A DESCRIPTIVE NOTE ON MALAGASY VERBAL COMPLEMENTATION AND THE BINDING HIERARCHY: With Special Reference to the Occurrence of the Complementizer fa

The aim of this paper is to describe verbal complementation in Malagasy and to consider how the Malagasy data reflect the "binding hierarchy" proposed by Givon (1980). It is shown that the Malagasy data provide support for the hierarchy and that the occurrence of the complementizer "fa" can be accounted for in terms of the strength of binding the main-clause verb. (Author/JL) ********************************************************************** Reproductions supplied by EDRS are the best that can be made from the original dolument. * *********************************************************************** A DESCRIPTIVE NOTE ON MALAGASY VERBAL COMPLEMENTATION AND THE BINDING HIERARCHY: With Special Reference to the Occurrence of the Complementizer fa Masuhiro Nomura University of California, San Diego Abstract: The aim of the present paper is to describe verbal complementation in Malagasy and to consider how the Malagasy data reflect the "binding hierarchy" proposed by Givon (1980). It will be shown that the Malagasy data provide support for the hierarchy and that the occurrence of the complementizer fa can be accounted for in terms of the strength of binding of the main-clause verb.

Mahafantatra aho fa handeha any NY izy.pres-know lsg(nom) that fm-go to NY 3sg(nom) 'I know that s/he will go to NY' b.
Name la azy handeha any NY aho, fa tsy nandeha izy past-allow 3sg(acc) fut-go to NY I, but NEG past-go 3sg(nom) 'I allowed him to go to NY, but he didn't' c.
Tsy nandeha aho fa avy ny orana.NEG past-go 1 sg(nom) because fall the rain 'I didn't go because it rained' In (4a), fa is used as a complementizer, introducing the subordinate clause of the verb mahafantatra 'know '. In (4b), fa is used as a conjunction 'but '. In (4c), fa is used to mean 'because'.
The aim of the present paper is to describe verbal complementation in Malagasy and to consider how Malagasy complementation reflects the "binding hierarchy" proposed by Givdn (1980,1990).Special attention is given to the consideration of the semantic conditions that govern the appearance offa as a complementizer.The organization of the paper is as follows: §2 describes three types of complement clauses in Malagasy and discusses what kinds of verbs can take each complement type.§3 shows how the three types of complement clauses reflect the binding hierarchy and how the occurrence offa may be accounted for.§4 comprises the conclusion.

Three Types of Verbal Complement Structures
This section describes three types of Malagasy verbal complementation and considers the syntactic/semantic characteristics of each type.
Type A Consider the following pair: (5) a.
Nilaza i Koto fa Oa an'i Soa izy past-say Koto that like ACC-Soa 3sg(nom) 'Koto said that he likes/liked Soli (5a) and (5b) represent direct speech and indirect speech, respectively.In (5b), Nilaza is the main-clause verb and i Koto is the main-clause subject.The fa-clause is the complement clause selected by the main-clause verb ni/aza.The sentence has the word order VSO, where 0 is a complement clause headed by fa. 4 I will call this complementation structure Type A. Let us take a look at other verbs that can take a Type A complement: (6) Mino aho fa handeha any NY izy.pres-believe lsg(nom) COMP fut-go to NY 3sg(nom) 'I believe that s/he will go to NY' (7) Mahafantatra aho fa handeha any NY izy.pres-know lsg(nom) COMP fm-go to NY 3sg(nom) 'I know that s/he will go to NY' (8) faly aho fa handeha any NY izy happy lsg(nom) COMP fm-go to NY 3sg(nom) 'I am happy that s/he will go to NY' (9) Nandre aho fa handeha any NY izy.

BEST COPY AVAILABLE
past-hear lsg(nom) COMP fut-eo to NY lsg(nom) 'I heard that s/he is going to NY' (10) Nahita aho fa namitaka ahy izy past-see lsg(nom) COMP past-cheat lsg(acc) 3sg(nom) I found out that s/he cheated me' (11) Nanapakevitra aho fa handeha any NY izy past-decide lsg(nom) COMP fut-go to NY 3sg(nom) 'I decided that s/he should go to NY' (12) Manaiky isika fa mahay mihira tsara indrindra izy agree 1p1(incl,nom) COMP able to sing good most 3sg(nom) 'We agree that s/he is the best singer' (13) Manantena i Koto fa hahita an'i Soa aho hope Koto COMP fut-see ACC-Soa lsg(nom) 'Koto hopes that I will see Soa ( 14) Mahatadidy aho fa nandeha any NY izy.remember lsg(nom) COMP past-go to NY 3sg(nom) 'I remember that s/he went to NY' (15) Nanadino aho fa nandeha any NY izy.past-forget lsg(nom) COMP past-go to NY 3sg(nom) 'I forgot that s/he went to NY'   In the above examples, fa is obligatorily present and cannot be omitted.The verb tia ('like'), however, can occur with or without fu : 5 ( 16) Tiako (fa) hianatra teny angilisy izy like-lsg(gen) (COMP) fut-study English 3sg(nom) '(Lit.)I like that s/he will study English' Examples ( 6)-( 16) suggest that the Type A complement clause is just like an independent/main clause in terms of VOS word-order.The complement-clause verb is tensed and tense agreement does not generally exist between the main-clause verb and the complement-clause verb: 6 (17) Mahafantatra aho fa Imandeha/nandehdhandehal any NY i Koto.pres-know lsg(nom) COMP pres/past/fut-go) to NY Koto 'I know that Koto (goes/went/will go} to NY' There are cases, however, where the Type A complement clause becomes less like an independent/main clause.We have so far seen examples where the main-clause subject and the complement-clause subject are non-coreferential; when the two are coreferential, the complement- clause subject may not overtly be expressed.'Compare, for example, ( 12) and ( 13) with ( 18) and ( 19), respectively: (18) Nanaiky isika fa hanomboka aloha (`` isika)  past-agree 1p1(incl,noin) COMP fut-start early 'We agreed that we should start early' (19) Manantena aho fa hahita an'i Koto ('aho) pres-hope lsg(nom) COMP fut-see Koto 'I hope that I will see Koto' To sum up the properties of Type A complementation:

BEST COPY AVAILABLE
(20) a.
The sentence has the VSO word order, with 0 being a complement clause headed by fa. b.
The complement clause has the same word order as the basic ma;.n clause (i.e., VOS). c.
The complement-clause subject is unrestricted in terms of coreference with the.main-clause subject.The complement-clause subject may not be expressed when it is coreferential with the main-clause subject. d.
The complementizerfa is obligatory except with the verb tia e.
The complement clause is tensed.No intrinsic tense agreement exists between the main-clause verb and the complement-clause verb.

Type B
It was observed in ( 18) and ( 19) that the complement-clause subject may not be expressed when it is coreferential with the main-clause subject.There is another way of coding the situation where the complement-clause subject/agent is coreferential with the main-clause subject/agent.Compare the following pair: (21) a.
Unlike Type A complementation, there exists tense agreement between the main-clause verb and the complement-clause verb; for example, it is reported by the consultant that the main- clause verb tense and the complement-clause verb mnse have to agree for verbs such as 'try', 'begin' and 'stop'.I I To sum up the syntactic characteristics of Type B complementation: (32) a.
The sentence has the VOS word order, where 0 is a complement clause b.
fa is optionally present for some verbs but not for others.c.
The complement clause lacks an overt subject.d.
The complement clause is tensed, and tense agreement exists between the mainclause verb and the complement-clause verb.e.
The complement-clause subject/agent is interpreted as coreferential with the mainclause subject/agent.
Type C Some verbs allow the following two types of complementation: Mino aho fa handeha any NY izy. (=( 6)) pres-believe lsg(nom) COMP fm-go to NY 3sg(nom) 'I believe that he will go to NY' b.
Namela azy {handeha/nandeha} any NY aho past-allow 3sg(acc) fut-go/past-go} to NY Isg(nom) 'I allowed him to go to NY/I let him go to NY' Note that fa is impossible in (49a) and (50a).Also, it seems to be the case that only nonactive counterparts can take a Type A complement, while only active counterparts can take a Type C complement.Compare ( 49) and ( 50) with the following: (51) a.
*Navelako azy handeha any NY past-allow-lsg(gen) 3sg(acc) fut-go to NY As the above examples show, the complement-clause verb of Type C complementation is tensed; however,.tenseagreement is observed between the main-clause verb and the complementclause verb.Let us see the following as an example: (53) Nanery azy {handehaPnandeha } any NY aho, fa tsy nandeha izy past-force 3sg(acc) { fut-go/*past-go } to NY Isg(nom), but NEG past-go 3sg(nom) 'I forced him to go, but he didn't' The past tense of the complement-clause verb implies that the causee actually went to NY, making it sound contradictory to add "he didn't go".The future tense, on the other hand, -does not have such an implication; therefore, it is not contradictory to add "he didn't go".Thus the complementclause verb tense serves to distinguish implicativity.
When the main-clause causative verb is in the future tense, it logically follows that only non -implicative future tense can appear in the complement-clause verb.This proves correct as the following example indicates: (54) Hanery azy { handehai*mandehaPnandeha} any NY aho fut-make him {fut-goPpresent-goPpast-go} to NY lsg(nom) 'I will make him go to NY' To sum up the characteristics of Type C complementation: (55) a.
The sentence has the word order VOS, with 0 being a complement clause.b.
The complement-clause agent is expressed in the accusative case.c.
The complement verb is tensed and tense agreement is observed.d.
fa cannot occur in this complementation. 3.
Malagasy Complementation and Givon's Binding Hierarchy Givon (1980Givon ( , 1990) ) proposes the concept of "binding" to capture systematic and iconic correlations between the semantics of complement-taking verbs and the syntactic structure of their complements.17 "Binding" is defined as "The stronger the influence exerted over the agent of the complement clause by the agent of the main-clause verb, by whatever means, the higher is the main-clause verb on the binding.scale" (Givon 1980:335).The basic claim is summarized as follows: (56) The higher a verb is on the binding scale, the less would its complement tend to be syntactically coded as an independent/main clause.(Givon 1980:337) Givon (1980,1990) classifies complement-taking verbs into three major classes as in (57) and demonstrates, with cross-linguistic evidence, that manipulative verbs and modality verbs occupy higher positions on the binding scale than cognition-utterance verbs: 18 (57) (i)   Cognition-utterance verbs ('say', 'think', 'know', etc.) (ii) Manipulative verbs ('order'.'cause', 'tell', 'force', etc.) (iii) Modality verbs ('want'.'succeed', 'begin', 'intend', 'try', etc.)The binding of the verb correlates negatively with the degree to which its complement appears syntactically similar to a main clause.More specifically, (60) The higher a verb is on the binding scale, (i) the less is the agent in its complement/embedded clause likely to exhibit the casemarking characteristic of main-clause subjects/agents/topics (ii) the less is the.verb of its complement clause likely to exhibit the tense-aspectmodality markings characteristic of main clauses (iii) the more is the verb in its complement clause likely to be predicate-raised, i.e. lexicalized as one word with the main verb (iv) the less the main clause and the subordinate clause are likely to be separated by a subordinator (or a physical pause) (GivOn 1980(GivOn :338, 1990:560-561) :560-561) It is now rather obvious that Malagasy verbs that take Type A, Type B, and Type C complementation described in §2 may correspond to cognition-utterance verbs, modality verbs, and manipulative verbs, respectively.This section aims to examine how Malagasy verbal complementation described in §2 reflects the syntactic ramifications ( 60).

The overall picture
The following is a summary chart that shows which Malagasy verb can take which complementation type.In the columns of Type A and Type B, it is indicated whether the occurrence of fa is obligatory.optional.or prohibited (indicated by 4, ('1), and *, respectively):19 10 BEST COPY AVAILABLE Using the above chart, let us consider what kinds of verbs take which types of complement clause.First we can observe that cognition-utterance verbs and non-active forms of manipulative verbs can take Type A com^lements, but modality verbs can't.
As for Type B complement, modality verbs typically take it, but manipulative verbs can't.It is well expected that higher cognition-utterance verbs on the binding scale (cf. ( 59) may take a Type B complement, but what is peculiar is that in addition to higher-positioned verbs such as 'hope', and 'want', verbs as low on the scale as 'say', 'believe', 'know' can take it: (62) Milaza fa handeha izy pres-say COMP fut-go 3sg(nom) 'S/he says that s/he will go' (63) Mino (fa) hianatra teny angilisy alio pres-believe (COMP) fut-study English lsg(nom) w/ fa: 'It is likely that I will study English' w/o fa: 'I believe that I will study English' 1 (64) Mahafantatra fa handeha any NY aho.pres-know COMP fut-go to NY I sg(nom) 'I know that I will go to NY' As for Type C complement, active forms of manipulative verbs can take it, but modality verbs can't.Again, higher cognition-utterance verbs on the binding scale are expected to take a Type C complement (and it is indeed the case with verbs like 'hope' and 'want), but verbs as low on the hierarchy as 'say' and 'know' can take it as well.
In sum, cognition-utterance verbs in Malagasy are peculiar in that they can participate in complementation patterns that stronger-binding verbs (manipulative/modality verbs) typically take.Except for this, we could say that there is a correlation between cognition-utterance verbs and Type A complement, modality verbs and Type B complement, and manipulative verbs and Type C complement.
Syntactic dimensions: case-marking.verb-form and co-lexicalization Let us next consider how (60 i-iii ) are reflected in the Malagasy complementation patterns.As for (60 i), it was observed in §2 that a complement-clause agent is expressed just like a mainclause subject in Type A complements, whereas it is not overtly expressed in Type B complements, and it is expressed as a main-clause object in Type C complements.
As for (60 ii), it was shown in §2 that no tense agreement generally exists between the main-clause verb and the complement-clause verb in Type A complementation, but there is tense agreement of some sort in Type B complementation and Type C complementation.te-hanaiky hanasa ny zaza Rasoa fa tsy afaka (Keenan 1976:279) want-agree wash the child Rasoa but not free 'Rasoa wants to agree to wash the child but (she) isn't free (to do so)' b.
It can be concluded from the above that syntactic ramifications (60 i-iii) are observed in the three cornplementation patterns of Malagasy.obligatorily in Type A complementation.except for the verb tia ('like', 'want'), which takes fa optionally.Notice that the verb 'want' is cross-linguistically situated higher on the binding scale than more typical cognition-utterance verbs such as 'say', 'believe', and 'know' (cf. (59)).
Furthermore, manipulative verbs when they take a Type A complement cannot have fa.Thus, we may generalize for Type A complementation that the higher a verb is on the binding scale, the less likely it is to take fa.
Let us next consider the occurrence of fa in Type B complementation.Chart (61) indicates that fa is impossible for modality verbs and stronger-binding cognition-utterance verbs such as mahatadidy ('remember'), manadino ('forget'), and tia ('want').For other cognition-utterance verbs, fa is either obligatory or optional in Type B complementation.This again supports the generalization that the higher a verb is on the binding scale, the less likely it is to take fa.As for Type C complementation, it is impossible to have fa, regardless of the verb class.
To conclude, we may propose the following generalization: (67) The higher a verb is on the binding scale.the less likely it is to take fa.
This observation shows that Malagasy complementation reflects (60 iv), namely, the higher a main-clause verb is on the binding.scale.the less the main clause and the subordinate clause are likely to be separated by a subordinator.

Conclusion
The aim of this paper has been two-fold: description of Malagasy verbal complementation and examination of the binding hierarchy with respect to the Malagasy data.For the former, this paper has classified Malagasy verbal complementation into three types, described structural/semantic properties of each type.and provided a list of verbs that take each complementation type.For the latter, it has been shown that Malagasy verbal complementation generally supports Givon's claims about syntactic reflections of the binding hierarchy, and that the occurrence of fa can be accounted for in icrius of the strength of binding of the main-clause verb.

NOTES
This is a revised version of my course paper for Field Methods instructed by Professor Suzanne Kemmer in Spring 1993 at University of California, San Diego.I am greatly indebted to Aurelien Rajoharison for patiently acting as consultant.I am also grateful to Professor Suzanne Kemmer for her comments on the course paper.to all the students who attended the course and Ron Sheffer for their help, and to a reviewer of KWPL for comments.Any remaining errors and inadequacies are, of course, my responsibility alone.
See Dyen (1971) for a succinct overview of the language.

2
The following abbreviations are used in this paper:  Richardson (1885:144) gives 'for', 'but', 'therefore', 'because', 'that' as the meanings of fa.Keenan (1976:274) notes that "We note that fa above is a very general sentential connective in Malagasy.It is often used with contrastive effect, like but in English, but also serves to introduce sentential complements of verbs of thinking, saying, etc., and as well 'serves as a largely contentless discourse connective." 4 See Keenan (1976:276-277) for his arguments for positing an underlying VOS order for these sentences.

5
There is a significant semantic difference between the sentence with fa and the one without fa: the sentence with fa means "I iike the fact that he is going to study English/I am happy that he is going to study English", whereas the sentence without fa means "I wish he would study English".The differenCe seems to concern factivity of the complement.6 I use the term "tense agreement" in the sense that the complement-clause verb tense is dependent in one way or another on the main-clause verb tense.cf.Givon (1990:531).Keenan (1976) calls this rule Equi-1 and notes that "It is not fully clear whether Equi-1 is obligatory"; see Keenan (1976:276 -27S) for more details.This construction corresponds to Keenan's (1976) Equi 2 and to Randriamasimanana's (1986) Equi 1. 9 Keenan (1976:278) and Randriamasimanana (1986:501) state that no complementizer can be present in Type B complement ("Equi-2" construction in Keenan and "Equi-1" construction in Randriamasimanana), but it seems to be the case that, at least for some verbs, the complementizer fa can optionally be present in Type B complements.This will be discussed more later below.lo The complementation of the verb tia ('like', 'want') is unlike other verbs in that the mainclause verb tense is expressed by the complement-clause verb: the replacement of miteny by niteny (past-speak) and hitcny (fut-speak) would make the sentence mean "I wanted to speak French" and "I will want to speak French", respectively.Randriamasimanana (1986:500) states that future tense marker is mandatory for Type B construction (="Equi 1" in his term), but it appears to be the case that future tense is not always mandatory.
fOr cognition-utterance verbs.Qv& (1980:345)  postulates the following scale and cites cross-linguistic evidence that the higher cognition-utterance verbs on the scale tend to develop either modality or manipulative senses:BEST COPYAVAILABLE

20
Let us next turn to (60 iii).As far as our data are concerned, there is no co-lexicalization of the complement-clause verb and the main-clause verb in either Type A, Type B, or Type C complementation.Since Type B complementation, however, has the main-clause verb and the complement-clause verb in sequence.it is conceivable that Type B complementation may develop co-lexicalization.The following examples max' be considered examples of co-lexicalization: complementizer fir Lastly, let us consider how (60 iv) is reflected in Malagasy complementation.It was observed in §2 and summarized in Chart (61) that cognition-utterance verbs generally take fa BEST COPY AVAIL ABLE .112 8