c) se- phoofolo
cl.7- animal
'animal-like'

(d) *se-mo-sadi
cl.7.-cl.1- woman









The first thing is to assume that ikemisetsa must belong to a different class of verbs.

However, the verbs rata 'like' and ikemisetsa 'intend' belong to the same semantic class so the

complement selection preferences cannot be attributed to class membership only. It turns out that

ikemisetsa 'intend' and hlophisetsa 'plan/prepare' are derived using the causative and

applicative extensions is and -el simultaneously. These two extensions when used together

require only one internal argument; either NP or IP but never together. If only one of the two

arguments is used, it has to be the infinitival argument (7f&g). All the other desiderative verbs

select both the infinitive and a proper noun as a complement.

(7) Noun phrases as complements5
(a) Bana ba rata [dijo] NP
2.children 2.SM like food
'The children like food'

(b) Bana ba rata [hoja]IP
2.children 2.SM like to eat
'The children like to eat'

(c) Bana ba hana [dijo tsa kgale] NP
2.children 2.SM refuse 10.food 10.POSSold
'The children refuse old food'

(d) Bana ba hana [hoja dijo tsa kgale] IP
2.children 2.SM refuse to eat 10.food 10.POSSold
'The children refuse to eat old food'

(e) Bana ba ikemisetsa [ho bina] IP
2.children 2.SM intend to sing
'The children intend to sing'

(f) *Bana ba ikemisetsa [ pina] NP
2.children 2.SM intend song
'* The children intend song'

(g) Bana ba ikemisetsa [ho bina pina] IP
2.children 2.SM intend to sing song
'The children intend to sing a song'


5 I have decided to use the infinitival clause without PRO in this section of the chapter in order to highlight the role
played by other complements of control verbs.









in Sesotho. Nominal infinitives show verbal properties such as negation and object agreement

which are not possible with other nouns. The structure in 2-1 is not able to accommodate the

required nodes associated with negation and agreement. If we take the nominal infinitive as some

form of a 'clause' then we need to distinguish between the nominal and clausal infinitive.

(a) DP (b) DP

D' D'

D NP D NP

0 N' 0 N'

N N

Dijo ho-ja

Figure 2-1. The noun in Sesotho

Visser 1989, proposes a three-way distinction in accounting for the infinitive in isiXhosa.

According to Visser, there are 'proper' class 15 nouns which would be represented by 2-1(a), the

nominal infinitive and the clausal infinitive. In Sesotho however, there are no observed 'proper'

class 15 nouns, instead we get the nominal infinitive and the clausal infinitive. These two are

represented in 2-2 (a&b).

(a) NP (b) IP

N' DP I'
A
N IP PRO I VP
I I A
0 D I' ho ja

PRO I VP
I A
Ho ja

Figure 2-2. The Nominal and clausal infinitive











(b) ba- sotho
cl.2- stem
'Sotho speaking people'

(c) Le- sotho
cl.5- stem
'A country of the Sotho speaking people'

(d) *ma- sotho
cl.6 stem
'Plural of Lesotho'

(e) Se- sotho
cl.7- sotho
'Sesotho language'

(f) *Di- sotho
cl.8- sotho
'Sesotho languages'

(g) bo- sotho
cl.14 stem
'Being a Sotho'

Examples (la) and (lb) indicate a singular and plural pair, the difference being expressed

through the prefix. Examples (Ic) and (Id) further indicate that some nouns do not follow the

expected singular and plural pairs due to semantic restrictions. Example (le) and (If) further

show that although the stem may appear with other noun classes prefixes it is very restricted in

plural forms. Example (Ig) represents a noun class in which the majority of nouns are abstract

nouns and for this reason they lack plural counterparts. As these data show, the noun class

prefixes cannot be freely prefixed to any noun stem. We will not go into the details of the

distribution of prefixes but will rather try and draw attention to the historical impact of these

irregularities of noun classification and how these may shed some light on the behavior of certain

noun classes. Although there are these irregularities with regards to singular and plural nouns,

the noun class system as a whole is fairly systematic. Let us now look at the noun classes in

Sesotho.









also thankful to Dr. Thomas Hurst, Mrs. Betsy Hurst and their family. I have learnt a lot about

the holidays in the United States from spending time in their home. Mrs. Hurst, having been a

student and a mom herself, shared some wonderful parenting tips. I also thank their children for

sharing some childhood stories with me. I am very thankful to their paying attention to my

personal comfort.

I thank my husband and my children. My children, Naledi, Tlhodiso and Mapule showered

me with love, support, understanding, and proved the much-needed diversion from academic life.

I am forever grateful for the free hugs, kisses, unplanned trips to the mall, birthday parties, and

playgrounds. Without their unconditional love, I would have not been able to stay at the

University of Florida. To my husband, Abuti Basie, without whom I would have not survived my

studies, I am deeply indebted. His friendship, support and love have kept me going.

I am very fortunate to have been able to engage in some discussions with Dr. Mchombo,

Dr. Salikoko Mufwene and the participants at the ACAL meeting held at University of Florida. I

am also indebted to Dr. Der Houssikian and Dr. Henderson for their advice and views on Bantu

syntax.

I would also like to sincerely thank my committee members, Fiona Mc Laughlin, Eric

Potsdam, Gary Miller (although he retired before I completed), Masungu Matondo and

Abdoulaye Kane. My spontaneous meetings (including one more than 10,000 miles away) with

Fiona Mc Laughlin were always very fruitful for solving riddles that I spent hours trying to

solve. I sincerely thank Dr. Mc Laughlin and Dr. Potsdam for spending all those hours correcting

a number of drafts.









infinitival clause supports in Sesotho is the future tense. The future tense morpheme is treated

with caution here since the same morpheme also works as a verb in other contexts1. The major

difference between the infinitival clause and finite verbs clauses is that the infinitival clause can

only be used with the future tense. Let us look at some examples.

(14) Future Tense

(a) Banna ba tla halef- is- a mme
2.men 2.SM- FUT angry- CAUS- FV 1. mother
'Men will make mother angry'

(b) Banna ba rata ho tla halef- is- a mme
2.men 2.SM- like INF- FUT- angry- CAUS- FV 1.mother
'Men like to come and make mother angry'

(c) Ba tla tla sebetsa
2.SM FUT come -work
'They will come and/to work'

(d) Ba tla ho tla tla -sebetsa
2.SM come INF FUT- come -work
'They come to work'/ 'they come to work (in the future)'

Example (14a) is an example of a finite clause in the future tense form. The future tense

morpheme tla is prefixed to the verb between the subject agreement morpheme and the verb.

Example (14b) on the other hand shows the use of the future tense morpheme with the infinitival

clause. However, it is important to note here that the future tense morpheme can also function as

a verb of motion as represented in sentences (14c&d). Additionally the same morpheme may

also be used to conjugate the two clauses in (14c) ba tla and sebetsa, and ba tla and ho sebetsa

in (14d). It is evident then from these examples that the future tense morpheme in the infinitival

clause may be used as a conjunction as well as marking a future action.




1 Doke (1957:188) claims that the infinitival clause inflects for aspect in Sesotho. His observation is based on the
use of the infinitive in isolation. Visser 1989 also notes that the use of the future tense in Xhosa functions as an
auxiliary verb, rather than tense: to come and the infinitival form

47









Thirdly, according to the Theta-criterion, mme is already receiving a theta-role from the main

verb it cannot receive two as this would be a violation of the Theta-criterion. The next problem

for this kind of bracketing has to do with Case. The object of the matrix clause mme receives

accusative Case as indicated by these examples.

(30) Mme o sebets-a lape-ng.
1.mother 1.SM work-FV home-LOC
'Mother works at home'

(31) Ke a mo hopotsa (mme)
1SG FOC 1.OM remind (mother)
'I remind her (mother)'

Example (30) shows the subject agreement for mme as o. However, in object position the

agreement associated with mme is mo (31). This suggests that mme in object position receives

accusative Case. An accusative case in a subject position would be a problem for the Case filter.

We noted earlier in the chapter that the clausal infinitive has some form of irrealis future tense.

This tense associated with the infinitive is 'defective' in that only some form of future tense is

supported. Examples (30) and (31) also indicate that the presence of subject agreement is an

indication of a Case assigner. Since the infinitive does not support subject agreement the lack of

Case in this position is supported. This would then force the object/subject to move into a Case

assigned position. This would leave us with an empty subject position of the embedded clause

violating both the EPP and the Theta-criterion.

However, if we go back to the initial proposal, where we posit PRO as the subject of the

infinitival clause we then satisfy the EPP, the Theta-criterion and the Case filter. The next step is

to see if the subject position of the infinitive satisfies the PRO theorem. According to the PRO

theorem, the subject of the infinitive must be a Caseless and an ungoverned position. We have

already established that the subject of the infinitive is a Caseless position due to the lack of SM.

This subject position of the infinitive also fails to be assigned Case by an external head such as

56









LIST OF FIGURES

Figure page

1-1 M D P and subject control ......................................................................... ....................26

1-2 M D P and object control ............................................................................ ................... 27

2-1 The noun in Sesotho ......................................... ................ 60

2-2 The Nominal and clausal infinitive........................ .............................................. 60

3-1 V erb structure in Sesotho ................................................................................. .... ..... 65









(2) Mosadi ha a sa tla mo bul- el- a kajeno
1.woman NEG 1.SM PROG- FUT 1.OM -open- APPL- FV today
'The woman will no longer open for him today'

(3) Mosadi ha a sa mo bul- etse
1.woman NEG 1.SM- PROG 1.OM open -APPL.PERF
'The woman is no longer opening for him'

Sentence (1) shows a future tense sentence with SM positioned between the subject and the

future tense morpheme. The OM is closest to the verb root. Sentence (2) on the other hand shows

that the SM for noun class 1 changes changes from o to a when the sentence is in the negative

form. We also note here that the progressive morpheme comes before the tense morpheme.

Sentence (3) shows that the perfective is marked post-verbally as opposed to the progressive

aspect.


NEG SM T/A OM Verb root- Verbal Extension(s)-T/A/FV

Figure 3-1. Verb structure in Sesotho

Putting these various positions together we come up with a template exemplified in Figure

3-1 which is a slight adjustment of Creissels 2005 representation.

3.3 Verbal Extensions

Verbal extensions play a central role in Bantu verbal morphology. This role extends to

syntax due to the interaction between these extensions and the arguments of the verb. Verbal

extensions occupy the position between the verb root and the final vowel (FV) in Sesotho.

Sesotho uses verbal extensions to derive new words and extend or alter the meaning and

argument structure of verbs. The verbal extensions in Sesotho are the causative /-is-/, the

applicative /-el-/, the reciprocal /-an-/ the passive /-w-/ and the neuter /-eh-/. These extensions

may be duplicated to express an intensified action. They may also combine to express different

actions. Two of these verbal extensions reduce valency (reciprocal and neuter) and two increase









Table 4-3. Achievement verbs

Verb Gloss Subject Object Control Causative Applicativ
Control Control Shift e
hopola remember Y N N Y Y
Lebala forget Y N N Y Y
tlohela refrain Y N N Y Y
phema avoid Y N N Y N
Hana decline Y N N Y Y
hlokomoloha neglect Y N N N N




Achievement verbs may also be used in other contexts where there are no control relations.

When used this way, these verbs select an NP as a complement. The examples in (14) illustrate

the different complement of achievement verbs.

(14) Achievement verbs and their complement

(a) Banana ba- hopola mme NP
2.girls 2SM- remember 1.mother
'The girls remember mother'

(b)Banana ba hopola ho- tsamaya IP/NP
2.girls 2.SM- remember INF go
'The girls remember to go/going'

(c)Banana ba- hopola hore ba tsamay- e IP
2.girls 2.SM- remember that 2.SM- go- SUBJ
'The girls remember that they (the girls) must leave'

(d)*Banana ba hopola mme ho- tsamaya NP IP/NP
2.girls 2SM- remember 1.mother INF -go
Intended: 'The girls remember mother leaving'

Sentence (14a) indicates that hopola 'remember' also selects an NP as a complement.

Sentence (14b) on the other hand shows that hopola 'remind' also selects an IP as a

complement. Achievement verbs also behave the same way as manipulative verbs with regards

to the subjunctive. However, unlike manipulative verbs, they have a strict requirement on the

subject of the subjunctive in that it can only refer to the subject of the matrix clause (14c).









possible if the object is a class 15 noun. The examples in (9) reveal the differences between class

15 nouns as objects and object nouns in other classes. In (9a), the object dijo 'food' triggers

object agreement as well as a focus morpheme on the verb when proposed. As sentence (9b)

demonstrates, this is not the case with class 15 nouns in object position. Instead we see in (9c)

that only the focus marker is used for emphasis unlike in (9a) where both OM and FOC are

present. Sentence (9d) illustrates that the focus marker is required when proposing the object.

(9) OM in proposing

(a) Dijo bana ba a di rata
l0.food 2.children 2.SM FOC 10.OM like
'As for food, children like it'

(b) *Ho ja bana ba ho rata
15. eat 2.chilrden 2.SM 15.OM like
Intended: 'As for eating, children like it'

(c) Ho ja bana ba a rata
15.eat 2.children 2.SM- FOC- like
'As for eating, children like it'

(d) *Hoja bana ba- rata
15.eat 2.children 2.SM like
Intended: 'As for eating, children like it'

Another way of emphasizing the object noun in Sesotho is by prefixing the OM to the verb

without proposing. In this instance the object is highlighted as the emphasis is on the OM. In the

presence of the OM the object noun becomes optional. Example (10b) once more shows that

class 15 nouns are not capable of triggering OM in object position.

(10) OM in emphasis

(a) Bana ba- a di rata (dijo)
2.children 2.SM- FOC- 10.OM like food
'Children like FOOD'

(b) *Bana ba a ho rata (hoja)
2.children 2.SM- FOC 15.OM- like 15.eat
Intended: 'Children like EATING'









require a plural subject or "actant" (Stiebels 2007), PRO must be interpreted as plural, despite

the fact that the controller Lisa is singular. The meaning of (16) is that Lisa wants that she and

some other people meet.

(16) Lisai wants [ PROi+j to meet in the evening]

(17) *Ngwanai o batla [PROi+j ho kopana]
1.child 1.SM- want to meet
Intended: 'The child wants to meet with someone'

A series of tests using verbs such as 'meet' and 'gather' reveal that Sesotho control verbs

do not allow partial control. The question that follows from this observation relates to how this

lack of partial control contributes to cross-linguistic descriptions of control. I explore this

question in Chapter 6.

The data used in this study come from written texts, transcriptions of spoken data, previous

linguistic descriptions and grammaticality judgments from me and other native speakers. The

written materials consulted in my study include a Sesotho classic novel, Chaka by Thomas

Mofolo, a contemporary novel, Kiriatshwana by Mkwanazi and Seyalemoya edited by Moeketsi

which consists of six radio dramas. Transcripts of two personal narratives also form part of

materials consulted. All data with morpheme ho were extracted from the sources. The

investigation of the data was guided by two questionnaires on control verbs, Stiebels et al. 2003

and Stiebels 2007.

1.3 Theoretical Background

This section discusses the theoretical background assumed throughout the dissertation. I

lay out the relevant aspects of the Principles and Parameters Theory that are important for

understanding what follows. The second part of this section lays out how the morphology, the

syntax and the lexicon interact.









Sentence (a) shows that PRO requires an antecedent in OC. This is not the case in NOC

((a) of NOC). Sentence (b) shows that PRO requires a local antecedent in OC. Conversely PRO

can have a non-local antecedent in NOC. The ungrammaticality of sentence (c) of OC shows

that the antecedent must c-command PRO. This is not a necessary condition as exemplified by

(c) of NOC. Under VP ellipsis exemplified by sentence (d), PRO permits a sloppy reading in OC

and only strict reading in NOC. Sentence (e) shows that NOC allows split antecedents whereas

OC does not. Bearing these properties of OC in mind, let us first examine how control verbs

interact with NPs and subjunctive clauses as complements.

5.3 Control Types

This section describes the control types observed in Sesotho, a relation that involves a co-

refemtial relation between two arguments, one being the controller the other being the controlled

element. The controller is defined as the argument of the control verb whose referent is

(improperly) included in the referents of an argument of the embedded verb (Stiebels et al.

(2003). The controller may be the subject or object of the control verb. As indicated by the word

improperly in the definition, the controller may also be an implicit argument of the matrix verb.

I explore these referential relations drawing from Sesotho data.

5.3.1 Subject Control

Subject control obtains when the subject of matrix clause is co-referenced with the null

subject of the embedded clause indicating a control relation. The majority of control verbs in

Sesotho show subject control. Although manipulative verbs also participate as subject control

verbs, I discuss them separately from the other verbs because they show a different pattern of

control relations from the rest. The examples in (2) illustrate the various verb classes that

participate in subject control.

(2) Subject control









data representing the different types of complement control structures attested in Sesotho. None

of the studies that I am aware of have discussed a typology of control in Bantu. The second

major contribution of this study presented in Chapter 5 is the discovery of lack of partial control

in Sesotho. Chapter 6 summarizes the findings of the study. Chapter 6 also points out areas that

require further investigation.









LIST OF TABLES

Table page

2-1 Sesotho noun classes ................................................... ............... .. ...... 35

2-2 N oun class 1 plural form action ......................................... .............................................36

2-3 N om inal properties of the infinitive................................................................................50

2-4 V erbal properties of the infinitive......................................................................... ...... 51

4-1 Desiderative verbs............. ... .. ....................... ......... 75

4-2 M manipulative verbs............ ........ .. ............... ...................80

4-3 Achievement verbs............ ........ .. ............... ...................84

4 -4 F active e v erb s ................................................................86

4-5 E xperiencer-obj ect verb s ......... ................. ........................................ ...........................89

5-1 R eferential properties of P R O ......................................... .............................................95

5-2 Properties of PRO in OC and NOC ..................................................... ..................95









Mugane, John. 1998. Gikuyu Morpho-syntax. In Trends in African Linguistics (2). Language
History and Linguistic Description in Africa, ed. by Ian Maddieson & Thomas
Hinnebusch. 239-248.

Mugane, John. 2003. Infinitive-gerunds as extrasequential hybrid constructions in Logoli. In
Trends in African Linguistics (5). Linguistic Typology and Representation of African
Languages. Africa World Press, INC.

Noonan, Michael.1985. Complementation. In Timothy Shopen (ed.) Language typology and
syntactic description. vol 2: Complex constructions. Cambridge University Press, 42-140.

Nurse, Derek and Gerard Philippson. 2003. The Bantu Languages. Routledge.

Nurse, Derek. 2006. Focus in Bantu: Verbal morphology and function, ZASPapers in
Linguistics 43:189-207.

Petter, Marga. 1998. Getting PRO under control, The Hague: Holland Academic Graphics.

Perez, C.H. 1986. Aspects ofcomplementation in three Bantu languages. Doctoral dissertation,
University of Wisconsin-Madison.

Polinsky, Maria, and Eric Potsdam. 2002. Backward control. Linguistic Inquiry 17:501-557.

Polinsky, Maria, and Eric Potsdam. 2006. Expanding the Scope of Control and Raising. Syntax
9(2):171-192.

Reinhart, Tanya, and Eric Reuland. 1993. Reflexivity. Linguistic Inquiry 24:657-720.

Rizzi, Luigi. 1986a. Null objects in Italian and theory of pro. Linguistic Inquiry 17(3):501-557.

Rizzi, Luigi. 1986b. On chain formation. In The syntax ofpronominal critics. ed. by Hagit
Borer, 65-95. New York: Academic Press.

Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press.

Roeper, Thomas. 1987. Implicit arguments and the head-complement relation. Linguistic Inquiry
18: 267-310.

Rosenbaum, Peter .1967. The grammar of English predicate complement constructions.
Cambridge, Ma.: MIT Press.

Rosenbaum, Peter .1970. A principle governing deletion in English sentential complementation.
In R. Jacobs & P. Rosenbaum (Eds.), Readings in English transformational grammar (pp.
20-29). Waltham, MA: Ginn.

Rugege, G. 1981. The infinitive in Kinyarwanda. Studies in African Linguistics. Supplement 8.

Sag, Ivan, and Carl Pollard. 1991. An integrated theory of complement control. Language
67:63-113.









restriction is that control verbs select the infinitive as a complement and each class of control

verbs has its own restrictions with regards to complements. This is only evident when we take

each class of verbs and compare its argument selection preferences between nouns, subjunctive

clauses and the clausal infinitive.


Table 4-1. Desiderative verbs

Verb Gloss Subject Object Control Causative Applicative
Control Control Shift
qeta decide Y N N Y Y
ikemisetsa intend Y N N Y Y
hlorela long Y N N Y Y
kgetha choose/prefer Y N N Y Y
batla want Y N N Y Y
rata like Y N N Y Y
rera plan Y N N Y Y
lebella expect Y N N Y Y
hana refuse Y N N Y Y
dumela agree Y N N Y Y
hloka need Y N N Y Y
tshaba afraid Y N N Y Y
hlophisetsa Plan/prepare Y N N Y Y
lalalabela aspire/yearn Y N N Y Y

Desiderative verbs when used in other contexts where there is no control relation select an

ordinary nominal noun phrase (NP) as a complement in Sesotho. This selectional property is

shared by the majority of subject control verbs although there are some differences. Let us look

at some examples to illustrate this behavior.

The examples in (7) show that the control verbs rata 'like' in (7a&b) and hana 'refuse' in

(7c&d) select ordinary NPs as well as clausal infinitival (IP) complements. However, verbs

such as ikemisetsa 'intend' only allow clausal infinitival complements (7e-g). Sentence (7e)

shows that ikemisetsa 'intend' works well with the clausal infinitive but this is not the case with

a noun (7f). Example (7g) on the other hand shows that the noun is only allowed as the object of

the infinitival clause.









(h) Bana ba hlophisetsa [ho bina pina] IP
2.children 2.SM plan to sing song
'The children plan to sing a song'

(i) *Bana ba hlophisetsa [ pina] NP
2.children 2.SM plan song
'The children plan a song'

Sentences (7h&i) show that hlophisetsa 'plan/intend' also selects a clausal infinitive as a

complement and not an NP. It appears then that desiderative verbs in Sesotho fall into two

categories based on their argument structure. Desiderative verbs that are bare (no causative or

applied extensions) select NP or IP but never both. Desiderative verbs that are derived using the

a combination of the causative and applicative only select an clausal infinitive as a complement.

These are exemplified in (8).


(8) Subcategorization frames for desiderative verbs
(a) rata, V [_ NP, IP] 'like'
(b) hana V, [ NP,IP] 'refuse'
(c) ikemisetsa V,[ IP] 'intend'
(d) hlophisetsa V, [ IP] 'plan'

The other complement choice that fits in with the subcategorization exemplified in (8) is

the subjunctive. As I have already mentioned, subjunctive complements are very common with

control verbs in Sesotho. The subjunctive as an embedded clause is introduced by a

complementizer hore 6. This complementizer translates to English "that" or "whether" in

interrogatives. In some environments, the complementizer is optional in Sesotho.The subjunctive

occurs with various tense and aspect morphemes. The examples in (9) illustrate the use of the

subjunctive with tense and aspect morphemes.


6There is yet another use of the subjunctive which is not related to control verbs which is worth pointing out.
The morpheme hore, as pointed out by du Plessis (1989), is used to introduce embedded questions. This form is
selected by interrogative predicates in Sesotho. This fact becomes relevant in that it accounts for the lack of
interrogative control verbs in Sesotho (discussed in Chapter 4) suggesting a natural class. Also, like in Xhosa, hore
is formed with the infinitival prefix ho 'to' and the verb re 'say'. This may be taken as further evidence that hore is
somehow related to the infinitive as proposed by du Plessis (1989).
77









(a)Titjherej o kgothal-ets-a banak PRO*j/k/*ARB ho- bapala
1.teacher 1SM- encourage- CAUS .APPL -FV 2.children INF- play
'The teacher encourages the children to play'

(b) Titjherei o kgothal- ets- a PRO*i/ARB ho bapala
1.teacher 1.SM- encourage- CAUS .APPL INF play
'The teacher encourages playing'

Sentence (5a) is an example of ordinary object control. Sentence (5b) on the other hand

shows that when the object is omitted the experiencer-object verb forces an arbitrary

interpretation. Let us bear these referential relations in mind as we look at another type of control

shift.

The second kind of control shift obtains when there is a shift in control from the object of

the matrix clause to the subject of the matrix clause when an object argument of the matrix

clause is omitted in control relations. Subclass 2 of manipulative verbs participates in this kind of

control shift. The examples in (6) illustrate

(6) Subclass 2 manipulative verbs and control shift

(a) Sekoloj se -qobella batswadik PRO*j/k/*ARB ho- reka dibuka
7.school 7.SM- forces 2.parents INF-buy l0.books
'The school forces/compels parents to buy books'

(b) Sekoloj se-qobella PROj//*ARB ho- reka dibuka
7.school 7.SM-force INF- buy l0.books
Intended: 'The school forces/comples (itself) to buy books'

Example (6a) demonstrates the use of a manipulative verb qobella 'force/compel' where

the object of the matrix clause is the controller of PRO. When the object of the matrix clause is

omitted, as in (6b), the control shifts to the subject of the matrix clause. This kind of shift is only

possible with subclass 2 of manipulative verbs.

The kind of interpretation that we observe with subclass 1 applicative and experience

object verbs in examples (4&5) is treated as an instance of non-obligatory control (NOC). In this

instance the reference of PRO is freed from the arguments of the main clause in that it can refer









Sentence (3a) is an example of a manipulative verb whereas (3b) is that of an experiencer-

object verb. In each sentence the null subject of the embedded clause is understood as having the

same reference as the object of the matrix clause. When the object of the matrix clause is omitted

we get two important distinctions between the two classes. These differences are better discussed

under a configuration where the control is shifted from one argument to the other which may or

may not be related to one of the arguments of the matrix clause. These verbs shed some

important light on the distinction between the nominal infinitive and the clausal infinitive in

Sesotho. Although it appears as if the two are interchangeable in other contexts, when used with

control verbs this is not the case. A close look at the patterns of control associated with object

control verbs in Sesotho shows that the distinction between nominal and clausal infinitive is

justified. Let us look at control shift with manipulative and experiencer-obj ect verbs.

5.3.3 Control Shift

Control shift refers to a phenomenon whereby "constructions with two matrix arguments,

normally agent and goal...in normal circumstances the controller of PRO is fixed either as agent

or goal, in special circumstances the controller shifts to the other argument" (Landau 2000:183).

In Sesotho, there are two kinds of control shift associated with two object control verb classes.

The first one involves a shift of control from the object argument of the matrix clause to the

subject argument of the matrix clause. The second kind involves a shift of control from the

object argument of the matrix clause to an arbitrary control. Let us look at two kinds of control

shift individually.

Arbitrary control obtains when a subset of manipulative verbs (subclass 1) is used without

an object in the matrix clause. This set of verbs is made up of verbs that are morphologically

causative or applicative. The examples in (4) illustrate a shift from object control to arbitrary

control.










There are a number of reasons why we do not get the interpretation where PRO is

controlled by both arguments of the matrix clause. First, the notion of whether partial control is

an instance of obligatory control continue to be debated, however, if we treat it as a case of NOC

in Sesotho PRO would have an arbitrary interpretation. We have already noted that the arbitrary

interpretation associated with PRO in Sesotho obtains when the object of the matrix clause is

omitted. Partial control requires both arguments of the matrix clause. This itself would be a

problem for Sesotho.

Secondly, we noted that the interpretation of PRO is determined by the MDP. For each of

the sentences in (12), the MDP rules out the subject of the matrix clause as the controller since

there object is the closest c-commanding NP. This is indeed the case as indicated by the notation

*i, where 'i' refers to the subject of the matrix clause. The third point relates to the requirement

of the embedded clause. We have already established that the embedded clause is IP, as such, the

argument selection requirements of the verb have to be met. Reciprocal verbs require a singular

plural so PRO must have a plural referent. As demonstrated by sentences (12a&b) this

requirement is not met because the MDP excluded the possibility of joint controllers. Sentences

(1 Ic&d) further reveal that the embedded verb requires the plural controller. In (1 Ic) the object

of the matrix clause is plural whereas in ( ld) it is singular.

Finally we noted earlier that in Sesotho, the subjunctive is also selected by control verbs.

Sentence (11e) shows that in order to get the reading where the subject of the embedded clause is

controlled jointly by the arguments of the matrix clause, the subjunctive is required. In this case

there is no instance of control.









ACKNOWLEDGMENTS

I would like to express my sincere thanks to the Department of Asian and African

Languages and Literatures, the Center for African Studies and the Linguistics Department at the

University of Florida for the many years of financial support and intellectual stimulation. The

intensity of the course work coupled with the academic excellence offered by my teachers and

colleagues instilled a deeper love for learning. I would also like to thank my Professors at

University of Cape Town, Sizwe Satyo, Rajend Mesthrie, Derek Gowlett, and Sandile Gxilishe

for instilling the love for languages and linguistics. I owe the completion of this study to all

those whom I had an opportunity to work with.

I am indebted to my brother Neo Motinyane who spent hours on the phone listening and

responding to my endless questions. His skepticism about certain constructions led me to think

of ideas I would have not thought about. I would also like to thank Tshediso Maweng for sharing

his wonderful stories with me. I am as well indebted to my mother. As a native speaker

consultant, she provided me with grammaticality judgments untainted by any formal

grammatical awareness and rebuking me when using "not Sesotho-like" sentences. I am perhaps

more indebted to her for showering me with the motherly love, friendship and support that

sustained me throughout my stay in the United States.

I would also like to extend sincere thanks to Dr. Hunt Davis (Uncle Hunt) and Mrs. Jeanne

Davis (Aunt Jeanne) for opening their home and hearts to my family. I would have never been

able to come to University of Florida without their support. They provided us with a home away

from home. I also extend sincere thanks to their children Richard and Jonathan for accepting a

few more members into their family. I would also like to thank Dr. Carter Gilbert and Mrs.

Nancy Gilbert for always taking care of us whenever Uncle Hunt and Aunt Jeanne were out of

town. I am forever grateful to these two families for their unconditional love and support. I am

4










(11) Subject, object agreement and verbal extensions

(a) Banna ba mo- halef- is- a
2.men 2.SM 1.OM ngry- CAUS- FV
'Men make him/her angry'

(b) Banna ba rata ho mo halef- is- a
2.men 2.SM- like INF- 1.OM- angry- CAUS- FV
'Men like to make him/her angry'

(c) *Banna ba rata ba ho mo halef- is- a
2.men 2.SM- like 2.SM- INF 1.OM- anger- CAUS- FV

Example (12b) shows that the infinitival clause can be modified by an adverb of manner

haholo as well as a locative mosebetsing in the same manner that ordinary verbs are modified as

indicated by sentence (12a).

(12) Verbal Modifiers

(a) Banna ba mo halef- is- a haholo mosebetsing
2.men 2.SM- 1.OM angry- CAUS- FV very work.at
'Men make him/her very angry at work'

(b) Banna ba rata ho mo halef- is- a haholo mosebetsing
2.men 2.SM- like INF 1.OM- angry- CAUS- FV very work.at
'Men like to make him/her very angry at work'

Transitive verbs require an object. The fact that the verb used in examples (12a) show OM

suggests that this is a possibility. These examples clearly indicate that the infinitival clause

accommodates a direct object.

(13) Direct objects

(a) Banna ba halef- is- a mme
2.men 2.SM angry- CAUS- FV 1.mother
'Men make mother angry'

(b) Banna ba rata ho halef- is- a mme
2.men 2.SM- like INF angry- CAUS- FV 1.mother
'Men like to make mother angry'

Ordinary verbs inflect for tense, aspect and mood. Tense in Sesotho is divided according to

time: past (recent and remote), present and future (recent and remote). The only tense that the

46









Satyo, Sizwe. 1985. Topics in Xhosa Verbal Extensions. Doctoral dissertation, University of
South Africa, Pretoria.

Scida, Emily. 2004. The Inflected Infinitive in Romance Languages. Routledge. New York and
London.

Sean. Madigan. 2008. Obligatory split control into exhortative complements in Korean.
Linguistic Inquiry. 39(3): 493-502

Sigursson, Hallad6r A. 1991. Icelandic Case-marked PRO and the licensing of lexical arguments.
Natural Language & Linguistic Theory 9:327-363.

Simpson, Jane and Joan Bresnan .1983. Control and Obviation in Walpiri. Natural Language &
Linguistic Theory, 1( 1): 49-64.

Speas, Margaret. 1991. Functional heads and inflectional morphemes. The Linguistic
Review.8:389-417.

Stiebels, Barbara, Szymon Slodowic, Yi-Chun Yang, 2003. Questionnaire for control Verbs. 2nd
Version, Berlin.

Stiebels, Barbara.2007. Towards a typology of complement control. In Barbara Stiebels (ed.),
ZAS Papers in Linguistics 47. Studies in complement control. 1-80. ZAS Berlin.

Terzi, Arhonto. 1997. PRO and null case in finite clauses. Linguistic Inquiry 14:335-360.

Van Der Spuy, Andrew. 1997. Syntactic traces in Zulu. In African Linguistics at the
Crossroads:Papers from Kwaluseni. Ed. By R.K. Herbert: 297-309.

Visser, Mariana. 1989. The syntax of the infinitive in Xhosa. Sni.,l African Journal of African
Languages: 9(4):154-185.

Vic Webb and Kembo-Sure (eds.) 2000. African Voices. An Introduction to the Languages and
Linguistics of Africa. Oxford University Press. Cape Town, South Africa.

Williams, Edwin. 1985. PRO and subject NP. Natural Language & Linguistic Theory. 3: 297-
315.

Williams, Edwin. 1987. Implicit arguments, the binding and control. Natural Language &
Linguistic Theory. 5: 151-180.

Wurmbrand, Susi. 1998. Infinitives, Doctoral dissertation, MIT, Cambridge, Mass.

Wurmbrand, Susi. 2001. Infinitives: Restructuring and Clause Structure. Mouton de Gruyter,
Berlin.









NOC Non-obligatory control

OC Obligatory control

OM Object agreement marker

PASS Passive

PC Partial control

PC Possessive concord (agreement)

PFV Perfective (aspect)

PL Plural

PRO Null subject of infinitival clauses

PROG Progressive (aspect)

PST Past tense

RC Relative concord (agreement)

RECP Reciprocal (verbal extension)

REL Relative

SBJV Subjunctive mood

SG Singular

SM Subject agreement marker

TNS Tense









nouns in class 7 (Gowlett 2005:620) 1. This observation is specifically supported by the

irregularities associated with the plural form of nouns in class 6. Let us look at how plural

formation works within the noun class system.

Table 2-1. Sesotho noun classes
Noun Prefix Noun Gloss
Class
1 mo- motho,morena person, chief
la 0- mme mother
2 ba- batho people
2a bo- bomme mothers
3 mo- motse homestead
4 me- metse homesteads
5 le- leru cloud
6 ma- maru, marena, clouds, chiefs
mahobe2
7 se- sefate tree
8 di- difate trees
9 0- kgomo, nku cow, sheep
10 di- dikgomo, cows, sheep
dinku
14 bo- bohobe bread
15 ho- ho ja3 to eat/eating
16 ho- hodimo up
17 mo- morao behind


The pattern of singularity and plurality is fairly stable amongst the noun classes in Sesotho.

The plural of class 1 mo- nouns is formed by prefixing the class 2 noun class prefix ba-. There

are however a few exceptions. There are some irregularities found with the plural of some nouns

found in classes 1. The plural form of mo-rena 'chief, for example, in class 1 is not the expected

*ba-rena. Instead the plural of this noun is derived by prefixing class 6 noun prefix ma-. This



1 An example of nouns being merged include the use of noun class 5 prefix with the less productive locative nouns:
le-ho-dimo- 'sky/heaven'; le-fa-tshe 'earth'
2 There also nouns that have two plural forms: le-siba (sg) 'feather' ma-siba (pl) 'attached feathers (i.e. on a bird) or
di-tshiba (pl) 'loose feathers'

3 The nouns in class 15 are written as two free morphemes in Sesotho orthography.

35









1.2 The Area of Investigation

Sesotho, also known as Southern Sotho, is a South Eastern Bantu language spoken by

around four and a half million people in South Africa. It is one of the 11 official languages of

South Africa. As an official language, it is used in education, media, and government and has a

literature which spans over a century. Sesotho is also a national and official language of Lesotho.

Like many other Bantu languages, Sesotho has rich nominal and verbal morphology. One

of the morphological features of Bantu languages is the elaborate noun class system. Sesotho has

17 noun classes which occur in pairs of singular and plural. Class membership is determined

partly phonologically and partly semantically. Semantic criteria hold for about four of the noun

classes, whereas two of the seventeen classes represent locative classes. Noun class 15 has nouns

derived from verbs. The prefix of this class, ho has the same phonological form as infinitive

morpheme in Sesotho as the following examples illustrate:

(4) Bana ba sesa lewatle
2.children 2.SM swim ocean
'The children swim in the ocean'

(5) Ho- sesa ho kgathatsa haholo.
15-swim 15.SM- tire very
'Swimming/to swim makes you/one very tired'

(6) Ke rata ho sesa
1SG.- like INF- swim
'I like to swim/swimming'

Sentence (4) exemplifies the use of the verb sesa 'swim' with a noun class two subject.

The same verb is used in sentence (5) as a noun class 15 derived noun, hence the subject

agreement (SM) associated with this class precedes the verb kgathatsa 'tire.' In sentence (6) the

same verb is used with infinitive morpheme as a complement of the verb rata 'like.' In this

instance the complement of the verb rata 'like' may be realized as a phrase 'to swim' or as a

noun 'swimming.' Such data raise the following questions: How does the prefix of noun class 15


16









Like other nouns in Sesotho, class 15 nouns can be modified by descriptive modifiers such

as adjectives, demonstratives and possessive pronouns. The examples in (8) illustrate. Nominal

modifiers such as adjectives and demonstratives agree with the nouns they modify in Sesotho.

Sentences (8a&b) indicate how the adjective agrees with the nouns it modifies. Sentences (8c&d)

are examples of nouns used with personal possessive concords. The same kind of agreement

observed in (8a&b) is also applicable in (8c&d). Examples (8e&f) furthermore show that the

agreement relationship between the demonstrative and the noun it qualifies is also required.

Class 15 nouns behave like other nouns in this respect.

(8) Descriptive Predicates

(a) Makwala a ma- be
6.cowards 6.SM 6.AC- bad
'Cowards are bad'

(b) Ho bapala ho ho- be
15.play 15.SM 15.AC- bad
'Playing is bad'

(c) Bokwala ba hae bo -mmo- laisitse
14.coward 14.PC 1.his 14.SM 1.OM- kill
'His cowardice cost him his life'

(d) Ho bapala ha hae ho fedile
15.play 15.PC- 1.his 15.SM end
'His playing(game) has ended'

(e) Makwala ana a ma be
6.coward 6.DEM 6.SM- 6.AC- bad
'These cowards are bad'

(f) Ho bapala hona ho ho- be
15.play 15.DEM 15.SM 15.AC- bad
'This playing is bad'

Object nouns also offer an interesting trait in Sesotho. Object nouns trigger object

agreement when the emphasis is placed on the object or when the object is proposed or omitted.

Whereas proposing is possible with class 15 nouns, it appears that object agreement is not









whose plural form falls into noun class 6. The majority of nouns in this class are abstract nouns,

resulting in most of these nouns lacking plural forms.

In addition to the complete loss of classes 11, 12 and 13, Sesotho has lost most of the

nouns that belong to locative classes. Classes 16 and 17 have a very small number of nouns.

These locative nouns are often referred to as adverbs of place (Doke & Mofokeng 1957:83).

Besides the fact that these noun classes have different prefixes, they all trigger the same

agreement morpheme on predicates. It is interesting to note here that this agreement morpheme

for the locative classes is ho. The number of nouns remaining in these classes as well as the

similar agreement morpheme associated with the locative classes has generated debates amongst

Bantuists regarding their status within the noun class system. Mchombo 1993 & 2004 for

example, questions the nominal status of the items in these classes in Chichewa.

One more class that has also attracted a lot of attention is noun class 15, the infinitival

class, which contains verbs in their infinitival form or verbal nouns. The prefix of this noun class

is ho. In Sesotho, the nouns found in this class are derived from verb stems. Whereas Nguni

languages such as Xhosa have proper nouns and nouns derived from verbs in class 15, at first

glance Sesotho does not show this two-way distinction. When 'nouns' in class 15 are subjected

to various tests, they reveal some form of ambiguity (see Creissels & Godard 2005 for

Setswana). I discuss this distinction later in this chapter. First let us examine nominal

morphology, specifically the formation of nouns from verbs.

2.2.2 General Nominal Derivation

The noun prefixes in Sesotho are central to nominal derivation and predicate agreement.

The noun prefix may be affixed to various lexical categories to derive new nouns. In the process

of noun formation the noun class prefixes generate new words while marking number and gender

(in this case noun class) at the same time. The process of nominal derivation is crucial in

37









'Mother promises the children to work for her'

(5) #Mmej o tshepisa banak PROj/*k ho ba- sebel-ets-a
1.mother 1.SM- promise 2.children INF 2.OM-work-APPL-FV
'#Mother promised the children to work for them'

(6) Mme o tshepisa PROj ho ba sebeletsa bana.
1.mother 1.SM- promise INF -2.SM- work-APPL-FV 2.children
'Mother promises to work for the children'

The introduction of the applicative suffix on the verb of the embedded clause as well as the

object agreement (OM) determines which NP controls PRO. In (4), the OM refers to the

recipient mme. This subject cannot be the recipient and agent of the same verb at the same.

However, we would expect (5) to be acceptable because the recipient is bana hence the OM

agrees with this NP, the agent is mme. The sentence is odd under the reading that PRO is co-

referenced with mme. The reason for (5) being odd is that this verb allows subject control, then

the clause containing PRO needs to get closer to the controlling NP, hence (6) is a perfect

sentence in Sesotho.

Two things come out of these data in (2-6). One is that the MDP seems adequate to

account for control relations in Sesotho complement control. The other is that an approach which

does not take into account the morphology associated with the main verb as well as the

embedded one would pose a challenge for the MDP.

6.3 Concluding Remarks

Most studies on control constructions have been based on the more widely studied

languages. The approaches to control mentioned in the previous section are based on many years

of research. As I hope to have shown throughout this study, Sesotho does not quite work well

with the various tests for empty categories for example. In Sesotho when you move the object

NP, the moved NP is replaced with the OM which keeps the connection, more like leaving a

trace. Sesotho control provides a window into languages that make use of verbal morphology to

116











Example (20) demonstrates object control. When both the subject and object of the matrix

clause are present the subject of the embedded clause (PRO) is co-referenced with the object of

the matrix clause. In the absence of the object in the matrix clause, the co-referent of PRO is

either arbitrary or an implied object as exemplified by sentence (21). The arbitrary reading

associated with an object is claimed to be a rare occurrence (Stiebels 2007:73). It is also

important to note that in the absence of object, the subject of the matrix clause is not understood

as the controller of the subject of the embedded clause. This is how the arbitrary reading is

associated with the object of the matrix clause. This is another distinguishing characteristic of

experience object verbs.

Table 4-5. Experiencer-object verbs
Verb Gloss Subject Object Control Causative Applicativ
Control Control Shift e
kgothaletsa cheer/applaud N Y N Y Y
kgotsofaletsa satisfy N Y N Y Y
hlonamisetsa sadden N Y N Y Y
thabisetsa amuse N Y N Y Y


Experiencer-object verbs behave in the same way as manipulative verbs with regards to the

complement selection. This is expected because manipulative verbs are also object control verbs.

Experiencer-object verbs occur more with the subjunctive regardless of whether they are used

together with the objects or not. The examples in (23) illustrate.

(23) Experiencer-object verbs with the subjunctive

(a) Mmuso o kgothaletsa bana hore ba bu-e Sesotho
3.government 3.SM- encourage 2.children that 2.SM- speak-SUBJ Sesotho
'The government encourages the children that they should speak Sesotho'

(b) Mmuso o khothaletsa hore ba bu-e Sesotho
3.government 3.SM- encourage that 2.SM speak-SUBJ Sesotho
'The government encourages (them) that they should speak Sesotho'

89









B. A pronominal is free in its governing category
C. An R-expression is free
(27) is bound by 3 if and only if:
(a) and 3 are co-indexed, and
(b) c-commands V
(28) is free only if it is not bound
(29) C-command (Chomsky 1986b:8)
A c-commands B iff A does not dominate B and every branching node that
dominates A also dominates B.

To illustrate how binding works, consider the following data from Davies & Dubinsky

(2004: 186) which illustrates the complementary distribution of anaphors and pronouns.

(30) The distribution of anaphors and pronouns
(a) The children soiled themselves.
(b) *The children soiled them.
(c) *The children claimed that the teacher scolded themselves.
(d) The children claimed that the teacher scolded them.

Sentence (30a) satisfies principle A of the binding theory in that the children c-commands

ri/,\hi'e'\ and the reflexive is bound within the minimal clause. Sentence (30b) violates

principle B of the Binding theory in that the pronominal must be free in the minimal clause but it

is not. In (30b) them is bound by the children under the indicated co-indexation. Sentence (30c)

violates principle A in that the antecedent, the children is in the matrix clause, whereas

//wr,\ei'e.\ is in the subordinate clause. Thus, lr/i/\h'e'\ is bound but not in its governing

category. Sentence (30d) is grammatical because it satisfies Principle B. The pronoun is free in

its minimal clause.

The Binding theory, Government theory, Case Theory, the Extended Projection Principle,

and the Theta Criterion are important for this study because of their interaction with Control

Theory. Control Theory determines the interpretation and distribution of PRO, the null subject of









Sentence (8a) is ambiguous because the complement of the verb rata 'like' is either NP/IP.

The difference between the two is that if PRO is co-indexed with the subject bana 'children'

then the complement is an IP (clausal infinitive). If PRO gets an arbitrary interpretation, the

complement is an NP. When we introduce modifiers such as possession, as in (8b) we note that

PRO can only get an arbitrary interpretation because this is the interpretation associated with

nominal infinitives. This arbitrary interpretation is also observed in (8c), when the nominal

infinitive is used as a subject.

5.4 Referential Dependencies

Another aspect of control that is regulated by a combination of class membership as well

as the morphology of the verbs is the referential relation between the controller and the

controlled. In this section I look at how the verbal extensions impact on referential dependencies.

Referential dependencies refer to the differences in control verbs regarding their potential

referential dependency between the controller (subject/object of control verb) and the controlled

(null subject of the infinitive, PRO) (Stiebels et al. 2003). The two major referential

dependencies are obligatory control (OC) and non-obligatory control (NOC). In this study, I

focus on exhaustive control (EC) and partial control.

Exhaustive control obtains when the subject or object of the matrix clause is understood as

the same as the null subject of the embedded infinitival clause. Partial control is a control

relation in which the interpretation of PRO properly includes the referent of the controller instead

of being identical to the controller. The examples we have used so far are examples of exhaustive

control which have a locality requirement between the controller and controlled. This locality

requirement is regulated by the MDP (Minimal Distance Principle) which determines the

interpretation of PRO. According to the MDP, PRO must be co-indexed with the closest NP that









linguistic typology (Landau 2003, 2006). Analyses of control in Bantu languages are in their

infancy so it is difficult at this point to indicate which direction my findings point. What I would

like to do at this point is show how the MDP, agreement specifically (OM) and verbal extensions

work together to determine the reference of PRO.

The MDP proposes that in obligatory control constructions the controller must c-command

PRO and that the controller must be the closest c-commanding NP. One set of the verbs that is

usually cited as a cause for concern for the MDP is the promise type set in which the subject of

the main clause controls across the object. The example in (1) illustrates:



(1) Samuelj promised Maryk [PROj/*k to cook dinner ]



Sentence (1) clearly indicates a violation of the MDP since the closest c-commanding NP

is Mary. Sesotho presents a slightly different problem. When tshepisa 'promise' is used without

the object, PRO is understood as having the same reference as the subject of the matrix clause.

However, when the object is introduced the results are mixed. PRO can either refer to mme or

bana in (2&3). This is indeed a problem for the MDP. However, we have already noted that

Sesotho uses the verbal morphology to account for concepts that can be dealt with syntax in

other languages.

(2) Mmej o tshepisa PROj ho- sebetsa
1.mother 1.SM promise INF work
'Mother promises to work'

(3) Mmej o tshepisa banak PRO?j/?k ho- sebetsa
1.mother 1.SM promise 2.children INF- work
'Mother promises children to work'


(4) Mmej o tshepisa banak PRO*j/k ho mo sebel-ets-a
1.mother 1.SM promise 2.children INF 1.OM- work-APPL-FV

115









'remember' are subject control verbs in Sesotho. However, when the causative affix is added as

in hopotsa 'remind', the verb becomes a manipulative verb, which in turn is an object control

verb. The only manipulative verbs that allow control shift from object to subject in Sesotho are

the ones that cannot take either the causative or the applicative extension since these are

inherently applicative. These verbs, although they may appear as f they contain an applicative

affix, this affix cannot be separated from the 'root' of the verb. Let us look at some examples.

(7) The applicative extension

(a) Ntate o- pheh-a dijo kantle
1.father 1.SM- cook--FV 10.food outside
'Father cooks the food outside'

(b) Ntate o- pheh- el- a bana dijo kantle
1.father 1.SM-cook- APPL-FV 2.children 10.food outside
'Father cooks the food outside'

(c) Ntate o- pheh- el- a kantle
1.father 1.SM- cook- APPL-FV outside
'Father cooks outside'

(d) *Ntate o pheh-el-a
1.father 1.SM- cook-APPL-FV
Intended 'Father cooks'

(e) Ntate o -pheh-is-a bana dijo kantle
1.father 1.SM-cook-CAUS-FV 2.children l0.food outside
'Father makes the children cook food outside'

(f) *Ntate o- pheh- is-a kantle
1.father 1.SM-cook -CAUS-FV outside
Inteded 'Father causes some cooking outside'


(g) Ntate o pheh-is- ets-a kantle
1.father 1.SM-cook- CAUS- APPL-FV outside
Inteded 'Father causes some cooking outside'


Example (7a) demonstrates the use of the transitive verb pheha 'cook' without the

applicative extension. In (7b) we note that when the applicative extension is added another









CHAPTER 4
SESOTHO CONTROL VERB CLASSES

4.1 Introduction

One of the questions raised in Chapter 1 seeks to answer the observed differences in

argument selection by verbs which share the same semantic class. I noted that verbs such as

hopotsa 'remind' and qophella 'insist' share some semantic similarities. The expectation in

this regard is that these verbs share the same syntactic behavior. I pointed out that this

expectation is not met as exemplified by these sentences.

(1) [matrix Mme i o hopotsa bana [subordinate PRO i ho sebetsa]]
1.mother 1.SM remind 2.children to work
'Mother reminds children to work'

(2) [matrix *Mme i o hopotsa [subordinate PRO i ho sebetsa]]
1.mother 1.SM remind to work
'Mother reminds (someone) to work'

(3) [matrix Ntate i o qophella [subordinate PRO i ho sebetsa]]
1.father 1.SM insist to work
'Father insists on working'

(4) [matrix Ntate i o qophella banaj [subordinate PRO *i/j ho -sebetsa]]
1.father 1.SM insist 2.children to- work
'Father insists that children work' Lit: 'Father forces children to work'


Sentences (1) and (4) show the expected similarities in syntactic behavior between the two

verbs. Sentences (2) and (3) show some differences in syntactic behavior. In what way do we

capture the similarities and differences? If argument structure can be derivable from the meaning

of verbs, what aspects of the verb's 'meaning' need to be captured in order to determine the

syntactic properties? This question is central to the ongoing debates on lexical entries and to

what extent they determine syntactic properties, an area which affects various areas of linguistic

research. In this chapter I deal with the interaction between semantics, morphology in the









dissertation. The infinitive is introduced by ho which is also the prefix of class 15 nouns. This

study compares the status of ho as an infinitival morpheme and ho as a noun class prefix in

Sesotho. I suggest that the morphology of the control verbs together with the referential relations

between the controller in the matrix clause and the subject of the infinitive help determine the

differences between nominal infinitive (class 15 nouns) and clausal infinitives.

This study also proposes that argument- structure changing verbal morphology is important

in the classification of control verbs in Sesotho. This morphology is also responsible for the

syntactic properties of control and directly accounts for the types of control observed in Sesotho.

Contrasting this analysis with a typology of control patterns from other languages, the lack of

partial control or split control is furthermore directly linked to argument-structure changing

morphology.









has resulted in a split between nouns that fall into class 2 ba- in plural and those that fall into

class 6 ma- for their plural form. It is also important to note that some of the nouns in class 1 that

take ma- as a plural prefix also have a singular form found in class 5. Let us look at some

examples.

Table 2-2. Noun class 1 plural formation
Noun class 14 Noun class 55 Noun class 6 Gloss
Mo-rena ---------- Ma-rena King/chief
Mo-fumahadi ---------- Ma-fumahadi Queen/wife
Mo-qhotsa Le-qhotsa Ma-qhotsa Xhosa speaking
Mo-swatsi Le-swatsi Ma-swatsi Swati speakers
---------Le-tebele Ma-tebele Ndebele speaker


Table 2-2 indicates that nouns that take ma- as a plural prefix are nouns associated with

speakers of other languages such as the Nguni languages of South Africa. The nouns

representing people who are regarded as 'socially closer' to the Sesotho speakers such as Mo-

tswana/Ba-tswana 'Tswana speaking persons' and Mo-pedi/Ba-pedi 'Pedi speaking persons'

use the plural prefix ba- associated with the Sesotho speaking people. The important thing to

note about these examples is that they may be used in a way which conveys speakers' attitudes

towards the people expressed through these nouns. These examples reinforce the suggestion that

a multitude of factors are considered in noun classification (see Mc Laughlin 1997 for this

approach).

A similar but slightly different pattern occurs with bo-hobe 'bread' in class 14. The plural

of bo-hobe 'bread' is also formed by prefixing class 6 noun class prefix ma-. Nouns in class 14

differ from those in class 1 in that the split is between nouns that lack the plural form and those




4 Nouns associated with people
5 This prefix tends to have a derogatory function when used with nouns that refer to people.

36









This is called the sloppy reading. Sentence (40) on the other hand shows a different pattern. The

omitted part in this case is a sebetse. Note however that the SM of the omitted verb still refers to

the original subject. This is called strict reading. What this tells us is that the subject of the

infinitive takes its interpretation from the nearest possible NP, in which case bana becomes the

closest. This is due to the fact that there isn't an SM occupying this position in the sense that (40)

does. This is another indication that the subject NP of the infinitive is indeed different from other

noun phrases.

2.6 Nominal and Clausal Infinitive

One of the questions that continue to be debated in Bantu literature concerns whether the

distinction between clausal infinitive and nominal infinitive is valid.2 In the preceding sections I

indicated that the nominal infinitive shows both nominal and verbal properties. I have also

demonstrated that the nominal infinitive occupies the positions that other NPs occupy. The

challenge lies in the instances where there seems to be syntactic ambiguity as presented in (41).

(41) Mme o- rata ho-ja.
1.mother -1.SM -like INF- eat
'mother likes to eat/eating'

Sentence (41) as it stands cannot give us much to say about whether ho-ja is a nominal or

clausal infinitive. Since the nominal infinitive seems to occur in all positions that other nouns

occupy in Sesotho one way would be to treat nominal infinitives like other nouns in terms of

syntactic structure. The ambiguity in (41) would be represented as in 2-1(b), a structure similar

to other nouns (2-1(a)).

The immediate problem with assuming the structure in 2-1(b) for the nominal infinitive is

that it does not capture the observed differences between the nominal infinitive and other nouns



2 Doke & Mofokeng 1957, Rugege 1971, Du Plessis 1982b, Visser 1989, Mugane 2003, Creissels & Godard 2005

59









IP

DP I'
A
Mmei I VP
A
o DP

Mme- VP IP

V1
DP I'
A
V DP PROi/j I VP

A ho A
kgothaletsa banaj
bala dibuka

Figure 1-2. MDP and object control

In 1-1 there is only one antecedent DP which c-commands PRO, Bana. Figure 1-2 presents

a different configuration in that there are two c-commanding DPs. However, since the closest c-

commanding antecedent is object of the matrix clause, PRO is then co-indexed with bana3.

Finally, the distribution of PRO is regulated by PRO Theorem stated in (34):

(34) PRO Theorem
PRO must be ungoverned

The PRO theorem restricts PRO to occurring in ungoverned and Caseless positions. For

our purposes, this includes the subject of infinitival complements. The PRO theorem follows

from the Binding Theory in the following way. Chomsky (1982) observed that binary features,

3It is important to note here that the MPD has always been challenged in generative grammar for its inability

to cater for Subject control verbs (see Chomsky 1981, Comrie 1984, Koster 1984, Larson 1990 Sag and Pollard
1991 amongst others).


































To my parents: Daniel Teboho and Limakatso Johanna Motinyane
Robalang ka kgotso Motaung le Mohlakoana.









(c) Mme o kgetha hore a tsamay- e
1.mother 1.SM- prefer that 1.SM leave- SUBJ
'Mother prefers that he/she/someone leaves'

Sentence (10a) shows an example of the subjunctive used where the subject of the

subjunctive is different from that of the matrix clause. Sentences (lOb&c) differ in that in (lOb)

the subject of the embedded clause is only understood as having the same reference as that of the

main clause (control). Sentence (lOc) on the other hand shows that in non-control the subject of

the subjunctive can be understood as having the same reference as that of the matrix clause but

not necessarily. The infinitive does not have the same morphology (such as SM) required to

make this distinction between the subject/object of the main clause and the subject of the

embedded clause. The subjunctive becomes the better option when this distinction is required.



4.2.2 Manipulative Verbs

Manipulative verbs express control or manipulation by the subject. Manipulative verbs in

Sesotho are the extended or the derived form of the bare desiderative verbs. The ability of a

desiderative verb to undergo derivation and act as a manipulative verb also depends on the

semantics of each verb. Manipulative verbs can be divided into three subclasses: those that are

morphologically causative, those that are morphologically applicative and those that are bare

forms inherently applicative. I have decided to merge the morphologically causative and

applicative control verbs here based on the shared syntactic property of object control.

I will label these two classes subclasses 1 and 2. Subclass 1 comprises of verbs that are

object control verbs with no possibility of control shift. Subclass 2 is made up of a small number

(from the ones I found) of verbs that are inherently applicative in form. They cannot be further

broken down in the same way as the verbs of the first sub-class. Finally subclass 2 of









CHAPTER 6
DISCUSSION AND CONCLUSION

6.1 Summary

In this study I embarked on a classification of control verbs in Sesotho. I identified 5

semantic classes: Achievement, active, desiderative which are subject control verbs and

experiencer-object and manipulative which are object control verbs. In order to capture all the

syntactic and morphological properties associated with these classes I established a typology of

control in Sesotho. As a way of understanding control, I concentrated on three major questions:

Question 1: What is the status of ho in Sesotho?

Question 2: Are there different kinds of controlled complements and, if so, what are they?

Question 3: What role does argument-structure changing verbal morphology play in control

phenomena?

The answers to these questions were answered in the various preceding chapters. In

Chapter 2 I explored the morpho-syntactic properties of nouns and verbs in Sesotho. I examined

the nominal infinitive (class 15 nouns) in relation to other noun classes. I showed that the

nominal infinitive demonstrates all properties associated with other nouns although it fails to

pluralize or trigger OM as an object. I compared this pattern to other noun classes (such as the

locative) that share the same prefix and agreement morphology. I also noted that the nominal

infinitive prefix ho is capable of deriving nouns from verbs, like other noun class prefixes.

However, unlike other noun derivations, the suffix of the derived noun is always a, which I

directly linked to the verbal morphology observed with nominal infinitives.

I also reviewed the treatment of the infinitive in the literature and showed that the various

scholars agree that the infinitive is both a noun and a verb at the same time. I took this a step

further by showing that nominal infinitives are more restricted as objects because they are









Machobane, 'Malillo 'Matsepo. 1996. The Sesotho Locative Alternation Verbs. .Snlu African
Journal of African Languages 16(1):8-15.

Machobane, 'Malillo 'Matsepo. 1997. Sesotho control verbs with applicative suffix. .Siuh
African Journal ofLinguistics 15(2): 59-64.

Machobane, Malillo Matsepo. 2001. Questions in Sesotho. Afrikanistische Arbeitspapiere
66:67-83.

Manzini, M. Rita. 1983. On control and control theory. Linguistic Inquiry 14:421-446.

Manzini, M. Rita, and Anna Roussou. 2000. A minimalist theory of A-movement and control.
Lingua 110: 409-447.

Marantz 1984. On the Nature of Grammatical relations. MIT Press, Cambridge Mass.

Mchombo, Sam. 1993. Theoretical aspects of Bantu grammar. Stanford, California: CSLI.

Mchombo, Sam .2004. The Syntax of Chichewa. Cambridge Syntax Guides Cambridge
University Press.

Mc Laughlin, Fiona. 1997. Noun classification in Wolof: When affixes are not renewed. Studies
in African Linguistics. Vol.26(1).

Mc Laughlin, Fiona. 2000. Consonant mutation and reduplication in Seereer-Siin. Phonology.
17:333-363.

Guthrie 1967 Guthrie, M. 1967/71. Comparative Bantu, vol. 1-4. Gregg International Publishers.

Miller, D. Gary. 2002. Non-finite structures in theory and change. Oxford University Press.

Miller, D. Gary. 2006. Latin Suffixal Derivatives in English and their Indo-European Ancestry.
Oxford University Press.

Mitchell Erica. 1994. When Agro is fused to Agrs: What morphology can tell us about functional
categories.MIT Working Papers in Linguistics. 111-130.

Mkwanazi. M.J. 1989. Kiriatshwana .S, nlhei n Sotho Poetry. Educum Publishers.

Moeketsi, R.H. (ed.).1996. Seyalemoya. An anthology of dramas written in Sesotho. Van Schaik
Pretoria.

Mofolo, Thomas. 1981. Chaka. Via Afrika.

Mohanan, K.P. 1985. Remarks on control and control theory. Linguistic Inquiry 16: 637-648.

Morimoto, Yukiko.2006. Agreement properties and word order in comparative Bantu. ZAS
Papers in Linguistics. 43:161-188.









derivation of control verbs, and argument structure and the role each plays in the classification of

control verbs in Sesotho.

4.2 Verb Classification Methodology

The question of the extent meaning plays in determining syntactic structure is debated in

many areas of linguistic theory. Within the area of Lexical Functional Grammar there is a move

towards the argument that the syntactic behavior of a word may be "fully" semantically

determined (See Levin and Rappaport Hovav 1991). Levin 1993 proposes that verb meaning is

key to its syntactic behavior in the same way that verbs that fall into the same classes according to

their shared behavior also show shared meaning components. In this framework, argument

structure may be derivable from the meaning of verbs.

Within the Minimalist framework there are also various approaches to the relation between

meaning and argument structure. One such approach claims that properties of control verbs

similar to the ones in (1-4) can be obtained through syntactic means (Hornstein 1999, 2003,

Boeckx & Hornstein 2006, amongst others). These are the scholars who treat control as an

instance of movement thereby doing away with PRO in obligatory control. Others claim that

there needs to be some reference to the semantic properties of the verb in question and that

semantic classes help to account for the different syntactic behaviors associated with control

(Cullicover & Jackendoff 2001, 2003, 2005, 2006). There are yet other scholars, while

supporting a different syntactic approach, who claim that purely semantic or purely syntactic

approaches to obligatory control are inadequate (Landau 2003, 2006). In this approach

morphological agreement and semantic tense is claimed to account for a cross-linguistic

typology of control. While my classification borders on a combination of a syntactic (control

class) and a semantic (semantic verb classes) approach, I also use verbal extensions

(morphology) as an extended means of classification.

72









forward control and backward control. In forward control, the controller is in the matrix clause

whereas the controlled is in the embedded clause. In backward control, the controller is in the

embedded clause whilst the controlled is in the matrix clause (Polinsky & Potsdam 2006,

Stiebels 2007). In another form of control, non-obligatory control, there is no obligatory co-

indexation between the subject of the subordinate clause and the subject or object of the main

clause. This study concentrates on obligatory control (OC) although I also discuss some aspects

of referential dependencies associated with non-obligatory control (NOC) where relevant.

Obligatory control relations observed cross-linguistically include subject and object

control. Subject control refers to a control relation in which the overt subject of the matrix clause

is identified with the covert subject of the embedded infinitival clause through co-indexation. In

object control, the overt object of the matrix clause is identified with the covert subject of the

embedded infinitival clause also through co-indexation. Implicit control on the other hand refers

to a control relation in which the controlling argument is not syntactically realized. Co-

indexation is not required in this instance. These are exemplified in (1) in this order: implicit,

object and subject control

(1) The controller8
(a) Mary signaled PRO to follow her
(b) Mary asked Johnj PROj to leave
(c) Maryi promised John PROi to leave

Sentence (la) is an example of implicit control. In this sentence the reference of PRO is

implied. In sentence (lb) PRO's controller is the object of the matrix clause, resulting in object

control. Sentence (Ic) is an example of subject control. In all the sentences in (1) the referential





8 I have decided to use English examples in this section in order to provide an example of typology of control. I use
Sesotho examples later in the chapter.









LIST OF REFERENCES

Adger, David. 2003. Core syntax: A Minimalist approach. New York: Oxford University Press.

Boeckx, Cedrick, and Norbert, Homstein. 2003c. Reply to "Control is not movement".
Linguistic Inquiry 34: 269-280.

Boeckx, Cedrick, and Norbert, Homstein. 2006. The virtues of control as movement. Syntax.
(9)2:118-130.

Bresnan, Joan. 1972. The theory of complementation in English Syntax. Doctoral dissertation,
MIT, Cambridge, Mass.

Bresnan, Joan. 1982. Control and Complementation. Linguistic Inquiry 13:343-43.

Bresnan, Joan and Sam Mchombo. 1986. Grammatical and anaphoric agreement.In CLS 22.
Vol.2 (278-297). Papers from the Parasession on Pragmatics and Grammatical Theory. ed.
by Anne, Farley, Peter, T. Farley and Karl, Mc Cullough. Chicago Lnguistic Society.
University of Chicago.

Brody, Michael. 1999. Relating syntactic elements: Remarks on Hornstein's Movement and
chains. Syntax 2:210-226.

Brody, Michael, and M. Rita Manzini. 1987. Implicit arguments. In Mental Representations: The
interface between language andreality. ed. by Ruth Kempson, 105-130, Cambridge:
Cambridge University Press.

Buell, Leston.2007. Semantic and formal locatives: Implications for the Bantu locative inversion
typology. SOAS Working Papers in Linguistics. 15:105-120.

Cecchetto, Carlo and Oniga Renato. 2004. A Challenge to Null Case Theory. Linguistic Inquiry
35 (1):141-149.

Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, MIT.

Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris.

Chomsky, Noam. 1982. Language and the Study of Mind. Tokyo: Sansyusya Publishing Co. Ltd.

Chomsky, Noam. 1986a. Barriers. Cambridge, Mass.: MIT Press.

Chomsky, Noam. 1986b. Knowledge ofLanguage. New York: Praeger.

Chomsky, Noam. 1995. The Minimalist Program. The MIT Press. Cambridge, Massachusetts,
London, England.

Chomsky, Noam. 2000. Minimalist inquiries: the framework. In Step by Step, ed. Roger Martin
and David Michaels and Juan Uriagereka, 89-155. Cambridge, Mass.: MIT Press.









Sentence (4) is an example of a basic transitive verb with two arguments. In sentence (5)

the addition of the causative extension introduces an additional argument. Example (6) indicates

that the order of the post-verbal arguments is fixed. The order of the agreement markers is also

fixed. The OM, when present is always closest to the verb. Example (7) demonstrates that only

the argument selected by the causative triggers object agreement. It is also this argument that is

fronted in passive formation as indicated in (8).

In control structures when the causative is used with the matrix verb, it changes the

syntactic and semantic properties of the verb thereby allowing a subject control verb to function

as an object control verb. As can be observed from the examples, in English the causative is

represented by a verb, 'make' which selects for an infinitival complement. In Sesotho, the

causative makes changes to the form of the verb. In so doing, a verb which is normally a subject

control verb may optionally function as an object control verb although this changes the meaning

of the verb. When used with the verb in the embedded clause, the causative has no direct effect

on control relations.

3.3.2 The Applicative

The applicative, also called applied extension, is represented by a suffix morpheme /-el/

with allomorphs [-el], [-il] and [-1]. The distribution of these allomorphs is phonologically

conditioned. The applicative suffix in Sesotho is associated with the introduction of benefactive

and locative roles. The benefactive role is more productive since Sesotho marks locative with

other morphemes. Another semantic role associated with the applicative in Bantu is the

instrumental role. The instrumental role in Sesotho is assigned by the instrumental morpheme ka.

Sentences (9-15) are some examples of the applicative in Sesotho.

(9) Banana ba pheha bohobe
2.girls 2.SM cook 14.bread
'Girls cook bread'









2.3.1 Distributional Properties of Nouns

Customary discussions of noun classes in Bantu tend to pay little attention to the

differences between class 15 nouns and the rest of the noun classes. This view tended to diminish

the importance of the many occurrences of class 15 nouns in syntactic positions which other

nouns do not occupy. Certain tests are used cross-linguistically to test for word-class

membership. These include the ability of that particular word to act as either a subject or an

object in a sentence, the ability to trigger subject or object agreement, the ability to be modified

by descriptive modifiers as well as being able to pluralize. Nouns in class 15 pass most of these

tests. Let us examine these various tests by starting with class 15 nouns as subjects and objects in

a sentence.

(7) Nouns as subjects and objects in Sesotho

(a) Makwala a timela kapele
6.coward 6.SM die quickly
'Cowards die quickly'

(b) Ho nanara ho a tshosa
15. sneak.up 15.SM- FOC- scary
'Sneaking-up (on a person) is scary'

(c) Bana ba rata ho-ja
2.children 2.SM like 15.eat
'Children like eating/to eat'

(d) Bana ba rata dijo
2.children 2.SM like 10.food
'Children like food'

Subjects trigger subject agreement (SM) in Sesotho. Sentence (7a) shows that the subject

noun makwala 'cowards' triggers agreement on the verb a-. We observe the same pattern with

class 15 nouns indicated by sentence (7b) where the noun ho nanara triggers the SM ho.

Sentence (7c) shows that class 15 nouns also act as objects in the same way that the object noun

in (7d) does.









BIOGRAPHICAL SKETCH

Mantoa Rose Smouse was born in Bloemfontein, South Africa. She is the second youngest

of six children. She grew up in Zastron, a small town on the borders of Lesotho, the Free State

and the Eastern Cape Province. She attended elementary school on a farm school near Zastron,

Diepfontein. She then moved on to do her middle school in the Eastern Cape, then called

Transkei. She completed her High School in 1989 at Lere La Thuto in Zastron. Mantoa earned

her Bachelor's Degree, Higher Education Diploma and Honor's Degree from the University of

Cape Town in 1994, 1995 and 1998 respectively and a Master's Degree from the University of

Florida in 2004.

Upon graduating in 1995, Mantoa started teaching at Rusternburg Girls' High School in

Cape Town. Whilst teaching at Rustenburg she enrolled to do her graduate studies at the

University of Cape Town. In 2000 she was awarded a Prestigious A.W. Mellon Scholarship

which afforded her an opportunity to pursue graduate studies at a university in the United States.

Mantoa is currently teaching at University of Cape Town in South Africa. She has been

married to Simon Smouse for 18 years. They are blessed with three children: Naledi, age 17,

Tlhodiso, age 13 and Mapule age 4.









verbs that also happen to share semantic and syntactic properties as well as a small overlap of

morphological properties. The common syntactic property shared by these verbs is that they are

all subject control verbs. The morphological property, the ability for verbs to interact with

derivational suffixes (in this case the causative and applicative), depends on the meaning of each

verb.4 The examples in (5&6) indicate the use of these verbs.

(5) Banaj ba ikem-is- ets- a PROj ho tsamaya
2.children 2.SM intend-CAUS-APP-FV INF go
'The children intend to go'

(6) *Banai ba ikem-is- ets- a mmej PROj ho- tsamaya
2.children 2.SM- intend-CAUS-APPL-FV 1.mother INF go
'The children intend for mother to go'

In sentence (5) the subject bana 'children' is co-indexed with the null subject of the

embedded infinitival clause PRO. In sentence (6) however, since verbs from this class only allow

subject control, the object of the matrix clause mme 'mother' is ruled out as the controller of the

embedded clause, in fact these verbs do not even allow an object in the main clause in control

relations.

Table 4-1 also shows that desiderative verbs can interact with the causative and applicative

extensions. The applicative extension introduces a benefactive role (an object) whereas the

causative extension introduces a patient role (object). The applicative form of the desiderative

verbs is realized as manipulative verbs in Sesotho. Sesotho control verbs select different kinds of

complements. The choice is between a noun phrase (including a nominal infinitive), a

subjunctive clause (introduced by hore) and a clausal infinitive. In Chapter 2 we established that

the nominal infinitive shares syntactic positions with ordinary nouns in Sesotho. However, there

are some restrictions when it comes to the clausal infinitives. One of the reasons for this

4 As I indicated in Chapter 3 this morphological process has to be treated with caution because there have been
instances in the literature (Machobane 1997) where the derived form of the verb has been mistakenly analyzed as
allowing control shift









(a) Mmek o holy-a PROk ho fihla bosiu
1.mother 1.SM hate -FV INF- arrive 14.night
'Mother hates to arrive at night'


(b)Mmek o hlo- ile PROk ho fihla bosiu
1.mother 1.SM hate- PERF INF arrive 14.night
'Mother hates to arrive at night'


(c) Ke k- thaba PROk ho utlw-a ditaba tseo
1SG -glad INF hear-FV 9.news 9.DEM
'I am glad to hear those news'


(d)Bananak ba thab- el- a PROk ho dula lape-ng
2.girls 2 .SM- glad.APPL-FV INF -stay home-LOC
'The girls are glad to stay at home'

Sentences (16a&b) contrast the use of the bare form of the verb (expressed in present tense

habitual) with the perfective counterpart indicating no change in meaning or control relations.

Sentences (16c&d) then again, indicate that other verbs can be used in the bare form or

applicative without any change in meaning or control relations7. These verbs however, are

distinguished from the other applicative verbs (such as manipulative verbs) in that they are

subject control verbs. Here again we see the need to make reference to both the semantic and

syntactic properties.

Table 4-4. Factive verbs

Verb Gloss Subject Object Control Causative Applicativ
Control Control Shift e
nyatsa loath/hate Y N N Y N
hloya dislike/hate Y N N Y N
rata like/love Y N N Y N
swabela sorry Y N N N Y
thabela glad Y N N N Y




7 Most of these verbs have nominal counterparts which also select infinitival complement such as 'ke maswabi ho
utlwa ditaba tseo' I am sorry to hear those news









occurring simultaneously as complements. The subcategorization frames associated with active

verbs are represented in (19).

(19) Subcategorization frames for active verbs

(a) Nyatsa, V, [ NP,IP] 'loathe/hate

(b) Thabela, V, [_ NP,IP] 'glad'

(c) Swaba, V, [ NP,IP] 'sorry'

The syntactic similarities that also overlap with some semantic similarities between some

desiderative verbs, and achievement verbs are evident in these subcategorization frames for

active verbs.

4.2.5 Experiencer-object Verbs

Experiencer-object verbs are verbs that express an experience or emotion encountered by

the object of the verb. Like the name suggests, these are object control verbs. However, in the

absence of an object, the control is shifted to an arbitrary control relation. I explore this

characteristic of experiencer-object verbs in Chapter 5. Unlike the other object control verbs

(manipulative) this set of verbs makes use of a combination of the causative and the applicative

form in their derivation. These two qualities distinguish this class of object control verbs from

the manipulative class whose morphology consists of either the causative or the applicative. Let

us look at some examples.

(20) Mmusoi o kgothaletsa banak PRO*i/k ho- bua Sesotho NP IP
3.government 3.SM- encourage 2.children INF- speak 7.Sesotho
'The government encourages children to speak Sesotho'

(21) Mmusoi o khothaletsa PRO*i/ARB ho -bua Sesotho IP
3. government 3.SM- encourages INF- speak Sesotho
'The government encourages speaking Sesotho'

(22) *Mmuso o- kgothaletsa bana NP
3.government 3.SM encourage 2.children
Intended: 'The government encourages children'









The conjunction used to join nouns in Sesotho is le 'and.' However, when joining verbs,

the verb that expresses the second and subsequent action is expressed either in the infinitival

form or the subjunctive form as indicated in the examples (18). Sentence (18) shows that in

coordinated verbs, the second verb is in the infinitival form. Sentence (19) on the other hand

shows that although the infinitive and subjunctive may overlap in their occurrences, the

subjunctive form cannot co-occur with the conjunction in the same way that the infinitive does.

(18) Re tla [reka] le [ho bala] dibuka
1PL FUT buy CONJ INF read books
'We will buy and read books'

(19) Re a reka (*le) re bal-e dibuka
1PL FOC buy (CONJ) 1PL read-SUBJ 10. books
'We buy books and (then) read them'

Ho may also be used as a preposition. In the example in (20) ho is used as a preposition

equivalent to English 'to'. This same use of ho is also used to express direction such as 'towards'

or 'in the direction of as in sentence (21). In sentence (22) ho is used withfihla (arrive) to

express 'until'. This use is clearly associated with the locative nouns in view of the fact that they

have a tendency to express location, time or manner.

(20) O re ho nna o ya masimo-ng
2SG say to 1SG 2SG. SM go fields-LOC
'You say to me you are going to the fields'

(21) Leba ho yena
go to 3SG
'Go to him/her'

(22) O -hanella ka dikobon-ng ho fihla nako ee?
2SG stay in blanket-LOC to arrive 9.time 9.DEM
'lit: You stay in the blankets until this time arrives'
'You stay in bed until this time?'

There is yet another very common use of ho in Sesotho associated with impersonal subject

or copulative. Sentence (23) expresses an instance of the use of ho as an impersonal copulative,









The examples provided in (2a-c) and (3a-c) show that noun formation involving the

adjectival and adverbial stems is less restricted. Examples (4a-d) however show that noun

derivation involving other nouns is more restricted. Overall the examples in (2)-(4) clearly

indicate that the formation of nouns through noun class prefixes is very productive in Sesotho.

These examples further reveal that noun derivation involving lexical categories other than verbs

does not cause any changes in the stems involved. Although this process is similar to the

derivation of nouns from verbs, there are some differences. Let us now turn to the derivation of

nouns from verbs in comparison.

2.2.3 Nouns Derived from Verbs

Sesotho has three major types of nouns derived from verbs. These three types are

distinguished based on their prefixes and suffixes. The noun class prefix determines the number

and class to which the derived noun belongs. The three basic suffixes associated with nouns

derived from verbs are -i;-o and a. Personal nouns generally have the prefix mo- and suffix-i

whereas the impersonal ones predominantly have the suffix-o and a variety of prefixes. The

suffix a is primarily associated with nouns derived through the prefixation of the class 15 noun

class prefix ho. This suffix, unlike the others occurs with both personal and impersonal derived

nouns. The examples below illustrate these facts.

The examples in (5a-e) show that one verb stem rat- 'like/love' creates different nouns

based on the different noun class prefixes. Although the deverbal nouns have different prefixes,

the relation in meaning between these nouns is preserved by the stem. The different noun

suffixes indicate that there has to be an agreement between the noun prefix and the suffix.

Examples (5d) and (5e) show noun formation involving verbal inflection. In this instance the

suffix is consistently -a regardless of the noun class prefix.









determine argument structure, in addition, in the case of control, this interface has various layers

which cannot just be restricted to lexical morphological processes. I represent the interfaces as in

(38).

(38) Lexicon.lexical class+verbal extensions argument structure- syntactic structure -control type



In Chapter 4 I discuss the role played by verbal morphology in determine argument

structure but also show how argument structure captures the generilizations about the behavior of

verbs. These generalizations would be missed if only lexical information was used to determine

the syntactic behavior of control verbs.

1.4 Organization of the Study

Chapter 1 has presented the basic empirical patterns to be discussed in the dissertation as

well as the theoretical background on which the discussion is based. The remainder of the

dissertation is organized as follows.

Chapter 2 presents a descriptive overview of grammatical properties of Sesotho morpho-

syntax such as noun classification and verbal agreement. Chapter 1 also outlines the various

functions associated with the morpheme ho in Sesotho. It explores the syntactic differences

between the nominal infinitive and the verbal infinitive in Sesotho.

Chapter 3 provides a description of verbal agreement in Sesotho as well as an examination

of verbal extensions. Chapter 4 presents patterns associated with control verb classes in Sesotho.

Chapter 4 also highlights how certain verbal extensions (causative and applicative) assist with

the further classification of these verbs. Chapter 4 closes with some differences between English

and Sesotho verb classes.

Chapter 5 presents patterns of control in Sesotho. It opens with a typology of complement

control observed cross-linguistically. The major contribution of Chapter 5 is the presentation of

30









(12) Manipulative verbs with the subjunctive

(a) Morena o hopotsa banna hore ba kopan-e
1.chief 1.SM remind 2.men that 2.SM- meet-SUBJ
'The chief reminds the men that they should meet'

(b) Morena o -hopotsa mme hore bana ba sebets-e
1.Morena 1.SM remind 1.mother that 2.children 2.SM- work-SUBJ
'The chief reminds mother that the children should work'

(c) Morena o hopotsa hore ba sebets-e
1.chief 1.SM- remind that 2.SM work-SUBJ
'The chief reminds (some people/them) that they should work'


Sentence (12a) shows the subjunctive used where its subject has the same reference as the

object of the matrix clause. Sentence (12b) is acceptable but sounds as if some information is

missing between the object of the matrix clause and the subject of the subjunctive. Sentence

(12c) is acceptable because the object of the matrix verb is implicit. It may or may not be the

same as the subject of the embedded clause.

Although I make reference to the subject of the subjunctive as having the same reference

as the subject or object of the matrix clause this should not be taken as control. The subjunctive

as a complement of control verbs allows a lexical noun and fully supports SM as well as tense

and aspect morphology, most of which are unavailable in control relations. The most important

difference between the subjunctive and the infinitive when used with control verbs is that there is

no co-indexing required since any kind of relation is expressed overtly through SM or a lexical

noun.

Finally, it is important to note that the traditional utterance/ communication verbs are also

classified as subclass 2 of manipulative verbs. The reason for merging these classes is due to the

shared semantic grouping as well as similar syntactic behaviors. It is also vital to note that

subclass 2 of the manipulative verbs shows the need for semantic categories since syntactic










(c) Ho se halef- is- e mme ho lok- ile
INF- NEG- anger- CAUS- NEG 1.mother 15.SM right- PERF
'Not making mother angry is (the) right (thing to do)'

(d) Se halef- e
NEG- angry- NEG
'Don't be angry'


The discussion so far indicates that the infinitival clause shows the majority of properties

associated with verbs as well as nouns in Sesotho. Tables 2-3 and 2-4 summarize these findings.

It is evident from this summary in Table 2-3 that the infinitive behaves like ordinary nouns

although there are minor differences. Table 2-4 further indicates that the infinitive in Sesotho has

some form of inflection although it does not share all the properties of finite structure. This

brings us to the question asked in Chapter 1 of the nature and identity of the clauses introduced

by ho. Is it a verb or a noun?


Table 2-3. Nominal properties of the infinitive
Property Nouns Infinitive
Prefix & stem Yes Yes
Plural Yes No
Subject Position Yes Yes
Object Position Yes Yes
Preposing Yes Yes
Trigger Subject Agreement Yes Yes
Trigger Object Agreement Yes No
Nominal Modifiers Yes Yes


The answer to this question has been a major area of investigation amongst Bantuists.

Guma (1971:158) points out that in Sesotho the infinitive is morphologically a noun, but "unlike

other nouns, it has both positive and negative forms... syntactically it may be both a noun and a

verb". Creissels & Godard (2005:1) indicate that while English has the gerund and infinitival

forms, in the Sotho languages the infinitive shows a "mixed morphology, exhibiting both verbal









infinitives in complement control structures. The two principles of interest are the Minimal

Distance Principle and the PRO Theorem, stated below. The former determines the interpretation

of PRO, which NPs it can be co-indexed with. It states that PRO must be coindexed with the

closest NP that c-commands PRO.

(31) Minimal Distance Principle (MDP) (Rosenbaum 1970)
PRO is bound by the closest c-commanding antecedent


To illustrate how the MDP works, let us look at our earlier examples renumbered for ease

of reference.

(32) [matrixBanai ba batla [subordinate PROi ho- sesa lewatle]]
2.children 2SM want INF- swim ocean
'The children want to swim in the ocean'

(33) [matrix Mmei o kgothaletsa banaj [subordinate PRO*i/j ho- bala dibuka]]
1.mother 1 .SM encourage 2.children INF- read l0.books
'Mother encourages the children to read books'


Sentences (32) and (33) may be represented as follows:

IP

DP I'
A
Banai I VP
A
ba DP
A
Bana V IP
batla
DP I'
A
PROi I VP
I A
ho sesa lewatle


MDP and subject control


Figure 1-1.











LIST OF ABBREVIATIONS

# Semantically anomalous

* Ungrammatical

1 First person

2 Second person

3 Third person

AC Adjectival concord (agreement)

APPL Applicative (verbal extension)

ARB Arbitrary

CAUS Causative (verbal extension)

COMP Complementizer

CONJ Conjunction

COP Copulative

CS Control shift

DEM Demonstrative

EC Exhaustive control

EXP Expletive

FOC Focus

FUT Future tense

FV Final vowel

INF Infinitive

LOC Locative

MDP Minimal Distance Principle

NEG Negation









do "syntax." We noted that Sesotho uses morphology together with syntax principles such as the

MDP to account for why a reference relation that requires a split between two antecedents (split

control) is not possible. We also noted that where the MDP would have been challenged the

morphology becomes of assistance.

I hope that this study will open up avenues for scholars of control regardless of their

approaches to control constructions. There are two issues that I feel need to be studied. One

concerns whether the control patterns observed in Sesotho are possible with other Bantu

languages, or whether they are varied. If they are varied, are there ways of determining the

variation by looking at the morphology?

On a more cross-linguistic level it would be interesting to see how the morphology

associated with verbal extensions would be written into features that would help determine the

referential dependencies of PRO. Finally, it would also be interesting to find out if all control

verbs in Sesotho take a clause with a subject or whether there are differences as proposed by

Wurmbrand 2001.










(11) Partial control and Sesotho verb classes

(a) Morena i o batla PRO i/*i+m/ARB ho- kopana
1.king 1.SM force INF- meet
Intended: 'The king wants to meet'

(b) Morenai o- lebala PROi/ *i+m /ARB ho -kopana
1.king 1.SM foget INF meet
'The king forgets to meet'

(c) Morenai o -nyatsa PROi/ *i+m ARB/ ho kopana
1.king 1.SM hates INF meet
'The king hates to meet (some people)'

(d) *Morenai o-lemosa PRO*i/ *i+m /*ARB ho kopana
l.king 1.SM- advise INF meet
Intended: 'The king advises some people to meet'

(e) Morenai o-lemosa bannaj PRO*i/j/*ARB/*i/j+m ho kopana
l.king 1.SM -advise 2.men INF- meet
'The king advices the men to meet'

(f) Morena i o hopotsa -bannaj PRO *i/j /*j/j+m/*ARB ho- kopana
1.king 1.SM force 2.men INF- meet
'The king reminds men to meet'

(g) Morenai o khothaletsa bannaj PRO*i/j/*ARB/*i/j+m ho -kopana
1.king 1.SM- encourage 2.men INF -meet
'The king encourages meeting (people to meet)'

The data in (11) demonstrate two things. First, the reading associated with PRO i+m is not

possible with all verb classes in Sesotho, where i represents one of the arguments of the matrix

clause and +m some other unspecified argument. All these sentences would be ungrammatical

under that reading. The second thing is that the morphology of the verbs (associated with

semantic verb classes) determines the interpretation of PRO. Note that for each of the sentences

(1 la-c), the sentence is only grammatical under PROARB because in this instance, kopana

'meet' is part of a nominal infinitive therefore the argument restrictions that are associated with

the reciprocal do not apply. However, this is only possible because these verbs also select the

nominal infinitive as a complement. Sentence (lid) confirms this claim because we noted









The first question that needs to be answered relates to the distinction between sentences

(10) and (11) in contrast with sentences (12) and (13). In control use, hopola takes one nominal

argument, the object. Hopotsa in contrast takes two nominal arguments in its control use, a

subject and a direct object. The grammaticality of sentence (11) bears evidence to this

requirement. Sentences (12) and (13) show the reverse of the contrast represented by (10) and

(11). The verb hopotsa 'remind' unlike hopola 'remember' requires two arguments. This contrast

is to be expected because these two verbs have different argument requirements, and as such

belong to different semantic classes.

The converse of this is found when looking at verbs that belong to the same semantic class

but have different selectional properties. Such verbs have been observed in many languages (see

Levin 1993). An example of this is observed in sentences (14) and (15). The verb hopotsa

'remind' shares the same class membership as qophella but behaves differently. In order to

make this conclusion about Sesotho a thorough investigation of the properties of these verbs is

required. In Sesotho, as observed in other languages, semantic class membership alone is not

adequate in determining the referential dependencies as exemplified by sentences (10)-(15).

Levin 2009 proposes root ontological type as a better way of classifying verbs in addition to

argument structure. In Chapter 4, I discuss the role of verbal morphology in determining

argument structure of verbs.

The final area of investigation relates to the typology of complement control relations

attested in Sesotho. Partial control has been cited as one of the major challenges to a movement

approach to control (Culicover and Jackendoff 2006). Partial control is a relation in which the

interpretation of PRO properly includes the referent of the controller instead of being identical to

the controller. Sentences (16) and (17) are examples of partial control. Since verbs like 'meet'









represented by pronouns provided the necessary agreement requirements (noun class) are met.

The pronoun can therefore be used to test whether a constituent is a NP or not. Sentences

(44&45) show that this is indeed possible with the nominal infinitive. Example (46) however,

illustrates that although (45) passed the pronoun test, the noun in (44) is not the same kind of NP

as the ones in (42). We then conclude that there is a difference between NP (noun classes 1-14)

and nominal infinitive. The next step involves the differences between the nominal infinitive and

the clausal infinitive. In (47) the same verb, Akgithaleia\ is used with two arguments, an NP and

NP/IP. Sentence (48) becomes ungrammatical when we replace the same construction ho bua.

Note however in this instance that when the pronoun nominal infinitive is used the sentence

becomes ungrammatical. This is a clear indication that ho bua in (44) and (47) cannot be the

same constituent.

The differences between ho bua in (44&47) depends on the verb that selects them and

other arguments of the verb. In (44), ho bua is an NP hence it can be replaced by a pronoun. In

(47) however ho bua is an IP. These different interpretations are directly linked to the semantic

classes to which the selecting verbs belong. I deal with these semantic classes in Chapter 4.

2.7 Summary

In this chapter I have developed the proposal that the infinitive in Sesotho behaves like

nouns and verbs at the same time. I have indicated that while the infinitive shows most of the

nominal properties it falls short as far as nominalization and plural formation are concerned. A

close examination of noun classes and noun formation indicates that the infinitive may function

as a noun, however like other noun classes that share the same SM, it does not follow the same

pattern as other nouns in noun formation. The infinitive retains the verbal suffix in noun

derivation, a property which I directly link to verbal characteristics.










There are various advantages associated with the adoption of the structures in 2-2. First, 2-

2(a) clearly suggest that ho-ja is a noun as indicated by NP. The second advantage is that

whereas the NP may take modifiers such as demonstratives and possessives, the structure also

allows the expansion to include the verbal properties associated with the infinitive. The next

question then is why we need two structures when one can do. A closer look at the distribution of

the nominal infinitive suggests that not all instances of the nominal infinitive are NP. Let us look

at some examples.

(42) Sekolo se -rek-is-a bana dibuka
7.school 7.SM-buy-CAUS-FV 2. Children 10.books
'The schools makes the children buy books'

(43) Sekolo se- rek-is-a bona tsona
7.school 7.SM- buy-CAUS- FV 2.PRON 10.PRON
'The school makes them buy them'

(44) Mmuso o- kgothaletsa ho- bua
3.government 3.SM-encourages INF- speak
'The government encourages speaking'

(45) Mmuso o -kgothaletsa hona
3.government 3.SM-encourages 2.PRON
'The government encourages it'

(46) *Mmuso o -kgothaletsa bana
3.government 3.SM-encourages 2.children
'The government encourages the children'


(47) Mmuso o -kgothaletsa bana ho- bua
3.government 3.SM-encourages 2.children INF -speak
'The government encourages the children to speak'

(48) *Mmuso o- kgothaletsa bona hona
3.government 3.SM-encourages 2.PRON 15.PRON
'*The government encourages the them it'

The data in (42-48) show that although the nominal and clausal infinitive overlap in their

distribution there are some instances where one form is excluded. Example (42) shows a non-

control verb used with two nouns bana and dibuka. Example (43) illustrates that nouns may be

61


































2010 Mantoa Rose Smouse









verbs are inherently applicative. Manipulative3 predicates that are inherently object control in

Sesotho, are semantically applicative. For most of these verbs, we cannot break the verb into a

verb root and an applicative suffix.

3.3.3 The Reciprocal

The reciprocal morpheme in Sesotho is /-an/. This affix adds a mutual relationship to the

arguments of the verb (do x to each other). The reciprocal affix requires two conjoined subject

noun phrases or a semantically plural subject. Unlike the causative and applicative, the reciprocal

reduces the number of arguments that a verb takes. The reciprocal deletes or eliminates the

internal argument. This affix is also sensitive to the semantics of the verb. It requires a verb that

is capable of expressing mutual relationship. Below are some examples of the reciprocal in

Sesotho.

(16) Mme o bon-a ngwana
1.mother 1.SM see-FV 1.child
'Mother sees a child'


(17) Mme le ngwana ba a bon- an- a
1.mother CONJ 1.child 2.SM FOC see- REC- FV
'Mother and child see each other'

(18) Ngwana o bon- an- a le mme
1.child 1.SM see- REC- FV CONJ 1. mother
'The child arranges to see mother'

(19) *Ngwana o bon- an- a mme
1.child 1.SM -
see- REC- FV 1. mother
'*Child see each other mother'

Sentence (16) is an example of an ordinary bare verb with two arguments. Example (17)

shows that the reciprocal requires singular conjoined subjects with a plural SM. If the subject is




3 See Chapter 4 for verb classes.









marker suffixed to a verb stem, as in arde 'to come', waalde 'to spend the night', etc. Conversely,

in Wolof, there is no class morphology on the noun, only on determiners agreeing with the noun,

and there is also no infinitival marker. This ties the presence of the infinitive to the presence of

noun class morphology. The next step is how to distinguish between the two.

In the discussion earlier in this chapter, we noted that the infinitive tends to be restricted

when used in object position. This is to be expected since this position is regulated by the

properties of the selecting verb. Since the concern of this study is with control verbs, the next

step is to look at the internal syntactic property of the infinitive as a complement of control

verbs.

2.5 PRO as the Subject of the Infinitive in Sesotho

Sesotho control verbs usually select a noun, the infinitive or subjunctive as a complement.

A typical embedded subjunctive clause contains a subject which may have the same referent as

the subject of the main clause but does not have to. This subject may be a lexical one or just a

subject agreement morpheme also known as pro. The subjunctive as a complement fulfills all the

required principles as outlined in Chapter 1. To illustrate let us look at this example:

(25) Subcategorization
a. Hopotsa: V[ NP (CP)/(IP)]
b. Sebetsa: V [ (NP/PP)]

(26) Ke hopotsa mme [cphore a sebets-e lapeng]
1SG- remind 1.mother that 1.SM work -SUBJ at home
'I remind mother to work' / 'I remind mother that she must work at home'

The subcategorization tells us that the verb hopotsa 'remind,' in addition to the external

argument requires an object NP and an optional CP or IP. In the main clause, the two theta roles

are assigned one to the subject Ke 'I' and one to the object of the matrix clause mme 'mother'. In

the embedded clause, the other theta role is assigned the subject a which is pro. As a

pronominal, pro must be free in its minimal clause. This condition is met because the possible

54









Doke, Clement, M. and Machabe Mofokeng. 1957. Textbook of S.in,,,lw Sotho Grammar,
London: Longman, Green and Company.

Du Plessis, J.A. 1982b. The analysis of the infinitive. .S,,hl African Journal of African
Languages. 2(2).

Du Plessis J. and Marianna Visser. 1992. Xhosa Syntax. Pretoria: Via Afrika.

Farkas, Donca, F. 1988. On obligatory control. Linguistics andPhilosophy 11: 27-58.

Gowlett, Derek. 2005. Zone S. In The Bantu Languages by Derek Nurse, Gerard Philippson
(eds.) Routledge.

Guma, Samson, Mbizo. 1971. An Outline Structure of.,iilheI ii Sotho. Shuter & Shooter.
Pietermaritzburg.

Haegeman, Liliane. 1991. Introduction to Government and Binding theory. Oxford: Basil
Blackwell.

Heine, Bemd and Derek Nurse. 2000. African Languages: An Introduction. Cambridge
University Press.

Henderson, Brent. 2007. Multiple agreement and inversion in Bantu. Syntax 9 (3):275-289.

Hornstein, Norbert. 1995. Logical Form: From GB to Minimalism. Oxford: Blackwell.

Hornstein, Norbert. 1997. Is the binding theory necessary? Ms., University of Maryland, College
Park.

Hornstein, Norbert. 1998. Movement and chains. Syntax 1:99-127.

Hornstein, Norbert. 1999. Movement and control. Linguistic Inquiry 30:69-96.

Hornstein, Norbert. 2000. On A-chains: A reply to Brody. Syntax 3: 129-143.

Hornstein, Norbert.2003. On Control. In Minimalist Syntax, ed. by Hendrick, Randall. Malden,
MA: Blackwell, pp 6-81.

Hvitfeldt Robert D. 1982. The Major Syntactic Structure of Xhosa. Doctoral dissertation,
University of Wisconsin-Madison.

Idiata, Daniel, F. 2005. What Bantu child Language tells about Noun classes. : LINCOM Studies
in African Linguistics 67.

Jackendoff, Ray and Peter W. Cullicover. 2003. The semantic basis of control in English.
Language 79:517-556.









'it.' Sentence (24), on the other hand shows the use of ho as an unspecified subject of a passive

sentence or an expletive also expressed as 'it' in English.

The examples in (23)& (24) show the various uses of ho in Sesotho which have not

received attention in the literature. A lot of issues remain unsolved regarding the question of

whether these are historically related or not. Synchronically, there seems to be some relation

between the morpheme used as a preposition and the locative nouns as discussed earlier in this

chapter. If we consider the use of ho to indicate a movement towards or away from some

location, it can be linked directly to the locative classes in Sesotho. Although the locative classes

have very few nouns left in Sesotho, their agreement marker is also ho and they are formed

(prefix +stem) in the same way as other nouns. The nominal infinitive may be a remnant of this

process in Sesotho.

(23) Ho lok-ile morena
COP fine-PERF 1.chief
'It is fine chief

(24) Ho robets-w-e mme yena o a sebets-a
Ho sleep-PASS-FV CONJ 3SG 1.SM FOC works-FV
Lit: The is some sleeping while he works'
'People are sleeping and he is working'

Taking into consideration the other instances of ho, although they may look remotely

related to the infinitive as we know it, we can begin to accept that the two uses of the infinitive

under discussion are also related. There is strong evidence to suggest that there is a relation

between the infinitival form and the nominal form of the infinitive. Mc Laughlin (personal

communication) points out that there is possible evidence for the relationship between noun class

markers and infinitival markers found in Pulaar (Fula), a northern Atlantic (Niger-Congo)

languages. In Pulaar, a language with more than twenty noun classes (although these are not

paired as in the Bantuist tradition), the noun class suffix, -de, also shows up as an infinitival









the number of arguments (the causative and applicative). When used together in one verb the

order in which these suffixes occur, follow the order observed in other Bantu languages: Neutar-

Causative-Applicative-Reciprocal- Passive (NCARP). Although the order is fixed, it is

sometimes altered to attain idiomatic expressions. Let us now look at the causative, applicative

and reciprocal individually as these are central to this study.

3.3.1 The Causative

The causative in Sesotho is realized as a suffix morpheme /-is/. The allomorphs are /-is/, /-

its-/ and /-ets-/. The causative in Sesotho translates to cause x to doy. It behaves slightly

different from other Bantu languages (especially Xhosa) in that the order of the internal

arguments is fixed. In other languages such as Xhosa and Zulu, inserting object agreement makes

it possible to switch the order of these arguments. This does not happen in Sesotho. When the

causative suffix is added, the argument selected by the cause becomes the closest to the verb.

The examples in (4-8) indicate these facts.

(4) Ntate o pheha dijo
1. father 1.SM cook l0.food
'Father cooks food'

(5) Ntate o pheh- is- a bana dijo
1.father 1.SM cook- CAUS- FV 2. children 10. food
'Father makes the children cook food'

(6) #Ntate o pheh- is- a dijo bana
1.father 1.SM cook- CAUS- FV l0.food 2.children
'Father makes the food to cook children'

(7) Ntate o ba pheh- is- a dijo
1.father 1.SM 2.OM cook- CAUS-FV l0.food
'Father makes them cook food'

(8) Bana ba pheh-is- w- a dijo ke ntate
2.children 2.SM cook- CAUS-PASS-FV l0.food COP 1.father
'The children are made to cook food by father'









Jacobson, Pauline. 1992. Raising without movement. In Control and grammar. ed. by
Richard Larson, Sabine latridou, Utpal Lahiri, and James Higginbotham, 149-194.
Dordrecht: Kluwer.

Katamba, Francis. 1993. Morphology. St. Martin's Press, New York.

Khoali, Benjamin Thakampholo.1991. A Sesotho tonal grammar, Doctoral dissertation,
University of Illinois at Urbana-Champaign.

Koopman, H.D and D. Sportiche.1991. The position of subjects. Lingua, 85,211-259.

Koster, Jan, and Robert May. 1982. On the constituency of infinitives. Language 58:117-143.

Koster, Jan. 1984. On binding and control. Linguistic Inquiry 15:417-459.

Kratzer, Angelika. 1996. Severing the external argument from its verb. In Johan Rooryck and
Laurie Ann Zaring (eds.). Phrase Structure and the Lexicon, 109-137. Dordrecht: Kluwer
Academic Publishers.

Landau, Idan. 2000. Elements of Control: Structure and meaning in infinitival Construction.
Dordrecht: Kluwer.

Landau, Idan. 2001. Control and extraposition: The case of super-equi. Natural Language and
Linguistic Theory 19:109-152.

Landau, Idan. 2004. The scale of finiteness and the calculus of control. Natural Language and
Linguistic Theory 22:4.

Landau, Idan.2006. Severing the distribution of PRO from Case. Syntax. 9(2):153-170.

Lasnik, Howard. 1995b. Verbal morphology: Syntactic Structures meets the Minimalist
Program, Ms., University of Connecticut, Storrs.

Lasnik, Howard. 1999. Minimalist Analysis. Blackwell Publishers.

Larson, Richard. 1991. Promise and the theory of control. Linguistic Inquiry 22:103-139.

Larson, Richard, Sabine Iatridou, Utpal Lahiri, and James Higginbotham (eds.). 1992. Control
and grammar. Dordrecht: Kluwer.

Levin, Beth. 1993. English Verb Classes and Alterations. A Preliminary Investigation. The
University of Chicago Press. Chicago and London.

Levin, Beth. 2009. Verb Typologies Revisite. Handout: Ghent, February 5-7, 2009

Lestrade,G.P. 1938. Locative Nouns and Formations in Sotho. Bantu Studies. Vol.XII (1).

Machobane, 'Malillo 'Matsepo. 1995. The Sesotho Locative Constructions. Journal of African
Languages and Linguistics 16(2):115-136.









the verb in the main clause because there is a maximal project IP which becomes a barrier to

government, a required condition for Case assignment. PRO becomes the only possible NP for

this position.

There are other tests that are used to compare the subject position of finite and non-finite

clauses. The common tests include Weak Cross Over (WCO) effect, interpretation under VP

ellipsis as well as the presence of an overt NP in the subject position. Sesotho has morphological

processes that compensate for moved elements. Using tests that involve movement such as WCO

become a challenge. Another challenge has to do with the fact that Wh-movement, which is used

in WCO test, is optional in Sesotho. Let us look at some examples to illustrate the point.

(32) Sello o rek-ile koloi.
1.Sello 1.SM buy-PERF 9.car
'Sello bought a car'

(33) Sello o rekile eng?
1.Sello 1.SM buy-PERF what
'What did Sello buy?'

(34) Ke eng eo Sello a e rek-ile-ng ?
COP what 9.REL 1.Sello 1.RM 9.OM boughtPERF-REL
'What is it that Sello bought?'

Sentence (32) is a perfective sentence. Example (33) shows that the object is questioned in-

situ whereas (34) shows an optional movement of the wh-phrase. Note however that the

movement of the wh-phrase across the verb and the subject leaves an OM and requires the

relative clause morphology on the verb and SM. The WCO effect obtains when a bound variable

is moved across another one (Terzi 1997). However, if one of the variables is PRO this effect is

escaped. Examples (35-38) demonstrate.

(35) Mme wa haevari o -a mvar2 rata.
1.mother 1.POSS his 1.SM FOC 1.OM love
'His mother loves him'











(8) [matrix Mmei o kgothaletsa banaj [subordinate PRO i/j ho bala dibuka]]
1.mother 1.SM- encourage 2.children to- read 10.books
'Mother encourages the children to read books'

(9) [matrix Ntatei o tshepisa banaj [subordinate PRO i/*j ho- sebetsa lapeng]]
1.father 1.SM- promise 2.children to -work home.LOC
'Father promises the children to work at home'

The other question relates to the nature of the subject position of the embedded clause

represented by PRO; what evidence do we have to suggest that this position is occupied? If it is

occupied, what role does this item in this position play? These questions are explored in Chapter

3 of this study.

The third major area of investigation relates to the referential dependencies between the

arguments of the main clause and the subject of the embedded clause. Let us consider more

examples:


(10) *[ matrix Mmei o hopola banaj [subordinate PRO*i/*j ho sebetsa]]
1.mother 1.SM- remember 2.bana to work
'*Mother remembers the children to work'

(11) [matrix Mme i o hopola [subordinate PRO i ho sebetsa]]
1.mother 1.SM remember to work
'Mother remembers to work'

(12) [matrix Mme i o hopotsa bana [subordinate PRO i ho sebetsa]
1.mother 1.SM remind 2.children to work
'Mother reminds children to work'

(13) [matrix *Mme i o hopotsa [subordinate PRO i ho sebetsa]]
1.mother 1.SM remind to work
'Mother reminds (someone) to work'

(14) [matrix Ntate i o qophella [subordinate PRO i ho sebetsa]]
1.father 1.SM insist to work
'Father insists on working'

(15) [matrix Ntate i o qophella banaj [subordinate PRO *i/j ho -sebetsa]]
1.father 1.SM- insist 2.children to- wor
'Father insists that children work' Lit: 'Father forces children to work'


]


k









A related but slightly different pattern is observed with aspect in Sesotho. Aspect marks

the phase or duration of an action. The progressive has been cited as an aspectual morpheme

associated with the infinitive in Sesotho, Xhosa and Setswana (Doke 1957, Visser 1989 and

Creissels & Godard 2005 respectively). As indicated by the examples in (15) the use of

progressive is restricted to certain environments. These examples illustrate.

(15) Aspect- Progressive

(a) Banna ba sa halef- is- a mme
2.banna 2.SM- PROG- angry- CAUS- FV 1.mother
'Men still make mother angry'

(b) #Banna ba rata ho sa halef- is- a mme
2.men 2.SM- like INF PROG- angry- CAUS- FV 1.mother
'Men like to still make mother angry'

(c) Banna ba sa rata ho halef- is- a mme
2.men 2.SM PROG- like INF angry- CAUS- FV 1.mother
'Men still like to make mother angry'

(d) Ho sa halef- is- a mme ha ho a loka
INF- PROG- angry- CAUS- FV 1.mother NEG- 15.SM FOC right
'To still make mother angry is not right'

Sentence (15a) shows the use of the progressive with an ordinary finite verb. The

progressive morpheme sa is prefixed to the verb between the SM and the verb. Sentence (15b)

indicates that the use of the progressive morpheme with the subordinate infinitival clause is less

acceptable. This is because the same sentence may be better expressed when the aspect is

associated with the main clause as in (15c). However, when the infinitival clause is used as a

subject the use of the progressive aspect is then acceptable in Sesotho.

Mood is expressed using various verb forms in Sesotho. A mood may either be finite or

non-finite. The non-finite mood forms in Sesotho are the infinitive and imperative. Non-finite

moods are traditionally regarded as verb forms that lack tense and agreement. So far we have

noticed that the infinitive in Sesotho shows very limited instances of tense and agreement. The




Full Text

PAGE 1

1 TOPICS IN SESOTHO CONTROL VERBS By MANTOA ROSE SMOUSE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSI TY OF FLORIDA 20 10

PAGE 2

2 20 10 Mantoa Rose Smouse

PAGE 3

3 To my parents : Daniel Teboho and Limakatso Johanna Motinyane Robala ng ka kgotso Motaung le Mohlakoana

PAGE 4

4 ACKNOWLEDGMENTS I would like to express my sincere t hanks to the Department of Asian and African Languages and Literatures, the Center for African Studies and the Linguistics Department at the University of Florida for the many years of financial support and intellectual stimulation. The intensity of the c ourse work coupled with the academic excellence offered by my teachers and colleagues instilled a deeper love for learning. I would also like to thank my Professors at University of Cape Town, Sizwe Satyo, Rajend Mesthri e, Derek Gowlett, and Sandile Gxilis he for instilling the love for languages and linguistics. I owe the completion of this study to all those whom I had an opportunity to work with. I am indebted to my brother Neo Motinyane who spent hours on the phone listening and responding to my endless questions. His skepticism about certain constructions led me to think of ideas I would have not thought about. I would also like to thank Tshediso Maweng for sharing his wonderful stories with me. I am as well indebted to my mother. As a native speaker co nsultant, she provided me with grammatical ity judgments untainted by any formal more indebted to her for showering me with the motherly love, friendship and support that sustained me throughout my stay in the United States. I would also like to extend si ncere thanks to Dr. Hunt Davis (U ncl e Hunt) and Mrs. Jeanne Davis (A unt Jeanne) for opening their home and hearts to my family. I would have never been able to come to University of Florida without their support. They provided us with a home away from home. I also extend sincere thanks to their children Richard and Jonathan for accepting a few more members into their family. I would also like to thank Dr Carter Gil bert and Mrs Nancy Gilbert for always taking care of us whenever Uncle Hunt and Aunt Jeanne were out of town. I am forever grateful to these two families for their unconditional love and support. I am

PAGE 5

5 also thankful to Dr Thomas Hurst, Mrs Betsy Hurst an d their family. I have learnt a lot about the holidays in the United States from spending time in their home. Mrs Hurst, having been a student and a mom herself, shared some wonderful parenting tips. I also thank their children for sharing some childhood stories with me. I am very thankful to their paying attention to my personal comfort. I thank my husband and my children. My children, Naledi, Tlhodiso and Mapule showered me with love, support, understanding, and proved the much needed diversion from aca demic life. I am forever grateful for the free hugs, kisses, unplanned trips to the mall, birthday parties, and playgrounds. Without their unconditional love, I would have not been able to stay at the University of Florida. To my husband, Abuti Basie, with out whom I would have not survived my studies, I am deeply indebted. His friendship, support and love ha ve kept me going. I am very fortunate to have been able to engage in some discussions with Dr. Mchombo, Dr Salikoko Mufwen e and the participants at th e ACAL meeting held at University of Florida. I am also indebted to Dr. Der Houssikian and Dr. Henderson for their advice and views on Bantu syntax. I would also like to sincerely thank my committee members, Fiona Mc Laughlin, Eric Potsdam, Gary Miller (a lthough he retired before I completed) Masungu Matondo and Abdoulaye Kane. My spontaneous meetings (including one more than 10,000 miles away) with Fiona Mc Laughlin were always very fruitful for solving riddles that I spent hours trying to solve. I since r el y thank Dr Mc Laughlin and Dr Potsdam for spending all those hours correcting a number of drafts.

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 LIST OF TABLES ................................ ................................ ................................ ........................... 8 LIST OF FIGURES ................................ ................................ ................................ ......................... 9 LIST OF ABBREVIATIONS ................................ ................................ ................................ ........ 10 ABSTRACT ................................ ................................ ................................ ................................ ... 12 1 INTRODUCTION ................................ ................................ ................................ .................. 14 1.1 Introduction ................................ ................................ ................................ .................. 14 1.2 The Area of Investigation ................................ ................................ ............................ 16 1.3 Theoretical Background ................................ ................................ ............................... 20 1.3.1 The Principles and Parameters Theory ................................ .............................. 21 1.3.2 The Intera ction between Semantics, Morphology and Syntax .......................... 28 1.4 Organization of the Study ................................ ................................ ............................ 30 2 THE SIGNIFICANCE OF HO IN SESOTHO ................................ ................................ ....... 32 2.1 Introduction ................................ ................................ ................................ .................. 32 2.2 Nominal Morphology ................................ ................................ ................................ ... 32 2.2.1 Noun Classes ................................ ................................ ................................ ...... 34 2.2.2 General Nominal Derivation ................................ ................................ .............. 37 2.2.3 Nouns Derived from Verbs ................................ ................................ ................ 39 2.3 Properties of t he Infinitive in Sesotho ................................ ................................ ......... 41 2.3.1 Distributional Properties of Nouns ................................ ................................ .... 42 2.3.2 Distributional Properties of Verbs ................................ ................................ ..... 45 2.4 Other Uses of Ho ................................ ................................ ................................ ......... 51 2.5 PRO as the Subject of the Infinitive in Sesotho ................................ ........................... 54 2.6 Nomin al and Clausal Infinitive ................................ ................................ .................... 59 2.7 Summary ................................ ................................ ................................ ...................... 62 3 ASPECTS OF SESOTHO MORPHO SYNTAX RELEVANT TO CONTROL .................. 64 3.1 Introduction ................................ ................................ ................................ .................. 64 3.2 The Verb in Sesotho ................................ ................................ ................................ ..... 64 3.3 Verbal Extensions ................................ ................................ ................................ ........ 65 3.3.1 The Causative ................................ ................................ ................................ ..... 66 3.3.2 The Applicative ................................ ................................ ................................ .. 67 3.3.3 The Reciprocal ................................ ................................ ................................ ... 69 3.4 Summary ................................ ................................ ................................ ...................... 70

PAGE 7

7 4 SESOTHO CONTROL VERB CLASSES ................................ ................................ ............. 71 4.1 Introduction ................................ ................................ ................................ .................. 71 4.2 Verb Classification Methodology ................................ ................................ ................ 72 4.2.1 Desiderative Verbs ................................ ................................ ............................. 73 4.2.2 Manipulative Verbs ................................ ................................ ............................ 79 4.2.3 Achievement Verbs ................................ ................................ ............................ 83 4.2.4 Factive Verbs ................................ ................................ ................................ ..... 85 4.2.5 Experiencer object Verbs ................................ ................................ ................... 88 4.3 Non control Verbs ................................ ................................ ................................ ....... 90 4.4 Summary ................................ ................................ ................................ ...................... 92 5 SESOTHO COMPLEMENT CONTROL PRO FILE ................................ ............................. 93 5.1 Introduction ................................ ................................ ................................ .................. 93 5.2 Properties of Control ................................ ................................ ................................ .... 93 5.3 Control Type s ................................ ................................ ................................ ............... 96 5.3.1 Subject Control ................................ ................................ ................................ .. 96 5.3.2 Object Control ................................ ................................ ................................ .... 97 5.3.3 Control Shift ................................ ................................ ................................ ....... 98 5.4 Referential Dependencies ................................ ................................ .......................... 105 5.4.1 Exhaustive Control ................................ ................................ ........................... 106 5.4.2 Partial Control ................................ ................................ ................................ .. 106 5.4.3 Split Control ................................ ................................ ................................ ..... 108 5.5 Summary ................................ ................................ ................................ .................... 110 6 DISCUSSIO N AND CONCLUSION ................................ ................................ .................. 111 6.1 Summary ................................ ................................ ................................ .................... 111 6.2 Implications for Control Relations ................................ ................................ ............. 114 6.3 Concluding Remarks ................................ ................................ ................................ .. 116 LIST OF REFERENCES ................................ ................................ ................................ ............. 118 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 125

PAGE 8

8 LIST OF TABLES Ta ble page 2 1 Sesotho noun classes ................................ ................................ ................................ .......... 35 2 2 Noun class 1 plural formation ................................ ................................ ............................ 36 2 3 Nominal properties of the infinitive ................................ ................................ ................... 50 2 4 Verbal properties of the infinitive ................................ ................................ ...................... 51 4 1 Desiderative verbs ................................ ................................ ................................ .............. 75 4 2 Manipulative verbs ................................ ................................ ................................ ............. 80 4 3 Achievement verbs ................................ ................................ ................................ ............. 84 4 4 Factive verbs ................................ ................................ ................................ ...................... 86 4 5 Experiencer object verbs ................................ ................................ ................................ ... 89 5 1 Referential properties of PRO ................................ ................................ ............................ 95 5 2 Properties of PRO in OC and NOC ................................ ................................ ................... 95

PAGE 9

9 LIST OF FIGURES Figure page 1 1 MDP and subject control ................................ ................................ ................................ ... 26 1 2 MDP and object control ................................ ................................ ................................ ..... 27 2 1 The noun in Sesotho ................................ ................................ ................................ ........... 60 2 2 The Nominal and clausal infinitive ................................ ................................ .................... 60 3 1 Verb structure in Sesotho ................................ ................................ ................................ ... 65

PAGE 10

10 LIST OF ABBREVIATIONS # Semantically anomalous Ungrammatical 1 First person 2 Second person 3 Third person AC Adjectival concord (agreement) APPL Applicative (verbal extension) ARB Arbitrary CAUS Causative (verbal extension) COMP Complementizer CONJ Conjunction COP Copulative CS Control shift DEM Demonstrative EC Exhaustive control EXP Expletive FOC Focus FUT Future tense FV Final vowel INF Infinitive LOC Loca tive MDP Minimal Distance Principle NEG Negation

PAGE 11

11 NOC Non obligatory control OC Obligatory control OM Object agreement marker PASS Passive PC Partial control PC Possessive concord (agreement) PFV Perfective (aspect) PL Plural PRO Null subject of infinitival clauses PROG Progressive (aspect) PST Past tense RC Relative concord (agreement) RECP Reciprocal (verbal extension) REL Relative SBJV Subjunctive mood SG Singular SM Subject agreement marker TNS Tense

PAGE 12

12 Abstract of Dissertation Presented to the Gradua te School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy TOPICS IN SESOTHO CONTROL VERBS By Mantoa Rose Smouse August 20 10 Chair: Fiona Mc Laughlin Major: Linguistics This study examines the phenomenon of forward complement control in Sesotho using Principles and Parameters Theory of syntax as a theoretical framework. It explores the interaction between syntax, morphology, and semantics through an examination of the role pla yed by argument structure changing verbal morphology (verbal extensions). This study suggests that this argument structure changing verbal morphology is central to understanding the different syntactic behaviors observed across and within verb classes and to the study of control in Bantu languages. Forward complement control is a relation of obligatory co indexation between the subject or object of the matrix clause and the subject of the subordinate clause. The subject of the subordinate clause is the n ull PRO which is understood as having the same reference as the pronounceable subject or object of the matrix clause. Sentence (1) illustrates control in Sesotho. (1) [Mme i o kgothaletsa bana j [PRO *i/j ho bala dibuka]] 1.mother 1.SM encourage 2.children INF read 10.books Control verbs select noun phrases, subjunctive clauses as well as the infi nitive in Sesotho. Only the infinitive participates in control relations However, an examination of the other complements of control verbs supports the classification of verbs in Sesotho proposed in this

PAGE 13

13 dissertation T he infinitive is introduced by ho wh ich is also the prefix of class 15 nouns. This study compares the status of ho as an infinitival morpheme and ho as a noun class prefix in Sesotho. I suggest that the morphology of the control verbs together with the referential relations between the contr oller in the matrix clause and the subject of the infinitive help determine the differences between nominal infinitive (class 15 nouns) and clausal infinitives. This study also proposes that argument structure changing verbal morphology is important in th e classification of control verbs in Sesotho. This morphology is also responsible for the syntactic properties of control and directly accounts for the typ es of control observed in Sesotho. Contrasting this analysis with a typology of control patterns from other languages, the lack of partial control or split control is furthermore directly linked to argument structure changing morphology.

PAGE 14

14 CHAPTER 1 INTRODUCTION 1. 1 Introduction This study explores the phenomenon of f orward c omplement c ontrol in Sesotho, a Bantu language spoken in South Africa. It investigates the interaction between morphology and syntax by analyzing the semantics of verbal extensions and the role they play in determining verb class membership. This morphology syntax interface evidenced in verbal morphology supports the idea that control in Sesotho cannot be understood simply as a syntactic process ; the semantics of verb s also play s an important explanatory role. Complement c ontrol is a relation of obligatory coindexation between the sub ject or object of the matrix clause and the subject o f the subordinate complement clause. In forward complement c ontrol (hereafter complement control) only the subject or the object of the matrix clause is overt The subject of the subordinate clause is n ull (PRO in sentences (1) (3)) but is understood as having the same reference as the subject or the object of the matrix clause. This is indicated by coindexation. The following examples illustrate (1) [ matrix Bana i ba batla [ subordinate PRO i ho sesa lewatle]] 2.children 2SM want INF swim ocean The c (2) [ matrix Mme i o kgothaletsa bana j [ subordinate PRO *i /j ho bala dibuka]] 1.mother 1.SM encourage 2.children INF read 10.books (3) [ matrix Ntate i o tshepisa bana j [ subordina te PRO i/*j ho sebetsa lapeng]] 1.father 1.SM promise 2.children INF work home.LOC

PAGE 15

15 Sentences similar to the ones in (1) to (3) have been an area of interest for at least four decades (since Rosenbaum 1967) Most of the previous work on control has been couched within the syntactic theory of Government and Binding (GB) and later the Principles and Parameters Theory (P&P) (Chomsky 1981, 2000 and Chomsky and Lasnik 1995) The control module of the Principl es and Parameters framework assumes that the subject of controlled clauses is the null element PRO. Various principles within the theory of control determine the interpretation and full distribution of PRO. In this study I use the Principles and Parameters Theory (henceforth P&P) of control as a technical and a familiar way to talk about control, although I will not be concerned with the theory of control per se. The reason for ado pting Principles and Parameters is that it is largely familiar and has had considerable success in accounting for cross linguistics patterns of control. The primary aim of this study is to provide a description and analysis of control verbs in Sesotho, us ing P&P as a framework. The secondary aim of this study is to classify control verbs and their properties in Sesotho, with a special focus on addressing the questions outlined below. Question 1 : What is the status of ho in Sesotho? Question 2 : Are there di fferent kinds of controlled complements and, if so, what are they? Question 3 : What role does argument structure changing verbal morphology play in control phenomena ? The remainder of the c hapter is organized as follows. Section 1.2 outlines the area of investigation. The analytic approach of this study is outlined in section 1.3. Additionally, section 1.3 provid es an overview of the modules of P&P that are used as a framework for analysis in this study. Section 1.4 presents an outline of the dissertation

PAGE 16

16 1. 2 The Area of Investigation Sesotho, also known as Southern Sotho, is a South Eastern Bantu language spoken by around four and a half million people in South Africa. It is one of the 11 official languages of South Africa As an official language, it is used in education, media and government and has a literature which spans over a century. Sesotho is also a national and official language of Lesotho. Like many other Bantu languages, Sesotho has rich nominal and verbal morphology One of the morphologic al features of Bantu languages is the elaborate noun class system. Sesotho has 17 noun classes which occur in pairs of singular and plural Class membership is determined partly phonologically and partly semantically Semantic criteria hold for about four of the noun classes, whereas two of the seventeen classes represent locative classes. Noun class 15 has nouns derived from verbs. The prefix of this class ho has the same phonological form as infinitive morpheme in Sesotho as the following examples illust rate: (4) Bana ba sesa lewatle 2.children 2.SM swim ocean The c (5) Ho sesa ho kgathatsa haholo. 15 swim 15.SM tire very /to swim (6) Ke rata ho sesa 1SG like INF swim /swimming Sentence (4) exemplifies the use of the verb sesa noun class two subject. The same verb is used in sentence (5) as a noun class 15 derived noun, hence the subject agreement (SM) associated with this class precedes the verb kgathatsa same verb is used with infinitive morpheme as a complement of the verb rata like In this instance the complement of the verb rata Such data raise the following questions: How does the prefix of noun class 15

PAGE 17

17 differ from the infinitive marker ? If there is a dif ference how is this difference represented syntactically ? Both these questions have been of interest to those who have interest in Bantu languages I explore these questions under four major areas of investigation. The first major area of investigation is associated with the nature of the infinitive in Sesotho and Bantu languages in general. E xamples (4) (6) illustrate that on the surface Sesotho appears to lack a distinction between the nominal infinitive (5) and the clausal infinitive (6) One way of get ting around this problem is by looking at the distribution of the nominal infinitives (class 15 derived nouns) and the clausal infinitive We need to determine whether t he clausal infinitive occupies all the positions occupied by a nominal NP or whether th ere is some overlap If it does then how do we determine the presence of control? This study, through exploring the selectional properties of control verbs, shows t hat while the nominal infinitive occupies all the possible positions associated with clausal infinitive the clausal infinitive is restricted. The questions related to the nature of the infiniti ve in Sesotho are discussed in C hapter 2 Chapter 4 explores the questions related to the selectional properties of control verbs. The second major area o f investigation is related to the previous one. While the focus of the preceding area is more morphological in nature, here we are looking at the syntactic nature of the clausal and nominal infinitives. In summarizing this particular area of investigatio n, let us look at our earlier examples (renumbered for easy reference) associated with the clausal infinitive. The first question in this area concerns the type of constituent associated with the subordinate clause in each of the sentences. Are these subo rdinate clauses represented in (7) (9) NPs, VPs or IPs? (7) [ matrix Bana i ba batla [ subordinate PRO i ho sesa lewatle]] 2.children 2SM want to swim ocean The c

PAGE 18

18 (8) [ matrix Mme i o kgothaletsa bana j [ subordinate PRO *i/j ho bala dibuka]] 1.mother 1.SM encourage 2.children to read 10 .books (9) [ matrix Ntate i o tshepisa bana j [ subordinate PRO i/*j ho sebetsa lapeng]] 1.father 1.SM promise 2.children to work home.LOC The other question relates to the nature of the subject position of the embedded clause represen ted by PRO ; what evidence do we have to suggest that this position is occupied? If it is occup ied, what role does this item in th is position play? These questions are explored in C hapter 3 of this study. The third major area of investigation relates to the referential dependencies between the arguments of the main clause and the subject of the embe dded clause. Let us consider more examples: (10) [ matrix Mme i o hopola bana j [ subordinate PRO *i /*j ho sebetsa ]] 1.mother 1.SM remember 2.bana to work *M other (11) [ matrix Mme i o hop o la [ subordinate PRO i ho sebetsa ]] 1.mother 1.SM remember t o work M (12) [ matrix Mme i o hopotsa bana [ subordinate PRO i ho sebetsa]] 1.mother 1.SM remind 2. children to work M (13) [ matrix *Mme i o hopotsa [ subordinate PRO i ho sebetsa]] 1.mother 1.SM remind to work M other reminds (someone) (14) [ matrix Ntate i o qophella [ subordinate PRO i ho sebetsa]] 1.father 1.SM insist to w ork F (15) [ matrix Ntate i o qophella bana j [ subordinate PRO *i/j ho sebetsa]] 1.father 1.SM insist 2.children to work F ather insists Lit: ather forces children to work

PAGE 19

19 The first question that needs to be answered relates to the distinction between sentences ( 10 ) and ( 11 ) in contrast with sentences ( 1 2 ) and ( 1 3 ). In control us e, hopola t akes one nominal argument, the object Hopotsa in contrast takes two nominal arguments in its control use, a subject and a direct object. The grammaticality of sentence (11) bears evidence to this requirement. Sentences (12) and (13) show the re verse of the contrast represented by (10) and (11). The verb unlike hop o la t wo arguments This contrast is to be expected because these two verbs have different argument requirements, and as such belong to different sem antic classes. The converse of this is found when looking at verbs th at belong to the same semantic class but have different selectional properties. Such verbs have been observed in many languages (see Levin 1993). An example of this is observed in sentenc es (14) and (15). The verb hop otsa shares the same class membership as qophella but behave s differently In order to make this conclusion about Sesotho a thorough investigation of the properties of these verbs is required In Sesotho, as observed in other languages, semantic class membership alone is not adequate in determining the referential dependencies as exemplified by sentences ( 10 ) (1 5 ). Levin 200 9 pro poses root ontological type as a better way of classifying verbs in addition to argument s tructure. In C hapter 4 I discuss the role of verbal morphology in determining argument structure of verbs. The final area of investigation relates to the typology of complement control relations attested in Sesotho. Partial control has been cited as one of the major challenges to a movement approach to control (Culicover and Jackendoff 2006). Partial control is a relation in which the interpretation of PRO properly includes the referent of the controller instead of being identical to the controller. Sente nces (16) and (17) are example s of partial control.

PAGE 20

20 the fact that the controller Lisa is singular. The meaning of (16) is that Lisa wan ts that she and some other people meet. (16) Lisa i wants [ PRO i +j to meet in the evening ] (17) *Ngwana i o batla [ PRO i +j ho ko pana] 1.child 1.SM want to meet Intended: The child wan ts to meet with someone do not allow partial control. The question that follows from this observation relates to how this lack of partial control contributes to cross linguistic descrip t ions of control I explore this question in C hapter 6 The data used in this study come from written texts, transcriptions of spoken data, previous linguistic descriptions and grammaticality judgments f r om me and other native speak ers. The written materials consulted in my study include a Sesotho classic novel, Chaka by Thomas Mofolo a contemporary novel, Kiriatshwana by Mkwanazi and Seyalemoya edited by Moeketsi which consists of six radio dramas. Transcripts of two personal narra tives also form part of materials consulted. All data with morpheme ho were extracted from the sources. The investigation of the data was guided by two questionnaires on control verbs, Stiebels et al. 2003 and Sti e bels 2007. 1. 3 Theoretical Background This section discusses the theoretical background assumed throughout the dissertation. I lay out the relevant aspects of the Principles and Parameters Theory that are important for understanding what follows. The second part of this section lays out how the mo rphology, the syntax and the lexicon interact.

PAGE 21

21 1.3.1 The Principles and Parameters Theory The Principles and Parameters theory of syntax is based on the notion of Universal Grammar (UG) as some system of principles, common to the species and available to each individual prior to Within P&P all languages obey these universal principles. The differences between languages are accounted for by differences in parameters rather th an principles. In this view of language the grammar has a lexicon (containing lexical entries with subcategorization information) and computational system (guided by X bar Theory). A derivation goes through four levels of representation: Deep Structure, Surface Structu re, Phonological Form (PF) and Logical Form (LF). There are various principles that apply at different levels of representation. Let us look at a derivation of a simple Sesotho sentence to illustrate the various aspects of the theory. (18) Bana na ba bapala mampatile. 2.girls 2.SM play hide and seek T In order to derive sentence (18), the computational system first selects the verb and its two arguments and various function al elements. These are combined into the D structure in (19), which adopts a version of the VP internal subject hyp othesis (Koopman and Sportiche 1991 and others) in which the external argument is base generated in the specifier of VP. (19) [ IP [ I [ VP Banan a bapala mmampatile]]] The g irls play hide. and seek The subject then moves to the specifier of IP, yielding the S structure in (20). I ignore PF and LF in the rest of the dissertation because it is not relevant to the issue of control (20) [ IP [ Banana i I [ VP ti bapala mmampatile]]]

PAGE 22

22 The structure is evaluated with respect to various principles to ensure well formedness. Two important principles are the Projection Principle and the Theta Criterion. The Projection Prin ciple is a requirement which ensures that the thematic structure associated with lexical items is saturated in the synta x (Haegeman 1991:63) and that the number of projected arguments is the same at all levels of representation. Simply stated, the Project ion P rinciple ensure s that lexical information is syntactically represented and prevents insertion of structure beyond D structure. Th is principle is stated formal ly as follows: (21) Projection Principle Representations at each syntactic level (i.e. LF and D S tructure and S Structure) are projected from the lexicon, in that they observe the subcategorization properties of lexical items. Chomsky further proposed that all clauses must have subjects, regardless of their lexical requirements. This requirement on c lauses together with the Projection Principle came to be known as the Exte nded Projection Principle (EPP) The Theta Criterion is responsible for the assignment and tracking of semantic roles such as agent, theme, goal and so forth. These theta roles are part of the lexical information contained in This assignment and distribution of theta roles derive s directly from the theta criterion. The theta criterion is stated formal ly as follows: (22) Theta Criterion Each argument bears one and on ly one theta role, and each theta role is assigned to one and only one argument (Chomsky 1981:36 ) The T heta C riterion together with the Projection Principle must be satisfied for the structures to be well formed. The sub categorization of some of the ver bs we have used in our examples thus far are given in (23).

PAGE 23

23 (23) Sub categorization 1 (a) Hopola : V, [_____NP/IP] (b) Batla : V, [______NP/IP] (c) H opotsa : V, [_____NP ,NP/IP ] (d) Hopola : Verb 1 NP 2 NP/IP i j (e) Batla : Verb 1 NP 2 NP/IP i j (f) Hopo tsa : Verb 1 NP 2 NP 3 NP/IP i j k The subcategorization frames in (23a c) represent the internal arguments selected by each of the verbs. The theta grids (d f ) shows the different theta roles represented by letters i, j and k assigned by the verbs. There are transformations that take place between the Deep Structure and the Surface Structure. W ithin GB, transformations such as passive movement are subsumed under Move Movement is optional but is constrained by various restrictions. One of the movemen ts that appl ied to our derivation in (19 ) is the movement of the subject from spec, VP t o spec, IP. This occurs because of Case theory which requires that all NPs be assigned Case by Surface Structure. This is stated formally in the Case Filter: (24) Every pho netically realized NP must be assigned (abstract) Case (Chomsky 1995: 111) 1 The external argument is excluded in this instance

PAGE 24

24 Case is assigned to NPs by certain heads under the structural relationship of Gove r nment stated in (25). In Sesotho for example, Spec I is a position where the subject NP is govern ed by and receives nominative C ase from I T he object is governed by and receives accusative C ase from V in a verb complement position 2 The Case Filter forces the movement of the subject to spec,IP in (19) otherwise the subject would not be assigned Ca se and the Case Filter would be violated, and the derivation would not succeed. (25) Government and m command (a) A governs B iff A m commands B and no Barrier intervenes between A and B. Maximal Projections are barriers to government. Governors are heads. (b) A m co mmands B iff i. A does not dominate B ii. B does not dominate A and iii. T he maximal projection of A dominates B (Chomsky 1986b:8) The various modules of GB interact to ensure well formedness. Let us now consider another module that relies on Government as defined in ( 25 ). The Binding Theory (BT) determines the distribution of anaphors, pronouns, and R expressions in relation to their potential antecedents ( Haegeman 1991:215 ). The principles regulating the interpretation and distribution o f these NPs as stated in ( 2 6 2 8 ) are referred to as t he B inding Principles The binding relations are expressed through co indexation of two constituents and the binding principle work s as a filter to determine licit and illicit co indexation s For our purposes, the governing categ ory can be understood as the minimal clause (26) The Binding theory (Chomsky 1981:188) A. An anaphor is bound in its governing category 2 I am ignoring the subject agreeme nt marker SM here for reasons of simplicity.

PAGE 25

25 B. A pronominal is free in its governing category C. An R expression is free (27) is bound by if and only if: (a) and are co indexed, an d (b) c commands (28) is free only if it is not bound (29) C command (Chomsky 1986b:8) A c commands B iff A does not dominate B and every branching node that dominates A also dominates B. To illustrate how binding works, consider the following data from Davies & Dub insky (2004: 186) which illustrates the complementary distribution of anaphors and pronouns. (30) The distribution of anaphors and pronouns (a) The children i soiled themselves i (b) *The children i soiled them i (c) *The children i claimed that the teacher scolded themselve s i (d) The children i claimed that the teacher scolded them i Sentence ( 30 a) satisfies principle A of the binding theory in that the children c commands themselves and the reflexive is bound within the minimal clause Sentence ( 30 b) violates principle B of t he Binding t heory in that the pronominal must be free in the minimal clause but it is not In ( 30 b) them is bound by the children under the indicated co indexation Sentence ( 30 c) violates p rinciple A in that the antecedent, the children is in the matrix c lause, whereas themselves is in the subordinate clause. Thus themselves is bound but not in its governing category. Sentence ( 30 d) is grammatical because it satisfies Principle B. The pronoun is free in its minimal clause. The Binding theory, Government t heory, Case Theory, the Extended Projection Principle, and the Theta Criterion are important for this study because of their interaction with Control Theory. Control Theory determines the interpretation and distribution of PRO, the null subject of

PAGE 26

26 infiniti ves in complement control structures. The two principles of interest are the Minimal Distance Principle and the PRO Theorem, stated below. The former determines the interpretation of PRO, which NPs it can be co indexed with. It sta tes that PRO must be coin dexed with the closest NP that c commands PRO. (31) Minimal Distance Principle (MDP) (Rosenbaum 1970) PRO is bound by the closest c commanding antecedent To illustrate how the MDP works, let us l ook at our earlier examples renumbered for ease of reference. (32) [ m atrix Bana i ba batla [ subordinate PRO i ho sesa lewatle]] 2.children 2SM want INF swim ocean The c (33) [ matrix Mme i o kgothaletsa bana j [ subordinate PRO *i/j ho bala dibuka]] 1.mother 1.SM encourage 2.children INF read 10.books Sentence s (32) and (33) may be represented as follows: IP DP Bana i I VP ba DP Bana V IP batla DP PRO i I VP ho sesa lewatle Figure 1 1 MDP and s ubject c ontrol

PAGE 27

27 IP DP Mme i I VP o DP VP Mme V P IP V DP V DP PRO *i /j I VP ho kgothaletsa bana j bala dibuka Figure 1 2 MDP and o bject c ontrol In 1 1 there is only one antecedent DP which c commands PRO, Bana Figure 1 2 presents a different configuration in that there are two c commanding DPs. However, since the closest c commanding antecedent is object of the matrix clause, PRO is then co indexed with bana 3 Finally, t he distr ibution of PRO is regulated by PRO Theorem stated in (34): (34) PRO Theorem PRO must be ungoverned The PRO theorem restricts PRO to occurring in ungoverned and Caseless positions. For our purposes this includes the subject of infinitival complements. The PRO theorem follows from the Binding Theory in the following way. Chomsky ( 1982) observed that binary feature s 3 It is important to note here that the MPD has always been challenged in generative grammar for its inability to cater for Subject control verbs (see Chomsky 1981, Comrie 1984, Koster 1984, Larson 1990 Sag and Pollard 1991 amongst others).

PAGE 28

28 [anaphor] and [pronominal] could be used to characterize the elements regulated by Binding Theory, yielding four categories. These combi nations are given below (Haegeman 1991:223) (35) NP types (a) [+anaphor, pronominal] anaphor (b) [ anaphor, +pronominal] pronominal (c) [ anaphor, pronominal] R expression (d) [+anaphor, +pronominal] PRO The combinations in ( 3 5 a) ( 3 5 c) correspond to anaphors, pronouns and referring expressions. PRO on the other hand has the features [+anaphor, +pronominal] represented in ( 3 5 d ) Considering the Binding Principles above, we can see that being [+anaphor] PRO would have to be bound (Principle A) and being [+pronominal] it would have to be free (Principle B) at the same time since anaphors must be bound and pronominals free In order to resolve the contradiction caused by the satisfaction of these two principles, PRO must not have a governing category. In such a case, it will vac uously satisfy both Principles A and B. This can occur if PRO is ungoverned, yielding the PRO Theorem Gov ernment and Binding principles, therefore determine the distribution of PRO. Given that Case is assigned under government, it also follows from the PR O Theorem that PRO will also not be assigned Case. 1.3.2 The I nteraction between S emantics, M orphology and S yntax The theroretical outline presented in the previous section makes the assumption that control is a purely syntactic process. The role assigne d to the meaning is reduced to subcategorization frames which encodes the information about the arguments associated with that particular verb or lexical item. This may be represented structurally as follows (36) 4 (36) 4

PAGE 29

29 syntactic type, and their hierachichal organization necessary for the mapping to the syntactic discussions that look at the relationship between the lexicon and argument structure continue to investigate whether argument structure is not redundant since it repeats information that is encoded in the lexicon. Bresnan 1995 points out that work in lexical semantics sh hierachichal organization of arguments in argument structure is in fact predictable from (1995:2), Levin 1993). (37) Bresnan 1995 interface model Lexical semantics Lexicon a structure Syntactic projection Initial syntactic structure (D structure) Syntactic Transformations Final syntactic structure (S Structure) Bresnan 1995 uses data from English and Chichewa to sho w that the verbal relation since these are lexical morphological processes. Bresnan therefore proposes that the initial syntactic phase should be abandoned. In a starting point because I look at the contribution of these verbal relation transformations to argument structure. I focus on the causative, the applicative and reciprocal in particular I propose that th ese lexi cal morphological processes indeed

PAGE 30

30 deter mine argument structure, in addition, in the case of control, this interface has various layers which cannot just be restricted to lexical morphological processes. I represent the interfaces as in (38). (38) Lexicon.lexical class+verbal extensions In C hapter 4 I discuss the role played by verbal morphology in determing argument structure but also show how argument structure captures the generilizatio ns about the behavior of verbs. These generalizations w ould be missed if only lexical information was use d to determine the syntactic be havior of control verbs. 1. 4 Organization of the Study C hapter 1 has presented the basic empirical patterns to be discussed in the dissertation as well as the theoretical bac kground on which the discussion is based. The remainder of the dissertation is organized as follows. Chapter 2 presents a descriptive overview of grammatical properties of Sesotho morpho syntax such as noun classification and verbal agreement Chapter 1 also outlines the various functions associated with the morpheme ho in Sesotho. It explores the syntactic differences between the nominal infinitive and the verbal infinitive in Sesotho. Chapter 3 provides a description of verbal agreement in Sesotho as well as an examination of verbal extensions Chapter 4 presents patterns associated with control verb classes in Sesotho. Chapter 4 also highlights how certain verbal extensions (causative and applicative) assist with the fu r ther classification of these ve rbs. C hapter 4 closes with some differences between English and Sesotho verb classes. Chapter 5 presents patterns of control in Sesotho. It opens with a typology of complement control observed cross linguistically. The major contribution of C hapter 5 is t he presentation of

PAGE 31

31 data representing the different types of complement control structures attested in Sesotho. None of the studies that I am aware of have discussed a typology of control in Bantu. The second major contribution of this study presented in Ch apter 5 is the discovery of lack of partial control in Sesotho. Chapter 6 summarizes the findings of the study. Chapter 6 also points out areas that require further investigation.

PAGE 32

32 CHAPTER 2 THE SIGNIFICANCE OF HO IN SESOTHO 2. 1 Introduction Control verbs in Sesotho select infinitival clauses as their complements. As we have observed in Chapter 1 the distinction between the infinitive and class 15 nouns is not a clear cut one. This issue is further complicated by the lack of phonological differences betwee n the morpheme ho as a noun class prefix and the morpheme ho as an infinitival morpheme. In order to identify whether a word is a noun or a verb we need to determine the morpho syntactic properties of nouns in Sesotho. The primary aim of this chapter is to explore the morpho syntactic properties associated with ho as a noun class prefix and ho as an infinitival morpheme. As a way of understand ing the dual nature of this morpheme, an evaluation of the structure of the noun and the morphological processes inv olved in noun formation becomes necessary F ocus will be mainly on deverbal nouns The remainder of Chapter 2 is dedicated to a comparison of the verbal and nominal functions of ho in Sesotho. 2. 2 Nominal Morphology Sesotho like many other Bantu language s has an elaborate noun class system. A typical noun in Sesotho is composed of a prefix and a stem. A Sesotho prefix may be represented in a template form as CV for the majority of classes and as in the case of classes 9 and 1(a) The prefix marks class membership as well as number Whereas the meaning of the noun depends primarily on the stem, the noun class prefix also carries some semantic content. The se examples show the different prefixes that can be used with the same stem to form different nouns. (1) Examples of noun prefixes used with sotho (a) mo sotho cl.1 stem A Sotho speaking

PAGE 33

33 (b) ba sotho cl.2 stem speaking (c) Le sotho cl.5 stem A country of the Sotho speaking (d) *ma sotho cl.6 stem P lu ral of (e) Se sotho cl.7 sotho (f) *Di s otho c l.8 sotho s (g) bo sotho cl.14 stem B eing a Examples (1 a ) and ( 1b ) indicate a singular and plural pair, the difference being expressed throu gh the prefix. Examples ( 1c ) and ( 1d ) further indicate that some nouns do not follow the expected singular and plural pairs due to semantic restrictions Example ( 1e ) and (1f) further show that although the stem may appear with other noun classes prefixes it is very restricted in plural forms. Example (1g) represents a noun class in which the majority of nouns are abstract nouns and for this reason they lack plural counterparts. As th ese data show the noun class prefixes cannot be freely prefixed to any no un stem. We will not go into the details of the distribution of prefixes but will rather try and draw attention to the historical impact of these irregularities o f noun classification and how these may shed some light on the behavior of certain noun classe s. Although there are these irregularities with regards to singular and plur a l nouns the noun class system as a whole is fairly systematic. Let us now look at the noun classes in Sesotho.

PAGE 34

34 2. 2.1 Noun Classes There are curently sixteen noun classes in Ses otho made up of singular and plural nouns The irregularities between singular and plural pairs pointed out in the previous section resulted in a proposal to split singular and plural forms These irregularities were initially observed during re construction of Proto Bantu numbering system (Doke & Mofokeng 1957). In the earlier system, Sesotho nouns were divide d into nine classes The singular and plural combined in the same class to make 5 classes. C lasses 14, 15 and the two locative classes were added as singular classes In the current system, classes 1 and 1a, 2 and 2a are the only classes that follow a clear semantic classification These classes contain nouns that refer to humans or personified animals. None of the other classes show this s emantic coherence although there are scholars who claim that clear semantic categories exist (see Guma 1971 for example). There are also Bantu scholars (Doke & Mofokeng 1957, Guma 1971 amongst others) who claim that noun classification is all based on sema ntic classes and prefixes. Current approaches in Niger Congo studies suggest that noun classification takes into account a multiple factors; semantics being one amongst others (see Mc Laughlin 1997). As I have already indicated in Sesotho, only a small num ber of classes show this clear semantic membership Table 2 1 represents the current classification of nouns in Sesotho. The classification in T able 2 1 is based on Meinhof f Bantu numbering system. The noun class prefixes are very stable and unifo rm amongst Bantu languages of Botswana, Lesotho, Mozambique, South Africa and Swazila nd (Gowlett 2005). Seso tho lacks classes 11, 12 and 13 although t hese classes are found in other Bantu languages. This gap in the classification occurred due to the merger of the nouns in class 11 with nouns in class 5 and those of 12 and 13 with

PAGE 35

35 nouns in class 7 (Gowlett 2005:620) 1 This observation is specifically supported by the irregularities associated with the plural form of nouns in class 6. Let us look at how plur al formation works within the noun class system. Table 2 1 Sesotho noun c lasses Noun Class Prefix Noun Gloss 1 m o motho,morena person, chief 1a mme mother 2 b a batho people 2a b o bomme mothers 3 m o motse homestead 4 m e metse homesteads 5 l e leru cloud 6 m a maru, marena, mahobe 2 clouds, chiefs 7 s e sefate tree 8 di d ifate t rees 9 kgomo, nku cow, sheep 10 di dikgomo, dinku cows, sheep 14 b o bohobe bread 15 h o ho ja 3 to eat/eating 16 h o hodimo up 17 m o morao behind Th e pattern of singularity and plurality is fairly stable amongst the noun classes in Sesotho. The plural of class 1 mo nouns is formed by prefixing the class 2 noun class prefix ba There are however a few exceptions. There are some irregularities found w ith the plural of some nouns found in classes 1 The plural form of mo rena for example, in class 1 is not the expected ba rena Instead t he plural of this noun is derived by prefixing class 6 noun prefix ma This 1 An example of nouns being merged include the use of noun class 5 prefix with the less productive locative nouns: le ho dimo le fa tshe 2 There also nouns that have two plural forms: le siba (sg) ma siba di tshiba 3 The nouns in class 15 are written as two free morphemes in Sesotho orthography.

PAGE 36

36 has result ed in a split betwee n nouns that fall into class 2 ba in plural and those that fall into class 6 ma for their plural form. It is also important to note that some of the nouns in class 1 that take ma as a plural pr efix also have a singular form found in class 5. Let us look at some examples. Table 2 2 Noun c lass 1 p lural f ormation Noun class 1 4 Noun class 5 5 Noun class 6 Gloss Mo rena ------------Ma rena King/chief Mo fumahadi ------------Ma fumahadi Queen /wife Mo qhotsa Le qhotsa Ma qhotsa Xhosa speaking Mo swatsi Le swatsi Ma swatsi Swati speakers ------------Le tebele Ma tebele Ndebele speaker Table 2 2 indicates that nouns that take ma as a plural prefix are nouns associated with speakers of other languages such as the Nguni languages of S outh Africa The nouns representing Mo tswana /Ba tswana Tswana speaking person /s and Mo pedi/Ba pedi use the plural prefix ba associated with the S esotho speaking people. The important thing to note about t hese examples is that they may be used in a way which convey s speakers attitudes towards the people expressed through these nouns These examples reinforc e the suggestion that a multitude of fact ors are cons idered in noun classification ( see Mc Laughlin 1997 for this approach). A similar but slightly different pattern occurs with bo hobe of bo hobe ma Nouns in class 14 differ from those in class 1 in that the split is between nouns that lack the plural form and those 4 Nouns associated with people 5 This prefix tends to have a derogatory function when used with nouns that refer to people.

PAGE 37

37 whose plural form falls into noun class 6. The majority of nouns in this class are abstract nouns, resulting in most of the se nouns lack ing plural forms. In a ddition to the complete loss of classes 11, 12 a nd 13, Sesotho has lost most of the nouns that belong to locative classes. Classes 16 and 17 have a very small number of nouns. These loca tive nouns are often ref er r ed to as adverbs of p lace (Doke & Mofokeng 1957:83). Besides the fact that these noun classes have different prefixes, they all trigger the same agreement morpheme on predicates. It is interesting to note here that this agreement morp heme for the locative classes is ho. The nu mber of nouns remaining in these classes as well as the similar agreement morpheme associated with the locative classes has generated debates amongst Bantuists regarding their status within the noun class system. Mchombo 1993 & 2004 for example, questions the nominal status of the items in these classes in Chichewa. One more class that has also attracted a lot of attention is noun class 15, the infinitival class which contains verbs in their infinitival form or verbal nouns The prefix of this noun class i s ho. In Sesotho, the nouns found in this class are derived from verb stems. Whereas Nguni languages such as Xhosa have proper nouns and nouns derived from verbs in class 15 at first glance Sesotho does not show this two way distinction class 15 are subjected to various tests they reveal some form of ambiguity (see Creissels & Godard 2005 for Se tswana ). I discuss this distinction later in this chapter. First let us examine nominal morphology, specifically the formation of nouns from verb s. 2. 2.2 General N ominal Derivation The noun prefixes in Sesotho are central to nominal derivation and predicate agreement. The noun prefix may be affixed to various lexical categories to derive new nouns. In the process of noun formation the noun class p refixes generate new words while mark ing number and gender (in this case noun class) at the same time The process of nominal derivation is crucial in

PAGE 38

38 understanding how the class 15 noun class is different from the other noun classes in Sesotho Below are examples that show the productivity of noun format ion I limit my examples to word formation that involves the noun class prefix. (2) Nouns derived from adjectives (a) mo holo cl.1 big/old (b) bo holo cl.14 big/big (c) se holo cl.7 old/big (3) Nouns derived from adverbs (a) bo hole cl.14 far (b) le ho dimo cl.5 cl. 16 up (c) di ka hare cl.10 P REP inside (4) Nouns derived from nouns (a) bo tho cl. 14 person being human (b) se sadi cl.7 woman (c) se phoofolo cl.7 animal (d) *se mo sadi cl.7. cl.1 woman

PAGE 39

39 The examples provided in ( 2 a c ) and (3a c) show that noun formation involving the adjectival and adverbial s tems is less restricted Examples (4 a d ) however show that noun derivation involving other nouns is more restricted Overall t he examples in ( 2 ) ( 4 ) clearly indicate that the format ion of nouns through noun class prefixes is very productive in Sesotho. The se examples further reveal that noun derivation involving lexical categories o ther than verbs does not cause any changes i n the stem s involved. Although this process is similar to the derivation of nouns from verbs, there are some differences. Let us now t urn to the derivation of nouns from verbs in comparison 2. 2.3 Nouns Derived f rom V erbs Sesotho has three major types of nouns derived from verbs The se three types are distinguished based on their prefixes and suffixes. The n oun class prefix determines the number and class to which the derived noun belongs. The three basic suffixes associated with noun s derived from verbs are i; o and a. Personal nouns generally have the prefix mo and suffix i whereas the impersonal ones predominantly have the suffix o and a variety of prefixes The s uffix a is primarily associated with nouns derived through the prefixation of the class 15 noun class prefix ho. This suffix, unlike the others occurs with both personal and impersonal derived nouns. The examples below i llustrate these facts. The examples in (5a e) show that one verb stem rat based on the different noun class prefixes. Although the deverbal nouns have different prefixes, the relation in meaning between these nouns is preserved by the stem. The different noun suffixes indicate that there has to be an agreement between the noun prefix and the suffix. Examples (5d) and (5e) show noun formation involving verbal inflection. In this instance the suffix is consistently a re gardless of the noun class prefix.

PAGE 40

40 The examples in (5e) and (6) are central to this study. Although this type of noun formation is observed elsewhere unlike other nouns, the final vowel of the verb remains unchanged. The various suffixes that we see wit h other deverbal nouns are not possible with class 15 nouns as indicated by the examples in (5e) and (6a). This pattern reveals that class 15 nouns are different from the other nouns with respect to noun formation although they employ the same strategy of prefixing the noun class prefix to verbs. (5) Nouns derived from the verb rata (a) mo rat i /*o cl.1 love Pers Suffix (b) le rat o /*i cl.5 love Imp Suffix (c) that o /*i cl.9 lov e Imp Suffix (d) mo rat uw a cl.1 love PASS FV (e) ho rat a /*i/*o cl.15 love FV (6) Nouns derived from deverbal nouns (a) ho nk a seabo cl.15 take FV part (b) mo nk a seabo cl.1 take FV part (c) mo nk i seabo cl.1 take Pers Suffix part

PAGE 41

41 These s imilarities and differences are crucial in un derstanding the dual function of class 15 nouns in Sesotho. Let us now look at the properties of morpheme ho in Sesotho and how it relates to the infinitive 2. 3 Properties of the Infinitive in Sesotho Earlier works in the studies of the infinitive tended to focus mainly on the nominal properties of the infinitive while neglecting the syntactic distribution. The infinitive in Bantu is often re (Creissels & Godard 2005 for Tswana Mugane 2005 on Logoli and Visser 1989 on Xhosa). In Sesotho t he infinitive functions as a noun, similar to the English gerund and as a verb in its infinitival form. In languages such as English, the infinitival form of the verb and the gerund are morphologically distinct. In Sesotho this morphological distinction is absent. This distinction is further eroded by the fact that the infinitival form and the noun class 15 prefix share the same phonological form. In additi on, the infinitive in Sesotho accommodates some form of future tense aspect and mood (TAM). Th ese properties are shared by many other Bantu languages. R ecent studies of the infiniti ve in Bantu such as Creissels and Godard 2005 Visser 1989 and Du Plessis 1982 show evidence supporting the idea that the infinitive inflects for TAM. ons about the dual nature of noun class 15. As I have already indicated, there is no morphologica l or phonological distin ction between the infinitive and noun class 15 nouns. The main question that follows from this is how one determines whether these are proper nouns, like all other noun classes, derived nouns or the infinitive (verbs) In the sec ti ons that follow, I explore the nominal a nd the verbal properties of noun class 15.

PAGE 42

42 2. 3 .1 Distributional Properties o f N ouns Customary discussions of noun classes in Bantu tend to pay little attention to the differences between class 15 nouns and the rest of the noun classes. This view tended to diminish the importanc e of the many occurrences of class 15 nouns in syntactic positions wh ich other nouns do not occupy Certain tests are used cross linguistically to test for word class membership. These includ e the ability of that particular word to act as either a subject or an object in a sentence, the ability to trigger subject or object agreement the ability to be modified by descriptive modifiers as well as being able to pluralize. Nouns in class 15 pass m ost of these tests Let us examine the se various tests by starting with c lass 15 nouns as subjects and objects in a sentence. (7) Nouns as subjects and objects in Sesotho (a) Makwala a timela kapele 6.coward 6.SM die quickly C o wards die q u (b) Ho nanar a ho a tshosa 15 sneak up 15.SM FOC scary S neaking (c) Bana ba rata ho ja 2.children 2.SM like 15 eat C hildren like eating /to eat (d) Bana ba rata dijo 2.children 2.SM like 10.food C Subjects trigger subject agreement (SM) in Sesotho. Sentence (7 a) shows that the subject noun makwala a We observe the same pattern with class 15 nouns indicated by sentence (7b) where the noun ho nanara triggers the SM ho Sentence (7c) shows that class 15 nouns also act as objects in the same way that the object noun in (7d) does.

PAGE 43

43 Like other nouns in Sesotho, class 15 nouns can be modified by descriptive modifiers such as adjectives, demonstratives and possessive pronouns. The examples in (8) illustrate. Nominal modifiers such as adjectives and demonstratives agr ee with the nouns they modify in Sesotho. Sentences (8a&b) indicate how the adjective agrees with the nouns it modifies. Sentences (8c&d) are examples of nouns used with personal possessive concords. The same kind of agreement observed in (8a&b) is also ap plicable in (8c&d). Examples (8e&f) furthermore show that the agreement relationship between the demonstrative and the noun it qualifies is also required. Class 15 nouns beh a ve like other nouns in this respect. (8) Descriptive Predicates (a) Makwala a ma be 6.cowards 6.SM 6.AC bad C (b) Ho bapala ho ho be 15 play 1 5.SM 15.AC bad P laying (c) Bokwala ba hae bo mmo laisitse 14.coward 14.PC 1. his 14.SM 1.OM ki ll (d) Ho bapala ha hae ho fedile 15 play 15.PC 1. his 15.SM end playing (game) (e) Makwala ana a ma be 6.coward 6.DEM 6.SM 6.AC bad T (f) Ho bapala hona ho ho be 15.play 15.DEM 1 5.SM 15.AC bad T Object nouns also offer an interesting trait in Sesotho. Object nouns trigger object agreement when the emphasis is placed on the object or when the object is preposed or om it t ed Whereas prepos ing is possible with class 15 nouns, it appears that object agreement is not

PAGE 44

44 possible if the object is a class 15 noun. The examples in (9) reveal the differences between class 15 nouns as objects and object nouns in other classes. In (9a), the object dijo object agreement as well as a focus morpheme on the verb when preposed. As sentence (9b) demonstrates, this is not the case with class 15 nouns in object position. Instead we see in (9c) that only the focus marker is used for emphasis unli ke in (9a) where both OM and FOC are present. Sentence (9d) illustrates that the focus marker is required when preposing the object. (9) OM in preposing (a) Dijo bana ba a di rata 10.food 2.children 2.SM FOC 10.OM like As for food c hildren like (b) *Ho ja bana ba ho rata 15. eat 2.chilrden 2.SM 15.OM like (c) Ho ja bana ba a rata 15.eat 2. children 2.SM FOC like (d) *Ho ja bana ba rata 15.eat 2. children 2.SM like I Another way of emphasizing the object noun in Sesotho is by prefixing the OM to the verb without preposing In this instance the object is highlighted as the emphasis is on the OM. In the presence of the OM the o bject noun becomes optional. Exampl e (10b) once more shows that class 15 nouns are not capable of triggering OM in object position. (10) OM in emphasis (a) Bana ba a di rata ( dijo ) 2.children 2.SM FOC 10.OM like food C (b) *Bana ba a ho rata ( ho ja ) 2.children 2.SM FOC 15.OM like 15.eat

PAGE 45

45 So far we notice that class 15 nouns pass all the distributional properties of nouns in subject position. In object position there are some restrictions. Class 15 nouns fail to trigger OM when used in this position. At this point we also need to remember what we noted earlier that class 15 nouns also fail to pluralize. The failure of class 15 nouns to pluralize and trigger OM could be attributed to what Creissels & Godard 2005 term semantic hierarchy. Creissels & Godard propose that in the hierarchy of semantic o bjects, class 15 nouns denote an abstract object, which should be viewed as an eventuality. An eventuality is a notion associated with verbs. This is not surprising because as I pointed out earlier, studies in Bantu languages show that class 15 nouns show nominal as well as verbal properties. Let us now turn to the verbal properties of these nouns. 2. 3 .2 Distributional Properties o f V erbs Bantu languages are well known for their rich verbal morphology Verbal characteristics relevant for the purposes of t his study include subject and object agreement, tense, aspect, mood, negation as well as the ab ility to take verbal extensions and verbal modifiers. I deal with verbal extensions in detail in C hapter 3. The first property that we will look at is the ability to clauses as infinitival clauses since we are looking at verbal properties associated with the infinitive. The examples in (11) illustrate some of the properties The fir st difference between the verb halefisa when used with ho is that it does not support subject agreement (SM) as demonstrated by (11a c). The OM is however supported. In the examples in (11a&b) the infinitival verb is extended by the causative extension. Th is is a common property of verbs in Bantu.

PAGE 46

46 (11) Subject object agreement and verbal extensions (a) Banna ba mo halef is a 2.men 2.SM 1.OM ngry CAUS FV M (b) Banna ba rata ho mo h alef is a 2.men 2.SM like INF 1 .OM angry CAUS FV M (c) *Banna ba rata ba ho mo halef is a 2.men 2.SM like 2.SM INF 1 .OM anger CAUS FV Example (12b) shows that the infinitival clause can be modified by an adverb of manner haholo as well as a locative m osebetsing in the same manner that ordinary verbs are modified as indicated by sentence (12a). (12) Verbal Modifiers (a) Banna ba mo halef is a haholo mosebetsing 2.men 2.SM 1.OM angry CAUS FV very work.at M en make him/her very angry at work (b) Banna ba rata ho mo halef is a haholo mosebetsing 2.men 2.SM like INF 1.OM angry CAUS FV very work.at M en like to make hi m /her very angry Transitive verbs require an object. The fact that the verb used in examples (12a) show OM suggests that this is a possibility These examples clearly indicate that the infinitival clause accommodates a direct object (13) Direct objects (a) Banna ba halef is a mme 2.men 2.SM angry CAUS FV 1.mother M en make mother angry (b) Banna ba rat a ho halef is a mme 2.men 2.SM like INF angry CAUS FV 1.mother M Ordinary verbs inflect for tense, aspect and mood. Tense in Sesotho is divided according to time: past (recent and remote), present and future (recent and remote). The only tense that the

PAGE 47

47 infinitival clause supports in Sesotho is the future tense. The future tense morpheme is treated with caution here since the same morph eme also works as a verb in other contexts 1 The major difference between the infinitival clause and finite verbs clauses is that the infinitival clause can only be used with the future tense Let us look at some example s (14) Future Tense (a) Banna ba tla halef is a mme 2.men 2.SM FUT angry CAUS FV 1. m other (b) Banna ba rata ho tla halef is a mme 2.men 2.SM like INF FUT angry CAUS FV 1.mother (c) Ba tla tla sebetsa 2.SM FUT come work (d) Ba tla ho tla tla sebetsa 2.SM come INF FUT come work to Example (14a) is an example of a finite clause in the future tense form. The future tense morpheme tla is prefixed to the verb between the subject agreement morpheme and the verb. Example (14b) on the other hand shows the use of the future tense morpheme with the infinitival clause However, it is important to note here that the future tense morpheme can also function as a verb of motion as represented in sentences (14c&d). Additionally the same morpheme may also be used to conjugate the two clauses in (14c) ba tla and sebetsa and ba tla and ho sebetsa in (14d). It is evident then from these examples tha t the future tense morpheme in the infinitival clause may be used as a conjunction as well as marking a future action. 1 Doke (1957:188) claims that the infinitival clause inflects for aspect in Sesotho. His observation is based on the use of the infinitive in isolation. Visser 1989 also notes that the use o f the future tense in Xhosa functions as an auxiliary verb, rather than tense: to come and the infinitival form

PAGE 48

48 A related but slightly different pattern is observed with aspect in Sesotho. Aspect marks the phase or duration of an action. The progre ssive has been cited as an aspectual morpheme associated with the infinitive in Sesotho, Xhosa and Setswana (Doke 1957, Visser 1989 and Creissels & Godard 2005 respectively). As indicated by the examples in (15) the use of progressive is restricted to cert ain environments. These examples illustrate. (15) Aspect Progressive (a) Banna ba sa halef is a mme 2.banna 2.SM PROG angry CAUS FV 1.mother (b) # Banna ba rata ho sa halef is a mme 2.men 2.SM like INF PROG angry CAUS FV 1.mother make mother angry (c) Banna ba sa rata ho halef is a mme 2.men 2.SM PROG like INF angry CAUS FV 1.mother mother angry (d) Ho sa halef is a mme ha ho a loka INF PROG angry CAUS FV 1. mother NEG 15.SM FOC right make mother angry Sentence (15a) shows the use of the progressive with an ordinary finite verb. The progressive morpheme sa is prefixed to the verb between the SM and the verb. Sentence (15b) indicates that the use of the progressive morpheme with the subordinate infinitival clause is less acceptable. This is because the same sentence may be better expressed when the aspect is associated with the main clause as in (15c). However, when the infinitival clause is used as a subject the use of the progressive aspect is then acceptable in Sesotho. Mood is expressed using various verb forms in Sesotho. A mood may either be finite or non finite. The non finite mood forms in Sesotho are the infinitive and imperative. Non finite moods a re traditionally regarded as verb forms that lack tense and agreement. So far we have noticed that the infinitive in Sesotho shows very limited instances of tense and agreement. The

PAGE 49

49 infinitival clause under investigation is expected not to accommodate other mood forms given that it is a mood form itself. Let us observe the following exampl es. (16) Mood Potential (a) Banna ba ka halef is a mme 2.banna 2.SM POT angry CAUS FV 1.mother (b) Banna ba rata ho ka halef is a mme 2.banna 2.SM like INF POT angry CAUS FV 1.mother Intended: make mother angry (c) Ho ka halef is a m me ha ho a loka INF POT angry CAUS FV 1. mother NEG 15.SM FOC right Intended: make mother angry The examples in (16) show the expected re sult. Sentence (16a) illustrates the potential mood form used with a finite verb. The potential morpheme ka is prefixed to the verb between the SM and the verb. When used with the infinitival clause, the sentence becomes less acceptable. Example (16c) furt her shows that using the potential mood with the infinitival clause in subject position does not change the situation. Finally, n egation in the indicative is expressed by morpheme ha and the final vowel changes to e In the indicative negative, the morphe me is inserted between the subject and the SM (17a). In the infinitive the negative morpheme is represented as se and the final verb also changes to e. Example (17b) shows that negation is possible with the infinitive as a subordinate clause. Negation is also possible when the infinitival clause is used in subject position (17c) This morpheme is also used as negation for the imperative as illustrated by (17d). (17) Negation (a) Banna ha ba halef is e mme 2.men NEG 2.SM angry CAUS NEG 1.mother (b) Banna ba rata ho se halef is e mme 2.men 2.SM like INF NEG angry CAUS NEG 1. Mother

PAGE 50

50 (c) Ho se halef is e mme ho lok ile INF NEG anger CAUS NEG 1.mother 15.SM right PERF (d) Se halef e NEG angry NEG The discussion so far indicates that the infinitival clause sh ows the majority of properties associated with verbs as well as nouns in Sesotho. Tables 2 3 and 2 4 summarize these findings. It i s evident from this summary in T able 2 3 that the infinitive behaves like ordinary nouns although there are minor differences Table 2 4 further indicates that the infinitive in Sesotho has some form of inflection although it does not share all the properties of finite structure. This brin gs us to the question asked in C hapter 1 of the nature and identity of the clauses introduc ed by h o Is it a verb or a noun? Table 2 3 Nominal properties of the infinitive Property Nouns Infinitive Prefix & stem Yes Yes Plural Yes No Subject Position Yes Yes Object Position Yes Yes Preposing Yes Yes Trigger Subject Agreement Yes Yes T rigger Object Agreement Yes No Nominal Modifiers Yes Yes The answer to this question has been a major area of investigation amongst Bantui s ts. other nouns, ing both verbal

PAGE 51

51 These statements are also supported by Mugane 2003 where he observes that the behavior of the infinitive and the gerund in Logoli is that of an ex tra sequential hybrid construction. Table 2 4 Verbal properties of the infinitive Property Finite verb Infinitive Subject Agreement Yes No Object Agreement Yes Yes Verbal Extensions Yes Yes Verbal Modifiers Yes Yes Direct Objects Yes Yes Tense Yes Yes (Restricted) Aspect Yes Yes (restricted) Mood Yes No Negation Yes Yes The conclusion that can be drawn from the discussion and the observation s from other scholars is that the infinitive is more restricted in object position. Most of the instances where the infinitive fails a morphological or syntactic test are when used as a n object. This is not surprising since verbs select their internal arguments As we will observe in Chapter 4 certain verbs have restrictions on the kind of arguments they sel ect. The infinitive is one such argument. The question that follows from this conclusion relates to the internal structure of the infinitive. I deal with these answering this question let us look at other uses of ho in Sesotho. 2. 4 Other Uses o f Ho The mor pheme ho in Sesotho has various other uses. In one of the uses, the morpheme is prefixed to noun stems where it functions as a noun class marker. While this is not evident in the orthography system, some nouns do show this property of ho In this section, I describe the use of ho as a conjunction coordinating verbs as a preposition, as a copulative and as an expletive or impersonal SM.

PAGE 52

52 The conjunction used to join nouns in Sesotho is le the verb that expresses the secon d and subsequent action is expressed either in the infinitival form or the subjunctive form as indicated in the examples (18) Sentence (18) shows that in coordinated verbs, the second verb is in the infinitival form. Sentence (19) on the other hand shows that although the infinitive and subjunctive may overlap in their occurrences, the subjunctive form cannot co occur with the conjunction in the same way that the infinitive does. (18) Re tla [ reka] le [ho bala] dibuka 1 PL FUT buy CONJ INF r ead books W e will buy and read books (19) Re a reka (*le) re bal e dibuka 1PL FOC buy (CONJ) 1P L read SUBJ 10. books W e buy books and ( then ) read H o may also be used as a preposition. In the example in (20) ho is used as a preposition ho is also used to express direc In sentence ( 2 2 ) ho is used with fihla (arrive) to This use is clearly associated with the locative nouns in view of the fact that they h ave a tendency to express locati on, time or manner. (20) O re ho nna o ya masimo ng 2SG say to 1SG 2SG. SM go f iel ds LOC (21) Leba ho yena go to 3SG o to him (22) O hanella ka dikobon ng ho fihla nako ee? 2SG stay in blanket LOC to arrive 9. time 9.DEM nkets until this time arrives There is yet another very common use of ho in Sesotho asso ciated with impersonal subject or copulative. Sentence ( 23 ) expresses an instance of the use of ho a s an impersonal copulative

PAGE 53

53 Sentence ( 2 4), on the other hand shows the use of ho as an unspecified subject of a passive sentence or an expletive The examples in (23 ) & ( 24 ) show the various uses of ho in Sesotho which have not received atten tion in the literature A lot of issues remain unsolved regarding the question of whether these are historically related or not. Synchronically, there seems to be some relation between the morpheme used as a preposition and the locative nouns as discussed earlier in this chapter. If we consider the use of ho to indicate a movement towards or away from some location, it can be linked directly to the locative classes in Sesotho. Although the locative classes have very few nouns left in Sesotho, their agreeme nt marker is also ho and they are formed (prefix +stem) in the same way as other nouns The nominal infinitive may be a remnant of this process in Sesotho. (23) Ho lok ile morena COP fine PERF 1. chief (24) Ho robets w e mme yena o a sebets a Ho sleep PASS FV CONJ 3SG 1. SM FOC works FV P eople are s leeping and he is working Taking in to consideration the other instances of ho although they ma y look remotely related to the infinitive as we know it we can begin to accept that the two uses of the infinitive under discussion are also related. There is strong evidence to suggest that the re is a relation between the infinitival form and the nominal form of the infinitive. Mc Laughlin (personal communication) points out that there is possible evidence for the relationship between noun class markers and infinitival markers found in Pulaar (F ula), a northern Atlantic (Niger Congo) languages. In Pulaar, a language with more than twenty noun classes (although these are not paired as in the Bantuist tradition), the noun class suffix, de, also shows up as an infinitival

PAGE 54

54 marker suffixed to a verb stem, as in arde 'to come', waalde 'to spend the night', etc. Conversely, in Wolof, there is no class morphology on the noun, only on determiners agreeing with the noun, and there is also no infinitival marker. This ties the presence of the infinitive to the presence of noun class morphology. The next step is how to distinguish between the two. In the discussion earlier in this chapter, we noted that the infinitive tend s to be restricted when used in object position. This is to be expected since this posit ion is regulated by the properties of the selec t ing verb. Since the concern of this study is wi th control verbs, the next step is to look at the internal syntactic property of the infinitive as a complement of control verbs. 2. 5 PRO as the Subject of the I nfinitive in Sesotho Sesotho control verbs usually select a noun, the infinitive or subj un ctive as a complement. A typical embedded subjunctive clause contains a subject which may have the same referent as the subject of the main clause but does not have to. This subject may be a lexical one or just a subject agreement morpheme also known as pro The subjunctive as a complement fulfills all the require d principles as outlined in C hapter 1 To illustrate let us look at this example: (25) Subcategorization a. Hopots a : V[_______NP (CP)/(IP)] b. Sebetsa : V [_______(NP/PP)] (26) Ke hopotsa mme [ CP hore a sebets e lapeng] 1SG remind 1.mother that 1.SM work SUBJ at home mother to w remind mother that she must work at home The subcategorization tells us that the verb hopotsa argument requires an object NP and an optional CP or IP. In the main clause, the two theta roles are assigne d one to the subject Ke mme In the embedded clause, the other theta role is assigned the subject a which is pro. As a pronominal, pro must be free in its minimal clause. This condition is met becaus e the possible

PAGE 55

55 binding antecedent for pro is mme The CP is a maximal projections which marks the minimal clause therefore mme can bind it When it comes to the infinitive as a complement of control verbs, there are major differences between the subjuncti ve and the infinitive. First, the subject of the embedded clause is understood as the same as the subject or object of the main clause In addition, n o overt lexical DP is allowed in the subject position of the embedded infinitival clause. Let us look at t hese examples. (27) Ke i hopotsa mme j [ IP PRO *i/j ho sebetsa lape ng] 1SG remind 1.mother INF work home LOC (28) *Ke i hopotsa mme j [ IP o *i /*j ho sebetsa lape ng] 1SG remind 1. mother 1.SM INF work home LOC for her to Sentence (27) indicates that the subject of the embe dded clause can only be co indexed with the object of the main clause. One would wonder why we need to posit a null NP and then co reference it with a nearby NP This is where we go back to the Extended Projection Principle ( EPP), the Theta criterion, Cas e theory and the PRO theorem. Let us see how these would work. EPP requires all clauses to have subjects. We have already established that the infinitive because of its verbal qualities may also function as a clause. As a clause it requires a subject. T he first option is to suggest that mme may perform this function in which case we would bracket the sentence as follows: (29) Ke i hopotsa [ IP mme j ho sebetsa lapeng] 1SG remind 1.mother INF work home LOC The first problem with sentence (29) is that we have already established that the infinitive in Sesotho does not show subject agreement, suggesting the absence of an overt lexical subject in that positi on. Secondly, we saw in sentence (28) that a lexical NP is not allowed in this position.

PAGE 56

56 Thirdly, according to the Theta criterion, mme is already receiving a theta role from the main verb it cannot receive two as this would be a violation of the Theta cri terion. The next problem for this kind of bracketing has to do with C ase. The object of the matrix clause mme receives accusative C ase as indicated by these examples. (30) Mme o sebets a lape ng. 1.mother 1.SM wo rk FV home LOC Mother (31) K e a mo hopotsa (mme) 1SG FOC 1. OM remind (mother) Example (30) shows the subject agreement for mme as o However, in object position the agreement associated with mme is mo ( 31) This suggests that mme in object position receives accusative Case An accusative case in a subject position would be a problem for the Case filter. We noted earlier in the chapter that the cla usal infinitive has some form of irrealis future tense supported. Example s (30 ) and (31) also indicate that the presence of subject agreement is an indicati on of a Case assigner. Since the infinitive does not support subject agreement the lack of Case in this position is supported. This would then force the object/subject to move into a Case assigned position. This would leave us with an empty subject positio n of the embedded clause violating both the EPP and the Theta criterion. However i f we go back to the initial proposal where we posit PRO as the subject of the infinitival clause we then satisfy the EPP, the Theta criterion and the Case filter The next step is to see if t he subject position of the infinitive satisfies the PRO theorem. According to the PRO theorem, the subject of the infinitive must be a Caseless and an ungoverned position. We have already established that the subject of the infinitive i s a Caseless position due to the lack of SM This subject position of the infinitive also fails to be assigned Case by an external head such as

PAGE 57

57 the verb in the main clause because there is a maximal project IP which becomes a barrier to government, a requi red condition for Case assignment PRO becomes the only possible NP for this position. There are other tests that are used to compare the subject position of finite and non finite clauses. The common tests include Weak Cross Over (WCO) effect, interpretati on under VP ellipsis as well as the presence of an overt NP in the subject position. Sesotho has morphological processes that compensate for moved elements. Us ing tests that involve movement such as WCO become a challenge. Another challenge has to do with the fact that Wh movement, which is used in WCO test is optional in Sesotho. Let us look at some examples to illustrate the point. (32) Sello o rek ile koloi. 1.Sello 1.SM buy PERF 9.car (33) Sello o rekile eng? 1.Sello 1.SM buy PERF what (34) Ke eng eo Sello a e rek ile ng _____? COP what 9.REL 1.Sel lo 1. RM 9.OM bought PERF REL ? Sentence (32) is a perfective sen tence. Example (33) shows that the object is questioned in situ whereas (34) shows an optional movement of the wh phrase. Note however that the movement of the wh phrase across the verb and the subject leaves an OM and requires the relative clause morphology on the verb and SM The WCO effect obtains when a bound variable is moved across another one (Terzi 1997). However, if one of the variab les is PRO this effect is escaped. E xamples (35 38) demonstrate. (35) Mme wa hae var1 o a mo var2 rata. 1.mother 1.POSS his 1.SM FOC 1.OM love

PAGE 58

58 (36) # Ke mang var 2 eo mme wa hae var1 a mo var 2 rata ng? COP who 1.REL 1.mother 1.POSS his 1.SM 1.OM like REL mother (37) PRO var1 ho bina ha hae var 2 ho a mo var3 kgathatsa. PRO INF sing 1.POSS his 15.SM FOC 1.OM bother (38) Ke mang var3 eo PRO var 1 ho bina ha hae var2 ho mo var3 kgathats ang? COP who 1.REL PRO INF sing 1.POSS his 15.SM 1.OM bother.REL Example (35) shows two bound variables represented as var1 and var2 If we move var2 over var1 in question formation as in (36) the result should be dubious. However due to the OM that stays put the sentence is strange but acceptable. In (37) one of the variables is PR O. If we move the questioned variable o ver PRO we do not get the WCO effect observed in (36). The test suggests that the PRO NP is different from oth er NP s su ch as mme wa hae The VP ellipsis test gives us a better picture. Under VP ellipsis we get a slo ppy interpretation when using the infinitive. In VP ellipsis the verb phrase is omitted, the omitted phrase is usually replaced by word such as too, as well in English. The two examples (39&40) show that we obtain a sloppy reading when the subject of the e mbedded clause is PRO and a strict reading when it is a lexical NP (40). (39) Sello o kopa ho sebetsa, le bana ba a 1.Sello 1.SM ask INF work, and 2.children 2.SM FOC (to work) (40) Sello o kopa hore a s ebetse, le bana ba a a sebetse) 1.Sello 1.SM ask that 1.SUBJ work, and 2.children 2.SM FOC ( 1 .SM work) Sello asks that he work s and the children ask too ( In (39) the omitted part is ho sebetsa of the coordinated sentence. The preceding sentence creates an environment that can help retrieve the omitted informa tion. Note that ho sebetsa if added to the second sentence takes the subject as bana as opposed to Sello the original subject.

PAGE 59

59 This is called the sloppy reading. Sentence (40) on the other hand shows a different pattern. The omitted part in this case is a sebetse Note however that the SM of the omitted verb still refers to the original subject. This is called strict reading. What this tells us is that the subject of the infinitive takes its interpretation from the nearest possible NP, in which case bana b ecomes the does. This is another indication that the subject NP of the infinitive is indeed different from o ther noun phrases 2. 6 Nominal and Clausal I nfinitive One of the questions that continue to be debated in Bantu literature concerns whether the distinction between clausal infinitive and nominal infinitive is valid 2 In the preceding sections I indicated that the nominal infinitive shows both nomin al and verbal properties. I have also demonstrated that the nominal infinitive occupies the positions that other NPs occupy. The challenge lies in the instances where there seems to be syntactic ambiguity as presented in (41). (41) Mme o rata ho ja 1.mother 1.SM like INF eat Sentence (41) as it stands cannot give us much to say about whether ho ja is a nominal or clausal infinitive Since the nominal infinitive seems to occur in all positions that other noun s occupy in Sesotho o ne way would be to treat nominal infinitives like other nouns in terms of syntactic structure. The ambiguity in (41) would be represented as in 2 1(b) a structure similar to other nouns (2 1(a)) The immediate problem with assuming t he structure in 2 1 (b) for the nominal infinitive is that it does not capture the observed differences between the nominal infinitive and other nouns 2 Doke & Mofokeng 1957, Rugege 1971, Du Plessis 1982b, Visser 1989, Mugane 2003, Creissels & Godard 2005

PAGE 60

60 in Sesotho. N ominal infinitive s show verbal properties such as negation and object agreement which are not possible with other nouns The structure in 2 1 is not able to accommodate the required nodes associated with negation and agreement. If we take the nominal infinitive as some infinitive. (a) DP (b) DP D NP D NP N N Dijo ho ja Figure 2 1 The noun in Sesotho Visser 1989, propose s a three way distinction in accounting for the infinitive in isiXhosa. According t which would be represented by 2 1(a) the nominal infinitive and the clausal infinitive. In Sesotho however, there are no observed class 15 nouns, instead we get the nominal infinitive and the clausal in finitive. These two are represented in 2 2 (a &b). (a) NP (b) IP DP N IP PRO I VP DP ho ja PRO I VP Ho ja Figure 2 2 The Nominal a nd c lausal i nfinitive

PAGE 61

61 There are various advantages associated with the adopti on of the structure s in 2 2 First, 2 2 (a) clear l y suggest that ho ja is a noun as indicated by NP. The second advantage is that whereas the NP may take modifiers such as demonst ratives and possessives, the structure also allows the expansion to include the verbal properties associated with the infinitive. The next question then is why we need two structures when one can do. A closer look at the distribution of the nominal infini tive suggests that not all instances of the nominal infinitive are NP Let us look at some examples. (42) Sekolo se rek is a bana dibuka 7.school 7.SM buy CAUS FV 2. Children 10.books (43) Sekol o se rek is a bona tsona 7.school 7.SM buy CAUS FV 2.PRON 10.PRON (44) Mmuso o kgothaletsa ho bua 3.government 3.SM encourages INF speak urages speaking (45) Mmuso o kgothaletsa hona 3.government 3.SM encourages 2.PRON (46) *Mmuso o kgothaletsa bana 3.government 3.SM encourages 2.children the child (47) Mmus o o kgothaletsa b a na ho bua 3.government 3.SM encourages 2. children INF speak children (48) *Mmuso o kgothaletsa bona hona 3.government 3.SM encourages 2.PRON 15.PRON The data in (42 48) show that although the nominal and clausal infinitive overlap in their distribution there are some instances where one form is excluded. Example (42) shows a non control verb used with two nouns bana and dibuka. Example (43) illustrates that nouns may be

PAGE 62

62 represented by pro nouns provided the necessary agreement requirements (noun class) are met. The pronoun can therefore be used to test whether a constituent is a NP or no t Sentences (44&45) show that this is indeed possible with the nominal infinitive. Example (46) however, illustrates that although (45) passed the pronoun test, the noun in (44) is not the same kind of NP as the ones in (42). We then conclude that there i s a difference between NP (noun classes 1 14) and nominal infinitive. The next step involves the differences between the nominal infinitive and the clausal infinitive. In (47) the same verb kgothaletsa is used with two arguments, an NP and NP/IP. Senten ce (48) becomes ungrammatical when we replace the same construction ho bua. Note however in this instance that when the pronoun nominal infinitive is used the sentence becomes ungrammatical. This is a clear indication that ho bua in (44) and (47) cannot be the same constituent. The differences between ho bua in (44&47) depends on the verb that select s them and other arguments of the verb. In (44), ho bua is an NP hence it can be replaced by a pronoun. In ( 47) however ho bua is an IP. These different i nterpretations are directly linked to the semantic classes to which the selecting verbs belong. I deal with these semantic classes in C hapter 4. 2. 7 Summary In this chapter I have developed the proposal that the infinitive in Sesotho behaves like nouns a nd verbs at the same time. I have indicated that while the infinitive shows most of the nominal properties it falls short as far as nominalization a nd plural formation are concerned. A close examina tion of noun classes and noun formation indicate s that the infinitive may function as a noun, however like other noun classes that share the same SM, it does not follow the same pattern as other nouns in noun formation. The infinitive retains the verbal suffix in noun derivation, a property which I directly li nk to verbal chara cteristics.

PAGE 63

63 I have also looked briefly at the behavior of the infinitive in relation to other verbs in Sesotho. This examination reveals that while the infinitive shows some form of inflection this is not the same type of inflection as fully fledged finite clauses. I have associated the l ack of SM in the infinitive to the lack of necessary morphology such as Case assigner due to some form of defective Tense morphology. Finally I have used tests to show that the subject of th e infinitive as co mplements of control verbs is different from the subject of other complements such as the subjunctive. In C hapter 3 I discuss the structure of verbs in Sesotho with special focus on the role of verbal extensions in determining the controller and control relations.

PAGE 64

64 CHAPTER 3 ASPECTS OF SESOTHO MORPHO SYNTAX RELEVANT TO CONTROL 3. 1 Introduction The goal of this chapter is to provide a description of some pertinent aspects of Sesotho morpho syntax operational in control structures. The main focus is on the general structure of the verb with special attention paid to verbal extensions. In order to understand the differences between the infinitive and subjunctive as complements of control verbs it becomes necessary to look at the basic sentence structure in Sesotho, in particular the role of subject agreement and object agreement. I start the chapter by providing an outline of basic Sesotho clause structure indicating the slots associated with inflectional and derivational morphemes. The remainder of the c hapter is dedicated to the description of Sesotho derivational morphemes (verbal extensions) paying particular attention to the ones that interact with control. 3.2 The Verb in Sesotho Sesotho verbs take various syllabic forms. There are monosyllabic, di s yllabic, vowel commencing and derived verb forms. Most ve rbs in Sesotho end with a vowel a which is normally referred to as a final vowel (FV) ( see Mchombo 2004 amongst others ). Verbal conjugation in Sesotho involves the use of prefixes, and suffixes. Pref ixes are usually associated with the subject and object agreement tense, aspect and mood. Suffixes usually represent derivational a ffixes such as verbal extensions Some forms of tense and aspect are also represented as suffixes While there are numerous morphological processes related to the verb, such as reduplication, I only show the position of inflectional and derivational affixes. Diagram (3 1) represents the morphological structure of the verb in Sesotho based on sentence s (1 3 ) (1) Mosadi o tla mo bul el a kajeno 1.Woman 1.SM FUT 1.OM open APPL FV today

PAGE 65

65 (2) Mosadi ha a sa tla mo bul el a kajeno 1 .wom a n NEG 1. SM PROG FUT 1. OM open APPL FV today a (3) M osadi ha a sa mo bul etse 1 woman NEG 1. SM PROG 1.OM open APPL. PERF woman is no longer opening for him Sentence (1) shows a future tense sentence with SM positioned between the subject and the future tense morpheme. The OM is closest to the verb root. Sentence (2) o n the other hand shows that the SM for noun class 1 changes changes from o to a when the sentence is in the negative form We also note here that the progressive morpheme comes before the tense morpheme. Sentence (3) shows that the perfective is marked po st verbally as opposed to the progressive aspect. NEG SM T/A OM Verb root Verbal Extension(s) T/A/FV Figure 3 1 Verb s tructure in Sesotho Putting these various positions together we come up with a templ ate exemplified in F igure 3 1 wh ich is a slight adjustment of Creissels 2005 representation 3. 3 Verbal Extensions Verbal extensions play a central role in Bantu verbal morphology. This role extends to syntax due to the interaction between these extensions and the arguments of the ver b. V erbal extensions occupy the position between the verb root and the final vowel (FV) in Sesotho. Sesotho uses verbal extensions to derive new words and extend or alter the meaning and argument structure of verbs. The verbal extensions in Sesotho are the causative / is /, the applicative / el /, the reciprocal / an / the passive / w / and the neuter / eh /. These extensions may be duplicated to express an intensified action. They may also combine to express different actions. Two of these verbal extension s reduce valency (reciprocal and neuter) and two increase

PAGE 66

66 the number of arguments (the causative and applicative). When used together in one verb the order in which these suffixes occur, follow the order observed in other Bantu languages: Neutar Causative Applicative Reciprocal Passive ( N CARP). Although the order is fixed, it is sometimes altered to attain idiomatic expressions. Let us now look at the causative, applicative and reciprocal individually as these are central to this study 3. 3.1 The C ausativ e The causative in Sesotho is realized as a suffix morpheme / is/. The allomorphs are / is/, / its / and / ets /. The causative in Sesotho translates to cause x to do y. It behaves slightly different from other Bantu languages (especially Xhosa) in that th e order of the internal arguments is fixed. In other languages such as Xhosa and Zulu inserting object agreement makes it possible to switch the order of these arguments. This does not happen in Sesotho. W hen the causative suffix is added, the argument se lected by the caus ee becomes the closest to the verb. The examples in ( 4 8 ) indicate these facts. (4) Ntate o pheha dijo 1. father 1. SM cook 10.food F (5) Ntate o pheh is a bana dijo 1.father 1. SM cook CAUS FV 2. children 10. food F (6) # Ntate o pheh is a dijo bana 1.father 1. SM cook CAUS FV 10 .food 2.children F ather makes (7) Ntate o ba pheh is a dijo 1.father 1. SM 2. OM cook CAUS FV 10.food F ather makes them cook food (8) Bana ba pheh is w a dijo ke ntate 2.children 2.SM cook CAUS PASS FV 10.food COP 1.father T

PAGE 67

67 Sentence ( 4 ) is an example of a basic tran sitive verb with two arguments. In sentence ( 5 ) the addition of the causative extension introduces an additional argument Exa mple ( 6 ) indicates that the order of the post verbal argument s is fixed. The order of the agreement markers is also fixed. The OM, when present is always closest to the verb. Example ( 7 ) demonstrates that only the argument selected by the causative triggers object agreement. It is also this argument that is fronted in passive formation as indicated in ( 8 ). In control structures when the causative is used with the matrix verb, it changes the syntactic and semantic properties of the verb thereby allowing a subject control verb to function as an object control verb. As can be observed from the examples, in English the causative is repre make causative makes changes to the form of the verb. In so doing, a verb which is normally a subject control verb may optionally function as an object control verb although t his changes the meaning of the verb When used with the verb in the embedded clause, the causative has no direct effect on control relations. 3. 3.2 The A pplicative The applicative, also called applied extension, is represented by a suffix morpheme / el/ wi th allomorphs [ el], [ il] and [ l]. The distribution of these allomorphs is phonologically conditioned. The applicative suffix in Sesotho is associated with the introduction of benefactive and locative roles. The benefactive role is more productive since Sesotho marks locative with other morphemes. Another semantic role associated with the applicative in Bantu is the instrumental role. The instrumental role in Sesotho is assigned by the instrumental morpheme ka Sentences ( 9 15 ) are some examples of the applicative in Sesotho. (9) Banana ba pheha bohobe 2.girls 2. SM cook 14. bread G

PAGE 68

68 (10) Banana ba pheh el a mme bohobe ifo 2.girls 2. SM cook APPL FV 1. mo ther 14. bread 5. fireplace G irls cook bread for mother (11) # Banana ba pheh el a bohobe mme 2.girls 2. SM cook APPL FV 14. bread 1. mother G (12) Banana ba bo pheh e l a mme ( bohobe ) 2.girls 2. SM 14. OM cook APPL FV 1. mother 14. bread (13) Banana ba mo pheh el a bohobe ( mme ) 2.girls 2. SM 1. OM cook APPL FV 14. bread 1. mother G irls cook bread for her (mother) (14) Mme o pheh el w a bohobe ke banana 1.mother 1.SM cook APPL PASS FV 14.bread COP 2.girls (15) # Bohobe bo pheh el w a mme ke banana 14.bread 14.SM cook APPL PASS FV 1.mother COP 2.girls The order of the internal arguments is generally fixed as indicated by ( 10 &1 1 ). However, an introduction of the OM as in ( 1 2 & 1 3 ) may change the order depending on the emphasis. If the emphasis is on the ben efactive/recipient then the direct object triggers an OM and becomes optional ( 12 ). When the emphasis is on the direct object, then the benefactive triggers OM and becomes optional ( 13 ). Note that the optional object is the one r epresented with an OM and becomes peripheral. Again like the causative, only the object selected by the applicative undergoes movement in passive formation ( 14 ) but not the object selected by the verb in its bare form ( 15 ). The applicative extension, unlik e the causative, plays a less prominent role in c ontrol structures. The applicative extension is associated with object control since most object control

PAGE 69

69 verbs are inherently applicative Manipulative 3 predicates that are inherently object control in Sesot ho, are semantically applicative. For most of these verbs, we cannot break the verb into a verb root and an applicative suffix. 3. 3.3 The R eciprocal The reciprocal morpheme in Sesotho is / an/. This affix adds a mutual relationship to the arguments of the verb (do x to each other). The reciprocal affix requires two conjoined subject noun phrases or a semantically plural subject. U nlike the causative and applicative the reciprocal reduces the number of arguments that a verb takes. The reciprocal deletes or eliminates the internal argument. This affix is also sensitive to the semantics of the verb. It requires a verb that is capable of expressing mutual relationship. Below are some examples of the reciprocal in Ses o tho. (16) Mme o bon a ngwana 1 .mother 1. SM see FV 1. child M (17) Mme le ngwana ba a bon an a 1.mother CONJ 1. child 2. SM FOC see REC FV M other and child see each (18) Ngwana o bon an a le mme 1.child 1. SM see REC FV CONJ 1. mother T (19) *Ngwana o bon an a mme 1.child 1. SM see REC FV 1. mother Sentence ( 16 ) is an exampl e of an ordinary bare verb with two arguments. Example ( 17 ) shows that the reciprocal requires singular conjoined subjects with a plural SM. If the subject is 3 See C hapter 4 for verb classes.

PAGE 70

70 singular then there must be a post verbal subject conjoined with the other subject ( 18 ) otherwise the sentence becomes ungrammatical ( 19 ). These properties of the reciprocal are crucial for an analysis of control verbs. In Sesotho, the reciprocal plays a role in determining referential dependencies. When a verb represents a collective action this is expressed with a verb that is inherently reciprocal in Sesotho. Collective verbs are usually used in expressions that show partial cont rol. As we have noted a reciprocal verb requires a plural or collective argument in Sesotho Becau se of this thematic requirement it becomes impossible to obtain partial control with collective verbs in Sesotho. This is a critical role since it determines the differences between Sesotho and other languages that show partial control. 3. 4 Summary In this chapter I described the internal structure of verbs in Sesotho in relation to inflection al and derivational affixes. In f l ectional affixes such as tense, a spect and agreement are prefixed to the verb although some aspectual and tense morpheme s occupy the position before and after the verb. Derivational affixes such as verbal extensions occur as suffixes and are positioned between the verb root and the final vowel I then represented this information about the verb in a linear format indicating the various positions associated with SM and OM. Finally I took a general look at verbal extensions and how their argument selection impacts on the overall sentence st ructure. In Chapter 4 I explore Sesotho control verb classes and their interaction with verbal extensions

PAGE 71

71 CHAPTER 4 SESOTHO CONTROL VERB CLASSES 4.1 Introduction One of the questions raised in C hapter 1 seeks to answer the observed differences in argumen t selection by verbs which share the same semantic class. I noted that verbs such as hopotsa and qophella this regard is that these verbs share the same syntactic behavior. I point ed out that this expectation is not met as exemplified by these sentences. (1) [ matrix Mme i o hopotsa bana [ subordinate PRO i ho sebetsa]] 1.mother 1 .SM remind 2.children to work (2) [ matrix *Mme i o hopotsa [ subordinate PRO i ho sebetsa]] 1.mother 1. SM remind to work (3) [ matrix Ntate i o qophella [ subordinate PRO i ho sebetsa]] 1.father 1.SM insist to work (4) [ matrix Ntate i o qophella bana j [ subordinate PRO *i/j ho sebetsa]] 1.father 1.SM insist 2.children to work Sentences (1) and (4) show the expected similarities in syntactic behavior between the two verbs. Sentences (2) and (3) show some differences in syntactic behavior. In what way do we capture the similarities and differences? If argument structure can be derivable from the meaning syntactic properties? This question is central to the ongoing debates on lexical entries and to what extent t hey determine syntactic properties, an area which affects various areas of linguistic research. In this chapter I deal with the interaction between semantics, morphology in the

PAGE 72

72 derivation of control verbs and argument structure and the role each play s in the classification of control verbs in Sesotho. 4.2 Verb Classification Methodology The question of the extent meaning plays in determining syntactic structure is debated in many areas of linguistic theory. Within the area of Lexical Functional Grammar t here is a move towards the argument determined (See Levin and Rappaport Hovav 1991). Levin 1993 proposes that verb meaning is key to its syntactic behavior in the same way that verbs that f all into the same classe according to their shared behavior also show shared meaning components. In this framework, argument structure may be derivable from the meaning of verbs Within the Minimalist framework there are also various approaches to the rela tion between meaning and argument structure. One such approach claims that properties of control verbs similar to the ones in (1 4) can be obtained through syntactic means (Hornstein 1999, 2003, Boeckx & Hornstein 2006, amongst others). These are the scho lars who treat control as an instance of movement thereby doing away with PRO in obligatory control. Others claim that there needs to be some reference to the semantic properties of the verb in question and that semantic classes help to account for the di fferent syntactic behaviors associated with control (Cullicover & Jackendoff 2001, 2003, 2005, 2006). There are yet other scholars, while supporting a different syntactic approach, who claim that purely semantic or purely syntactic approaches to obligatory control are inadequate (Landau 2003, 2006). In this approach morphological agreement and semantic tense is claimed to account for a cross linguistic typology of control. While my classification borders on a combination of a syntactic (control class) and a semantic (semantic verb classes) approach, I also use verbal extensions (morphology) as an extended means of classification.

PAGE 73

73 In order to explore the phenomena of control in Sesotho, an understanding of the semantic and syntactic properties associated with verbs is necessary before drawing any conclusions. If verb classes are viewed as a means of organizing the lexicon, capturing shared verb behavior as well as a way of identifying grammatically relevant elements of meaning (Fillmore 1970) then using a comb ination of approaches becomes necessary. As a starting point, I assume that verbs that share some syntactic behavior such as subject control, share some elements of meaning. I have also used the already established semantic verb classes (Stiebels 2006) in grouping verbs in Sesotho Secondly, I assume that a further look at the meaning of the verbs within broader syntactic classes and their interaction with verbal morphology results in further sub classifications. I did this by extracting verbs and other pr edicates whose selectional properties included the infinitival morpheme ho. The various verbs associated with ho as a complement were tested to establish their ability to induce control and referential dependencies. Having realized that some of the verb c lasses are either morphologically causative or applicative, I used these morphemes as a further classifying factor. The syntactic properties (subject /object control, control shift) were then mapped onto the semantic verb classes resulting in a typology of referential relations associated with each control verb class complement in Sesotho. I end the chapter with a section that discusses verbs that have been identified as potential control verbs in other languages but are not control verbs in Sesotho. For t he purpose of exemplification in this chapter, I take the traditional view of PRO as the subject of the infinitival clause. I discuss the theoretical significance of the control verbs classes in Chapter 6 4.2.1 Desiderative Verbs Desiderative verbs are verbs that express desire, aspiration or what is needed or wanted. In Sesotho, desiderative verbs are subject control verbs. Table 4 1 represents a semantic class of

PAGE 74

74 verbs that also happen to share semantic and syntactic properties as well as a small over lap of morphological properties. The common syntactic property shared by these verbs is that they are all subject control verbs. The morphological property, the ability for verbs to interact with derivational suffixes (in this case the causative and appli cative), depends on the meaning of each verb 4 The examples in (5&6) indicate the use of these verbs. (5) Bana j b a ikem is ets a PRO j ho tsamaya 2.children 2 .SM intend CAUS APP FV INF go (6) *Bana i ba ikem is ets a mme j PRO j ho tsamaya 2.children 2.SM intend CAUS APPL FV 1.mother INF go In sentence (5) the subject bana indexed with the null subject of the embedded infinitival clause PRO. In sentence (6) however, since verbs from this class only allow subject control, the object of the matrix clause mme is ruled out as the controller of the embedded clause, in fact these verbs do not even allow an object in the main clause in control relations. Table 4 1 also shows that desiderative verbs can interact with the causative and applicative extensions. The ap plicative extension introduces a benefactive role (an object) whereas the causative extension introduces a patient role (object). The applicative form of the desiderative verbs is realized as manipulative verbs in Sesotho. Sesotho control verbs select dif ferent kinds of complements. The choice is between a noun phrase (including a nominal infinitive), a subjunctive clause (introduced by hore ) and a clausal infinitive. In C hapter 2 we established that the nominal infinitive shares syntactic positions with o rdinary nouns in Sesotho. However, there are some restrictions when it comes to the clausal infinitives. One of the reasons for this 4 As I indicated in Chapter 3 this morphological process has to be treated with caution because there have been instances in the literature (Machobane 1997) where the derived form of the verb has been mistakenly analyzed as allowing control shift

PAGE 75

75 restriction is that control verbs select the infinitive as a complement and each class of control verbs has its own restric tions with regards to complements. This is only evident when we take each class of verbs and compare its argument selection preferences between nouns, subjunctive clauses and the clausal infinitive. Table 4 1 Desiderative v erbs Desiderative verbs when used in other contexts where there is no control relation select an ordinary nominal noun phrase (NP) as a complement in Sesotho. This selectional property is shared by the majority of subject control verbs although there are some differences. Let us look at some examples to illustrate this behavior. The examples in (7) show that the control verbs rata in (7a&b) and hana in (7c&d) select ordinary NPs as well as clausal infinitival (IP) complements However, verbs such as ikemisetsa g). Sentence (7e) shows that ikemisetsa a noun (7f). Example (7g) on the other hand shows that the noun is only allowed as the object of the infinitival clause. Verb Gloss Subject Con trol Object Control Control Shift Causative Applicative qeta decide Y N N Y Y ikemisetsa intend Y N N Y Y hlorela long Y N N Y Y kgetha choose/prefer Y N N Y Y batla want Y N N Y Y rata like Y N N Y Y rera plan Y N N Y Y lebella expect Y N N Y Y h ana refuse Y N N Y Y dumela agree Y N N Y Y hloka need Y N N Y Y tshaba afraid Y N N Y Y hlophisetsa Plan/prepare Y N N Y Y lalalabela aspire/yearn Y N N Y Y

PAGE 76

76 The first thing is to assume that ikemisetsa must belong to a different class of verbs. However, the verbs rata and ikemisetsa complement selection preferences cannot be attributed to class membership only. It turns out that ikemise tsa hlophisetsa are derived using the causative and applicative extensions is and el simultaneously. These two extensions when used together require only one internal argument; either NP or IP but never together. If only one of the two arguments is used, it has to be the infinitival argument (7f&g). All the other desiderative verbs select both the infinitive and a proper noun as a complement. (7) Noun phrases as complements 5 (a) Bana ba rata [dijo] NP 2.children 2.SM like food (b) Bana ba rata [ ho ja] IP 2.children 2.SM like to eat (c) Bana ba hana [dijo tsa kgale] NP 2.children 2.SM refuse 10.food 10.POSS old (d) Bana ba hana [ ho ja dijo tsa kgale] IP 2.children 2.SM refuse to eat 10.food 10.POSS old (e) Bana ba ikemisetsa [ho bina] IP 2.children 2.SM intend to sing (f) *Bana ba ikemisetsa [ pina] NP 2.children 2.SM intend song (g) Bana ba ikemisetsa [ho bina pina] IP 2.children 2.SM intend to sing song 5 I have decided to use the infinitival clause without PRO in this section of the chapter in order to highlight the role played by other complements of control verbs.

PAGE 77

77 (h) Bana ba hlophisetsa [ho bina pina] IP 2.children 2.SM plan to sing song (i) *Bana ba hlophisetsa [ pina] NP 2.children 2.SM plan song Sentences (7h&i) show that hlophisetsa complement and not an NP. It appears then that desiderative verbs in Sesotho fall into two categories based on their argumen t structure. Desiderative verbs that are bare (no causative or applied extensions) select NP or IP but never both. Desiderative verbs that are derived using the a combination of the causative and applicative only select an clausal infinitive as a compleme nt. These are exemplified in (8). (8) Subcategorization frames for desiderative verbs (a) rata, V [__ NP, IP] (b) hana V, [___NP,IP] (c) ikemisetsa (d) hlophisetsa The other complement choice that fits in with the s ubcategorization exemplified in (8) is the subjunctive. As I have already mentioned, subjunctive complements are very common with control verbs in Sesotho. The subjunctive as an embedded clause is introduced by a complementizer hore 6 This complementizer interrogatives. In some environments, the complementizer is optional in Sesotho.The subjunctive occurs with various tense and aspect morphemes. The examples in (9) illustrate the use of the subjunctive with ten se and aspect morphemes. 6 There is yet another use of the subjunctive which is not related to control verbs which is worth pointing out. The morpheme hore as pointed out by du Plessis (1989), is used to introduce embedded questions. This form is selected by interrogative predicates in Sesotho. This fact becomes relevant in that it accounts for the lack of interrogat ive control verbs in Sesotho (discussed in C hapter 4 ) suggesting a natural class Also, like in Xhosa, hore is formed with the infinitival prefix ho re hore is somehow related to the in finitive as proposed by du Plessis (1989).

PAGE 78

78 (9) The subjunctive (a) Letsatsi le a tjhaba [le dikel e] 5.sun 5SM FOC rise and set SUBJ (b) sa tsebe [ hore ke tla re ng] 1 SG NEG know that 1SG FUT say NEG (c) Ke tseba hore ba ne ba tsama ile maobane 1SG know that 2.SM PAST 2.SM left PERF yesterday The example in (9a) shows the use of the subjunctive where it indicates a second action. When performing this function, the subjunctive is not introduced b y complementizer hore. Instead it has the same function as the infinitive in this regard. Sentence (9b) shows the type of subjunctive that we are concerned with in this section. In sentence (9b) the subjunctive is used in the future form whereas in example (9c) it is used with past perfective. This is the major difference between the subjunctive and the infinitive although they appear to perform the same function. Desiderative verbs select freely between the infinitive and the subjunctive. The subjunctive is the preferred choice when the subject of the embedded clause is different from the subject or object of the main clause. When the subject of the embedded clause is the same as the subject or object of the main clause then the infinitive becomes a better option. The examples in (10) illustrate. (10) The subjunctive and desiderative verbs (a) Mme o kgetha hore [ba tsamay e] 1.mother 1.SM prefer that 2.SM leave SUBJ (b) Mme k o kgetha [PRO k ho tsamaya] 1.mother 1.SM prefer to leave

PAGE 79

79 (c) Mme o kgetha hore a tsamay e 1.mother 1.SM prefer that 1.SM leave SUBJ Sentence (10a) shows an example of the subjunctive used where the subject of the subjunctive is different from that of the matrix clause. Sentences (10b&c) differ in that in (10b) the subject of the embedded clause is only understood as having the same reference as that of the main clause (control). Sentence (10c) on the other hand shows that in non control the subject of the subjunctive ca n be understood as having the same reference as that of the matrix clause but not necessarily. The infinitive does not have the same morphology (such as SM) required to make this distinction between the subject/object of the main clause and the subject of the embedded clause. The subjunctive becomes the better option when this distinction is required. 4.2.2 Manipulative Verbs Manipulative verbs express control or manipulation by the subject. Manipulative verbs in Sesotho are the extended or the derived fo rm of the bare desiderative verbs. The ability of a desiderative verb to undergo derivation and act as a manipulative verb also depends on the semantics of each verb. Manipulative verbs can be divided into three subclasses: those that are morphologically causative, those that are morphologically applicative and those that are bare forms inherently applicative. I have decided to merge the morphologically causative and applicative control verbs here based on the shared syntactic property of object control. I will label these two classes subclasses 1 and 2. Subclass 1 comprises of verbs that are object control verbs with no possibility of control shift. Subclass 2 is made up of a small number (from the ones I found) of verbs that are inherently applicative in form. They cannot be further broken down in the same way as the verbs of the first sub class. Finally subclass 2 of

PAGE 80

80 manipulative verbs, unlike subclass 1 allows subject and object control thereby participating in control shift. The following examples s how these patterns: Table 4 2 Manipulative verbs (10) Manipulative verbs (a) Morena i o dumella basadi j PRO *i / j ho sebetsa [NP IP] 1.king 1.SM allow 2.women INF work (b) *Morena i o dumella PRO i h o sebetsa [IP] 1.king 1.SM allow INF work (c) Morena i o qophella PRO i ho sebetsa [IP] 1.king 1.SM insist IN F work (d) Morena i o qophella banna j PRO i/j ho sebetsa [NP IP] 1.king 1.SM insist men INF work The king insists Sentence (10a) illustrates that verbs in subclass 1 can only function as object control verbs. This is expressed through the co indexation of the matrix object and the null subject of the embedded clause. The same sentence also reveals t hat the verb does not support subject control Verb Gloss Subject Control Object Control Control Shift Causative Applicativ e hopotsa remind N Y N Y N tlohedisa stop N Y N Y N Lebatsa forget N Y N Y N lemosa advice N Y N Y N hanela forbid N Y N N Y thibela prevent N Y N N Y bolella tell N Y N N Y dumella allow/let N Y N N Y hanela refuse N Y N N Y laela order N Y N N Y qobella force/compel Y Y Y N N hatella press Y Y Y N N qophella insist Y Y Y N N tshepisa promise Y Y Y N N kopa ask Y Y Y Y Y

PAGE 81

81 as observed in (10b). This is demonstrated with the co indexation of the matrix subject and the null subject of the embedded clause which results in ungrammaticality. Sentence (10b) further exemplifies that sub class 1 of the manipulative verbs obligatorily requires an object in the main clause. Sentences (10c*d) on the other hand show that subclass 2 manipulative verbs allow both subject and object control. However, in the presence of an object, the control is shifted from the subject to the object. The two classes represented in T able 4 2 coincide with the different argument structures associated with manipulative verbs. Once more we see the morphology of the verb providing that necessary distinction which de termines argument structure. Manipulative verbs that are either causative or applicative require two arguments, an NP and an IP. Manipulative verbs that are inherently applicative require and NP and IP, however, the NP is optional in this instance. The su bcategorization for this class may be represented as in 11. (11) Subcategorization frames for manipulative verbs (a) Hopotsa, V, [____NP IP] (b) dumella, V, [____NP IP] (c) Qophella, V, [____ (NP)IP] Manipulative verbs also select the subjunctive as a complement. Although there is this choice between the subjunctive and the infinitive as with desiderative verbs, manipulative verbs prefer the subjunctive in all aspects. For example, when manipulative verbs are used with an object and the subjunctive, the sentence sounds bet ter if the object of the matrix clause refers to the subject of the embedded clause (the subjunctive). However, when only the subject is used with the manipulative verbs then the subject of the embedded subjunctive can refer to some other subject without m aking the sentence sound odd. The examples in (12) illustrate.

PAGE 82

82 (12) Manipulative verbs with the subjunctive (a) Morena o hopotsa banna hore ba kopan e 1.chief 1.SM remind 2.men that 2.SM meet SUBJ (b) Morena o hopotsa mme hore bana ba sebets e 1 .Morena 1.SM remind 1.mother that 2.children 2.SM work SUBJ (c) Morena o hopotsa hore ba sebets e 1.chief 1.SM remind that 2.SM work SUBJ S entence (12a) shows the subjunctive used where its subject has the same reference as the object of the matrix clause. Sentence (12b) is acceptable but sounds as if some information is missing between the object of the matrix clause and the subject of the s ubjunctive. Sentence (12c) is acceptable because the object of the matrix verb is implicit. It may or may not be the same as the subject of the embedded clause. Although I make reference to the subject of the subjunctive as having the same reference as th e subject or object of the matrix clause this should not be taken as control. The subjunctive as a complement of control verbs allows a lexical noun and fully supports SM as well as tense and aspect morphology, most of which are unavailable in control rela tions. The most important difference between the subjunctive and the infinitive when used with control verbs is that there is no co indexing required since any kind of relation is expressed overtly through SM or a lexical noun. Finally, it is important to note that the traditional utterance/ communication verbs are also classified as subclass 2 of manipulative verbs. The reason for merging these classes is due to the shared semantic grouping as well as similar syntactic behaviors. It is also vital to note t hat subclass 2 of the manipulative verbs shows the need for semantic categories since syntactic

PAGE 83

83 behavior only would fail to capture the similarities shown available through semantic class belonging. 4.2.3 Achievement Verbs Achievement verbs are those ver bs that express the bringing about of an achievement, success, triumph or lack thereof. Achievement verbs in Sesotho are subject control verbs with no option of shifting to object control. Although these verbs allow the derivation with the causative and the applicative, in this case they are used in their bare, non derived forms. The derived forms of these verbs are realized as belonging to the manipulative verb class. Recall that desiderative verbs are also subject control verbs, and that these also pa rticipate in subject control only. The sentences in (13) exemplify achievement verbs. (13) Achievement verbs (a) Bana k ba lebala PRO k ho tsoha hoseng 2.children 2.SM forget INF wake up morning (b) *Bana k ba lebala mme j PRO k /*j ho tsoha hoseng 2.children 2.SM forget mother INF wake up morning Note that while the causative form of the achievement verbs occur as manipulative verbs ( ho p o la hopotsa ; lebala lebatsa ; tlohela tlohedisa applicative forms of th ese verbs do not participate in the same manner ( hopotsa hopolela feel ; lebala leballa forgive ; tlohela tlohella form of the desiderative form is realized as manipulative verbs whereas the causa tive form is not? Now we observe that the other set of manipulative verbs is formed by the derived forms, in this case causative form of the achievement verbs. This is another example of how the morphological property determines class and syntactic proper ties.

PAGE 84

84 Table 4 3 Achievement v erbs Achievement verbs may also be used in other contexts where there are no control relations. When used this way, these verbs select an NP as a complement. The examples in (14) illustrate the different complement of achievement verbs (14) Achievement verbs and their complement (a) Banana ba hopola mme NP 2.girls 2SM remember 1.mother (b) Banana ba hopola ho tsamaya IP/NP 2.girls 2.SM remember INF go irls remember to go/ going (c) Banana ba hopola hore ba tsamay e IP 2.girls 2.SM remember that 2.SM go SUBJ (the girls) (d) *Banana ba hopola mme ho tsamaya NP IP/NP 2.girls 2SM remember 1.mother INF go Intended: leaving Sentence (14a) indicates that hopola Sentence (14b) on the ot her hand shows that hopola complement. Achievement verbs also behave the same way as manipulative verbs with regards to the subjunctive. However, unlike manipulative verbs, they have a strict requirement on the subject of the subjunctive in that it can only refer to the subject of the matrix clause (14c). Verb Gloss Subject Control Object Control Control Shift Causative Applicativ e hopola remember Y N N Y Y Lebala forget Y N N Y Y tlohela refrain Y N N Y Y phema avoid Y N N Y N Hana decline Y N N Y Y hlokomoloha negl ect Y N N N N

PAGE 85

85 Although it is possible to use the subjunctive with the achievement verbs, the infinitive is the preferred choice. The contrast between (14b&c) highlights that achieveme nt verbs require the subject of the subjunctive to have the same reference as that of the matrix verb. This pattern is necessary. This may be the reason for achievement verbs working better with the infinitive since they require strict co referencing. The subcategorization frames for achievement verbs is exemplified in 15. (15) Subcategorization frames for achievement verbs (a) Hopola (b) Tlohela (c) Hana As can be noted in the translation, the complement may be treated as nominal or a clausal complement. This is one syntactic property that ties this class together Another semantic property observed in this class is that these verbs tend to have meanings that make them overlap with desiderative verbs. In addition, it is important to note that these verbs share a similar argument structure to bare desiderative verbs 4.2.4 Factive Verbs Factive verbs express a certain reality or an accepted fact. Factive verbs are subject control verbs in Sesotho. In addition factive verbs appear mainly in their bare forms although a few have the applicative form. Factive verbs tend to be used more in their stative/perfective form as indicated by verbs such as with the applicative can also be used in their bare form without effecting a change of meaning or affec ting the control relations. Here are some examples: (16) Factive verbs and their complements

PAGE 86

86 (a) Mme k o holy a PRO k ho fihla bosiu 1.mother 1.SM hate FV INF arrive 14.night (b) Mme k o hlo ile PRO k ho fihla bosiu 1.mother 1.SM hate PERF INF arrive 14.night (c) Ke k thaba PRO k h o utlw a ditaba tseo 1SG glad INF hear FV 9.news 9.DEM (d) Banana k ba thab el a PRO k ho dula lape ng 2.girls 2 .SM glad. APPL FV INF stay home LOC Sentences (16a&b) contrast the use of the bare form of the verb (expressed in present tense habitual) with the perfective counterpart indicating no change in meaning or cont rol relations. Sentences (16c&d) then again, indicate that other verbs can be used in the bare form or applicative without any change in meaning or control relations 7 These verbs however, are distinguished from the other applicative verbs (such as manipul ative verbs) in that they are subject control verbs. Here again we see the need to make reference to both the semantic and syntactic properties. Table 4 4 Factive v erbs 7 Most of these verbs have nominal counterparts which also select infinitival complement such as Verb Gloss Subject Control Object Control Control Shift Causative Applicativ e nyats a loath/hate Y N N Y N hloya dislike/hate Y N N Y N rata like/love Y N N Y N swabela sorry Y N N N Y thabela glad Y N N N Y

PAGE 87

87 Table 4 4 show that while the use of the causative applicative is possible with factive verbs this is only limited to a few of these verbs. One other characteristic of factive verbs is that when used with the applicative, and two arguments are used, this results in the formation of a purposive clause. The sentences in (17) illustrate. (17) Factive verbs and clauses of purposive (a) Ntate o rat el a bana ho sebetsa 1.father 1.SM like APPL FV 2.children INF work (b) *Ntate o rat el a bana hore ba sebetse 1.father 1.SM li ke APPL FV 2.children that 2.SM work SUBJ Factive verbs pattern with achievement verbs in that they prefer the infinitive as a complement. Factive verbs do not allow the use of a noun and the subjunctive to gether eliminating the need for a difference between the object of the matrix clause and the subject of the subjunctive. The examples in (18) illustrate. (18) Factive verbs and the subjunctive (a) Mme o hloyile hore a fihl e bosiu 1.mother 1.SM hate that 1.SM arrive SUBJ at.night (b) *Mme o hloyile bana hore ba fihl e bosiu 1.mother 1.SM hate 2.children that 2.SM arrive SUBJ at.night (c) Mme o hloyile hore ba na ba fihl e bosiu 1.mother 1.SM hate that 2.childre 2.SM arrive SUBJ at.night Example (18a) shows that with fa ctive verbs the subject of the subjunctive can refer to the subject of the matrix clause or any other subject not expressed in the matrix clause. Sentence (18b) on the other hand is ungrammatical because factive verbs do not allow an NP and IP

PAGE 88

88 occurring si multaneously as complements. The subcategorization frames associated with factive verbs are represented in (19). (19) Subcategorization frames for factive verbs (a) Nyatsa, (b) Thabela, (c) Swaba, The syntactic similarities that also overlap with some semantic similarities between some desiderative verbs, and achievement verbs are evident in these subcategorization frames for factive verbs. 4.2.5 Experiencer object Verbs Experiencer object ve rbs are verbs that express an experience or emotion encountered by the object of the verb. Like the name suggests, these are object control verbs. However, in the absence of an object, the control is shifted to an arbitrary control relation. I explore thi s characteristic of experiencer object verbs in Chapter 5 Unlike the other object control verbs (manipulative) this set of verbs makes use of a combination of the causative and the applicative form in their derivation. These two qualities distinguish th is class of object control verbs from the manipulative class whose morphology consists of either the causative or the applicative. Let us look at some examples. (20) Mmuso i o kgothaletsa bana k PRO* i /k ho bua Sesotho NP IP 3.go vernment 3.SM encourage 2.children INF speak 7.Sesotho (21) Mmuso i o khothaletsa PRO i /ARB ho bua Sesotho IP 3. government 3.SM encourages INF speak Sesotho (22) *Mmuso o kgothaletsa bana NP 3.government 3.SM encourage 2.children

PAGE 89

89 Example (20) demonstrates object control. When both the subject and object of the matrix clause are present the subject of the embedded clause (PRO) is co referenced with the object of the matrix clause. In the absence of the object in the matrix clause, the co referent of PRO is either arbitrary or an implied object as exemplified by sentence (21). The arbitrary reading associated with an object is claimed to be a rare occurrence (Stiebels 2007:73). It is also important t o note that in the absence of object, the subject of the matrix clause is not understood as the controller of the subject of the embedded clause. This is how the arbitrary reading is associated with the object of the matrix clause. This is another distingu ishing characteristic of experiencer object verbs. Table 4 5 Experiencer object verbs Experiencer object verbs behave in the same way as manipulative verbs with regards to the complement selection. This is expected because manipulative verbs are also object control verbs. Experiencer object verbs oc cur more with the subjunctive regardless of whether they are used together with the objects or not. The examples in (23) illustrate. (23) Experiencer object verbs with the subjunctive (a) Mmuso o kgothaletsa bana hore ba bu e Sesotho 3.government 3.SM encourage 2.children that 2.SM speak SUBJ Sesotho (b) Mmuso o khothaletsa hore ba bu e S esotho 3.government 3.SM encourage that 2.SM speak SUBJ Sesotho Verb Gloss Subject Control Object Control Control Shift Causative Applicativ e kgothaletsa cheer/applaud N Y N Y Y kgotsofaletsa satisfy N Y N Y Y hlonamisetsa sadde n N Y N Y Y thabisetsa amuse N Y N Y Y

PAGE 90

90 (c) Mmuso o kgothaletsa hore ho buu w e Sesotho 3.government 3.SM encourage that 15 speak PASS SUBJ Sesotho Sentence (23a) shows the subjunctive used where its subject has the same reference as the matrix subject although this i s not a necessary condition since the SM ba can refer to some other noun not expressed in the matrix clause. In (23b) the subject of the subjunctive may be related to the omitted object but not the expressed subject. Example (23c) expresses an impersonal s ubject or an expletive which is used with the passive. This is different from (23b) in that the subject of the matrix clause may be included as a referent. The subcategorization frames for experiencer object verbs pattern with those of sub class 2 of mani pulative verbs. The differences between experiencer object verbs and manipulative verbs lie in the interpretation of PRO. Experiencer object verbs have the following subcategorization frames: (24) Subcategorization frames for experiencer object verbs (a) Kgothalets a, (b) Hlonamisetsa, (c) Thabisetsa, Experiencer object verbs selects an NP and an IP, however, when only one is se lected it has to be the IP. In C hapter 5 I explore the d ifferences between these verb classes further. 4.3 Non control V erbs The verb classes presented in the preceding sections are classes that have been classified as control verbs in many languages. There are however, other verb classes that act as control v erbs in other languages but are not control verbs in Sesotho. The verb classes outlined in this section

PAGE 91

91 are those that have been proposed for other languages but do not function as control verbs in Sesotho. These are interrogative, perception, propositiona l and modal verbs. Interrogative verbs that induce control are generally associated with wh movement. Sesotho does not show wh movement. In instances of bi clausal interrogative sentences, the embedded clause is introduced by a complementizer which select s the subjunctive form o f the verb. As we have seen in C hapter 3, most of the control verbs also select the subjunctive form as a complement in non control relations. also do not participate in contro l in Sesotho. These verbs act like other ordinary verbs. Proposition or attitude verbs also do not occur with infinitival complement in Sesotho. Again, like the interrogative verbs, these verbs select the subjunctive form. There is one verb however, nahana This is the only verb from this class that allows control which makes its membership questionable or it suggests a move from the use of the infinitive to the subjunctive form. Modal verbs are restricted to raising i n Sesotho. Although these verbs select an infinitival clause, they also allow the insertion of an expletive in the matrix clause. However, when an expletive is inserted in the matrix clause, the embedded clause is a finite clause. The example below is used to illustrate this behavior. (25) Ke tshwanela [ho sebetsa.] 1SG must to work (26) Ho tshwanela [hore ke sebets e] It necessary that 1SG work SUBJ Although other c ontrol verbs also select the subjunctive form as a complement, the ability to insert an expletive in this manner is limited to this class of verbs. Another thing to mention is xiliary verb

PAGE 92

92 in Sesotho otherwise when used as main verbs such as achievement verbs. 4. 4 Summary The verb classes presented in this chapter are similar to verbs classes presented in other languages of the world (Noonan 1985; Pollard & Sag 1994; Wurmbrand 2001). However, there are three issues that came out of this chapter. The first concerns the role of morphology in determining argument structure which in turn determines control verb classes. As we have already noted the applicative form of the desiderative verbs is realized as manipulative verbs whereas the causative form of desiderative verbs is realized as achievement verbs. The combination of the causative and the applicative extensions gives us a class of e xperiencer object verbs. The second issue, also morphological, relates to the use of the verbal extensions in explaining control relations between the controller and the controllee. The verbs that are either causative or applicative in form tend to functi on primarily as object control verbs. The final point has to do with the fact that arbitrary control is normally associated with the subject especially when there is an implicit or an implied argument. Sesotho is one language that allows association of ar bitrary control with the object.

PAGE 93

93 CHAPTER 5 SESOTHO COMPLEMENT C ONTROL PROFILE 5 1 Introduction The primary goal of this chapter is to present the complement control patterns observed in Sesotho. Complem e nt control refers to a relation of obligatory coind exation between the subject and the object of the matrix clause and the subject of the subordinate complement clause. The secondary goal of this chapter serves to address the question of the role played by argument structure changing verbal morphology in c ontrol phenomena. In Chapter 3 I discussed the verbal morphology associated with control verbs in Sesotho. One of the areas discussed relates to argument structure changing verbal morphology. In this chapter I explore the role played by argument structure changing verbal morphology in accounting for the differentiated behavior of verbs that belong to the same class I also examine how this argument changing morphology i mpact s on the types of control attested in Sesotho. In so doing, I also explore how con trol relations help distinguish between the nominal infinitive and the clausal infinitive Chapter 5 is organized as follows: In the first part of this chapter I define terminology associated with control providing illustrative examples. I then go on to di scuss control types attested in Sesotho followed by an examination of referential dependencies. For each of the sections I try to determine whether class membership is an adequate determinant of referential properties in Sesotho. 5 2 Properties of Contr ol In C hapter 1 I defined complement control as a relation of obligatory co indexation between the subject or object of the matrix clause and the subject of the subordinate complement clause. The referential relations of the controlled element (PRO) are d eterm ined by those of the controller. This is called referential dependency. The two major types of obligatory control are

PAGE 94

94 forward control and backward control. In forward control, the controller is in the matrix clause whereas the controllee is in the em bedded clause. In backward control, the controller is in the embedded clause whilst the controllee is in the matrix clause (Polinsky & Potsdam 2006, Stiebels 2007). In another form of control, non obligatory control, there is no obligatory co indexation be tween the subject of the subordinate clause and the subject or object of the main clause. This study concentrates on obligatory control (OC) although I also discuss some aspects of referential dependencies associated with non obligatory control (NOC) where relevant. Obligatory c ontrol relations observed cross linguistically include subject and object control Subject control refers to a control relation in which the overt subject of the matrix clause is identified with the covert subject of the embed ded inf initival clause through co indexation In object control, the overt object of the matrix clause is identified with the covert subject of the embedded infinitival clause also through co indexation Implicit control on the other hand refers to a control rela tion in which the controlling argument is not syntac tically realized. Co indexation i s not required in this instance. These are exemplified in (1) in this order: implicit, object and subject control (1) The controller 8 (a) Mary signaled PRO to follow her (b) Mary as ked John j PRO j to leave (c) Mary i promised John PRO i to leave Sentence (1a) is an example of implicit control. In this sentence the reference of PRO is implied. In sentence (1b) s controller is the object of the matrix clause, resulting in object control Sentence (1c) is an example of subject control In all the sentences in (1) the referential 8 I have decided to use English examples in thi s section in order to provide an example of typology of control. I use Sesotho examples later in the chapter.

PAGE 95

95 configuration is such that PRO is understood as either the subject or object of the matrix clause. This is called exhaustive control. The other control relations are partial control and split control. In partial control, the subject of the matrix clause is included as one of the referents of PRO In split control, the su bject of the infinitival clause refers to tw o distinct arguments. Stiebels 200 7 provides the r epresentation in T able 4 1 to explain the refere ntial properties of PRO. In t he configuration in 5 1, k refers to PRO, i refers to a subject argu ment of the matrix clause, whereas j refers to an object argument of the matrix clause These properties in 5 1 are spread between obligatory control (OC) and non obligatory control (NOC) (Williams 198 7 ). Split control is a phenomenon which is restricted to NOC 9 The differences between OC and NOC are based on the phenomena in Table 5 2 (adapted from Hornstein 2 007: 114): Table 5 1 Referential p roperties of PRO Table 5 2 Properties of PRO in OC and NOC Obligatory Control (OC) Non Obligatory Control (NOC) (a) *It was expected PRO to shave himself. (a) It was believed that PRO shaving was important. (b) *John thinks that it was expected PRO to shave himself. (b) John i thinks that it is believed that PRO i shaving himself is important. (c) Joh himself. (c) i campaign believes that PRO i keeping his sex life under control is necessary for electoral success. ( d ) John expects PRO to win and Bill does too.(=Bill win) (d) John thinks th at PRO getting his resume is crucial and Bill does too. ( e ) *John i told Maryj i PRO i+j to wash themselves/eachother (e) John i told Mary j [that [PRO i+j washing themselves/each other] would be fun]]. 9 This position is still debated, see Madigan 2008 on split control in Korean. Type of control Subject Control Object Control Exhaustive k=i k=j Partial k i k j Split k= i+j

PAGE 96

96 Sentence (a) shows that PRO requires a n antecedent in OC. This is not the case in NOC ((a) of NOC) Sentence (b) shows that PRO requires a local antecedent in OC. Conversely PRO can have a non local antecedent in NOC The ungrammaticality of sentence (c) of OC shows that the antecedent must c command PRO. This is not a necessary condition as exemplified by (c) of NOC. Under VP ellipsis exemplified by sentence (d), PRO permits a sloppy reading in OC and only strict reading in NOC. Sentence (e) shows that NOC allows split antecedents whereas OC does not. Bearing these properties of OC in mind let us first examine how control verbs interact with NP s and subjunctive clauses as complement s 5 3 Control Types This section describes the control types observed in Sesotho, a relation that involves a co referntial relation between two arguments, one being the controller the other being the controlled element. The controller is defined as the argument of the control verb whose referent is (improperly) included in the referents of an argument of the embedd ed verb (Stiebels et al. (2003). The controller may be the subject or object of the control verb. As indicated by the word improperly in the definition, the controller may also be an implicit argument of the matrix verb. I explore these referential relati ons drawing from Sesotho data. 5 3 .1 Subject C ontrol Subject control obtain s when the subject of matrix clause is co referenced with the null subject of the embedded clause indicating a control relation. The majority of control verbs in Sesotho show subjec t control. Although manipulative verbs also participate as subject control verbs, I discuss them separately from the other verbs because they show a different pattern of control relations from the rest. The e xamples in ( 2 ) illustrate the various verb clas ses that participate in subject control. (2) Subject control

PAGE 97

97 (a) Desiderative : Ke i rata PRO i ho 1SG like I NF go hear what Rampou .. (b) Achievement : Ke i sitwa PRO i ho ba hlalosetsa 1SG unable INF 2. OM explain (c) Factive : Ke i swabela PRO i ho utlwa ditaba tse bohloko 1SG sorry INF hear 10. news 10.R M sad For all the sentences in ( 2 ), the subject of the matrix clause is co referenced with the subject of the embedded infinitival clause. The subject of the embedded clause in each case can only refer to the subject of the matrix clause. These are examples of exhaustive subject control. Let us look at object control verbs. 5 3 .2 Object C ontrol Object control obtains when the null subject of the infinitive in the embed ded clause is co referenced with the object of the matrix clause. Object control verbs are manipulative and experiencer object verbs in Sesotho. There are very few observed o bject control verbs in Sesotho. This limitation may be linked to the morphologica l make up of these verbs. These verbs are derived by suffixing the causative extension together with the applicative. Each verbal extension is semantically restricted to occur with certain verbs, when the se two affixes are put together on one verb the pool for the combined affix is smaller. The examples in ( 3 ) illustrate. (3) Object Control (a) Bashanyana i ba qobella dikgomo j PRO *i / j ho kena 2.boys 2.SM force 10.cows INF enter (b) Titjhere i o kgothaletsa bana k PRO *i / k ho bapala 1.teacher 1.SM encourage 2.childern INF play

PAGE 98

98 Sentence ( 3 a) is an example of a manipulative verb whereas ( 3 b) is that of an experiencer object verb. In each sentence the null subject of the embedded clause is understood as having the same reference as the object of the matrix clause. When the object of the matrix clause is omitted we get two important distinctions between the two classes. These differences are better discussed under a configuration where the control is shifted from one argument to the other which may or may not be related to one of the arguments of the matrix clause These verbs shed some important light on the distinction between the nominal infinitive and the clausal infinitive in Sesotho. Although it appears as if the two are interchangeable in other contexts, when used with control verbs this is not the case. A close look at the patterns of control associated with object control verbs in Sesotho shows that the distinction between nominal and clausal infinitive is justified. Let us look at contr ol shift with manipulative and experiencer object verbs. 5.3.3 Control S hift normal circumstances the controlle r of PRO is fixed either as ag ent or goal, in special circumstances the controller shifts to the other In Sesotho, there are two kinds of control shift associated with two object control verb classes The first one involves a s hift of control from the obje ct argument of the matrix clause to the subject argument of the matrix clause. The second kind involves a shift of control from the object argument of the matrix clause to an arbitrary control. Let us look at two kinds of control shift individually. Arbi trary control obtains when a subset of m anipulative verbs (subclass 1) is used without an object in the matrix clause. This set of verbs is made up of verbs that are morphologically causative or applicative. The examples in (4) illustrate a shift from obj ect control to arbitrary control

PAGE 99

99 (4) Manipulative verbs and arbitrary control (a) Titjhere j o thib el a bana k PRO *j/k/*ARB ho sebetsa 1.teacher 1.SM prohibit APPL FV 2.children INF work (b) Titjhere j o thib el a PRO j / ARB ho sebetsa 1.teacher 1.SM prohibit APPL FV INF work The t eacher prohibits (c) Titjhere j o lemo s a bana k PRO j /k/*ARB ho sebetsa 1.teacher 1.SM advise CAUS FV 2.c hildren INF work (d) Titjhere j o lemo s a P RO j / ARB ho sebetsa 1.teacher 1.SM advise CAUS FV INF work Intended: (someone/ anyone) In sentence ( 4a ) the verb thibela which has an applicative morphology is used with two arguments in the matrix clause The object argument of the matrix cla use is the controller of PRO. In sentence (4b) we note that for the same verb, when the object argument of the matrix clause is omitted, the referen t of PRO becomes arbitrary (marked as ARB ) Sentence (4c) is an example of a causative manipulative verb, l emosa used with two arguments in the matrix clause. Here we note that when the object of matrix clause is omitted (4d) the referent of PRO is not arbitrary as expected instead the sentence is ungrammatical These data are a challenge to my verb c lasses in that the two verbs thibela lemosa to the same class What could be responsible for the differences in referential dependencies? Before we answer this question let us look at how experiencer object verbs and the rema inder of manipulative verbs fit into the picture. The pattern observed with the applicative manipulative verbs in Sesotho is also possible with experiencer object verbs in Sesotho. Let us look at some examples. (5) Experiencer object verbs and arbitrary inte rpretation

PAGE 100

100 (a) Titjhere j o kgothal ets a bana k PRO *j/k / *ARB ho bapala 1.teacher 1SM encourage CAUS .APPL FV 2. children INF play (b) Titjhere i o kgothal ets a PRO *i /ARB ho bapala 1.teacher 1.SM encourage CAUS .APPL INF play Sentence ( 5 a) is an example of ordinary object control. Sentence ( 5 b) on the other hand shows that when the object is omitted the experiencer object verb force s an arbitrary interpretation. Let us bear these referential relations in mind as we look at another type of control shift. The second kind of control sh ift obtains when there is a shift in control from the object of the matrix clause to the subject of the matrix clause when an object argument of the matrix clause is omitted in control relations. Subclass 2 of manipulative verbs participates in this kind o f control shift. The examples in (6) illustrate (6) Subclass 2 manipulative verbs and control shift (a) Sekolo j se qobella bats w adi k PRO *j/k/*ARB ho reka dibuka 7.school 7.SM forces 2.parents IN F buy 10.books /compels (b) Sekolo j se qobella PRO j/ /*ARB ho reka dibuka 7.school 7.SM force INF buy 10.books Intended: /comples (itself) to buy Example (6a) demonstrates the use of a manipulative verb qobella the object of the matrix clause is the controller of PRO. When the object of the matrix clause is omitted, as in (6b), the control shifts to the subject of the matrix clause. This kind of shift is only possible with subclass 2 of manipulative verbs. Th e kind of interpretation that we observe with subclass 1 applicative and experiencer object verbs in examples (4&5) is treated as an instance of non obligatory control ( NOC ) In this instance the reference of PRO is freed from the arguments of the main clause in that it can refer

PAGE 101

101 to some other argument These data in ( 4 and 5 ) answer the question r aised in C hapter 2 of how to distinguish between the nominal infin itive and the clausal infinitive. The answer to this question lies not only in the internal structure of the infinitival complement or its position in a sentence, but also on the semantics of the control verb specifically the morphological makeup of that verb As we can observe from the English glosses, whenever the infinitive does not have its subject co referenced with the subject or object of the matrix clause it takes a nominal form (gerundive). This use is similar to the use of the nominal infinitive as a subject or a class 15 noun. What element of the control verb is responsible for the distinction between object to subject control shift and object to arbitrary control shift? In ( 6a ) the object control verb, qobella is co indexed with PRO because it is the closest c commanding NP. However, when th is object is omitted the control shifts from the object to the subject since it is the closest c commanding NP This is not the configuration we get in the case of arbitrary control. The initial thought woul d be that qobella is a manipulative verb wher e as kgothaletsa is an experiencer object verb. As observed with ( 4b ) thibela shift in the same manner that qobella does. In ( 4d ) we do not even get any kind of control shift with lemosa The question that we need to ask then is why only a subset of manipulative verbs allow s control shift. The answer to this question may lie with the morphology of these verbs Verbal extensions are used to extend the meaning of verbs but also to derive new verbs. The extensions that are associated with increasing the number of arguments are the applicative and the causative. In many instances, the introduction of a causative affix changes the verb class c ompletely. For example, I pointed out in C hapter 4 that achievement verbs such as hopola

PAGE 102

102 o. However, when the causative affix is added as in hopotsa n is an object control verb. The only manipulative verbs that allow control shift from object to subject in Sesotho are the one s that cannot take either the causative or the applicative extension since these are inherently applicative. These verbs, althou gh they may appear as f they contain an applicative Let us look at some examples. (7) The applicative extension (a) Ntate o pheh a dijo kantle 1.father 1.SM cook FV 10.food outside (b) Ntate o pheh el a bana dijo kantle 1.father 1.SM cook APPL FV 2.children 10.food outside (c) Ntate o pheh el a kantle 1.father 1.SM cook APPL FV outside (d) *Ntate o pheh el a 1.father 1.SM cook APPL FV Intended (e) Ntate o pheh is a bana dijo kantle 1.father 1.SM cook CAUS FV 2.children 10.food outside (f) *Ntate o pheh is a kantle 1.father 1.SM cook CAUS FV outside Int causes some cook ing (g) Ntate o pheh is ets a kantle 1.father 1.SM cook CAUS APPL FV outside Example (7a) demonstrates the use of the transitive verb pheha ithout the applicative extension. In (7b) we note that when the applicative extension is added another

PAGE 103

103 argument is introduced, a benefactor. Sentence (7c) shows that an applicative verb may be used without these two arguments. It is important to note tha t when this verb is used in this manner, the applicative takes a different meaning in that it requires a locative role. This is further emphasized by the fact that when the locative argument becomes ungrammatical ( 7d). We observe a similar pattern with the causative extension. In example (7e) the verb pheha In this instance the argument introduced by the causative affix, bana ic role form the one associated with the applicative. Sentence (7f) shows th e distinction b et we en the argument structure of the applicative and causative verb. This is the distinction that accounts for why applicative verbs can be used without the object of the matrix clause in control relations but causative verbs cannot. If we take another example using the applicative e xtension we note that e xperiencer object verbs in Sesotho fail to shift to subject control because they contain a combination of the ca usative and the applicative. Experiencer object verbs fail to shift control because the verbal morphology on the matrix verb requires an object. However, w hen this object is omitted we get an instance of arbitrary control Sentence (7g) demonstrates two things. The first one is that when the two extensions, the causative and applicative are used together, it is possible to omit the object of the matrix clause. The second one is that, although the object of the matrix clause is omitted, it is somehow impl ied. These data support the claim that I made earlier about the interactions between the morphology of verbs and their syntactic behavior. We observed earlier in the absence of the object in the matrix clause, the co referent of PRO with experiencer objec t verbs is either

PAGE 104

104 arbitrary or an implied object. Sentence (7g) demonstrates this connection between the verbal extensions and syntactic behavior. This use of the applicative subclass 1 and experiencer object verbs to induce arbitrary control relations is responsible for the distinction between the use of ho as an infinitival morpheme (PRO j/k ho) and ho as a nominal morpheme which is equivalent to class 15 noun class prefix (PRO ARB ho ). 10 This distribution accounts for why the nominal infinitive is restrict ed when used as an object of control verbs When used as a subject, the nominal infinitive lacks a reference or a controller, and therefore gets the PRO ARB reading. This is the similar reading we get with ho when it is used as an impersonal subject (explet ive). However when it is used in object position as a complement of control verbs, we get PRO ARB and PRO subject/object depending on the verb that selects it as well as its other arguments. In Sesotho, all control verb classes covered in this study allow the clausal (PRO subject/object ) and the nominal (PRO ARB ) infinitive. Subclass 2 of manipulative verbs and the causative verbs are the verbs that do not select the nominal infinitive. The examples in (19) illustrate. (8) Different interpretation of the subjec t of the infinitive i n Sesotho (a) Bana j ba rata PRO j / ARB ho robala 2.children 2.SM like INF sleeping The c hildren like to (b) Bana j ba rata PRO *j / ARB ho robala ha mme motsheare 2.children 2.SM like INF sleep 15.POSS 1.mother during.day (c) PRO ARB Ho robala ho a thusa. INF sleep 15.SM FOC help Sleeping 10 Visser 1989 made a similar observation regarding the nature of th e infinitive in isiXhosa. See C hapter 2 for an elaborate treatment of the infinitive in Sesotho.

PAGE 105

105 Sentence (8a) is ambiguous because the complement of the verb rata The difference between the two is that if PRO is co indexed with the subject ban a then the complement is an IP (clausal infinitive). If PRO gets an arbitrary interpretation, the complement is an NP. When we introduce modifiers such as possession, as in (8b) we note that PRO can only get an arbitrary interpretation because t his is the interpretation associated with nominal infinitives. This arbitrary interpretation is also observed in (8c), when the nominal infinitive is used as a subject. 5 4 Referential Dependencies Another aspect of control that is regulated by a combina tion of class membership as well as the morphology of the verbs is the referential relation between the controller and the controllee. In this section I look at how the verbal extensions impact on referential dependencies Referential dependencies refer to the differences in control verbs regarding their potential referential dependency between the controller (subject/object of control verb) an d the c ontrollee (null subject of the infinitive, PRO) (Stiebels et al. 2003 ). The two major referential dependenc ies are o bligatory control (OC) and non obligatory c ontrol (NOC). In this study, I focus on e xhaustive c ontrol ( EC ) and p artial c ontrol. Exhaustive control obtains when the subject or object of the matrix clause is understood as the same as the null subjec t of the embedded infinitival clause. Partial control is a control relation in which the interpretation of PRO properly includes the referent of the controller instead of being identical to the controller. The examples we have used so far are examples of exhaustive control which have a locality requirement between the controller and controllee. This locality requirement is regulated by the MDP (Minimal Distance Principle) which determines the interpretation of PRO. According to the MDP PRO must be co inde xed with the closest NP that

PAGE 106

106 c commands it. Let us look at some examples of referential dependencies and how they inter act with argument changing verbal morphology 5 4 .1 Exhaustive C ontrol Exhaustive control is a referential dependency in which the refe rents of the c ontroller and PRO overlap completely. This can i i In Sesotho, all verb classes allow exhaustive control as long as the lexical requirement of the embedded verb is met. Sentences ( 1 0) and ( 1 1) illustrate EC with t he subject and the object. (9) Badisana j ba kgetha PRO j ho tsamaya 2.shepherds 2.SM prefer INF go (10) Badisana j ba qobella bashanyana k PRO j /k ho tsamaya 2.shepherds. 2.SM force 2.boys INF go Going back to our MDP notation, we note that the nearest c commanding NP in ( 1 0 ) is badisana and all the referents of PRO are included in badisana PRO cannot possibly refer to some other implicit NP. In ( 1 1 ) on the other hand, the nearest c commanding NP is ba shanyana Again this is the only NP that PRO is co refer enced with. These two examples are instances of exhaustive control. 5 4 .2 Partial C ontrol Partial control presents a completely different picture. In partial control relations, the interpretation of PRO properly includes the referent of the controller ins tead of being identical to the controller. We represent partial control as NP j j + in line with the MDP. Collective verbs gather kopana are normally used as tests for partial and split control. In Sesotho, kopana is inherently recipr ocal. As indicated in C hapter 3 t he reciprocal requires a plural subject or conjoined singular subjects. As the examples in ( 22 2 4 ) show that Sesotho does not allow partial control.

PAGE 107

107 (11) Partial control and Sesotho verb classes (a) Morena i o batla PRO i /*i + m /ARB ho kopana 1. king 1. SM force INF meet T he king wants (b) Morena i o lebala PRO i / *i+m /ARB ho kopana 1.king 1.SM foget INF meet (c) Morena i o nyatsa PRO i / *i+m ARB/ ho kopana 1.king 1.SM hates INF meet (d) *Morena i o lemosa PRO i / *i+m /*AR B ho kopana 1.king 1.SM advise INF meet (e) Morena i o lemosa banna j PRO i /j/*ARB/*i/j+m ho kopana 1.king 1.SM advise 2.men INF meet (f) Morena i o hopotsa banna j PRO i /j /*j /j + m /*ARB ho kopana 1. king 1. SM force 2.men INF meet he king (g) Morena i o khothaletsa banna j PRO i /j/*ARB/*i /j +m ho kopana 1.king 1.SM encourage 2.men INF meet The da ta in (11) demonstrate two things. First, the reading associated with PRO i+m is not possible with all verb classes in Sesotho, where i represents one of the arguments of the matrix clause and +m some other unspecified argument All these sentences would b e ungrammatical under that reading. The second thing is that the morphology of the verbs (associated with semantic verb classes) determines the interpretation of PRO. Note that for each of the sentences (11a c), the sentence is only grammatical under PRO ARB because in this instance, kopana the reciprocal do not apply. However, this is only possible because these verbs also select the nominal infinitive as a complement. Sentence (11d) confirms this claim because we noted

PAGE 108

108 earlier that the causative manipulative verbs only select an IP as a complement. Sentences (11e g) also support the sensitivity of the interpretation of PRO to the verbal argument structur e. These sentences do not allow the arbitrary interpretation because in this instance the complement is an IP. This in turn rules out the partial control interpretation because the embedded verb kopana needs to have its argument structure requireme nts fulfilled. Since this requirement is not met we do not get a partial interpretation. 5 4 .3 Split C ontrol Split control is a referential dependency whereby two arguments of the control verb jointly control PRO (Landau 2000). In Sesotho, generally obje ct control verbs look like they allow split control as long a pronoun that refers to the matrix subject is included in the embedded clause. In In Sesotho, we can use mmoho Again, mmoho together Let us look at some examples. (12) Split control (a) Bana i ba dumella mme j PRO i+j/j ho kop an a mmoho 2.children 2. SM allow 1.mother INF meet REC FV together T he children allowed mother (b) *Morena i o tshepisa monna j PRO i /*i +j ho sebetsa mmoho 1.king 1.SM promise 1.man INF work together Intended: (c) Mme i o kopa bana j PRO i /*i +j /j ho hlatsw an a mmoho 1.mother 1.SM ask 2.children IN F wash APPL FV together (d) Mme i o tshepisa ngwana j PRO i /*i +j /j ho hlatsw an a mmoho 1.mother 1.SM promise 1.child INF wash REC FV together Mother (e) Mme o bolella ngwana hore ho hlapa mmoho ho monate 1.mme 1.SM promise 1.child that INF wash together fun 15 fun i/ i+j / j washing togethe

PAGE 109

109 There are a number of reasons why we do not get the interpretation where P RO is controlled b y both arguments of the matrix clause. First, t he notion of whether partial control is an instance of obligatory control continue to be debated, however, if we treat it as a case of NOC in Sesotho PRO would have an arbitrary interpretation. We have already noted that the arbitrary interpretation associated with PRO in Sesotho obtains when the object of the matrix clause is omitted. Partial control requi res both arguments of the matrix clause. T his itself would be a problem for Sesotho. Secondly, we noted that the interpretation of PRO is determined by the MDP. For each of the sentences in (12), the MDP rules out the subject of the matrix clause as the co ntroller since there object is the closest c commanding NP. This is indeed the case as indicated by the notation The third point relates to the requirement of the embedded clause. We have already e stablished that the embedded clause is IP, as such, the argument selection requirements of the verb have to be met. Reciprocal verbs require a singular plural so PRO must have a plural referent. As demonstrated by sentences (12a&b) this requirement is not met because the MDP excluded the possibility of joint controllers. Sentences (11c&d) further reveal that the embedded verb requires the plural controller. In (11c) the object of the matrix clause is plural whereas in (11d) it is singular. Finally we note d earlier that in Sesotho, the sub junctive is also sel ected by control verbs. Sentence (11e) shows that in order to get the reading where the subject of the embedded clause is controlled jointly by the a rguments of the matrix clause, the subjunctive is req uired. In this case there is no instance of control

PAGE 110

110 As can be observed from the data Sesotho does not allow the arguments of the control verb to control PRO jointly. In Sesotho, only one argument of the control verb may control PRO. Again we conclude that the argument structure requirements of the embedded verb ha ve an influence on referential dependencies. 5 5 Summary In this chapter I investigated the role played by argument structure changing verbal morphology in accounting for the differences in r eferential dependencies observed across and within verb classes. I then show how this verbal morphology impacts on a typology of control attested in Sesotho. I also explored how control relations help distinguish between the nominal infinitive and the clau sal infinitive. I subsequently examined control types in Sesotho. I discovered that object control verbs that allow control shift from object to subject are inherently applicative. Those that allow the shift from object control to arbitrary control are mor phologically applicative or have a combination of causative and applicative morphology. These are the verbs that help to distinguish between the nominal infinitive and the clausal infinitive. Finally, I established that the MDP and the argument structu re of the embedded clause is responsible for the lack of partial control and split control in Sesotho.

PAGE 111

111 CHAPTER 6 DISCUSSION AND CONCL USION 6.1 Summary In this study I embarked on a classification of control verbs in Sesotho. I identified 5 semantic cla sses: Achievement, factive desiderative which are subject control verbs and experiencer object and manipulative which are object control verbs. In order to capture all the syntactic and morphological properties associated with these classes I establish ed a typology of control in Sesotho. As a way of understanding control, I concentrated on three major questions: Question 1 : What is the status of ho in Sesotho? Question 2 : Are there different kinds of controlled complements and, if so, what are they? Quest ion 3 : What role does argument structure changing verbal morphology play in control phenomena? The answers to these questions were answered in the various preceding chapters. In C hapter 2 I explored the morpho syntactic properties of nouns and verbs in Ses otho. I examined the nominal infinitive (class 15 nouns) in relation to other noun classes. I showed that the nominal infinitive demonstrates all properties associated with other nouns although it fails to pluralize or trigger OM as an object. I compared t his pattern to other noun classes (such as the locative) that share the same prefix and agreement morphology. I also noted that the nominal infinitive prefix ho is capable of deriving nouns from verbs, like other noun class prefixes. However, unlike other noun derivations, the suffix of the derived noun is always a, which I directly linked to the verbal morphology observed with nominal infinitives. I also reviewed the treatment of the infinitive in the literature and showed that the various scholars agree that the infinitive is both a noun and a verb at the same time. I took this a step further by showing that nominal infinitives are more restricted as objects because they are

PAGE 112

112 selected by control verbs and these verbs impose their requirements on the types of complements th ey select. This is explored in C hapter 4 which I summarize below. I also used the Theta criterion, the Case filter, EPP and the PRO theorem to determine the subject of the infinitive in Sesotho. These principles were used to highlight the differences between the subjunctive and the infinitive while showing that control is only possible with the infinitive in Sesotho. In C hapter 3 I examined aspects of Sesotho morpho syntax operational in control phenomen a I paid attention to agreement (SM and OM) and argument structure changing verbal morphology (verbal extensions). I looked at the SM and OM in Sesotho sentence structure to set the stage for the argument that although Sesotho shows OM in the infinitive, the SM and OM are different in that agreement of the infinitive to the lack of Tense which is associated with finite clauses. I then moved on to look at verbal extensions. I focused mainly on the causative, the applicative and the reciprocal. I then extended the argument structure associated with these verbal extensions to the control patterns associated with control verb classes. I proposed that when the causative is used with control verbs it changes that part icular verb from a subject control verb class to an object control verb class. I also suggested that although the applicative is also associated with changing a subject control verb to an object control class, it performs a slightly different function f r om the applicative. For example, I show that the applicative is responsible for distinguishing between verbs that induce control shift, associated with the interpretation of PRO. The reciprocal gives a slightly different picture in that it is not used with the control verbs. However, it is useful in determining referential dependencies. I used the reciprocal as a diagnosis

PAGE 113

113 for partial and split control in Sesotho. Putting all these proposals together, this chapter addressed the questions of the role played by argument structure changing verbal morphology in control phenomen a as well as the question of the status of ho in Sesotho. In C hapter 4 I approached the classification of control verbs by looking at whether a correlation exists between certain verb clas ses and syntactic properties. I did this by mapping the syntactic properties onto verbal morphology. The main finding is that the variation in referential dependencies observed within verb classes can be directly accounted for through verbal morphology. I control and these are also associated with a lack of shift from subject to object. Object control verbs on the other hand are either causative, applicat ive or both. As we obs erved in C hapter 3 these are responsible for introducing a benefactive and patient role. It is therefore not surprising that these verbs are object control verbs. Another important distinction within object control verb classes is that the subdivisions ar e based on whether or not these verbs participate in control shift. This is a property also associated with verbal morphology. I closed Chapter 4 with an evaluation of verbs that are control verbs in other languages but not in Sesotho. I proposed that the associated with the lack of wh movement in Sesotho, a movement which is associated with these verbs in languages that allow wh movement. I also proposed that a close examination of these verbs r eveals that they select the subjunctive rather than the infinitive in Sesotho. This further supports my observation that certain control verbs prefer the subjunctive over the infinitive. In C hapter 5 I set out to establish a typology of control in Sesotho. I used this analysis of the types of control attested in Sesotho to further explore the differences between the nominal infinitive and the c lausal infinitive. As noted in C hapter 2 the nominal and clausal infinitive

PAGE 114

114 share the same properties with minor d ifferences. This comparison in C hapter 2 highlighted the similarities between the two but did not really explain why the two should be treated differently. Through an exploration of the referential dependencies as well as the morphological composition of c ontrol verbs, I illustrated that the distinction between the nominal and clausal infinitive rests with referential dependencies associated with PRO as the subject of the infinitive. For example, I showed that object control verbs that are applicative or ha ve a combination of causative and applicative give the interpretation of PRO ARB which is an equivalent of the nominal infinitive. Because PRO ARB has no c commanding referent, it can occur in subject position, but PRO that requires an antecedent is restri cted to object position. Another major contribution associated with control verb classes and the role played by verbal extensions is that these two directly determine types of control. Subject control is common with bare verbs, object control is associated with the causative and applicative, and the reciprocal is responsible for determining referential relation s but also responsible for determining potential control relations. Sesotho lacks partial control and split control because the morphology associated with these types of control is not compatible with Sesotho verbal morphology. Chapter 5 has directly addressed questions (1) and (3) above. 6.2 Implications for Control Relations Recent studies on control may be divided into three areas: Control may be vi ewed primarily as a syntactic phenomenon (Chomsky and Lasnik 1993, Hornstein 1999, Boeckx and Hornstein 2003 amongst others). Others view semantics as a central component in understanding control (Sag and Pollard 1991, Levin 1993 & 2007, Cul l icover and Jac kendoff 2001 & 2005 amongst others). There is yet another line of thought which supports the idea that a purely semantic or purely syntactic approach to obligatory control is not adequate, but rather a combination of morphological agreement and semantic te nse is better suited to account for cross

PAGE 115

115 linguistic typology (Landau 2003, 2006). Analyses of control in Bantu languages are in their infancy so it is difficult at this point to indicate which direction my findings point. What I would like to do at this p oint is show how the MDP, agreement specifically (OM) and verbal extensions work together to determine the reference of PRO. The MDP proposes that in obligatory control constructions the controller must c command PRO and that the controller must be the clo sest c commanding NP. One set of the verbs that is usually cited as a cause for concern for the MDP is the promise type set in which the subject of the main clause controls across the object. The example in (1) illustrates: (1) Samuel j promised Mary k [PRO j/* k to cook dinner ] Sentence (1) clearly indicates a violation of the MDP since the closest c commanding NP is Mary. Sesotho presents a slightly different problem. When tshepisa the object PRO is understood as having the same ref erence as the subject of the matrix clause. However, when the object is introduced the results are mixed. PRO can either refer to mme or bana in (2&3) This is indeed a problem for the MDP. However, we have already noted that Sesotho uses the verbal morph ology to account for concepts that can be dealt with syntax in other languages. (2) Mme j o tshepisa PRO j ho sebetsa 1.mother 1.SM promise INF work (3) Mme j o tshepisa bana k PRO ? j / ? k ho sebetsa 1.mother 1.SM promise 2.children INF work (4) Mme j o tshepisa bana k PRO* j /k h o mo se b e l ets a 1.mother 1.SM promise 2.children INF 1.OM work APPL FV

PAGE 116

116 (5) #Mme j o tshepisa bana k PR O j/* k ho ba sebel ets a 1.mother 1.SM promise 2.children INF 2.OM work APPL FV (6) Mme j o tshepisa PRO j ho ba sebeletsa bana. 1.mother 1.SM promise INF 2.SM work APPL FV 2.children The introduction of the applicative suffix on the verb o f the embedded clause as well as the object agreement (OM) determines which NP controls PRO. In (4), the OM refers to the recipient mme. This subject cannot be the recipient and agent of the same verb at the same. However, we would expect (5) to be accepta ble because the recipient is bana hence the OM agrees with this NP, the agent is mme. The sentence is odd under the reading that PRO is co referenced with mme. The reason for (5) being odd is that this verb allows subject control, then the clause contain ing PRO needs to get closer to the controlling NP, hence (6) is a perfect sentence in Sesotho. Two things come out of these data in (2 6). One is that the MDP seems adequate to account for control relations in Sesotho complement control. The other is that an approach which does not take into account the morphology associated with the main verb as well as the embedded one would pose a challenge for the MDP. 6.3 Concluding Remarks Most studies on control constructions have been based on the more widely studi ed languages. The approaches to control mentioned in the previous section are based on many years of research. As I hope to have shown throughout this study, Sesotho does not quite work well with the various tests for empty categories for example. In Sesot ho when you move the object NP the moved NP is replace d with the OM which keeps the connection more like leaving a trace. Sesotho control provides a window into languages that make use of verbal morphology to

PAGE 117

117 ology together with syntax principles such as the MDP to account for why a reference relation that requires a split between two antecedents (split control) is not possible We also noted that where the MDP would have been challenged the morphology becomes of assistance. I hope that this study will open up avenues for scholars of control regardless of their approaches to control constructions. There are two issues that I feel need to be studied. One concerns whether the control patterns observed in Sesotho are possible with other Bantu languages, or whether they are varied. If they are varied, are there ways of determining the variation by looking at the morphology? On a more cross linguistic level it would be interesting to see how the morphology associat ed with verbal extensions would be written into features that would help determine the referential dependencies of PRO. Finally, it would also be interesting to find out if all control verbs in Sesotho take a clause with a subject or whether there are diff erences as proposed by Wurmbrand 2001.

PAGE 118

118 LIST OF REFERENCES Adger, David. 2003. Core syntax: A Minimalist approach New York: Oxford University Press Linguistic Inquir y 34: 269 280 Boeckx, Cedrick, and Norbert, Hornstein. 2006. The virtues of control as movement. Syntax (9)2:118 130 Bresnan, Joan. 1972. The theory of complementation in English Syntax. Doctoral dissertation, MIT, Cambridge, Mass Bresnan, Joan. 1982. Control and Complementation. Linguistic Inquiry 13:343 43 Bresnan, Joan and Sam Mchombo. 1986. Grammatical and anaphoric agreement.In CLS 22. Vol.2 (278 297). Papers from the Parasession on Pragmatics and Grammatical Theory. ed. by Anne, Farley, Peter, T. Farley and Karl, Mc Cullough. Chicago Lnguistic Society. University of Chicago chains. Syntax 2:210 226 Brody, Michael, and M. Rita Manzi ni. 1987. Implicit argumen ts. I n Mental Representations: The interface between language and realit y ed. by Ruth Kempson, 105 130, Cambridge: Cambridge University Press Buell, Leston.2007. Semantic and formal locatives: Implications for the Bantu locative inversion typology. SOAS Working Papers in Linguistics 15:105 120. Cecchetto, Carlo and Oniga Renato. 2004. A Challenge to Null Case Theory. Linguistic Inquiry 35 (1):141 149 Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, MIT Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris Chomsky, Noam. 1982 Language and the Study of Mind Tokyo: Sansyusya Publishing Co. Ltd. Chomsky, Noam. 1986a. Barriers Cambridge, Mass.: MIT Press Chomsky, Noam. 1986b. Knowledge of Language New York: Prae ger Chomsky, Noam. 1995. The Minimalist Program. The MIT Press. Cambridge, Massachusetts, London, England Chomsky, Noam. 2000. Minimalist inquiries: the framework. In Step by Step ed. Roger Martin and David Michaels and Juan Uriagereka, 89 155. Cambridg e, Mass.: MIT Press

PAGE 119

119 Chomsky, Noam and Howard Lasnik.1995. The theory of principles and parameters. In The Minimalist Program by Chomsky N The MIT Press. Cambridge, Massachusetts, London, England Chomsky, Noam, and Howard Lasnik. 1977. Filters and contr ol. Linguistic Inquiry. 8:425 504 Comrie .1984. Language Universals and Linguistic Typology. University of Chicago Press Creissels, Denis. 1996. Conjunctive and disjunctive verb forms in Setswana. South African Journal of African Languages; Vol.16(4):109 116 Creissels, Denis. 2005. Tswana verb morphology and the Lexical Integrity Principle. Paper presented at Fifth Mediterr anean Morphology Meeting, Freju Creissels, Denis and Daniele Godard. 2005. The Tswana infinitive as a mixed category. Proceedings o f the HPSG05 Conference, University of Lisbon, CSLI Publications Culicover, Peter W., and Ray Jackendoff. 2001. Control is not movement. Linguistic Inquiry 32 :49 3 512 Culicover, Peter W., and Ray Jackendoff. 2003. The semantic basis of control. Language 79.3:517 556 Culicover, Peter W., and Ray Jackendoff. 2005. Simpler Syntax. Oxford: Oxford University Press Culicover, Peter W., and Ray Jackendoff. 2006. Turn over control to the semantics. Syntax 9(2): 131 152 Culicover, Peter W., and Wendy Wilkins. 19 84. Locality and linguistic theory. New York: Academic Press Culicover, Peter W., and Wendy Wilkins. 1986. Control, PRO, and the Projection Principle. Language 62:120 153 Davies, William D. and Stanley Dubinsky. 2004. The Grammar of Raising and Control. A Course in Syntactic Argumentation. Blackwell Publishing Davies, William D. and Stanley Dubinsky (Guest eds.). 2006. The place, range, and taxonomy of control and raising. Syntax 9:2 : 111 117 Demuth, K atherine 1995. Questions, relatives, and minimal p rojection. Language Acquisition 4 : 49 71 Demuth, Katherine. 1990. Locatives, impersonals and expletives in Sesotho. The Linguistic Review 7: 233 249. Demut h, Katherine and Sheila Mmusi. 1997. Presentational focus and thematic structure in comparative Bantu Journal of African Languages and Linguistics 18:1 19.

PAGE 120

120 Doke, C lement, M. and M achabe Mofokeng. 1957. Textbook of Southern Sotho Grammar, London: Longman, Green and Company Du Plessis, J.A. 1982b. The analysis of the infinitive. South African Journal of A frican Languages. 2(2). Du Plessis J. and Marianna Visser .19 9 2. Xhosa Syntax. Pretoria: Via Afrika Farkas, Donca, F.1988. On obligatory control. Linguistics and Philosophy 11: 27 58 Gowlett, Derek. 2005. Zone S. In The Bantu Languages by Derek Nurse, G rard Philippson (eds.) Routledge Guma, S amson, M bizo 1971. An Outline Structure of Southern Sotho. Shu ter & Shooter. Pietermaritzburg Haegeman, Liliane. 1991. Introduction to Government and Binding theory Oxford: Basil Blackwell Heine, Bernd and Derek Nurse. 2000. African Languages: An Introduction Cambridge University Press Henderson, Brent. 200 7 Multiple agreement and inversion in Bantu. Syntax 9 (3) : 275 289 Hornstein, Norbert. 1995. Logical Form: From GB to Minimalism. Oxford: Blackwell Hornst ein, Norbert. 1997. Is the binding theory necessary ? Ms., Unive rsity of Maryland, College Park Hornstein, Norbert. 1998. Movement and chains. Syntax 1:99 127 Hornstein, Norbert. 1999. Movement and control. Linguistic Inquiry 30:69 96 Hornstein, Norbert 2000. On A chains: A reply to Brody. Syntax 3: 129 143 Hornstein, Norbert.2003. On Control. In Minimalist Syntax ed. by Hendrick, Randall. Malden, MA: Blackwell, pp 6 81 Hvitfeldt Robert D. 1982. The Major Syntactic Structure of Xhosa. Doctoral disse rtation, University of Wisconsin Madison Idiata, Daniel, F. 2005. What Bantu child Language tells about Noun classes. : LINCOM Studies in African Linguistics 67 Jackendoff, Ray and Peter W. Cullicover. 2003. The semantic basis of control in English. Lan guage 79:517 556

PAGE 121

121 Jacobson, Pauline. 1992 Raising without movement. In Control and grammar ed. by Richard Larson, Sabine Iatridou, Utpal Lahiri, and James Higginbot ham, 149 194. Dordrecht: Kluwer Katamba, Francis. 1993. Morphology New York Khoali, Benjamin Thakampholo.1991 A Sesotho tonal grammar Doctoral dissertation, University of Illinois at Urbana Champaign Koopman, H.D and D. Sportiche.1991. The position of subjects. Lingua, 85,211 259 Koster, Jan, and Robert May. 1982. O n the constituency of infinitives. Language 58:117 143. Koster, Jan. 1984. On binding and control. Linguistic Inquiry 15:417 459 Kratzer, Angelika. 1996. Severing the e xternal a rgument from its v erb. In Johan Rooryck and Laurie Ann Zaring (eds.). Phrase Structure and the Lexicon 109 137. Dordrecht: Kluwer Academic Publishers Landau, Idan. 2000. Elements of Control: Structure and meaning in infinitival Construction. Dordrecht: Kluwer Landau, Idan. 2001. Control and extraposition: The case of super equ i. Natural Language and Linguistic Theory 19:109 152 Landau, Idan. 2004. The scale of finiteness and the calculus of control. Natural Language and Linguistic Theory 22:4 Landau, Idan.2006. Severing the distribution of PRO from Case. Syntax. 9(2):153 170 Lasnik, Howard. 1995b. Verbal morphology: Syntactic Structures meets the Minimalist Program, Ms., Un iversity of Connecticut, Storrs Lasnik, Howard. 1999. Minimalist Analysis. Blackwell Publishers Larson, Richard. 1991. Promise and the theory of contro l. Linguistic Inquiry 22:103 139 Larson, Richard, Sabine Iatridou, Utpal Lahiri, and James Higginbotham (eds.). 1992. Control and grammar. Dordrecht: Kluwer Levin, Beth. 1993. English Verb Classes and Alterations. A Preliminary Investigation. The Universi ty of Chicago Press. Chicago and London. Levin, Beth. 2009. Verb Typologies Revisite Handout: Ghent, February 5 7, 2009 Lestrade,G.P.1938. Locative Nouns and Formations in Sotho Bantu Studies. Vol.XII (1) Machobane, 'Malillo 'Matsepo. 1995. The Sesotho Locative Constructions. Journal of African Languages and Linguistics 16(2):115 136

PAGE 122

122 Machobane, 'Malillo 'Matsepo. 1996. The Sesotho Locative Alternation Verbs. South African Journal of African Languages 16(1):8 15 Machobane, 'Malillo 'Matsepo. 1997. Seso tho control verbs with applicative suffix. South African Journal of Linguistics 15(2):59 64 Machobane, Malillo Matsepo. 2001. Questions in Sesotho. Afrikanistische Arbeitspapiere 66:67 83 Manzini, M. Rita. 1983. On control and control theory. Lingu istic Inquiry 14:421 446 Manzini, M. Rita, and Anna Roussou. 2000. A minimalist theory of A movement and control. Lingua 110: 409 447 Marantz 1984. On the Nature of Grammatical relations MIT Press, Cambridge Mass Mchombo, Sam. 1993. Theoretical aspe cts of Bantu grammar Stanford, California: CSLI Mchombo, Sam .2004. The Syntax of Chichewa. Cambridge Syntax Guides Cambridge University Press Mc Laughlin, Fiona. 1997. Noun classification in Wolof: When a ffixes are not renewed. Studies in African Ling uistics. Vol.26(1) Mc Laughlin, Fiona. 2000. Consonant mutation and reduplication in S e ereer Siin. Phonology. 17:333 363 Guthrie 1967 Guthrie, M. 1967/71. Comparative Bantu vol. 1 4. Gregg International Publishers Miller, D. Gary. 2002. Non finite stru ctures in theory and change. Oxford University Press Miller, D. Gary. 2006. Latin Suffixal Derivatives in English and their Indo European Ancestry Oxford University Press Mitchell Erica 1994. When Agro is fused to Agrs: What morphology can tell us abou t functional categories. MIT Working Papers in Linguistics. 111 130 Mkwanazi M.J. 1989. Kiriatshwana Southern Sotho Poetry. Educum Publishers Moeketsi R.H. (ed.) .1996. Seyalemoya An anthology of dramas written in Sesotho. Van Schaik Pretor ia Mofolo, T homas. 1981. Chaka Via Afrika Mohanan, K.P. 1985. Remarks on control and control theory. Linguistic Inquiry 16: 637 648 Morimoto, Yukiko 2006. Agreement properti es and word order in comparative Bantu ZAS Papers in L inguistics 43 : 161 188

PAGE 123

123 Mugane, John. 1998. Gikuyu Morpho syntax. In Trends in African Linguistics (2). Language History and Linguistic Description in Africa, ed. by Ian Maddies on & Thomas Hinnebusch. 239 248 Mugane, John. 2003. Infinitive gerunds as extrasequential hybrid constructions in Logoli. In Trends in African Linguistics (5). Linguistic Typology and Representation of African Languages. Africa World Press, INC. Noonan, Michael.1985. Complement ation. In Timothy Shopen (ed.) Language typology and syntactic description. vol 2: Complex constructions. Cambridge University Press, 42 140 Nurse, Derek and Gerard Philippson. 2003. The Bantu Languages. Routledge Nurse, Derek. 2006. Focus in Bantu: Ver bal morphology and function, ZAS Papers in Linguistics 43:189 207 Petter, Marga. 1998. Getting PRO under control, The Hague: Holland Academic Graphics Perez, C.H. 1986. Aspects of complementation in three Bantu languages. Doctoral dissertation, Univer sity of Wisconsin Madison Polinsky, Maria, and Eric Potsdam. 2002. Backward control. Linguistic Inquiry 17:501 557 Polinsky, Maria, and Eric Potsdam. 2006. Expanding the Scope of Control and Raising. Syntax 9(2):171 192 Rei n hart, Tanya, and Eric Reul and. 1993. Reflexivity. Linguistic Inquiry 24:657 720 Rizzi, Luigi. 1986a. Null objects in Italian and theory of pro. Linguistic Inquiry 17(3):501 557 Rizzi, Luig i. 1986b. On chain formation. I n The syntax of pronominal clitics. ed. by Hagit Borer, 65 95. New York: Academic Press Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press Roeper, Thomas. 1987. Implicit arguments and the head complement relation. Linguistic Inquiry 18: 267 310 Rosenbaum, Peter .1967 The grammar of Engl ish predicate complement constructions. Cambridge, Ma.: MIT Press Rosenbaum, Peter .1970 A principle governing deletion in English sentential complementation. In R. Jacobs & P. Rosenbaum (Eds.), Readings in English transformational grammar (pp. 20 29). W altham, MA: Ginn Rugege, G.1981. The infinitive in Kinyarwanda. Studies in African Linguistics. Supplement 8. Sag, Ivan, and Carl Pollard. 1991. An integrated theory of complement control. Language 67:63 113

PAGE 124

124 Satyo, Sizwe. 1985. Topics in Xhosa Verbal Extensions. Doctoral dissertation, Unive rsity of South Africa, Pretoria Scida, Emily. 2004. The Inflected Infinitive in Romance Languages. Routledge. New York and London Sean Madigan 2008. Obligatory s plit c ontrol into e xhortative c omplements in Korean Linguistic Inquiry 39 (3): 493 502 Sigursson, Halladr A. 1991. Icelandic Case marked PRO and the licensing of lexical argume nts. Natural Language & Linguistic Theory 9:327 363 Simpson, Jane and Joan Bresnan .1983. Control and Obviation in Walpiri. Natural Language & Linguistic Theory 1( 1): 49 64 Speas Margare t 1991. Functional heads and inflectional morphemes. The Lingu istic Review. 8:389 417 Stiebels, Barbara, Szymon Slodowic, Yi Chun Yang, 2003. Questionnaire for control Verbs. 2 nd Version, Berlin Stiebels, Barbara.2007 Towards a typology of complement control. In Barbara Stiebels (ed.), ZAS Papers in Linguistics 47 Studies in complement control 1 80. ZAS Berlin Terzi, Arhonto. 1997. PRO and null case in finite clauses. Linguistic Inquiry 14 :335 360 Van Der Spuy, Andrew.1997. Syntactic traces in Zulu. In African Linguistics at the Crossroads:Papers from Kwaluseni Ed. By R.K. Herbert: 297 309 Visser, Mariana. 1989. The syntax of the infinitive in Xhosa. South African Journal of African Languages: 9(4):154 185 Vic Webb and Kembo Sure (eds.) 2000. African Voices. An Introduction to the Languages and Linguistics o f Africa. Oxford University Press. Cape Town, South Africa Williams, Edwin. 1985. PRO and subject NP. Natural Language & Linguistic Theory. 3: 297 315 Williams, Edwin. 1987. Implicit arguments, the binding and control. Natural Language & Linguistic The ory. 5: 151 180 Wurmbrand, Susi. 1998. Infinitives, Doctoral dissertation, MIT, Cambridge, Mass. Wurmbrand, Susi. 2001. Infinitives: Restructuring and Clause Structure Mouton de Gruyter, Berlin.

PAGE 125

BIOGRAPHICAL SKETCH Mantoa Rose Smouse was born in Bloem fontein, South Africa. She is the second youngest of six children. She grew up in Zastron, a small town on the borders of Lesotho, the Free State and the Eastern Cape Province. She attended elementary school on a farm school near Zastron, Diepfontein She then moved on to do her middle school in the Eastern Cape then called Transkei. She completed her High School in 1989 at Lere La Thuto in Zastron. Mantoa earned the University of Cap e Town in 1994, 1995 and 1998 respectively Florida in 2004 Cape Town. Whilst teaching at Rustenburg she enrolled to do her g raduate studies at the University of Cape Town. In 2000 she was awarded a Prestigious A.W. Mellon Scholarship which afforded her an opportunity to pursue graduate studies at a university in the United States. Mantoa is currently teaching at Universi ty of C ape Town in South Africa. She has been married to Simon Smouse for 18 years. They are blessed with three children: Naledi, age 17, Tlhodiso, age 13 and Mapule age 4.