The Causative Construction in Miskitu

Ken Hale

Introduction

Miskitu, a language of the Misumalpan family, is spoken on the Atlantic Coast of

Nicaragua and Honduras. Although its lexicon and basic grammatical properties
have been documented to some extent (e.g., Heath and Marx (1961); Salamanca
and Scott (1985); Salamanca (1988)), detailed study of the syntax of Miskitu is
in its infancy. The present study, necessarily descriptive in nature, has the
modest purpose of presenting certain facts concerning the Miskitu causative, a
construction which displays certain interesting, superficially paradoxical,
properties by comparison with the causative of more amply documented
languages.

The sentences in (1) below exemplify the Miskitu causative construction, and
they illustrate the property of the Miskitu causative which has led us to use the
term "insubordination" in reference to it (in Aviles et al. (1987)):

(l)a. Yang mita tuktan ba yab-rika kauhw-bia

I AG child DEF cause-FC fall-FUT
'I will make the child fall.'
b. Yang (taim bani) tuktan ba yab-ri kauhwi-sa.
I (time all) child DEF cause-NFC falling-PRES
'I (always) make the child fall.'
c. Yang mita tuktan ba yab-ri kauhw-an.
I AG child DEF cause-NFC fall-PAST
'I made the child fall.'

The "insubordinate" character of this construction, and the interest which we

believe that it has for linguistic theory, can be brought out best by first
considering certain relevant aspects of the English sentences used to translate
the Miskitu -- these translations are the nearest semantic and functional English
190

equivalents to the corresponding Miskitu and, we contend, the Miskitu and

English constructions are linguistically comparable, albeit different in ways which
will become clear momentarily.

The construction we are examining belongs to a set of structures appropriately

termed "causative", involving not only the verb yab-aia (cause-INF) 'to cause,
give' (with suppletive object-incorporating partners ai-k-aia, mai-k-aia, wan-k-aia
'to cause me, you, us (incl)), but also the approximately synonymous mun-aia 'to
cause', and the verb swi-aia 'to let, allow, leave'. These correspond rather
closely to the English verbs make and let, and to their counterparts in any
number of languages (e.g., Spanish hacer, dejar, French faire, laisser, Dutch
laten, etc.), though interesting and important parametric differences between the
languages abound, to be sure.

In the English construction, as represented in (1) above, the "causative verb"

(i.e., make or let) -- or more accurately, the auxiliary of that verb -- bears the
morphology which realizes the tense of the "causative event" as a whole. The
"effect" verb (i.e., fall, in (1)) is uninflected, appearing in what is known as the
"bare infinitive" (where the causative verb is make, let, at least). Details of
morphology aside, the same would be true, in essence, in the corresponding
sentences of Spanish, French, and other Inda-European languages. It is generally
held for these languages that what we have called the "effect" verb here heads a
predicate expression which is, at some level of syntactic representation, a
complement to the causative verb. In any case, the "effect" verb is dependent,
and its tense is dependent, or bound to that of the causative verb, one
morphological realization of a bound tense being the infinitive. The causative
verb itself has the characteristics of a main verb -- in the translation of (la-c),
therefore, the tense of the causative verb is free.

By contrast, in the Miskitu construction, it is the effect verb whose tense is

"free", while the tense of the causative verb is dependent. Thus the effect verb
in (1) can inflect for the full range of tenses typical of main clause verbs in
Miskitu. The causative verb appears in the "connective" form (an extension of
the term conexivo used in the CIDCA grammar, following Heath (1927)), which
distinguishes just two tenses, future connective (FC), as in (la), and non-future
connective (NFC), as in (lb-c). This is a dependent form, and, accordingly, does
not appear in root clauses. Rather, it appears as a non-final verb form in a
variety of Miskitu constructions, of which the causative is just one.
 191

While the non-future connective happens to be homophonous with the past tense
of Miskitu, it must be distinguished from that. since in the construction under
consideration here its temporal reference is dependent upon the tense of the
...
·-
.!
_ final verb (i.e., the effect verb, in this case), as can be seen from the
translations of (lb-c). Just as there is some reduction of tense distinctions in

' the connective, so also there is some minor reduction in the category of person.
In the future connective, the person categories are reduced to two -- third (-ka)
and non-third (-r(i)ka). In the non-future connective, as in the homophonous
past tense, the full three-person system is realized (i.e., first -ri, second -ram,
third -(a)n). This picture is a slight simplification inasmuch as the first person
inclusive, in its suffixal realization (in the connective and elsewhere), is identical
to the third person -- hence, "third" above is to be understood as embracing the
first inclusive as well.

The reason behind our use of the term "insubordination" should now be clear. In
the causative construction, the "effect" verb, generally realized as a "dependent"
or "subordinate" verb, in languages in which the construction has been
extensively studied, displays the morphological characteristics of a "main" verb in
Miskitu -- it is therefore "insubordinate", by comparison. at least. It remains to
be seen, of course, what the precise nature of this Miskitu construction is. In
particular, to what extent is it linguistically proper to liken the Miskitu
construction to that represented by its typical Inda-European translation?
Is this really a causative construction, or does it belong to another construction
type altogether? We will not be able to answer these questions fully here. Our
purpose is rather to present data relevant to the study of the construction and
to its position within the context of Miskitu grammar generally.

1. Further evidence for "insubordination"

In the previous section, evidence from tense inflection was used to argue that,
in the Miskitu causative construction, it is the effect verb, not the causative
verb, that assumes the grammatical characteristics of a main clause predicator,
reversing the asymmetry found in English. In this section we consider some
additional data supporting this argument. Here again, we will use the expository
device of contrasting the Miskitu causative with that of English.
192

1.1. Negation in the causative construction.

Simplifying matters somewhat, sentence negation is effected in Miskitu by means

of elements attached to the verb root (in the non-future) or placed after the
inflected verb (in the future), as in the following examples:

(2)a. Yang plap-ras.

I run-NEG
'I don't/ didn't run.'
b. Yang plap-amna apia.
I run-FUTl NEG
'I will not run.'

In the negative of a causative construction, this mqrphology appears in

association with the effect verb, not the causative verb:

(3)a. Yang mita tuktan ba yab-rika kauhw-bia apia.

I AG child DEF cause-FC fall-FUT NEG
'I will not make the child fall.'
b. Yang tuktan ba yab-ri in-ras.
I child DEF cause-NFC cry-NEG
'I don't/didn't make the child cry.'

In this respect, as expected, Miskitu contrasts with English, where the negative
morphology is adjoined to the auxiliary of the main verb.
The behavior of negation parallels that of tense inflection perfectly. If we
consider the causative construction to be the linguistic realization of a
conceptual structure corresponding to a single event, grammatically speaking,
despite the appearance of two verbs, then it is to be expected that just one
tense will be attributed to that event. Similarly, the polarity (e.g., negative) of a
sentence depicting a single causative event is expected to be realized just once.
It should not be surprising, therefore, if languages differed in the manner in
which tense and negation are realized morphosyntactically in the causative
construction. English realizes these categories in the clause headed by the
causative verb, while Miskitu realizes them in the clause headed by the effect \
verb.
 193

1.2. Causative constructions in the infinitive.

In addition to structures of the sort represented by (1) above, Miskitu also

possesses constructions which exhibit the more usual characteristics of standard
}'•
1f complementation. Among these more standard structures are those in which the
verb of the complement clause appears in the infinitive. Among the predicators
taking infinitival complements are English-based verbal expressions want k-aia
(want be-INF) 'to want' and trai kaik-aia (try verb-INF) 'to try', as exemplified
by the sentences of (4) below. These sentences also exemplify the fact that the
infinitival complement may either precede or follow the main verb:

{4)a. Yang [Bilwi ra w-aia] want sna.

I [Port to go-INF] want be-PRESl
'I want to go to Port (- Puerto Cabezas).'
b. Yang want sna [Bilwi ra w-aia].
I want be-PRES [Port to go-INF]
(same translation as (a))
c. Yang [truk kum atk-aia] trai kaik-ri.
I [car one buy-INF] try see-PASTl
'I tried to buy a car'.
d. Yang trai kaik-ri [truk kum atk-aia].
I try see-PASTl [car one buy-INF]
(same as (c))

The main verbs in these constructions, both in Miskitu and in their English
counterparts, have the property of "selecting" the infinitival morphology on their
complements -- this is in concert with the "control relation", in which the
subject of the complement is bound to that of the main clause. H a causative
construction appears as the complement of a verb which "selects" the infinitive,
then we might expect the infinitival morphology to be realized differently in
Miskitu and English, given the observations we have made to this point. And this
is in fact the case, as (5) illustrates:

(5) Yang want sna [tuktan ba mun-rika kauhw-aia]

I want be-PRES [child DEF cause-FC fall-INF]
'I want to make the child fall.'
194

In English, the infinitival morphology selected by the main verb is realized on

the causative verb (i.e., make). In Miskitu, on the other hand, it is realized on
the effect verb (i.e., fall). This is in keeping with the pattern which has
emerged in general in relation to the causative construction which, in Miskitu,
gives prominence to the effect verb (as opposed to the causative verb itself) in
matters having to do with realizing morphology germane to the construction as a
whole.

As an aside, it should be pointed out that imposing infinitival morphology on the

causative has the effect of obliterating the person marking on the effect verb,
even though the person category associated with that verb is, of course, distinct
from that of the main verb. This is consistent with the fact that the infinitival
goes along with "control", and the subject of the causative construction is
indeed controlled by the subject of the main verb. However, it is unusual in
Miskitu for the infinitival morphology to be used in situations where the person
category of the dependent verb is distinct from that of the main verb. It is,
however, not surprising that it should happen here, since the phenomenon
observed in the causative is the natural "working out" of genuine principles of
Miskitu grammar -- to wit, subject control, infinitival selection, and the
principle of effect-verb prominence in the causative. It is, however, possible to
employ an alternative resolution of these matters according to which the future,
rather than the infinitive, is used -- in fact, this is preferred over the pattern
in (5), whieh is regarded as ungrammatical by some speakers:

(6) Yang want sna [tuktan ba mun-rika kauhw-bia]

I want be-PRES [child DEF cause-FC fall-FUT3]
'I want to cause the child to fall'.

Here, the person category appropriate to the effect verb (i.e., third person) is
preserved. This use of the future is usual, quite apart from the causative, in the
complement of a verb of the type represented by want k-aia 'to want' wherever
the subject of the complement is distinct from that of the main verb. In this,
the future tense and the infinitive function in a complementary pattern
comparable to that of the subjunctive and the infinitive in, say, the Romance
languages. Properly speaking, therefore, it should be said of the main verb in (5)
and (6), and of other members of its class, that they select a bound irrealis
tense in their complements, this being realized as the infinitive in the "control"
 195

case (i.e., where the subject of the complement is bound and non-overt), and as
the future or subjunctive (depending on the language) in the "obviative" case
(i.e., where the subject is free, not controlled). This more accurate
characterization of the selection properties of the verb of (5) explains the
appearance of the FC morphology (rather than NFC morphology) on the causative
verb of that sentence, incidentally.
It is perhaps relevant to point out that the use of the future, over the
infinitive, is strongly preferred in purposive constructions of the type
represented by (7) below:

(7) Witin bu-an yang ra ai mun-ka kauhw-amna dukiara.

he rise-PAST me ACC me cause-FC fall-FUTl PURP
'He stood up in order to make me fall'.

This is not a control construction, in the sense of (5), and the irrealis tense
morphology -is induced, not by a matrix verb, but rather by the purposive
postposition dukiara (PURP) 'for'.

We have considered three morphosyntactic structures (tense, negation, and

infinitival morphology) which indicate that languages (English and Miskitu in this
case) may differ in the manner in which the asymmetry inherent in the relation
between the verbs implicated in the causative construction may be realized. We
will attempt now to give a preliminary characterization of the Miskitu
construction in the context of Miskitu grammar generally.

2. The grammar of the Miskitu causative construction

In English, and in other Inda-European languages as well, the morphosyntax of

the causative construction belongs for the most part to the syntactic system of
comp.lementation. Thus, the effect verb generally heads a clause which bears the
complement relation to the causative verb; and, typically, the effect verb is in
the infinitive, or at least a dependent tense form.

Miskitu, like English, has the infinitival complementation construction, as in (4-

5). But this construction is not normally used in expressing the causative
relation in Miskitu. The properties of infinitival complementation are utterly
196

distinct from the properties of the favorite Miskitu causative construction. In

the complementation construction, in Miskuti as in English, the verb of the
complement cause is clearly dependent, while that of the main clause is not
(provided, of course, it is not itself subordinated within a larger structure)--
thus, the tense of the complement clause is bound, while that of the main clause
is free. By contrast, in the Miskitu causative, as we have seen, the direction of
the dependency asymmetry is reversed, and it is not at all clear to what extent
it makes sense to speak of the effect verb as heading the complement of the
causative verb.

The "effect portion" of the Miskitu causative is unlike a complement clause not
only in respect to the grammatical asymmetries observed in previous sections but
also in respect to linear ordering -- an infinitival complement may precede or
follow the main verb, while the effect verb must always follow the causative
verb.

The ordering relation just noted for the Miskitu causative, together with the
morphosyntactic properties which characterize it, is directly relevant to
determing its proper position within the grammar of the language. The
properties of the causative are not unique to that construction in Miskitu. They
are also characteristic of at least two other constructions involving "verbs in
sequence" -- namely, (i) the serial verb construction, and (ii) the protasis-
apodosis construction. These are exemplified, respectively, in (8a) and (8b)
below:

(8)a. Yang truk-ki-i atk-ri w-an.

I car-CONST-my transact-NFC go-PAST
'I sold my car off.
b. Tuktan ba waitla ra dim-ka, yang kik-amna.
child DEF my:house to enter-FC, I laugh-FUTl
'When( ever) the child enters my house, I'll laugh.'

In the first of these, the verbs are in close sequence, approximating the tight
organization typical of verbal compounding. In fact, sequences of this type
figure prominently in the lexicon of Miskitu. The second sentence represents
one of several ways in which the protasis-apodosis relation is realized in
Miskitu. The morphology in these sentences is identical to that of the causative
 197

-- in particular, the first verb appears in the connective, while the final verb
appears in the full tensed form. It is reasonable to argue, therefore, that the
Miskitu causative belongs to this system of ''verb sequencing", rather than to
the complementation system.

Before continuing this line of thought, it should be mentioned that there is more
to the morphology of Miskitu verb sequencing than our examples so far would
suggest. Miskitu, like many Native American languages, has a "subject obviation"
or "switch reference" system. In the examples so far cited, the subjects of the
two verbs are distinct. This is the "obviate" subject relation, and it is in the
obviative, typically, that the connective morphology (i.e., FG and NFC) is used.
When the subjects of verbs in sequence are identical, i.e., in the "proximate"
relation, the non-final verb appears in a participial form marked by means of the
suffix i (called the participio presente in the CIDGA grammar and glossed PRP
here). This participial form is purely dependent and does not vary for categories
of person or tense. Proximate constructions are illustrated in (9) below (taken
from the CIDGA grammar, pp. 142-4), the first representing the so-called serial
verb construction, and the second representing the protasis-apodosis structure:

(9)a. Witin aras ha alk-i wilk-an.

he horse DEF catch-PRP tie-PAST3
'He caught and tied the horse.'
b. Yang baha lawan-ka aiwan-i, mahka utla ra wa-ri.
I that song-CONST sing-PRP, imm. house to go-PASTl
'Having sung that song, I immediatly went to the house.'

In light of its superficial morphosyntactic properties, we will assume that the

Miskitu causative belongs properly to the system which defines the grammar of
verbs in sequence, i.e., the system exemplified in (8) and (9).

3. The Miskitu causative as a serial verb construction

The causative construction can be viewed as involving a conceptual structure

embodying two subevents, the cause and the effect. Miskitu and English agree
in realizing the first of these by means of a "causative" verb (e.g., yab-aia, mun-
198

aia in Miskitu, make in English); and they agree, of course, in realizing the
effect subevent by means of a predication (generally headed by a lexical verb, as
Miskitu kauhw- or English fall in (1)). But the two languages differ according
to their structural expression of the causal relation. In English, as in Indo-
European languages generally, this is realized syntactically by means of the
head-complement relation, according to which . the effect portion of the
conceptual structure is represented by a clause embedded as the complement of
the causative verb. The syntactic constituent corresponding to the effect is
thus syntactically subordinate to the causative verb -- the latter, therefore,
heads a matrix clause in relation to the structure as a whole.

In Miskitu, by contrast, the causal relation is realized in surface syntax by

means of the so-called serial verb construction, in which the causal verb appears
in sequence with the predication representing the effect portion of the
causative conceptual structure. It is not clear, as yet, what the syntax of
serialization is, precisely, though a number of highly promising suggestions have
been made in the literature (e.g. Awoyale (1987); Dechaine (1988); etc.).

On the basis of the tense inflections in the Miskitu causative, it is evident that
the clause realizing the effect portion of the causative serial verb construction
is not morphosyntactically subordinate in Miskitu surface structure. This follows
since the tense of the effect clause is free, while that of the causative verb is
bound. This is a common circumstance in the type of clausal serialization found
in languages which possess fully developed subject obviation, or switch reference,
systems (cf. Jeanne (1978); Finer (1985a,b)), a system which exists in Miskitu
and to which the causative belongs (cf. Hale (1988); Salamanca (1988)).

In the analyses of subject obviation developed by Jeanne (1978) and Finer

(1985a,b ), the final clause is in a structurally superior position in relation to the
prior clauses in sequence; the final complementizer, accordingly, commands all
prior ones. If, in the sense relevant for the Binding Theory (Chomsky (1981)),
its tense also commands all prior tenses, then the final tense can bind the prior
tenses, accounting for the asymmetry observed in the Miskitu causative, as well
as other clausal serialization constructions. Abstractly, this would suggest that
the surface syntactic structure of the Miskitu causative, as a representative of •
this variant of the serial clause construction, would have a form embodying
appoximately the command relations present in (10) below, where IP is the
 199

phrasal category projected by the functional head of the sentence (i.e.,

Infl(ection), or I), in which tense (as well as agreement) is located:

(10)
IPef
.....···...
........ ······...
IPea Ief

Here the causative clause (!Pea) is subordinate to the effect clause (!Pet). The
functional head of the latter (let) commands that of the causative clause (lea)
and, accordingly, lef is in the proper position to bind lea. Assuming, then, that
the connective (together with the present participle) of Miskitu is an
"anaphoric" tense, its appearance as lea in (10) is appropriate since, in that
position, it can be properly bound, as required by the Binding Theory.

The structural relations embodied in (10), with modifications of detail to

accommodate the complementizer (cf. Finer (1985)), may be correct for some
subject obviation structures, perhaps for most of them, it cannot be correct for
the Miskitu causative. While it would account for the observed asymmetry in
tense dependencies, the structure is completely wrong in what it implies about
the subject of the effect clause.

From the sentences of (1), it is evident that the subject of the effect verb (i.e.,
kauhw- 'fall' in that case) is treated as the surface object of the causative verb.
While the causative does not select the subject of the effect verb, in the sense
of assigning it a semantic role, it nonetheless exerts upon it the influence
otherwise associated with the structural relation of government. Specifically, the
causative may assign case to the subject of the effect clause, a fact which is
reflected not only by overt accusative case morphology {ACC), but also by the
fact that the subject may be represented by proclitic object agreement
morphology in the first and second persons, and by the fact that the subject of
the effect clause may appear displaced to the position before the causative verb,
in conformity with the standard direction of government in this verb-final
language. All of this is illustrated in (11) below:
200

(11) Man yang ra taim bani ai mun-ram kauhwi-sna.

You me ACC time all me cause-NFC2 fall-PRESl
'You always make me fall.'

If the subject of the effect clause is governed by the causative verb, then the
causative constrution must, at some level of syntactic representation, have a
structure in which the command relations are exactly the reverse of those
obtaining in (10), and in particular, a structure in which the effect clause (IPef)
is a complement of the causative verb (Vea), which therefore governs the effect
clause and its subject (NP*):

(12)
I Pea
..······· ·······...
....../ VPca ··.i.~a

IPef
N.P·~···· ·······~~f
Vea

In structure (10) above, the functional head of the effect clause -- i.e., the
inflection associated with the effect verb (Ief) -- commands the causative verb
and its inflection (lea). In (12), this is reversed, so that lea commands Ief. These
two structures contradict one another, but each is a legitimate representation of
grammatical relations -- structure (10) correctly represents the command
relations required in order for Ief (containing the free tense) to bind lea
(containing the bound tense), and (12) represents the command structure
required for Vca to govern the effect clause (IPef) and, therefore, its subject.

Under the grammatical assumptions just advanced, it is evident that the syntactic
structure of the causative construction must ·partake of both command relations.
It cannot do so at one and the same time, of course, but the situation
represented by the causative is duplicated in an interesting way by such pairs as
the following:
 201

(13)a. Yang [Ingglis Ian tak-i] ta krik-ri.

(I [English learn-PRP) start-PASTl]
'I started learning English.'
b. Yang ta krik-i [Ingglis Ian tak-ri).
(I start-PRP [English learn-PASTl])
'I started learning English.'

These are alternative realizations of an inceptive conceptual structure

(informally, and approximately, X start activity Y). In (13a), this conceptual
structure is realized by means of a complementation construction, while in (13b),
it is realized as the serial construction. That is to say, ceteris paribus, (13a)
corresponds to (12) above, while (13b) corresponds to (10), the structure
associated with the causative. In the former, the "activity clause" (bracketed) is
an embedded participial, while in the latter, it is the final, and finite, clause.

It is unlikely that the two realizations of the inceptive exemplified in (13) are
related transformationally, since if they were, we would expect, contrary to fact,
that all serial constructions would exhibit this "alternation". But the idea is
nonetheless suggestive, since a transformational relation between (10) and (12)
would enable us to solve the "causative contradiction" in a straightforward
manner.

The analysis would be as follows. The so-called Projection Principle (Chomsky

(1981)) determines, for the causative, that the structure depicted in (12) is the
initial projection from the lexicon, i.e., the d-structure. This follows, since it is
in (12) that the lexical selection properties of the causative verb are
represented. In particular, it is in (12) that the causative verb properly governs
its complement, the effect clause. And let us assume that the effect clause is
finite, not participial. The fact that (12) does not correctly represent either the
observed surface ordering of verbs or the command relations required by the
tenses is not a problem, if we assume, as we must, that these latter
requirements are matters of derived structure, i.e., of s-structure, not d-
structure. The generalized transformational rule "Move-alpha" (Chomsky (1981))
will, we propose, apply to move the effect clause Ipef -- i.e., the complement of
the causative verb -- rightward to a raised adjunct position from which it can
bind the anaphoric tense inherent in the inflection (lea) of the causative clause:
202

(13)
I Pea

................ ············...

IPef<i> Vea

The coindexed trace (IPef<i>) is in accordance with the theory of syntactic

movement. While the derived structure (13) involves adjunction to the causative
clause, and therefore does not produce a structure identical to (10), it is
nevertheless true in (13) that IPef (the maximal projection of the inflectional
element Ief, embodying the "free" tense) commands, and therefore potentially
binds, the tense of the causative clause, as required, given that the latter is
anaphoric. The movement is therefore "forced" by the Binding Theory -- i.e.,
the free tense must be enabled to bind the anaphoric one; this is impossible at
the ct-structure level of representation.

Final remarks

The "complement-raising" hypothesis just outlined is, in fact, just one of several
analyses which must be explored in reaching an adequate understanding of the
causative constructions of Miskitu. Unfortunately, it is the only one which we
will have the opportunity to examine here (but see Dechaine (1988), for a non-
tranformational analysis of the closely similar Haitian causative). In favor of a
transformational analysis is the fact that it conforms both to the Projection
Principle, requiring that the selectional properties of the causative verbs be
represented at all levels of syntactic representation, and to the surface
structure reality that the Miskitu causative is realized as a serial clause
construction, of a type that occurs independently in the language. The
causative, originating as a conventional complementation structure, assumes the
surface form of a serial construction at the s-structure level of representation.
 203

There are some technical details which must be dealt with in developing the
complement-raising hypothesis. We will deal only with the most pressing of
• these, the first having to do with the surface syntactic position of the subject
of the effect verb, the second having to do with the possibility of extraction out
'\ of the clauses of the causative construction.

Generally, as we have observed in examples given so far, the subject of the

complement clause remains behind when the effect clause is moved. The derived
structure (13), however, does not indicate this. The fact is that the subject may
or may not remain behind. Thus, both of the following are possible causatives:

(14 )a. Witin yang ra ai swi-n skul ra wa-ri.

he me ACC me let-NFC3 school to go-PASTl
'He let me go to school.'
b. Witin ai swi-n yang skul ra wa-ri
he me let-NFC3 I school to go-PASTl
'He let me go to school.'

In (14a), the subject of the raised effect clause remains in the position preceding
the causative verb -- the latter governs it, assigns case to it, and in addition
registers it in object agreement (realized by means of the proclitic first person
object marker ai). In (14b), on the other hand, the subject of the effect clause
remains internal to that clause, moving with it and, as expected, receiving
nominative case as befitting its surface position. It is, however, registered in
object agreement on the causative verb.

In order to account for the more common of these two possibilities, that
represented by (14a), we must assume either that a projection lower than IP can
be moved, thereby excluding the subject, allowing it to remain behind; or else
we must assume that the subject can first move to an adjunct position peripheral
to the effect clause, allowing that clause to move without necessarily taking the
subject with it. The second of these possibilities seems most likely (cf. Massam
(1985)), in part because there does not exist in Miskitu a pre-sentential position
to which arguments can move -- this is the position occupied by the relative NP
in one type of relative clause. This could be the required adjunct position to
which the subject may be removed prior to extraction of the clause, provided we
can guarantee -- for the causative but not for the relative -- that only the
204

subject is a candidate for the proposed peripheralization. We will leave this

issue without taking a definite position on it, moving now on to a brief
consideration of content question formation in relation to the causative
construction.

It is possible to question the subject of the causative verb, as in (15):

(15) Ya mita mai mun-an ai pruk-ram?

who AGTyou cause-NFC3 me hit-PASTI
'Who made you hit me?'

And it is possible also to question a constituent of the effect clause, as shown

by (16) below, where the object of the effect clause is questioned:

(16) Dia mai mun-an pi-ram?

what you cause-NFC3 eat-PASTI
'What did he (or they) make you eat?'

This is not surprising if we assume that question formation applies to the

causative structure represented in (12), in which the effect clause appears as a
complement to the causative verb. It would be surprising, however, if it applied
to the derived structure (13), in which the effect clause is not governed, given
well-known constraints preventing extraction from adjuncts (cf. Huang (1982);
Chomsky (1986)). Although details of question formation in Miskitu remain to be
investigated, we take (16) to be evidence that the effect clause, at some level of
syntactic representation, is not an adjunct.

The complement-raising hypothesis outlined above succeeds in expressing the

grammatical relations required by the Projection Principle, but it fails to capture
what is perhaps the most important fact about the Miskitu causative
constructions -- namely, the fact that their form is precisely that of the serial
clause constructions which exist independently in the language within the
generalized subject obviation (or switch reference) system. An adequate treatment
of the Miskitu causative, and its closely similar counterpart in the sister
Misumalpan language Sumu (cf. Norwood (1987), Hale (1988)), must come to grips
with this basic fact.
 205

Notes

* We are pleased to dedicate this brief essay to Wim de Geest, whose work on
complementation has been so valuable to the field. We wish only that our work
were more worthy of Wim and of the standards he represents in his own research.