Professional Documents
Culture Documents
WS 2013/14
Hauptseminar "Grammaticalization"
bei Martine Bruil, M.A.
Lukas Denk
M.A. Allgemeine und vergleichende Sprachwissenschaft
Matrikelnummer: 1509398
Content
Abbreviations....................................................................................................................................... 2
Introduction.......................................................................................................................................... 3
1 Form.................................................................................................................................................. 3
1.1 Definition................................................................................................................................... 3
1.2 Valence-Change and voice.........................................................................................................4
1.2.1 Antipassive and passives.................................................................................................... 5
1.2.2 Antipassive and deobjectives............................................................................................. 6
1.2.3 Morphological difference between valence changing and voice categories......................6
2 Distribution........................................................................................................................................6
2.1 Geographical Correlation...........................................................................................................7
2.2 Correlation with alignment........................................................................................................ 7
3 Function and functional correlates.................................................................................................... 8
3.1 Semantic/pragmatic function..................................................................................................... 9
3.2 Functional correlates................................................................................................................10
3.2.1 Aspectual correlation........................................................................................................10
3.2.2 Meaning shift....................................................................................................................11
3.2.3 Counterfactualness........................................................................................................... 11
3.3 Structural function................................................................................................................... 12
3.3.1. Feeding the pivot.............................................................................................................12
3.3.2 Relative clauses................................................................................................................13
4 Grammaticalization......................................................................................................................... 14
4.1 Sources ....................................................................................................................................14
4.1.1 Reflexives ........................................................................................................................14
4.1.2 Do/ make.......................................................................................................................... 17
4.1.3 Denominatives (Rgyalrong)............................................................................................. 18
4.2 Inner-antipassive grammaticalization: pragmaticalization and proper grammaticalization ...19
4.3 Further development: split-alignment......................................................................................21
5 Summary: Grammaticalization map: diachronic and synchronic correlation of functions.............23
References.......................................................................................................................................... 25
1
Abbreviations
1 first person
2 second person
3 third person
A agentive; subject of the transitive clause
ABS absolutive case
ACC accusative
AP antipassive
BEN beneficiary
CAUS causative
CLI nominal class N°1
COMPL completive
DAT dative case
DEOBJ deobjective
DEF definite
ERG ergative case
EVD evidential
FIN final vowel
FOC focus
FUT future
HUM human
INDEF indefinite
IMPERF imperfective
INST intrumental case
INTERR interrogative
INTR intransitive
IO recipient; indirect object
LOC locative
NF non-future
NHUM non-human
NOM nominative case
NP nominal phrase
O objective, patient; direct object of a transitive clause
OBL oblique case
PASS passive
PERF perfective
PL plural
POSS possessive
PRES present
PST past
PV preverb
REFL reflexive
S single argument; subject of an intransitive clause.
SG singular
TAM tense/aspect/mood
TOP topic
TR transitive
V verb
VTR transitive verb
VINTR intransitive verb
2
Introduction
The antipassive is a voice-construction which derives transitive sentences to propositional identical
intransitive ones, treating the actor as the single argument and omitting the undergoer of the action,
or expressing the undergoer by a peripheral construction.
The antipassive construction may have several functions: focussing the actor, backgrounding the
undergoer, aspectual function, lexical function and others.
The main aim of this paper is to present some grammaticalization paths of antipassives. In particu-
lar, I want to examine the genesis and development of antipassives, that means, the diachronic link
to other functions and categories.
By introducing the functions and functional correlates, I hope to contribute in the theory of gram -
maticalization paths for antipassives, since what is synchronically related, may be also diachroni-
cally "near".
In this paper, various examples of different languages are presented to hypothesize a general nature
and development of antipassives. Of course, I won't list all the phenomena around antipassive, but
should pick interesting clues to build a preliminary sketch. This sketch is a closed representation
within the boundaries of this paper, but can be seen as a starting point to add more links and path -
ways through future research.
1 Form
1.1 Definition
We shall begin with a wide-accepted, concrete, but strict definition of antipassives (Dixon 1994:
146)
(1) Antipassive
(a) applies to an underlying transitive clause and forms a derived intransitive;
(b) the underlying A NP becomes S of the antipassive;
(c) the underlying O NP goes into a peripheral function, being marked by a noncore case, preposition, etc.; this
NP can be omitted, although there is always the option of including it;
(d) there is some explicit formal marking of an antipassive construction (same preference and possibilities as
for passive).
Dixon's (1994) definition is based on the distinction of fundamental grammatical relations (A, S,
O). These relations are a generalization of semantic and syntactic criteria: S stands for the single ar-
gument in an intransitive clause (i.e. the single argument of an antipassive construction) and does
not reveal whether the argument is an actor or an undergoer of the event; A (agentive, agent) and O
3
(objective, patient) are restricted to transitive clauses and are connected to semantics, since A stands
for the actor of the event and O for the undergoer. Syntactically, S is the intransitive subject, A the
transitive subject and O the transitive object of the clause.
If we want to paraphrase the operation in english, we have to search a translation, where the derived
object is followed by a preposition: The man hit at the woman. Nevertheless, this english sentence
can't be seen as an antipassive construction, since it lacks an antipassive morpheme which would in -
transitivize the clause.
4
Table 1: Operations affecting valence of (in)transitive clauses
A O
adding CAUSATIVE APPLICATIVE
backgrounding PASSIVE ANTIPASSIVE
removing ANTICAUSATIVE DEOBJECTIVE
merging REFLEXIVE
Describing the antipassive merely by Table 1, the main operation done to transitive clauses through
antipassivation would be "backgrounding of the O".
In the semantic sense, the term 'anti-passive' points to a concept opposed to the passive. The S of a
passive construction denotes an undergoer (O) and the peripheral argument denotes an actor (A),
whereas the S of an antipassive construction denotes an actor (A) and the peripheral argument de -
notes an undergoer (O). This functional symmetry is also maintained in the definition in Dixon
(1994: 146), where all the qualities which define the antipassive also define the passive by merely
interchanging the A and O function (compare examples 1 and 4):
(4) Passive
(a) applies to an underlying transitive clause and forms a derived intransitive;
(b) the underlying O NP becomes S of the passive;
(c) the underlying A NP goes into a peripheral function, being marked by a noncore case, preposition, etc.; this
NP can be omitted, although there is always the option of including it;
(d) there is some explicit formal marking of a passive construction (generally, by a verbal affix or else by a pe -
riphrastic element in the verb phrase – such as English be...-en – although it could be marked elsehwere in
the clause)
With the further analyzing of the form and function of antipassives I will include some comments
about whether the relation of antipassive and passive construction is balanced, or if there are some
asymmetries regarding the distribution, usage, functional correlates and especially their diachronic
evolution.
5
1.2.2 Antipassive and deobjectives
Another construction close to antipassive are the so-called deobjectives. The deobjective is a (mor-
phological) device which causes the deletion of the O:
b) I-ku-an.
DEOBJ-drink-1SG.INTR
‘I drink.’
Here, the causative morpheme (valence change) is nearer to the stem than the passive one.
2 Distribution
Among causatives, applicatives and passives, antipassives are the least distributed valence-affecting
category2. Table 2 shows valence-affecting categories /voice systems found in 174 languages of the
2 Nichols does not distinguish between voice and valence-change, so antipassive and deobjectives may be subsumed
6
world (Nichols 1992).
Antipassives are frequent in Australia, North-America and the Caucasus, but the generalization is
quite imprecise. If we want to find a correlate for the construction, geographical parameters (like a
linguistic area) turn out to be less prolific.
7
guages should have a device to treat the A (marked argument) as S (unmarked argument).
Cooreman (1994:50) states: "The antipassive is a construction typical for ergative languages and oc -
curs along with ergative constructions as a morphosyntactic alternative for the same transitive
proposition."
Her "functional typology of antipassives" (Cooremann 1994) is indeed based solely on ergative lan-
guages (19 in all) although she admits that "de-transitivized constructions from accusative lan-
guages (...) have also been characterized as antipassives by other scholars (Givón 1984; Health
1976; and Postal 1997)" (Cooremann 1994:82).
Nichols (1994) calculated statistically this question. She admits that the most strong correlation of
voice is with alignment, but she discovers that the antipassive construction is more related to object-
centring systems (such as agreement morphemes on the verb) than with alignment (Nichols
1994:154):
"Generalizing over these patterns, we can state that antipassives are associated with languages in which the O
is central to clause morphosyntax: ergative languages, where the O takes the unmarked case, and those accu -
sative languages, in which there is agreement with the direct object (=O). It is thus observationally true, as
traditionally maintained, that the antipassive is associated with ergativity; but the causal factor is evidently
not the ergative alignment per se but the special structural status of the O which is basic to ergativity".
Although passive-constructions seem to be rare in ergative languages (cf. Nichols 1994:153) such
as antipassives in accusative languages, the main connection between voice and other structural
phenomena has to be found in the particular treatment of A and O.
In the last two sections we have dealt with how an antipassive construction is built up, how it is for -
mally derived (example 1) and which structural correlates can be pointed out statistically (section
2). This section deals with the functional justification for the construction; the antipassive's function
will be referred to pragmatic, semantic or syntactic motivations.
Furthermore we will look at how many functions can be expressed in one language by an antipas -
sive construction. It is important to consider these functional correlates in order to make assump-
tions about the diachronic development of antipassives. The relation between synchronic functions
and grammaticalization paths will be clarified in section 4.
I will categorize the functions based on Cooreman (1994) into semantic/pragmatic and structural
function.
8
3.1 Semantic/pragmatic function
The antipassive function can be seen as a sum of other functions. In its pure pragmatical use, the
proposition (semantic content) of an active clause should not be changed by its antipassive deriva-
tion. Then, the antipassive construction is used to foreground the actor of the event and to back -
ground the undergoer. However, Cooreman (1994) notes that every antipassive operation also goes
along with a semantic change on the content. She subsumes her observations in a semantic/prag-
matic definition of antipassives (Cooreman 1994:51):
"The antipassive which is used for semantic/pragmatic reasons is best described as indicating a certain degree
of difficulty with which an effect stemming from an activity by A on an identifiable O can be recognized."
Based on this "certain degree of difficulty" between the A and the O, antipassive constructions
cover both pragmatic and semantic functions. As an example for pragmatic motivations, the antipas -
sive serves to focus the A of the proposition. This 'focus-antipassive' is found in Mayan languages
and even functions when A is questioned3:
Close to the focussing function of the A is also the non-specificity of the O. As Cooreman (1994:
52) states, the antipassive is used when the O is more or less unidentifiable. In Mam, another Mayan
language, "the O must be unknown, implied, or non-specific and may not be explicitly referred to in
the clause". A similar case holds for Bezti. In the antipassive construction, the O is often absent, but
overtly, it can only be expressed in the plural instrumental, and has to be non-specific (cf. Coore-
man 1994: 53):
b) is Xorilarad ̃xol̃cʰa
brother.ABS sheep.PL.INST cut.AP
'Brother occupies himself cutting up sheep.'
3 In (7), the O is omitted, and in (8) it appears as an unmodified nominal phrase. This antipassive construction would
then not be prototypical according to Dixon (1), as the O is not expressed by peripheral constructions (preposition,
oblique case marking).
9
c) *is Xorid ̃sol̃cʰa
brother.ABS sheep.SG.INTR cut.AP
'Brother occupies himself cutting up a sheep.'
Cooreman (1994:56) explains why the antipassive and deobjective are very closed within the prag-
matic/semantic function:
"When an O is indefinite, non-specific, generic, or even non-referential, its identity is of relatively low im -
portance to the discourse, and hence the O is easily deleted."
Cooreman (1994:69) explains that the strong correlation between antipassive and imperfective has
to do with the low identifiability of the O, since
"in events which have no discernable point of beginning or end, the focus of attention tends to be more on the
process, or the activity itself as described by the verb, rather than on its effect on the O, or even on the iden-
tity of the O. Since it is the activity itself which is more important, the identification of the O is marginal for
the purpose of the discourse and is relegated to the background."
In my opinion, focussing the A has a lot to do with focussing the action, since in the imperfective, in
the 'unaccomplished' action, the A, as the bearer or the initiator of the event, is the only or first ac -
tant to refer to.
10
We can compare the closeness between antipassives and imperfective aspect to the closeness be-
tween passives and perfective aspect. An example could give us Latin. The voice category of par-
ticiples is bound to that of aspect/tense. There is either the 'perfect passive participle' or the 'imper-
fective active participle', but no 'perfect active participle' and no 'imperfective passive participle'.
We could explain this by applying analogy to the other explanation: the view of an event as "perfec-
tive" could be connected to the result of the action, and the affected argument (O) is nearer to the re -
sult (in time and space) than the A. Diachronic evidence for shifts between voice and aspect (section
4) will reinforce this hypothesis.
In K'iche', the antipassive morpheme can be used in order to derive a new lexical meaning for the
verb: 'to drop' shifts to 'to abort', 'to wash' can shift to 'to wash clothes' or 'to menstruate' and 'to hit'
becomes 'to fight' (cf. Cooreman 1994:58). Those shifts go along with an imperfective connotation
of the event (cf. loc. cit.).
3.2.3 Counterfactualness
The functional domain of the antipassive in Yukulta is quite differentiated:
Cooreman (cf. loc. cit) speaks about 'counterfactualness' as the motivation for the several functions
covered by the antipassive in Yukulta. The link between counterfactualness and the main
pragmatic/semantic function of antipassives lies in the "reduction in realization potential" (cf.
Cooreman 1994:63). So by negating a non-past event, there is possibility for it to happen in later
time. The same holds for the 'personal desire' of the speaker expressed by the 'irrealis desideratives',
11
where the accomplishment of the event is less foregrounded than the wish of the event to take place.
When we connect the 'reduction in realization potential" with the aspectual connotations posed in
3.2.1, we might build a chain reaching to one core function of the antipassive, namely to mark the
unidentifiability of the O. The 'reduction in realization potential' implies that the event is not seen as
accomplished but understood as more "imperfective", and imperfective events are less O-centered
than perfective events.
In (12), we have a transitive clause conjoined with an intransitive clause. That the omitted S in the
intransitive clause is correferent with Oscar and not with Bill is due to the 'syntactic pivot' A/S (Fo-
ley and Van Valin 1984), which allows A to be correferent only with A or S in complex sentences.
The syntactic pivot is build up from those grammatical relations which are treated the same. In Eng-
lish, the syntactic pivot is S and A, since those functions are treated the same (syntactically preced-
ing the verb) and O is treated otherwise (syntactically: following the verb).
If, however, we want to clarify that the omitted/pronominal S-Element is not correferential with the
A of the former clause, but with O, then English has to apply a passive construction:
12
(13) Bill was hit by Oscar and Ø(/he) went away.
O(S) V A(Periphery) S V
*Oscar hit Bill and OSCAR went away
= Oscar hit Bill and BILL went away.
In (13), the O is treated as an S to "feed" the pivot – correference is therefore only allowed through
syntactic pivots.
The same holds even for some languages whose syntax patterns ergatively, i.e., where the syntactic
pivot is S/O. Here, an antipassive construction is needed to feed the pivot in favor of refer-
ence-tracking.
Since in (14c), the omitted correferent 'people' is in A function and hence not part of the syntactic
pivot S/O, the antipassive construction changes the participant from function A to S in order to
make it correferential with O. In this sense, the structural antipassive in syntactic ergative languages
(i.e. with pivot S/O) is merely the mirror-technique to the passive we use in a syntactic accusative
language like English.
A structural phenomenon which only holds for ergative languages is the use of the antipassive to
modify the A by an antipassive construction. In contrast, the analogical restriction, where the O has
to be relativized by a passive construction, hasn't been observed in accusative languages. We can
see an example from K'iche', where the antipassive is compulsory for the relativization of A:
13
The reason why relativized As have to be antipassivized in ergative languages but relativized Os are
not passivized in accusative languages can be explained by the 'Absolutive Hypothesis (Fox 1987,
cf. Cooreman 1994:80 f.): Relative clauses are normally restricted to the S/O pivot, but as the accu-
sative languages have in addition the S/A pivot, the unmarked A is easy to relativize. In ergative
languages, the syntactic pivot S/O coincides with the natural S/O pivot to build up relative clauses,
and the marked A stands alone as an over-marked referent, which is less accessible for relativizing.
The antipassive construction allows the A to behave like S in order to maintain the syntactic pivot
S/O which is the accessible pivot for relative clauses.
4 Grammaticalization
We finally come to our main question: How can antipassive constructions come into being, which
lexical items are often used especially to form antipassive meaning, what is the further develop-
ment?
With describing the domain in which antipassives can be used, we have already shown which func-
tions correlate with the prototypical antipassive function. The idea is that the already described syn-
chronic functional scope can represent possible pathways of change. Especially the relation between
the pragmatic/semantic and the structural antipassive is even a diachronic one. In this section, I will
not present all the possible sources which have established antipassives, but concentrate at few ones
in order to focus on the further development of antipassives and the diachronic differentiation of
their functions.
4.1 Sources
4.1.1 Reflexives
The reflexive voice has been claimed to be one of the most frequent sources for the antipassive. In
many languages of Australia we find homophonous morphological marking for antipassives and re-
flexives (cf. Terill 1997:71). Consider following sentences:
14
bayi djaŋgayma-rri-nju
CLI.NOM eat-REFL-NFUT
'he eats [someone/thing]' OR 'he eats himself'
(Dixon 1972:90)
Although there are no historical documents which could prove the pathway from reflexive to an-
tipassive, Terill (1997:84) argues that the hypothesis, that the reflexive could have "developed from
out of a chronologically and functionally prior antipassive" is "less plausible", because
"the reflexive construction is extremely widespread in Australia, while the antipassive is not. Furthermore,
there are no languages with an antipassive marked by verbal morphology which do not also have a reflexive
marked by verbal morphology."
The theory behind the development is that first, in the prototypical reflexive, A and O are correfer-
ential. That causes the clause to be reanalyzed not more as a transitive clause, but as an intransitive,
exhibiting only one argument because there is only one referent. Once intransitivized, the speaker
can chose whether the S takes the semantics of the A or the O. We deduce, that if the S is A-cen-
tered, we could obtain a construction similar to the deobjective, with the former reflexive mor-
pheme (which has been correferential with O and A) representing hence the deobjective morpheme.
Up to then, the clause is understood as an event which has got an initiator or bearer, but no result or
affected participant. Therefore, a new O can be introduced, but the intransitive structure of the verb
does not allow this O to be treated like a core argument, the direct object, but instead to be ex-
pressed by a peripheral construction (preposition, oblique case). The former intransitive deobjective
morpheme is then reanalyzed as an antipassive morpheme.
Terrill visualizes the function change of the morphology X, whose formal shape remains untouched:
15
(18) proposed pathway of change from reflexive to antipassive (Terill 1997:83)
construction A
===> morphology - X
construction A construction A/B function-reflexive
morphology -X ===> morphology -X
function- reflexive functions- reflexive
- antipassive
construction B
===> morphology- X
function- antipassive
Terrill (1997:84) explains the coexistence of new and older functions by reasoning that "the new
construction has a separate function from the source function, and therefore the new construction
does not threaten the existence of the old construction."
A reflexive source for an antipassive/ deobjective construction is also found in Slavic languages,
and also in Lithuanian:
Reflexive sources are not only develop into antipassive constructions, but can, in contrast, just as
well become passives. Spanish is a language, where the reflexive se4 expanded its function to form
passives:
(20) Spanish
(a) Se lava.
REFL.3SG wash.3SG
‘S/he washes him/herself.' 'it is being washed.'
The motivation for reanalyzing the reflexive as a passive is situated by the same base as I have de-
4 Genetically related to si in Lithuanian, Latin se, Indo-European *-s(w)e. However, the Lithuanian suffix -si does
not agree in Person like the Spanish reflexive pronouns/clitics (me, te, se).
16
ducted for antipassives: Once the clause is understood as intransitive, the single argument has to
take a semantic role: in the case of the passive it is the S with the semantics of O (In the antipassive
it is the S with the semantics of A).
b) Sòrô-n pátá-ndì.
people.PL-DEF cut-AP
‘The people harvested (the crops).’
In (22), we find a quite similar suffix ndí (high tone) which marks the causative:
The source of this suffix which covers two very different functions can be traced back to the lexical
verb do/make, as Creissels (2012:17) proposes:
[A] Proto-West-Mande root *tin ‘do’ can be reconstructed, and the hypothesis I would like to propose here is
that the antipassive suffixes -ndì (Soninke) and -rí (Mandinka) and the causative suffixes -ndí (Mandinka,
Soninke) and -ní (Bozo) result from the grammaticalization of *tin ‘do’ in causative and antipassive pe-
riphrases.
Following the reconstruction, one can imagine an antipassive periphrases with 'do' in example
(21b): 'The people did the cutting'. At once, the connection between the antipassive (deobjective)
and the causative turns out to be very close, formally distinguished by the omitting/ adding the O.
5 The antipassive construction presented in 21b is understood as a deobjective construction in Haspelmaths (2004)
terms. Jacques (2014:10) says that "the patient cannot be reintroduced with oblique case in Japhug antipassive
constructions, except for some ditransitive verbs ".
17
4.1.3 Denominatives (Rgyalrong)
A recent study (Jacques, 2014) discovers the pathway from denominatives (derivations from nouns
to verbs) to antipassives. Japhug Ryalrong (Sino-Tibetan) has two kind of antipassives: the one re-
ferring to a (deleted) human entity (sɤ) and the other to a (deleted) non-human (rɤ):
In contrast, we have the same morphemes to derive verbs from nouns (denominatives):
We notice, that the denominal derivation is quite inconsistent: first, the correlation between sɤ- and
rɤ- with the properties [±human] is not visible. In addition, we cannot say whether the denominal
derivation creates transitive (24a-c) or intransitive (24d) verbs. Regarding the historical link be-
tween the denominals and antipassive prefixes, Jacques (cf. 2014:20) proposes the following stages.
The bare lexical root can be attached to possessive like tɤ- in (24b, d) to form nouns. The denominal
suffix, indeed, converts the lexical meaning into a "property stative verb". For instance, the lexical
item -sat 'kill(ing)' can be derived to sɤ-sat 'to have a propensity to kill, to have killing power.' From
this stage on, we have a verb, which can be further "reanalyzed as an action verb due to some am -
biguous contexts" as 'to kill people'. Rather of speaking about an implicit object which refers to
18
'people', we have to see the sɤ- prefix not as an agreement prefix, but a verbalizer which denotes
that the action carried out by the single argument (S) in the clause is restricted to human entities.
The denominal prefix rɤ-, indeed, is mostly attached to transitive actions, like ɕpʰɤt 'mending', mak-
ing them intransitive rɤ-ɕpʰɤt ‘to patch (clothes)’. The connotation of the deleted O's property [±hu-
man] arose from the original function of the denominal prefixes: the derivation rɤ-/rɯ- denoted
something like 'production/action' and sɤ- 'property of the verb base noun to which humans can be
susceptible' (cf. Jacques 2014:20).
This development caused the two functions (denominative/antipassive) to adapt a slightly different
form/behavior; for instance, the antipassive and denominative suffixes have their own position in
the verbal template. "Antipassive prefixes occur after the causative prefix, but before the tropative
and denominal prefixes", which are closer to the stem (cf. Jacques 2014:13).
That means, whereas the antipassive prefix solidified two functions (human/non-human), the de-
nominational prefix remained bearing a less consistent derivational meaning (See examples in 24).
If we try to look at the sources for the 'pragmatic/semantic' antipassive, we are looking at how lin-
guistic material changes its former meaning and structure in order to create new constructions with
pragmatic/semantical function. In this sense, the term "grammaticalization" would sound mislead-
ing, as the new function is rather a pragmatical device than a grammatical, especially if this new
pragmatic device has arose by lessening the former grammatical function (compare the grammatical
morpheme 'reflexive' becoming a pragmatically used antipassive morpheme).
For Diewald (2008:384), there is no problem with subsuming the phenomenon of 'pragmaticaliza-
tion'6 under the general domain 'grammaticalization':
"The need to introduce the term pragmaticalization has been shown to be the outcome of an attempt to preserve the
domains of "grammar" and "pragmatics" as clearly distinct domains. It has been argued that a more comprehensive
notion of "grammar", which encompasses "pragmatic" functions, makes the latter term dispensable, or – if one
6 In Diewald 2008, the notion 'pragmaticalization' is applied to explain the diachronic development of several modal
particles in German like aber, eben, ruhig into discourse markers.
19
prefers to keep it – restricts it to naming a specification of the superordinate process of grammaticalization"
To distinct the several instances of pragmaticalization and to keep the term apart from grammatical-
ization should not be the aim of this paper. Rather, I will use these terms to describe the rising of the
pragmatic/semantic antipassive (more pragmatically motivated) and its diachronic transition into the
structural antipassive (properly grammatically motivated).
The status of the antipassive in several Australian languages (described in 3.3.1 and 4.1.1) could
give us evidence for this pragmatical-into-grammatical pathway.
As exemplified in 4.1.1, the reflexive morpheme is also used in an antipassive construction making
it an antipassive morpheme. Terrill has given us an understandable explanation for why the change
happened from reflexive to passive. Analogically to the explanation for the direction of grammati-
calization from reflexives to antipassives, Terrill (cf. 1997:81 f.) argues, that the structural antipas-
sive function (the pivot-feeding function, compare 4.1.1) has evolved from the pragmatic function
and not the other way round, because there are many Australian languages with semantic/pragmatic
antipassives unaccompanied by structural antipassives.
But in addition, she has a more elaborated hypothesis for this development. In order to do so, the re-
flexive construction, the pragmatic antipassive and the syntactic antipassive are listed successively
with their respective functions, the prototypical one(s) in bold. The functions "defocussed object,
less agentive subject, semantic intransitivity" are found in the reflexive and the pragmatic antipas-
sive, but to the reflexive they are peripheral, whereas it is the main function for the pragmatic an -
tipassive. The pragmatic antipassive has lost the function 'A is coreferential to O', but gained an-
other peripheral functions/constructions 'A now an S, the single argument in the clause, and the O
has to be expressed as an oblique case'. This latter function, in turn, has become the main function
to the next stage, the syntactic antipassive, which has also gained the 'clause coordination on S/O
pivot' as an additional discourse function.
20
Thus, the figure in (25) not only represents in which constructions the reflexive/antipassive mor-
pheme is used and what its functions are, but also clarifies how those functions and applications
have evolved diachronically.
It is necessary to emphasize that the morpheme has remained the same, but the construction not, be-
cause in the structural antipassive the O has to be mentioned (in the oblique case), whereas in the
pragmatic/semantic antipassive it can be either expressed or omitted.
We can now understand the historical connection between reflexives and antipassives, and gained a
look at the inner-antipassive grammaticalization, but what we have also gained is the idea that
grammaticalization is not only restricted to the process from "lexical forms to grammatical forms"
or "from grammatical forms to even more grammatical form" (Heine&Kuteva 2002:2), but that a
grammatical construction (A=O) may gain pragmatic function (defocussed object, less agentive
subject, semantic intransitivity), pragmatical function then changes again to structural function
(A=>S, O=>obl) and that this structural function can be applied again to pragmatic functions like
reference-tracking (clause coordination on S/O pivot).
In order to accomplish the development, this section shows to what grammatical form an antipas-
sive may evolve. I choose the example of (Old) Georgian (Kartvelian), to illustrate how the antipas-
sive clause expanded its function to act as an active clause. This functional shift brought new
changes to the grammatical structure, causing the split-alignment system.
Georgian shows an accusative pattern in the imperfective aspect and an ergative one in the imper-
fective:
IMPERFECTIVE PERFECTIVE
A ABS/NOM ERG
S ABS/NOM ABS
O DAT ABS
IO DAT DAT
21
Consider following examples:
(26) Old Georgian (Kartvelian) (Schulze 2010:47 f.)
a) k'ac-i mšier-sa mi-s-c-em-s p'ur-s
Mann-NOM/ABS hungry-DAT PV-3SG:IO-give-IMPERF-3SG:A bread-DAT
'The man gives bread to the hungry one.'
We notice: while the second sentence (26b) assigns a case to each grammatical function (ergative,
dative, absolutive), the first sentence (26a) shows a formal syncretism regarding O and IO (indirect
object). In addition, we find more morphological material on the former verb (miscems vs. misca).
Hewitt (1995) glosses it as a 'thematic suffix'; we will call it imperfective, in order to connect
quickly the synchronic and diachronic view of this morpheme. When this suffix isn't followed by
other tense/mood/aspect suffixes, then the reading of the action is 'present'. If it is absent – that
means, the zero-marked perfective is present, and no other TAM suffixes follow, then the reading is
'(narrative) perfective past.' Otherwise, the suffix -eb can be understood as an imperfective mor-
pheme:
Schulze (2010) hypothesizes that the imperfective marker -eb was once an antipassive marker
which served to detransitivize the clause. Therefore, the active construction in (26a) is equal to a
former antipassive construction, where the verb had the antipassive marker, the A was not more in
the ergative case but had changed to S function, and the O was marked by an oblique case, the da -
tive, like the indirect object (recipient). The comparison of other Kartvelian languages allowed
scholars to reconstruct this 'stem formation element as *(w)ew (cf. loc. cit). The pragmatic (cogni-
tive) motivation behind the function shift antipassive > imperfective and ergative (active) > perfec-
tive could be explained in terms of the close correlation between the imperfective and antipassive
function related to in section 3.2.1.
During the process, the morpheme *(w)ew lost its antipassive function as it filled a gap to mark the
imperfective aspect, leaving the unmarked active verbal form (which governed an ergative pattern)
to function as the opponent aspect, the perfective.
But the process went further: by introducing new tense markers like future (27) or preverbial mark-
22
ers covering aspectual functions (26), "the aspectual opposition became more and more obscured"
(cf. Schulze 2010:54). The analysis of recent-day Georgian is right as Hewitt (1995) describes it:
-eb is not more a productive derivation, but is better viewed as a thematic element. What has re-
mained out of this process is the curious fact that the case marking of NPs depends on tense/aspect
(split-alignment).
23
Figure 1: Summarized functional and diachronic relations between antipassives and other cate-
gories presented in this paper
Explanation: terms in squares represent constructions, the terms without squares lexical items or
functions. The link between each term represents a functional closeness, if the link is an arrow then
it represents, beside the functional closeness, a diachronic change presented in this paper.
24
References
Aldai, G., 2000. "Split ergativity in Basque: the pre-Basque antipassive-imperfective hypothesis."
Folia Linguistica Historica 21:31-98.
Cooreman, A., 1994. "A functional typology of antipassives." In: Fox, B., Hopper, P. (Eds.), Voice:
Form and Function. Benjamins: Amsterdam. 49-87.
Creissels, D., 2012. "The origin of antipassive markers in West Mande languages." 45th Annual
Meeting of the Societas Linguistica Europaea, Stockholm. 29 August - 1 September 2012. Available
online: http://www.deniscreissels.fr/public/Creissels-West_Mande_antipassives.pdf [retrieved
17.02.2014.]
Dürr, M., 1987. Morphologie, Syntax und Textstrukturen des (Maya)Quiché des Popol Vuh. Bonn:
Holos.
Dixon, R.M.W., 1972. The Dyirbal language of North Queensland. Cambridge: Cambridge
University Press.
Foley, W.A. and Van Valin, R.D., 1984. Functional syntax and universal grammar. Cambridge:
Cambridge University Press.
Fox, B., 1987. "The Noun Phrase Accessibility Hierarchy reinterpreted: Subject primacy or the
Absolutive Hypothesis." Language 63. 856-870.
Haspelmath, M., Müller-Bardey, T., 2004. Valency change. In Booij, Geert; Lehmann, Christian;
Mugdan, Joachim (eds.) Morphology: A Handbook on Inflection and Word Formation. Vol. 2.
(Handbücher zur Sprach- und Kommunikationswissenschaft). Berlin: de Gruyter. 1130-1145.
Heath, J., 1976: "Antipassivization: a functional typology." Proceedings of the second annual
meeting of the Berkeley Linguistics Society. 202-211.
Heine, B., Kuteva, T., 2002. World Lexicon of Grammaticalization. Cambridge University Press,
Cambridge.
Hewitt, B. G., 1995. Georgian: A Structural Reference Grammar. [London Oriental and African
Language Library]. Amsterdam: John Benjamins J.
25
Jacques, G., 2014. "Denominal affixes as sources of antipassive markers in Japhug Rgyalrong".
Lingua 138. 1-22.
Klimas, A., 1985. Some remarks on the Lithuanian reflexive verbs. Lituanus 31. Available online:
http://www.lituanus.org/1985/85_2_05.htm [retrieved: 15.02.2014].
Larsen, T.W., 1987. "The syntactic status of ergativity in Quiché". Lingua 71. 33-59.
Nichols, J., 1992. Linguistic Diversity in Time and Space. The University of Chicago Press:
Chicago.
Polinsky, M., 2013. Antipassive constructions. In: Dryer, M.S., Haspelmath, M. (Eds.), The World
Atlas of Language Structures Online. Max Planck Digital Library: Munich.
Pye, C., 1989. "The Focus Antipassive in Quiché Mayan" Kansas University Working
Papers in Linguistics.
Schulze, W., 2010. The Grammmaticalization of Antipassives. LMU Munich / UMB: Banská
Bystrica. Available Online: http://schulzewolfgang.de/material/antipas2.pdf [retrieved: 17.02.2014].
Van Valin, JR, R. (1980) On the distribution of passive and antipassive constructions in universal
grammar. Lingua 50. 303-32.
Plagiatserklärung
Ich habe die Arbeit selbständig verfasst, keine anderen als die angegeben Quellen und Hilfsmittel
benutzt und bisher keiner anderen Prüfungsbehörde vorgelegt.
___________________
26