Volume 16 Issue 2 (2018) DOI:10.1349/PS1.1537-0852.A.494 Note: Linguistic Discovery uses Unicode characters to represent phonetic symbols. Please see Optimizing Display for requirements to accurately reproduce this page. The
Development of Agent-demoting Passives in Malayic*
The di-passive in Malay/Indonesian and
other Malayic languages is clearly a European-type agent demotion passive; the
agent, nevertheless, unexpectedly exhibits grammatical properties which are
cross-linguistically typical of arguments and unusual for adjuncts. We explain
the unusual properties of di-passive agents by arguing for the analysis of the
di-passive construction as deriving historically from a Philippine-type voice
construction in which the non-subject agent remains an argument. Based on
evidence from modern and classical varieties of Malayic, we propose a pathway
of historical changes whereby a Philippine-type non-active construction
developed into a European-type agent demoting passive. 1. Introduction In
this paper we examine the grammatical characteristics of di-passives both in Standard Indonesian and in several colloquial
Malayic varieties spoken in central Sumatra.[1]
We shall argue that the characteristics of di-passives
are typologically surprising in passive constructions, but the unusual features
of this construction are explained if they are viewed as remnants of an earlier
‘object voice’ construction that, we claim, was the historical source for the di-passive. Some background on
Austronesian voice may be helpful before we turn to the details of Indonesian
and the central Sumatran varieties. A
key typological characteristic of conservative western Austronesian languages
like Seediq, Malagasy, and Tagalog is that these
languages manifest a distinctive voice system which has come to be known as the
‘Philippine type’ voice system.[2] In a
Philippine type voice system the choice of voice determines which argument
occurs as the surface subject. No voice is morpho-syntactically basic. A
salient characteristic of this sort of voice system, one which, as we shall
see, distinguishes Philippine voice systems from European type passives, is
that in ‘non-active’ voices, demotion of the agent to adjunct does not occur.
In LFG terminology, agent arguments are retained as core arguments in
transitive constructions with non-agent subjects. The
Tagalog sentences in (1) exemplify the Philippine type voice system. In each
sentence a different argument occurs as surface subject. In each case, verbal
morphology indicates which argument is subject. In cases where the agent is not
the subject it remains an argument of the verb.
As
was already mentioned above, the Philippine type voice system can be contrasted
with voice systems often referred to as ‘European type’ voice systems. In
languages with a European type voice system, the ‘unmarked’ voice is the active
voice. This voice is literally unmarked in that no special morphology is used
to indicate the occurrence of the construction. Furthermore, the agent
corresponds to the surface subject of the clause. In ‘European’ voice systems
the active is contrasted with a 'marked' passive voice, in which the agent is
either omitted or it appears as an adjunct rather than as a core argument of
the verb.[3]
Notably,
both the agent and the patient are obligatory arguments in 2, but the agent is
omitted or occurs in an optional prepositional phrase in 3 while the patient
remains as the sole core argument of the verb. In 3 not only is the agent
‘demoted’ to adjunct, but the patient is ‘promoted’ from object to subject. In
works from the 1980’s (e.g. Comrie 1988, Perlmutter & Postal 1984 inter alia) the promotion of the object
to subject is often treated as the defining characteristic of the passive;
however, many more recent works (e.g. Roeper & van Hout
1999, Arka & Manning 1998, Kiparsky
2013, Wunderlich 2012), treat demotion of the agent as primary. Valency-decreasing
operations reduce the number of syntactically active arguments; they apply
directly on the Θ-structure. [The schema below] shows the passive, which binds
the highest argument existentially, so that that it remains unexpressed. (The
event argument is irrelevant here and therefore ignored.) PASS
[...
lx
VERB(x, ... )] = ... ∃x VERB(x, ...) (Wunderlich 2012:2231) From
this perspective, the promotion of the object to subject in examples like 2 is
viewed as secondary, and, in some frameworks, is attributed to the fact that
passive verbs share with adjectives the property of failing to case mark their
objects (thereby motivating the promotion of the object to surface subject so
it will receive case). Thus, the promotion of the patient to subject may be
viewed as a strategy to avoid stranding the patient in an uncasemarked
position, rather than a core characteristic of the passive construction.
Evidence for this approach is presented by Roeper & Van Hout
1999 based on data from nominal passives. In nominal passives the agent is
demoted to adjunct but the patient need not be
promoted to subject (because the preposition of is present to case-mark the object).
Compare
4 with its active nominal counterpart in 5 below.
In
the nominal passive in 4 the agent (the
enemy) has been demoted to an adjunct but the patient (the city) remains in object position and is not promoted to subject
(because the city is case-marked by of). We adopt this perspective, and,
like the authors cited above, we take the demotion of the agent to adjunct/non-core
oblique to be firmly established as the critical characteristic of passive
constructions. The
voice system of conservative Austronesian languages contrasts sharply with
European type voice. As noted by Himmelmann (2005
inter alia), not only are the voice systems in conservative Austronesian
languages symmetrical (i.e. no voice is morpho-syntactically basic), but, more
critically for our purposes, in these voice systems, while non-active voices
involve the selection of a non-agent argument as surface subject, the demotion
of the agent to adjunct does not occur, and agent arguments are retained as
core (but non-subject) arguments in transitive constructions with non-agent
subjects. We shall refer to non-active voices that do not involve demotion of
the agent to non-argument status as instances of ‘object voice’ rather than
‘passive’.[4] 2. Mixed voice systems in Indonesian-type languages Early work on Indonesian (like Chung 1976) treated examples
like 6 and 7 as simply two different passive constructions,
and did not distinguish between passive and object voice (except as two
variants of the passive). There is, however, an emerging consensus among Austronesianists that Indonesian-type languages exhibit a
‘mixed’ voice system (Arka & Manning 1998 inter
alia). These languages, on the one hand, retain (a reduced version of) a
‘Philippine’ type object voice construction, while on the other hand, they also
appear to exhibit a European type passive, in which the agent argument is
‘demoted’ to adjunct. This passive construction is marked with the prefix di- in Standard Indonesian, and, on this
view, is to be contrasted with object voice.
There
are a number of reasons to view 6 and 7 as different constructions. In the di-passive, as in 6, the agent is
optional and occurs after the verb (either as a bare NP or in a prepositional
phrase) while in object voice, as in 7, the presence of the agent is obligatory
and occurs pre-verbally. The fact that the agent is optional in 6 but
obligatory in 7 is in itself strong evidence that the agent is an argument in
object voice but not in the di-passive.
Furthermore, Arka and Manning have shown that the
distribution of reflexives provides yet stronger evidence that the agent is a
core argument in 7 but not in 6: In the object voice the agent is able to
antecede reflexives, including reflexives in surface subject position.
In
contrast, in di-passives the agent
cannot be the antecedent for a reflexive in subject position.
The
sentences of 9 are ungrammatical despite the fact that the agent is a bare NP
and not a prepositional phrase. In
Arka and Manning it is taken as definitional that a
nominal that can antecede a reflexive must be an argument. Members of our
research group have argued elsewhere with respect to Toba Batak that similar
facts follow naturally from c-command (Cole
and Hermon 2008). The same arguments presented there apply to
Indonesian as well. We assume that in object voice constructions the agent is
generated in a high position outside the VP (as the specifier of vP) and, thus, c-commands the arguments generated within
the VP (like the patient, before it moves to subject position). In contrast, in
passives we assume a Larsonian structure in which
adjuncts are generated as ‘lowest complements’ inside the VP, hence lower than patients.[5]
We further assume that reflexive dirinya is possible iff it has a
potential c-commanding antecedent at some stage. (10)
Object Voice (11) Di-passive[6] In object voice, the agent c-commands the patient prior to
the movement of the patient to specifier of TP. Thus, it is predicted that the
agent will be able to be the antecedent for a reflexive in specifier of TP. In
the passive construction, however, the agent is lower than the patient at all
stages in the derivation. Thus, it is predicted on the basis of the structures
proposed that in passives the agent will not be a possible antecedent for a
reflexive in subject (specifier of TP) position. Therefore, given our
(relatively standard) assumptions, the reflexive facts reported by Arka and
Manning provide very strong evidence that the agent is an argument in the
object voice but that it is an adjunct in di-passives.[7] The conclusion that the agent in di-passives is an adjunct is further supported by evidence from
nominalization. In SI (but not in all colloquial varieties) predicates with a
single argument can be nominalized (Sneddon et al. 2010). The basic pattern for
nominalization is illustrated in 12 below. The predicates teriak ‘scream’ and membeku
‘freeze’ take a single argument, and are able to undergo nominalization.
Predicates with more than one argument, like mencuri ‘to steal’ and membekukan ‘to freeze something’, cannot
undergo nominalization.
Nominalization can be used to test whether
the di-passive agent is an argument
or an adjunct. If the agent is an argument in this construction, the di-passive predicate will have two core
arguments (agent and patient), thus it is predicted that nominalization will
not be possible. If the agent is an adjunct, and therefore the predicate has a
single argument (patient), nominalization is predicted to be possible. As
Sneddon and colleagues (2010) demonstrates, di-passives
allow nominalization, thus supporting the conclusion that the agent is an
adjunct.
In contrast, predicates in the object voice do not undergo
nominalization.
These facts follow directly from Arka and Manning’s analysis
which claims that the object voice is a predicate with two core arguments,
where the agent is a core argument, whereas the di- passive is a predicate with a single argument (patient) with
the agent (optionally) appearing as an adjunct. Thus, the facts regarding nominalization lead to the same
conclusions as Arka and Manning’s Binding facts. Object voice and di-passives differ in that the agent in
object voice remains a core argument while in di-passives the agent has been demoted to adjunct. Given the evidence we have presented above, we shall
therefore assume in what follows that Arka and Manning’s characterization of
the diffeences between passive and object voice is correct: In object voice
constructions the agent remains an argument and is not demoted to adjunct
status while in di-passives the agent
has been demoted to non-argument status. 3. Some peculiarities of the di- passive in Standard Indonesian We shall now turn to some apparent counter examples to the
generalization argued for in the previous section. As described in the previous
section, the di- passive is simply a
European type passive, so it would be expected to fit the prototype for such
constructions. In fact, as we shall show, the construction has a variety of
properties that receive no explanation if the di-passive is simply a European type passive with the syntactic
properties described above, in which the agent is considered to be in a very
low position in the VP. We shall first show some properties of di- passives in Standard Indonesian that
fail to fit the typological profile of a European type passive. We shall then
consider some additional characteristics found in traditional Malayic languages
of central Sumatra that provide further evidence that the description of the di-passive as a European passive is an
oversimplification. 3.1 Complementary distribution In
Standard Indonesian the di-passive
and object voice are in (near) complementary distribution, suggesting the
possibility that the two constructions should be viewed as two realizations of
a single construction in different contexts, either in contemporary Indonesian
or at some earlier historical stage of the language. Speakers almost invariably
use object voice with first and second person agents
and passive with third person non-pronominal agents.
While
some overlap occurs with third person pronominal agents, it is plausible to
claim that speakers view the two constructions as contextual variants of a
single construction rather than as two independent,
contrasting constructions. Furthermore, this pattern can be interpreted as
indicating that verbs in these two constructions select for the agent and stand
in an agreement relationship with the agent as well, with one verb form (di-) agreeing with third person agents
and a second verb form (Ø) agreeing with first
and second person agents.[8]
Such an analysis (with some additional refinements to handle the overlap
between the two constructions) would be unproblematic from a typological
perspective with regard to an object voice construction because in the object
voice the agent is analyzed as an argument of the verb, and verbs frequently
select for and agree with their arguments. It is, however, problematic, with
regard to the passive since selection for and
agreement with an adjunct is typologically uncommon and is not expected to
occur. 3.2. Adjacency requirements We will turn next to a second grammatical requirement shared
by arguments of verbs and agents of di-passives,
but not by adjuncts. Verbs in Indonesian require adjacency between the verb and
its direct object (shown in boldface).
There
is no adjacency requirement for adjuncts. This is true even when the adjunct is
itself a noun phrase, as is the case with regard to bulan ini in (21)-(22). Turning
now to di-passives, much like the
direct object of the active verb in (21)-(22) above, the nominal agent (i.e. agent
DP not agent PP) in the di-passive
must be adjacent to the verb.
Examples
23-24 show that (noun phrase) agents of di-passives
are like direct objects of active verbs in requiring adjacency to the verb. In
contrast, adjuncts display no adjacency requirement, even when the adjunct
consists of a noun phrase like hari ini (and, hence, might, presumably, require case
marking). This distribution would be easily explained if passive agents were
analyzed as (non-subject) arguments rather than as adjuncts. 3.3. Relativization A
related distribution is observed with regard to relativization and similar processes of movement to an A’ (non-argument) position. Standard
Indonesian (as opposed to many urban colloquial varieties) is well known as an
example of a language that allows subject but not object relativization (Keenan & Comrie 1977). In this construction as well, NP agents in di-passives pattern with direct objects
rather than with adjuncts. In the following examples, the arrow indicates a
possible movement path. An X on the arrow indicates that the movement path is
not possible.
Di-passive agents in
Kerinci varieties show the same distribution of
argument-like properties as do passive agents in Standard Indonesian. First, in
Kerinci the di-passive
and the object voice are in complementary distribution (not in near
complementary distribution, as in Standard Indonesian).
Second,
the di-passive verb selects for and
agrees with the features of the agent.
The nominal agent must be adjacent to the passive verb.
This
is similar to the situation in the active, in which the direct object must be
adjacent to the active verb.
It
is also similar to the object voice, in which the agent must also be
immediately adjacent to the verb.
We
can summarize what we have shown so far as follows: The di-passive construction in Kerinci
displays many of the same attributes as Standard Indonesian (but in some cases
the argument-like properties of the agent are more clear
cut than in Standard Indonesian). In fact, the argument-like properties of di-passive agents are attested to a greater
or lesser extent across Malay varieties and registers. These properties should
be seen as characteristic of the di-passive
construction in general, and therefore, highly pertinent to understanding the
construction’s historical development. We
shall now turn to ways that nominal agents in the di-passive construction in Kerinci display argument-like properties that go beyond those found in Standard
Indonesian.[9] 4.1. Only nominal agents are possible
in di-passives We saw earlier that in Standard Indonesian passive agents can
either be noun phrases or prepositional phrases.
In
contrast, in object voice only nominal agents are possible and prepositional
phrase agents are not.
Example
38 shows that a PP agent is not possible in object voice, regardless of whether
the agent is preverbal (as in 37b) or postverbal (as in 37c).[10]
Turning
to Kerinci, unlike SI, passive agents can only be
nominal.
The
restriction of passive agents to NPs rather than PPs more closely resembles the
pattern seen in object voice in Standard Indonesian than the pattern seen in
the di-passive. 4.2. Obligatory agents It will be remembered that in Standard Indonesian the
presence of the agent is optional in di-passives.
In fact, in Standard Indonesian the agent is frequently omitted.
This
is to be contrasted with object voice, in which the occurrence of the agent is
obligatory.
In
the Tanjung Pauh variety of
Kerinci, however, the presence of the agent is
obligatory in di-passives.
The
requirement that an agent must occur resembles the pattern observed in object
voice, in which the agent is an argument of the verb, but contrasts with the
pattern observed in Indonesian passives, in which the optionality of the agent
suggests that the agent is an adjunct rather than an argument. 4.3. Ablaut marking object
registration We will turn next to an aspect of Kerinci
morphology that has no counterpart in Standard Indonesian, object registration
morphology. In many Kerinci varieties, roots from
diverse lexical categories exhibit two (or more) morphological
forms distinguished by the phonological shape of their final rime (see among
others: Usman 1978, Prentice & Usman 1978, Mckinnon
2011, Mckinnon, Cole and Hermon 2011). As the glosses
for the following examples suggest, so-called secondary forms (also referred to
as ‘oblique’ or ‘ablaut’ forms in the literature and glossed as O forms)
exhibit multiple functions, while the primary (basic forms) are glossed as A
forms, following the glossing conventions used in previous work on this topic.
Among other functions,
the occurrence of the secondary
form of the
root functions as morphological registration of the head/complement relationship, marking on the
head the presence of objects
and other nominal complements. The relationship between the head
and complement is similar to
that of agreement,
but (unlike agreement in e.g. Romance
languages) without co-variation in person/number-features. (The term ‘agreement’
rather than ‘nominal complement registration’ was employed in our earlier work.)
What is relevant
here is that
active verbs with nominal direct objects must occur
in the secondary form.
In
contrast, intransitive verbs and verbs with clausal objects occur in the
primary form.
With
this background, we turn now to the form of the verb in Kerinci
passives. In passives, like transitive active constructions, the verb must be
in the secondary form.
The
fact that passive verbs occur in the secondary form would be predicted if the
agent in di-passives, like the direct
object in the active, were an argument of the verb, but it is unexpected if the
agent is an adjunct. 5. Evidence that passive agent is not an argument in Kerinci. We have observed a number of ways that passive agents in Kerinci exhibit properties typical of arguments rather than
adjuncts. These include the grammatical properties observed in SI, but also
include additional properties not found in SI. For example, we saw that in SI di-passive agents are adjunct-like in
that they are optional constituents and can be realized as either noun phrases
or as prepositional phrases; in Kerinci this
constituent is obligatory and can only be a noun phrase, not a prepositional
phrase. This might appear to suggest that di-passive
agents in Kerinci not only resemble arguments, but
rather that they are arguments. This,
however, is not the case. There is evidence which, in our view, shows
definitively that the agent in the Kerinci di-passive must be analyzed as an
adjunct synchronically. Following
Arka and Manning (1998), if an agent is able to
antecede a reflexive, this provides definitive evidence for its status as a
core argument of the verb. As we saw in SI, the agent in the object voice
behaves as core argument because it can antecede reflexives in subject
position. In contrast the agent in the SI di-passive
cannot antecede a reflexive surface subject; thus it
is not a core argument of the verb. Analogous facts are found in Kerinci: the agent in the object voice can bind a reflexive
subject 48, while the agent in the di-passive
cannot 49. This means that the agent of the di-passive
agents does not retain core argument status.
We
therefore conclude that despite the fact that Kerinci
di-passive agents display multiple
properties associated with arguments, in contemporary Kerinci
these phrases are adjuncts rather than arguments. This leaves open the question
of how the argument-like characteristics are to be explained. 6.
A diachronic account To review the general
picture, we have shown that in both Standard Indonesian and in Kerinci, object voice agents are arguments by all criteria,
but di-passive agents display a
mixture of argument-like and adjunct-like characteristics. The characteristics
of agents in Standard Indonesian and Kerinci are
summarized in the following chart.
There
is strong structural evidence (based on the ability to antecede subject
reflexives and based on nominalization) that di-passive agents are synchronically adjuncts despite the fact that
they display a number of characteristics that would be more natural if they
were arguments. Such a mix of characteristics makes it difficult to develop a
consistent synchronic analysis of the di-passive,
especially in the context of a framework like Minimalism, that eschews the use
of language specific and construction specific rules. From a synchronic
perspective, the voice systems of both Standard Indonesian and Kerinci include a number of typologically unexpected
‘irregularities’ that do not follow from general principles and must be
memorized. While exceptions of all sorts are problematic for synchronic
analysis and may undercut the ability of the analyst to find interesting
synchronic generalizations, from a diachronic perspective
exceptions can be valuable tools for detecting traces of earlier stages
of the language. In particular, archaic irregularities can be more helpful than
contemporary regular forms for reconstructing the stages through which the present day language developed. We would like to argue that
the agent-like properties of both Standard Indonesian and Kerinci
constitute archaisms of the sort that provide an entrée into the syntax of
earlier stages of Malayic. If
the argument-like properties of the agent in the di-passive construction are viewed as archaisms, a promising
approach is to claim that at some previous stage of Malayic the di-passive agent did not simply display
some argument-like properties, but it was
an argument, presumably with the full range of characteristics seen in present
day object voice. Such a scenario would fit our general picture of how Malayic
languages developed because we know that the current mixed voice system found
in Indonesian type languages is a departure from the earlier Philippine-type
voice system still found in the Austronesian languages of the Philippines (and
in most of the languages found in Taiwan). Previous research, however, has not
shown concrete evidence for an earlier stage in which Indonesian languages (and
specifically Malayic) conformed to the Philippine voice system without the
admixture of a European type passive voice. Our claim is that the argument-like
properties of the di-passive in both
Standard Indonesian and in Kerinci constitute this
sort of concrete evidence. These archaisms, then, are the ‘missing link’ that
connects Indonesian type languages and the earlier Austronesian pattern. If
we are correct in interpreting the argument-like properties in Malayic as
indications of an earlier fully symmetrical voice system, it is useful to map
out in more detail how, under our account, some of the other characteristics of
the di-passive might have evolved. At
the earliest stage, the di-passive
was presumably part of an object voice paradigm. Stage 1 As
we envisage the system, at this stage the voice system of Malayic (or perhaps
some earlier ancestor language) looked roughly like that of today’s Toba Batak.
The core voice system consisted of an active indicated by the active prefix and
an object voice consisting of a paradigm of verbal prefixes agreeing with the
agent. The prefixes presumably corresponded to today’s, ku- ‘1’, kau- ‘2’ and di- ‘3’ and
were in effect agreement markers indicating the person of a postverbal nominal agent.[12] The
nominal agent was obligatory in the third person (with the usual caveats for
contexts in which the agent stood in an anaphoric relationship with an overt
nominal found earlier in the discourse). First and second person nominal agents
were optional and typically did not occur overtly, but
were understood on the basis of the first and second person prefixes. In
the object voice, the agent occurred immediately after the verb, and displayed
similar characteristics to contemporary object voice agents without displaying
any split between the properties of first/second person and third person
agents. Stage 2 The
second stage is roughly like Stage 1 except that clarificatory phrases
providing additional information regarding the agent were possible. These had
the following form.
In
examples following the form of (51), the prepositional phrase stood in an
appositional relationship with the pronominal nominal agent -nya.[13]
The structure proposed in (51) is still found in some Sumatran Malayic varieties.
Examples from Tapus and Rantau
Panjang are provided (though we would not wish to claim that overall these
varieties remain in Stage 2.):
(The cognate to -nya ‘3’ in
Indonesian is incorporated into the verb form is (53b) but not in (54a)).
The structure seen in (52) and (53) typically occurs when the
agent is third person in these languages. We assume that this is the case
because no clarification would be needed in the case of first
or second person agents. Thus a split occurs
between the first and second person agent construction, in which no post verbal
nominal agent and no clarificatory phrase is usually found, and the third
person agent construction, in which a post verbal agent is obligatory and an
additional clarificatory description of the agent is possible though not
required. Examples
similar to (52) and (53) are found not only in contemporary Sumatran dialects but
also in Classical Malay. For example, the following examples were found by
searching the Malay Concordance Project database (http://mcp.anu.edu.au), a
corpus of Malay texts currently containing 165 texts and 5.8 million words.
According
to our analysis, at Stage 2 the construction is still an object voice
construction. The pronoun -nya (or its counterpart) is an obligatory argument
constituting the agent, and the prepositional phrase is simply appositional to
-nya. Stage 3 In
the third stage speakers reanalyze the prepositional phrase as the agent rather
than as appositional. The pronoun -nya (or its analog) is interpreted as a linker (i.e. as an
element that is not meaningful in this consruction)
rather than as the agent.
In
structure 56 the agent is in a prepositional phrase rather than a noun phrase.
On the assumption that the default interpretation for prepositional phrases
modifying the VP is adjunct rather than argument, the prepositional phrase
agent in (55) is interpreted as an adjunct, and thus, the construction exhibits
the key characteristic of a passive construction.
This
construction co-exists with a second construction in which -nya does not occur (i.e. -nya is optional).
Stage 3+ We
would suggest that subsequent to the reinterpretation of (55) as a passive
construction, by analogy all di-Verb-Stem
constructions were reinterpreted as passives; that is the NP agent in (56) was
reanalyzed as being an adjunct. Thus, in contemporary Sumatran varieties, as in
Standard Indonesian, sentences with the form of 57 manifest key grammatical
properties of a passive voice construction rather than those of an object voice
construction. Stage 4 In
this stage the use of nya-
as a linker was dropped in some varieties (like SI). Some traditional Sumatran
Malayic languages (like Tapus and Rantau
Panjang) did not undergo this change, and continued to
maintain the use of a cognate of –nya with PPs. With the exception of dropping -nya as a linker,
the construction otherwise remained essentially the same, resulting in the
contemporary Standard Indonesian passive voice. 7.
Further predictions of the analysis Our historical
proposal accounts for several peculiar and previously unexplained properties of
the di-passive construction. These
include the argument-like properties of di-passive agents, the existence of two
agent ‘constructions’ (i.e. PP and NP agents), the existence of a pronominal
‘linker’ (non-meaningful use of -nya) in some traditional Malay varieties (e.g. Tapus and Rantau Panjang) and the
fact that di-passive and object voice
constructions are in near complementary distribution in the standard language
and in full complementary distribution in Kerinci. We would now like to discuss two additional
predictions that follow from our analysis. The first prediction is that, in some Malayic varieties, Stage 3+
might not have occurred. In particular, the analogical process, whereby non-PP
agents were reanalyzed as adjuncts, may not have applied uniformly to all NP
agents in di- passives. In such cases
what in Standard Indonesian are instances of NP agent passives might have
remained object voice constructions. In such a variety, it would be expected
that at least some NP passive agents would be arguments, and therefore would be
able to serve as antecedents for reflexive subjects. This prediction turns out
to be correct for pronominal agents with -nya in Standard Indonesian, as
was shown by Arka and Manning.
Similar facts are true for Sumatran Malayic
languages like Kerinci as well.
These examples show that there exists a
subset of di-passives that remain
instances of object voice. This state of affairs is not unexpected if di-passives derive from an earlier
object voice construction, but it is unexplained on other accounts. Our account makes a second prediction with
regard to the types of agents which are possible in object voice. Recall that
one claim of our account is that at Stage 1 the person marked agent forms that
precede the verb in object voice (the analogs of ku-, kau- and di- in
contemporary Standard Indonesian) were historically prefixes or clitics, rather
than free pronouns. In modern-day Malay varieties, we find considerable
variation regarding which forms may appear as agent in the object voice. In
conservative Kerinci varieties, as we have seen, only
1st and 2nd person pronouns are permitted as agents in
object voice. Non-pronominal forms, as well as 3rd person pronouns
are not permitted in this position.
However, unlike the
construction we posit at Stage I, Kerinci does allow
full (non-clitic) pronouns to occur as agents in object voice.
Many other varieties exhibit fewer
restrictions regarding which forms can occur as agent in object voice. For
example, in contemporary Indonesian, unlike European languages, first and
second person nominals are not restricted to a closed, pronominal word class.
Rather, in context, proper names can be used freely to refer to the speaker and
addressee (examples adapted from Sneddon 1996), as seen in the following active
sentences.
Both proper names and free pronouns are
grammatical in Standard Indonesian object voice.
These facts suggest that in contemporary,
Standard Indonesian, the correct synchronic analysis is one in which Tini, saya, and ku- occupy a syntactic position to the left of the
verb phrase, for example, the position of specifier of vP in some contemporary models. In other words, Standard Indonesian and (to a
lesser extent) conservative Sumatran varieties like Kerinci exhibit an object voice construction in which restrictions on agents have been
‘relaxed’ from the earlier state of affairs, as illustrated below.
From our historical account it follows that
varieties which exhibit less restrictions on the properties of object voice
agents developed from an earlier, more conservative variety in which agents
were clitic or affix agreement markers. Thus, we predict that varieties which
allow full pronominal forms will also permit clitic pronominals.
We also predict that varieties which allow full NP agents in object voice, will
also permit full pronouns. As far as we know, both of these predictions are
borne out: It appears to be the case that there are conservative village
varieties which restrict the preverbal form to the clitic,
but there are no varieties (as far as we know) in which only the full form of
the pronoun is used and the clitic form is excluded.
Likewise, we are not aware of any varieties in which NP agents are permitted
but which do not allow full pronouns. 8. Conclusions We would like to review what we view to be the advantages of
our analysis of the origin of the di-passive
in Malayic. First of all, our account explains why there should be a split
between, on the one hand, first and second person agents, that remained arguments
in modern varieties of Malayic, and third person agents that have become
adjuncts. Our analysis claims that the passive developed for third person
agents because only the third person was easily compatible with a clarificatory
by-phrase, which later was
reinterpreted as a passive agent, transforming the structure from an object
voice construction to a passive. No similar clarificatory phrase occurs easily
with first and second person agents, so no change took
place for first and second person agents. Secondly,
the analysis explains why agents in the passive construction have a variety of
properties typical of arguments rather than adjuncts. These are properties
retained from an earlier stage in the development of the construction; in
effect, archaisms. Third,
the conclusions drawn in this paper on the basis of the internal reconstruction
of Malayic and the comparison of Malayic varieties are broadly compatible with
comparative analyses of Western Malayo-Polynesian languages carried out by Kikusawa (2012). We shall not review the arguments
presented by Kikusawa, but that work makes proposals
similar to those found
here to explain the relationship between passive and ergative marking in a
variety of Austronesian langauges. If Kikusawa’s proposals are correct, it would add plausibility
to the specific claims made here. To
conclude, the sequence of changes proposed here illustrates what we consider to
be a plausible path toward the acquisition of a passive by a previously
symmetrical voice system. To the best of
our knowledge, no alternative explanations have been proposed for the seemingly
chaotic list of properties associated with the passive and object voice
constructions in Malayic languages. While it is possible that future
alternatives to our analysis will be proposed, the current analysis at the
least demarcates the grammatical characteristics of Malayic that demand
explanation and proposes a series of changes that constitute a plausible
historical account for the characteristics. We would like to emphasize that the
study of local Sumatran Malayic varieties played a key role in our analysis. It
is important to understand that there are in fact hundreds of such languages
that have not been described at all. Our expectation is that if our analysis is
correct, the future study of these varieties will fill in still unattested
steps between object voice and passive. For instance, our analysis predicts
that we might find a dialect in which nominal agents in the di- construction have all the properties
of arguments, while PP agents will be adjuncts. Another
important issue is what light Malayic can shed on other Austronesian languages
spoken in Indonesia. Passive constructions with characteristics like those we
have described in Malayic are found outside of Malayic (e.g. Balinese, Batak,
etc.). This raises many new questions that are beyond the scope of this paper. Assuming
the correctness of our analysis for Malayic, did the changes discussed here
take place in Malayic or in some earlier common ancestor of the set of
languages displaying Indonesian type voice systems? What is the role of
language contact and grammatical borrowing in the spread of this construction?
More generally, what are the consequences of these proposals for the genetic
grouping of Indonesian type languages? We hope that an investigation of these
questions in non-Malayic Indonesian-type languages will assist us in finding
the answers to these questions. References ______.
Malay concordance project database.
Online: http://mcp.anu.edu.au. Adelaar,
Alexander. 2005. Much ado about di-. Bijdgragen tot de Taal-,
Land- en Volkenkunde 161.127-142.
Online: http://www.jstor.org/stable/pdf/27868203.pdf. Arka,
I Wayan, and Christopher D. Manning. 1998. Voice and grammatical relations in
Indonesian: A new perspective. Proceedings of the LFG98 Conference, ed.
by Mirriam Butt and Tracy Holloway King, 1-17.
Stanford, CA: CSLI Publications. Online:
http://web.stanford.edu/group/cslipublications/cslipublications/LFG/LFG3-1998/lfg98arkamanning.pdf. Arka,
I Wayan. 2003. Voice systems in the Austronesian
languages of Nusantara: Typology, symmetricality and undergoer orientation. Linguistik Indonesia 21.113-139. Chung,
Sandra. 1976. On the subject of two passives in Indonesian. Subject
and topic, ed. by Charles Li, 57-98. New York: Academic Press. Online: https://www.researchgate.net/publication/240154392_On_the_subject_of_two_passives_in_Indonesian. Cole,
Peter and Gabriella Hermon. 2008. VP Raising in a VOS Language. Syntax
11: 144 - 197. Comrie, Bernard. 1988. Passive and voice. Passive and voice, ed. by Masayoshi Shibatani,
9-24. Amsterdam: John Benjamins. Dardjowidjojo, Soenjono. 1978. Sentence patterns of Indonesian.
Honolulu: University of Hawaii Press. Foley,
William A. 1998. Symmetrical Voice systems and precategoriality
in Philippine languages. Paper presented at the Workshop on Voice and
Grammatical Functions in Austronesian Languages, LFG '98 conference, Brisbane,
Australia. Online:
http://www.sultry.arts.usyd.edu.au/LFG98/austro/download/download.htm. Himmelmann, Nikolaus P. 2005. The
Austronesian languages of Asia and Madagascar: typological
characteristics. The Austronesian
languages of Asia and Madagascar, ed. by Karl Alexander
Adelaar and
Nikolaus Himmelmann,
110-181. London and New York: Routledge. Online:
http://www.ling.nthu.edu.tw/download/himmelmann-2005b.pdf. Keenan,
Edward L., and Bernard Comrie. 1977. Noun
phrase accessibility and universal grammar. Linguistic Inquiry 8.63-99. Online:
http://lingo.stanford.edu/sag/L222B/papers/KeenanComrie.pdf. Kikusawa, Ritsuko.
2012. Optional ergative marking and the emergence of passive structures in
Austronesian languages. Objectivization
and subjectivization: A typology of voice systems, ed.
by Wataru
Nakamura and Ritsuko Kikusawa,
115-155. Osaka: National Museum of Ethology. Online:
https://minpaku.repo.nii.ac.jp/index.php?action=pages_view_main&active_action=repository_action_common_download&item_id=2540&item_no=1&attribute_id=18&file_no=1&page_id=13&block_id=21. Kiparsky, Paul. 2013. Towards a null theory of the passive. Lingua 125.7-33.
Online: https://www.researchgate.net/publication/256806528_Towards_a_null_theory_of_the_passive. Mckinnon, Timothy A. 2011. The
morphophonology and morphosyntax oj Kerinci word shape alternations. Newark: University of
Delaware dissertation. Mckinnon, Timothy A., Peter Cole and Gabriella
Hermon. 2011. Object agreement and 'pro-drop' in Kerinci Malay. Language
715-730. Perlmutter, David, and Paul Postal. 1984. The 1-advancement
exclusiveness law. Studies in relational grammar 2,
ed. by David Perlmutter and Carol Rosen, 81-125. Chicago: University of Chicago
Press. Online: https://www.researchgate.net/publication/244443341_The_1Advancement_Exclusiveness_Law. Prentice, David John, and Amir Hakim
Usman. 1978. Kerinci sound-changes and phonotactics. Second
International Conference on Austronesian Linguistics: Proceedings, Fascicle I:
Western Austronesian (Pasific Linguistics C-61), ed. by Stephen A. Wurm
and Lois Carrington, 121-163. Canberra: Australian National University. Roeper, Thomas, and Angeliek van Hout. 1999. The impact of nominalization on passive,-able and middle: Burzio’s
generalization and feature-movement in the lexicon. MIT Working papers in
Linguistics 35.185-211.
Online:
https://www.researchgate.net/publication/46596875_The_Impact_of_Nominalization_on_Passive-able_and_Middle_Burzio's_Generalization_and_Feature-MNovement_in_the_Lexicon. Sneddon, James Neil. 1996. Indonesian: A comprehensive grammar. London
and New York: Routledge. Sneddon, James Neil. 2003. The Indonesian language: Its history and role in modern society.
Sydney: UNSW Press. Sneddon, James Neil. 2006. Colloquial Jakartan Indonesian. Canberra:
Pacific Linguistics. Sneddon, James Neil; Alexander Adelaar;
and Dwi Noverini Djenar. 2010. Indonesian
reference grammar (2nd Edition). New South Wales: Allen &
Unwin. Usman,
Amir Hakim. 1978. Struktur bahasa Kerinci, dialek Sungai Penuh: Fonologi, morphologi, and sintaksis: Laporan penelitian. Jakarta:
Pusat Pembinaan dan Pengembangan Bahasa, Departement
Pendidikan dan Kebudayaan. van den Berg, René. 2004.
Some notes on the origin of Malay di-. Bijdragen tot de Taal-, Land- en Volkenkunde 160.532-554. Online: Wunderlich, Dieter. 2012. Operations on argument structure. Semantics: An International Handbook of
Natural Language Meaning (Vol 3), ed. by Claudia Maienborn,
Klaus von Heusinger, and Paul Portner,
2224-2259. Berlin: Mouton de Gruyter. Wolff, John U. 2001. A historical account of the origin of
the Malay verbal affix di-nya. Paper presented at the
5th International Symposium on Malay/Indonesian Linguistics,
Leipzig, Germany. [1]These varieties of colloquial
Malayic, which are spoken in small villages in the so-called ‘Malay Heartland’,
the area around the border between Jambi Province and West Sumatra. It is from
this region that many scholars believe Malayic languages initially proliferated
(e.g. Sneddon 2003), developing into prominent regional languages, and
eventually spreading to the outer reaches of the Indonesian Archipelago, as the
official national languages of Malaysia, Indonesia, Singapore, and Brunei Darussalam.
The varieties that are the focus of this paper are to be distinguished from the
colloquial Malay/Indonesian spoken in major cities of Indonesia like Jakarta,
which result from a combination of Batawi Malay,
local languages like Javanese and Sundanese, the southern Chinese varieties
spoken in Jakarta, and so forth. [2]A notable exception of great
diachronic interest is Rukai, which does not exhibit
the complex voice system found in other Formosan Austronesian languages. [3]Some linguists working in the
framework of Lexical Functional Grammar categorize the passive agent as an
oblique non-core argument. See Arka (2003) inter alia. [4]Other names for this construction
are ‘patient focus’, ‘objective voice’ and, in the case of Indonesian, Passive
2 and passif semu (Dardjowidjojo 1978). To some extent the issue is more than
terminological. The choice of voice does not seem to be connected in any way
with ‘focus’ as the term is typically used by linguists, so the term ‘focus’
seems misleading, and ‘passive’, which we in fact used in earlier works, is
misleading in its own way, so we hope that the term ‘object voice’ will become
standard. [5]According to Larson (1988), both arguments
and adjuncts are structurally complements. Adjuncts
are analyzed as ‘lowest complements’, i.e. for Larson these adjuncts are lower
in the tree than direct and indirect objects. [6]See Mckinnon,
Cole and Hermon (2011) for a discussion of the motivation in Malayic for this
and similar Larsonian structure. [7]There is one exception to this
generalization with respect to di-passives,
which will be discussed below. [8]This is a slight over
simplification. In Standard (or near Standard) Indonesian object voice is
possible with third person pronominal agents (as well as first and second
person agents) while di-passives are
grammatical with all third person agents, regardless of whether they are
pronominal. Thus, the distribution in Standard Indonesian is a near
complementary distribution. We shall argue below that the (near)
complementarity has a historical rather than a synchronic explanation. [9]We have in fact seen one such case
already. In Kerinci object voice and di-passive are in full complementary
distribution, while in Standard Indonesian they are in near complementary
distribution. We shall return to this topic below. [10]A postverbal
agent would be ungrammatical in (37a) as well. [11]It is not completely ungrammatical
to omit a third person agent, but the absence of an overt agent is interpreted
as having a phonologically third person null personal pronoun (‘him’, ‘her’) as
agent. This is in fact an instance of a more general process of ‘prodrop’: the
omission of a noun phrase is interpreted as though there were a third person
referential pronoun present . In contrast, in SI the
absence of an explicit agent in Indonesian is interpreted as an indefinite
agent, similarly to English, e.g. the
book has been sold around the world. See Mckinnon,
Cole and Hermon (2011) for details. [12]As a matter of convenience
we refer to the prefix in question as di- and so
forth, but in fact we
take no position on whether for example the morpheme at that time was di-, ni- or yet another form, with the
same caveat with respect to the first and second person prefixes. There are a
variety of proposals in the literature regarding the etymology of the
modern-day prefix di- (for an brief overview of these proposals see Adelaar
2005). We do not take any position on this issue, since we see the etymology of
the prefix di- as an issue which is
orthogonal to our claim that the di-passive
originated from an object voice construction containing a 3rd person
agent. [13]The analogous morpheme in many
varieties is realized as -e rather
than -nya.
Our interest is in the syntactic structure rather than in reconstructing the
morpheme in question. | | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||