Download Relativization strategies in Thulung Rai Aimée Lahaussois Histoire

Relativization strategies in Thulung Rai
Aimée Lahaussois
Histoire des théories linguistiques (UMR 7597, CNRS/Univ Paris Diderot)
[email protected]
Abstract
Three morphemes are found in relativization scenarios in Thulung Rai (Eastern Nepal, TibetoBurman, Kiranti subgroup): a finite nominalizer –m (along with allomorph –mim), and two
participial markers, -pa and –ma.
While –m is the apparent primary morpheme used for relativization (it occurs with finite verbs),
with an allomorphic distribution between –m and –mim based on syllable length and tense, the
fact that relativization on A is not allowed with –m has promoted relativization with participials
as an alternate strategy. The result is a system of mixed strategies taking into account both the
semantico-syntactic role of the relativized noun and the tense of the verb in the relative clause.
This presentation describes the distribution which has been found through narrative materials and
elicitation, and which is more nuanced than had previously been described (Lahaussois 2003).
Introduction
My goal in this article is to present the relativizing strategies used in Thulung Rai. There are
three strategies available, but they each have different distributions based on the role of the head
of the relative clause (henceforth RC) in the matrix sentence.
Section 1 of this article provides an overview of the language. Section 2 presents the types of
nominalization patterns found in Sino-Tibetan languages and how they manifest in Thulung.
These patterns are relevant to the topic at hand because the three relativization strategies used in
Thulung, nominalizer -m and participial markers -pa and -ma, also participate in other
nominalizing functions. Section 3 presents generalities about relativization in Thulung. Section
4 then discusses the relativization of core arguments (S, A and P) and the constraints on various
strategies. Section 5 presents the relativization of non-core arguments: indirect objects, locative,
instrumental, comitative, ablative/allative, and genitive. Section 6 presents the typological
background against which the data should be examined, notably the accessibility hierarchy from
Keenan and Comrie 1977 and the question of how Thulung's ergative split plays into patterns of
relativization.
1. Overview of the language
Thulung Rai is a Tibeto-Burman language of the Kiranti subgroup, spoken in Solu district in
Eastern Nepal by several thousand speakers. It is an exclusively oral language which was the
subject of descriptive work by Allen (1975).
From a typological point-of-view, Thulung is an extensively suffixing language with an
ergative split, with 2nd person plural and 3rd person A arguments taking ergative case-marking,
and human P arguments marked with -lai. P marking is optional on non-human animates and
elsewhere (Lahaussois 2003a, 2003b). Thulung is also characterized by complex verbal
morphology, with distinct sets of intransitive and transitive markers encoding person and number
of up to two arguments (Lahaussois 2011).
1
2. Standard Sino-Tibetan Nominalization pattern
Thulung follows the same pattern found throughout the Sino-Tibetan area, with nominalizing
morphology covering a number of functions. These patterns have been documented widely
(Matisoff 1972, DeLancey 1989, Noonan 1997, Genetti et al 2008, Yap et al 2011, among others)
and given the label Standard Sino-Tibetan Nominalization (SSTN) (Bickel 1999). The following
varied functions have been found to share the same nominalizing morphology :
attributive/genitive marking, relativization, nominalization both at the clausal and sentential
levels. The Thulung morphemes which participate in SSTN are -pa, -ma, and -m, and the range of
functions they each cover will be discussed in turn.
2.1 Main functions of active participle marker -pa
a) Active participle marker
The main function found for the morpheme -pa is as an active participial marker. This is
exemplified in (1) and (2).
(1) luŋ
tsʌi
tsjapljapkai rwak-pa
kwa-jy
lus-ta
stone CONTR ONOM
call-ACT.PTCP mud-low.LOC come.out-PST
'The stone jumped down into the mud making a falling sound.'
(2) gu ŋo seʈ-ɖa
lʌk-pa,
rok-pa
3SG fish kill-PURP go-ACT.PTCP come-ACT.PTCP
'Coming back, having gone fishing.'
b) Complementizer
The morpheme -pa is also used in complementation, subordinating the clause it marks to the
main verb of the sentence. This is seen in example (3) and (4) which respectively use the main
verbs 'to pretend' and 'to know', with the -pa marked clause functioning as a complement to the
verbs.
(3) [ŋo seʈ-ɖa
lʌk-pa]
li-sa
fish kill-PURP go-ACT.PTCP pretend-2SG.IMP
'Pretend that you are going fishing.'
(4) [u-ri-tsip
bu-tsi rwak-pa]
tha by-ɖy-m
pʌtshi
3POSS-sibling-DU be-3DU say-ACT.PTCP know do-3SG>3SG.PST-NMLZ after
'After she learned that he had two sisters.'
2.2 Main functions of the passive participle marker -ma
a) Passive participle marker
The main function of -ma is to form passive participles, as seen in (5).
(5) ʣamke khok-ʣɵl-ma,
nem wo sjaŋ-ʣɵl-ma,
bae ʦhim-ma,
food cook-put-PASS.PTCP house also clean-put-PASS.PTCP floor sweep-PASS.PTCP
dys-thal-la
ʔe
become-AUX-PST HS
2
‘In the house the cooked meal, the cleaned house, the swept floor were ready.’
b) Nominalization of comitative-marked NPs
A very wide-spread use of -ma is to nominalize comitative-marked nouns, describing a
characteristic of the noun in question. This is illustrated in (6) and (7).
(6) khomdzɵl-num-ma
goiter-COM-PASS.PTCP
'the one with goiter'
(7) jum-num-ma
power-COM-PASS.PTCP
'the one with power'
2.3 Main functions of nominalizer -m
The morpheme -m is best described as the standard nominalizer in Thulung, as it is the default
morpheme for many kinds of nominalizations. It undergoes allomorphy conditioned by tense: -m
is used with past forms, and allomorph -mim with non-past forms.1
a) Expression of perfect aspect
Perfect aspect is expressed through a construction with an -m nominalized verb and copula. This
is shown in (8) and (9).
(8) meram leledym-mim-ka
ɖyl-miri-m
ba-iɖa ʔe
DEM
boogeyman-PLU-ERG care.for-3PL.PST-NMLZ be-PST HS
'Those boogeyman had cared for them.'
(9) bwa dzam-ra lʌs-ta-m
bu
pig feed-PURP go-PST-NMLZ be
'She has gone to feed the pig.'
b) Standard nominalizations
The nominalizer -m is used for the nominalization of entire sentences (example 10), of
clauses (11), as well as for complementation (12), to mark attribution (13), (14), and for verbal
nominalization2 (15.)
(10) [homu-ku bwi theptso dys-ta-m]
ʔe
bear-GEN head flat
become-3SG.PST HS
'(It is said that) the bear's head became flat.'
The conditioning is in fact a little more complicated than simply tense, as it also takes into account syllable
length of the marked verb. Non-past verbs which are bisyllabic or more can be nominalized with either -m or
-mim, whereas past verbs, which are always at least bisyllabic, can only be nominalized with -m.
1
2
Note that the infinitive form of verbs is formed in -mu and not in -m.
3
(11) [suk-khep dzes-i-mim]
tsʌŋra
three-times speak-1PI-NMLZ after
'After we say it three times...'
(12) [bhansa go khok-to-m]
rwak-ta
food 1SG cook-1SG>3SG.PST-NMLZ say-PST
'I cooked the food, she said.'
(13) oram bhante-m
this where-NMLZ
'This one from somewhere we don't know.' [lit. this where's]
(14) baja-ra-m
khole ku-ku
mam
earth-LOC-NMLZ all
water-GEN mother
'all the springs of the earth'
(15) pi-mim re o ne
eat-NMLZ FOC this TOP
'This is to eat.'
3. Relativization
In addition to the functions described in section 2, the morphemes -pa, -ma and -m all participate
in relativization in Thulung.
While some languages have a number of possibilities for whether the RC is internally- or
externally-headed, or pre- or post-nominal (Genetti et al 2008: 128), this is not the case for
Thulung. No matter which relativizing morpheme is used, Thulung relatives are always
prenominal and externally headed, as can be seen in (16), where the relative clause (bracketed)
precedes the head noun3 dzam.
(16) [go khok-to-m]
dzam brʌpa bai-ɖa
1SG cook-1SG>3SG.PST-NMLZ rice good be-PST
'The food I cooked was good.'
Headless relative clauses are also common, such as (17), where the head noun (which
could be mytsy, 'person') is ommitted from the matrix clause.
(17) [ama tsape ɖupe be-pa,
dukhʌ be-pa]
rok-a
1POSS food drink make-ACT.PTCP hardship do-ACT.PTCP come-2SG.IMP
'Come, person who makes my food, who works hard!'
Generally we do not find abstract head nouns or resumptive pronouns in Thulung RC's.
Sometimes, however, a demonstrative is used resumptively as a place to attach a case marker,
clarifying the head noun's role in matrix clause. This can be seen in (18), where demonstrative
3
I use the traditional term 'head noun' throughout this paper, but Dixon (2009: 317ff) recommends 'common argument' in its
stead.
4
mɵ acts resumptively for the missing head noun, and takes the expected case marker given the
head noun's role in the matrix:
(18) [mytsy pe-pa
Lamkane rwa-ki-m]
mɵ-ku
des-ra...
man eat-ACT.PTCP Lamkane call-1PI-NMLZ DEM-GEN country-LOC
'[the man eater we call Lamkane]'s country '
Unlike what has been described for a number of languages (Dixon 2009:341, Yap et al
2011), Thulung does not have restrictions on tense/aspect in RC's. When the relativization
strategy is to use the nominalizer -m, the result is a fully finite RC, with person/number and
tense/aspect markers as they would appear in an independent clause.
4. Relativization on core arguments (S, A, P)
Depending on the argument which is relativized upon in a sentence, different strategies will
apply, with different distributions of the relativizers -pa, -ma and -m.
4.1 Relativization on S
When the head noun is the S of the relative clause, the most frequently found strategy is the use
the active participial marker -pa. The nominalizer -m can also be used, but with restrictions. The
participial relative clause in -pa, being non-finite, has no tense marking morphology. Arguments
must be clearly identifiable, as verb morphology cannot be used to identify the arguments. The
default interpretation of the relative clause is as non-past, but interpretation as past is possible if
called for by the context.
(19) kruk-pa
gupsy
roar-ACT.PTCP tiger
'the roaring tiger'
(20) on-pa
myʦy
run-ACT.PTCP person
'the running person'
In (21), the interpretation is as past, because the adverb forces such a reading.
(21) malo gʌk-pa
ʦɵʦʦɵ
just be.born-ACT.PTCP child
'the just-born child'
The other available strategy involves using the standard nominalizing morpheme -m.
Because this strategy results in finite relative clauses, the resulting clauses code for tense, as seen
in (22).
(22) kruk-ta-m
gupsy
roar-PST-NMLZ tiger
'the tiger who roared'
5
The possibility of forming non-past equivalents is restricted: the reading of non-past forms is, in
the absence of aspectual markers, generic. Note that the non-relative sentences equivalent to the
ungrammatical relative clauses without aspectual marking (such as habitual, continuous, or
progressive) are also considered odd.
(23) *kru-mim gupsy
roar-NMLZ tiger
'the tiger who roars'
With added aspectual morphology, however, relativized non-past forms become acceptable.
(24) jem-ta-si-mim
myʦy (*jem-si-mim)
stand-CONT-REFL-NMLZ person (*stand-REFL-NMLZ)
'the person who is standing '
With stative verbs, on the other hand, the non-past form can be relativized, as in (25).
(25) me-lwasi-mim gupsy
NEG be.visible-NMLZ tiger
'the invisible tiger'
In sum, there are two possible strategies for relativization on an S argument:
The most common in narrative is with -pa, resulting in a non-finite relative clause with a default
non-past reading. The less common, but still perfectly grammatical strategy, is to relativize with m, a strategy mostly found with past relative clauses.
4.2. Relativization on A
When the head noun is the A argument of the relativized clause, there is a single relativization
strategy available: the use of -pa.
Because of the non-finite nature of participles, tense is not coded in these relatives.
As with relativization on S, the default interpretation is non-past, unless context leads to a
different reading.
(26) [make-m
sɵmdilwa sɵ-pa]
mytsy
long.ago-NMLZ story
tell-ACT.PTCP person
'the person telling an old story'
(27) maːke khli-pa
mytsy
grain plant-ACT.PTCP human
'the person planting grains'
Relativization on a pronominal head noun acting as A is possible:
(28) ma:ke khli-pa
gana-lai go lwan-ni
grain plant-ACT.PTCP 2SG-DAT 1SG see-1SG>2SG
'I saw the you who was planting grains.'
6
(29) make-m
sɵmdilwa sɵ-pa
go
long.ago-NMLZ story
tell-ACT.PTCP 1SG
'The I who is telling an old story...'
A past reading will result from the use of temporal adverbs which fix the RC as past (30),
or from the context of the matrix clause (31):
(30) make
tsɵtstsɵ-lo gupsy lwa-pa
ŋami
long.ago child-TEMP tiger see-ACT.PTCP old.woman
'the woman who saw a tiger when she was a child'
(31) meram nem bʌnebe-pa
mytsy si-ɖa
DEM
house make-ACT.PTCP person die-3SG.PST
'The person who built that house is dead.'
Any attempts to produce a nominalized finite equivalent for an RC on a A role are
rejected, or, alternately, corrected by turning them into complement clauses. Complement
clauses, while they look similar to finite RC's, do not gap the head noun in the clause. This is a
crucial syntactic test in distinguishing between pre-head RC's (the only type in Thulung) and
various types of nominalizations which share the same morphology (see Genetti et al. 2008:127).
(32) [meram myʦy-ka
ma:ke khli-y-mim]
go la-u
DEM
person-ERG grains plant-3SG>3SG-NMLZ 1SG see-1SG>3SG
‘I see that that person is planting grain’
While it is sometimes unclear whether one has a headless relative clause or a complement
clause, the verbal agreement markers can provide a clue. In determining whether it is possible to
relativize in -m on a head noun made up of the pronoun gana ('2SG'), I was given sentences (33)
and (34):
(33) *ma:ke khlin-na-m
gana-lai go lwan-ni
grain plant-2SG-NMLZ 2SG-DAT 1SG see-1SG>2SG
'I saw the you planting grains.'
(34) gana ma:ke khlin-na-m
go la-uto
2SG grain plant.PST-2SG-NMLZ 1SG see-1SG>3SG.PST
'I saw you planting grains.'
While (34) could plausibly have been a relative clause with the head noun preposed (as 2SG
cannot take ergative marking due to the person-based split described in section 1, it is not easily
identified as the patient of the matrix or the agent of the relative clause), the difference in
agreement markers shows that it is a complement clause, as can be determined from the
1SG>3SG marker, which would have been 1SG>2SG if gana were the head of the relative and
the patient of the matrix.
7
Relativization on an A argument is thus only possible with a single strategy: participial
relativization with -pa; the tense is non-past by default, unless context makes it otherwise; the RC
is necessarily non-finite.
4.3 Relativization on P
When the head noun is the P argument of the relative clause, all three nominalizing morphemes
can be used for relativization, with certain constraints.
The two most widely found strategies are to use the standard nominalizer -m and the
passive participial marker, -ma. Relativizing with -m results in a fully finite clause, which can
optionally include agents, as in (35) but need not, as in (36).
(35) [khomda-ka by-ɖy-m]
kaijo ʈhakro
Khomda-ERG do-3SG>3SG.PST-NMLZ comb brush
'the comb and brush which Khomda made'
(36) py-mim dzam
eat-NMLZ food
'the food (he) eats'
The other commonly available strategy, namely relativization with the passive participial
marker -ma, has a restriction: it cannot cooccur with an agent in the RC (see (39)).
(37) mijakma rwa-ma
ritsɵ
Miyakma call-PASS.PTCP sister
'The sister called Miyakma'
(38) lwam-ma
mytsy
see-PASS.PTCP person
'the seen person'
but
(39) *go lwam-ma
mytsy
1SG see-PASS.PTCP person
'the person seen by me'
The last strategy which we find is much more limited: active participle marker -pa can sometimes
be used for P relativization, but only when the verb context makes it very clear that the only
possible interpretation is as a P (such as a non-animate head noun) and the relative clause
describes an object with a specific use.
(40) pe-pa
dzam
eat-ACT.PTCP food
'food for eating'
(41) ma:ke khli-pa
je
grain plant-ACT.PTCP field
'the field for grain-planting'
8
(42) khok-pa
sodzam
cook-ACT.PTCP pot
'a pot for cooking'
5. Relativization on non-core arguments
In this section, I discuss the arguments other than S, A and P which are potentially accessible to
relativization. Indirect objects deserve special treatment as there are two broad types of
ditransitive verbs in Thulung and they must be considered separately in how they undergo
relativization.
The adjunct categories which will be discussed are locative, instrumental, comitative,
ablative/allative, genitive. Objects of comparison, which can be relativized upon in some
languages, are inaccessible in Thulung. The general rules concerning relativization on adjuncts
are that all three strategies are usually possible: finite relativization with –m is always possible.
Relativization with –pa is usually possible, with a generic reading or a non-past reading
concerning a specific event, and agents can be encoded for this type of relative clause;
relativization with –pa is blocked in situations where the interpretation defaults to an A or an S
role for the head noun: this happens with all comitatives and with genitives where the verb in the
relative clause is transitive. Relativization with –ma is usually possible, with a past reading, but
is blocked for relative clauses with a genitive role on the head noun. It appears that with two
arguments present, the head noun and the additional argument in the relative clause, a possessive
interpretation of the head noun is too far from the default reading and this role is therefore
blocked to relativization with this strategy. These scenarios will be illustrated in the relevant
sections below.
5.1 Indirect object
While we find indirect objects in the accessibility hierarchy presented in Keenan and Comrie
(1977), the concept of indirect object must in fact be nuanced for languages which have different
types of ditransitive verbs. Japhug, for instance (Jacques 2016) has a clear distinction between
indirective and secundative verbs: for the former, the theme is marked like the P of a
monotransitive verb, whereas for the latter, it is the recipient which is treated like a
monotransitive P, and in Japhug this distinction can be seen from the verbal agreement marking,
which encodes agent and patient: when the recipient’s person and number marking appears as
part of the verb agreement morphology, the verb is secundative.
A similar distinction can be made in Thulung as well: A and P marking on the verb makes
it possible to see whether the theme or the recipient is being indexed and whether the verb is to be
considered indirective or secundative. Example (43) uses both an indirective and a secundative
verb:
(43) go nem-ra
khotle-lai
ʣam pheʈ-pu-ma
gwak-pumi
1SG house-LOC everyone-DAT food serve-1SG>3SG-CONJ give-1SG>3PL
'At my house I serve everyone food and give it to them.'
The agreement marking shows that the first verb phenmu ‘serve’ takes ʣam ‘food’ as its patient
(as indicated by the 1SG>3SG marker –pu), indicating that it is an indirective verb, whereas the
second verb gwa:mu ‘give’ indexes a 3pl patient concordant with khotle ‘everyone’, showing that
it is a secundative verb.
9
Note that -lai in Thulung, borrowed from the Nepali dative-marker, is used to mark both
animate patients and recipients. It is thus found marking the recipients of both indirective and
secundative verbs, even though other languages, such as Japhug, do not have overt marking on
secundative recipients.
For both indirective and secundative verbs, two strategies are available for relativization
on the recipient: -ma and –m. The following examples show relative clauses including
secundative verb gwa:mu, ‘give’ ((44) and (45)) and indirective verb so:mu ‘pay’ ((46) and
(47)). The fact that the strategies for relativization are the same for both verb types shows that
there is no difference to be made in Thulung with respect to relativization. The distinction
between the verb types nonetheless needs to be taken into consideration when examining the
relativization strategies employed by different languages.
(44) gana dzam gwak-nami-m
mytsy
2SG food give-2SG>3PL-NMLZ person
'the people you gave food to'
(note 2SG>3PL marking on verb, and no indexing of dzam ‘food’)
(45) dzam gwak-ma
mytsy
food give-PASS.PTCP person
'the person food was given to'
(46) go khʌuluŋ sok-to-m
myʦy-mim
1SG money pay-1SG>3SG.PST-NMLZ person-PLU
'the people I paid money to'
(note 1SG>3SG marking on verb, indexing money as P, not people)
(47) khʌuluŋ sok-ma myʦy-mim
money pay-PASS.PTCP person-PLU
'the people who were paid money'
Relativization with -pa leads to a different interpretation, namely that the head noun
argument is the agent of the relative clause, and is therefore blocked for relativization on an
indirect object.
(48) dzam gwak-pa
mytsy
meal give-ACT.PTCP person
*the person rice was given to
'the person who gave rice'
Indirect objects are in fact closer to core arguments, in that they are part of the verb’s
argument structure and cannot be deleted in the same way as other adjuncts.
5.2 Locative
For a head noun which would be locative case-marked if the relative clause were a main clause,
all three relativization strategies are possible.
10
For adjunct relativization, the difference between the active and passive participle
strategies seems to be neutralized to a matter of tense, with -pa coding non-past events (as in (49)
and (50)) and -ma coding past events (as in (51) and (52)). Relativization with -m being finite,
tense is coded through verb agreement markers ((53), (54)).
(49) go baʈ-pa
nem
1SG live-ACT.PTCP house
'the house I live in'
(50) go bhramu
phoʈ-pa
je
1SG buckwheat sow-ACT.PTCP field
'the field I will plant buckwheat in'
(51) je
dzyl-ma
ʣyl-khop
clothes place-PASS.PTCP put-place
'the (specific) place the clothes were put'
(52) tsɵttsɵ-mim si:siʈ-ma
iskul
child-PLU learn-PASS.PTCP school
'the school where the children studied (but don’t study anymore as they are grown)'
(53) go bu-ŋu-mim
nem
1SG live-1SG-NMLZ house
'the house I live in'
(54) go bopsesi ʣyl-to-m
ʈhal
1SG fruit
place-1SG>3SG.PST-NMLZ plate
'the plate I put fruit on'
For all of the relatives in the examples above it is possible to make the equivalent independent
sentences in which the head noun from the RC becomes an adjunct with locative marking -ra.
5.3 Instrumental
When the head noun is an instrumental, all three strategies can be used. One of the challenges is
that as the case marker is missing, some of the relative clauses below could potentially be on a
locative rather than instrumental role (see for example (62), which is to be read as instrumental
but could potentially have been interpreted as a locative). The examples must be checked against
their case-marked independent sentence equivalents in order to ensure that the role is indeed an
instrumental taking marker -ka.
Note that the main difference between active and passive participial strategies is one of
tense: non-past for the former (see (55)), past for the latter ((56)-(58)) but that encoding of the
agent remains, as with core arguments, blocked for -ma relative clauses. Finite relative clauses
with -m can contain not only explicit agents but also indication of the person and number of agent
and patient via the agreement marking on the verb ((59)-(62)).
(55) khadzem-ra rjak-pa
rjakhop
book-LOC write-ACT.PTCP pen
11
'the pen (which will be) used to write in a book'
(56) khadzem-ra rjak-ma rjakhop
book-LOC write-PASS.PTCP pen
'the pen (which was) used to write in a book'
(57) dʌksa phʌl-ma
betho
tree cut-PASS.PTCP machete
'the knife (which was) used to cut the tree'
(58) khlea jal-ma
dylymʦa
dog strike-PASS.PTCP stick
'the stick (which was) used to hit the dog'
(59) ʦɵʦʦɵ-ka khlea jal-ly-m
dylymʦʌ
child-ERG dog strike-3SG>3SG.PST-NMLZ stick
'the stick the child beat the dog with'
(60) go dʌksa phʌl-to-m
betho
1SG tree
cut-1SG>3SG.PST-NMLZ machete
'the machete I used to cut the tree'
(61) khadzem-ra rjak-to-m
rjakhop
book-LOC write-1SG>3SG.PST-NMLZ pen
'the pen I wrote in a book with'
(62) go dzam pheʈ-pu-mim
ʈhal
1SG rice serve-1SG>3SG-NMLZ plate
'the plate I use to serve rice'
5.4 Comitative
For comitatives, only two of the three strategies are available:–ma and –m can both be used.
Relativization in –ma results, as with other oblique arguments, in an interpretation as a past and
completed event, as in (63). As for relativization in -m, it requires that the verb be non-singular,
in order to obtain the comitative reading. This is seen in examples (64 - 66).
(63) ʦamsiʈ-ma
ŋopsɵ
play-PASS.PTCP friend
'the friend I played with'
(64) khojo khari be-ttsoko-m
mytsy
always work do-1DE.PST-NMLZ person
'the person I always work with (and who always works with me)' (lit. the we always work
together person)
(65) tsamsiʈ-tsi-m ŋopsɵ
play-2DU-NMLZ friend
12
'the friend you played with (and who played with you)' (lit. the we played together friend)
(66) kaʈhmanɖu-ra lamdiʈ-ʦoko-m
ŋopsɵ
Kathmandu-LOC walk-1PE.PST-NMLZ friend
'the friend I walked to Kathmandu with' (lit. the we walked to Kathmandu friend)
It is not possible to relativize with -pa, as the interpretation will default to A (67) or S (68).
(67) hellolo khari be-pa
mytsy
everyday work do-ACT.PTCP person
*the person we work with every day
'the person who works every day'
(68) ʦamsiʈ-pa
ŋopsɵ
play-ACT.PTCP friend
*the friend I play with
'the friend who is playing'
5.5 Ablative/allative
Thulung has a case marker for the ablative, -laŋka (or variant -lam) (69), but the allative is
unmarked and inferred from the context (70).
(69) go del-laŋka lʌ-ŋro
1SG village-ABL go-1SG.PST
'I went from the village.'
(70) go del
lʌ-ŋro
1SG village go-1SG.PST
'I went to the village.'
While all three relativization strategies are found with ablative/allative head nouns, there is a
considerable amount of ambiguity concerning their relativization: as the case marker disappears
in the process of relativization, it is not clear whether one has an ablative or allative head noun.
With some verbs, the case is clear from the context. In (71) below, it is difficult to imagine
plausible context for fetching water TO a pond, and thus the ablative interpretation is the only
realistic option; similarly in (73), the only plausible context for the relative is that the head noun
is the tree that the person is dropped FROM, thus with an ablative interpretation. Note that in the
following examples, -m relativization is shown, but in each case, -ma or -pa relativization could
have been substituted (after making necessary changes, such as removing the agent for -ma
relatives).
(71) gui ku
hʌr-i-m
kudin
1PI water fetch.water-1PI-NMLZ pond
'the pond we fetched water from'
13
(72) go je
thyr-to-m
del
1SG clothes send-1SG>3SG.PST-NMLZ village
'the village I sent clothes to'
(73) gu-ka
myʦy-lai
al-ly-m
dʌksa
person-DAT make.fall-3SG>3SG.PST-NMLZ tree
'the tree he made him fall from'
3SG-ERG
With directional motion verbs, which incorporate the direction of movement (go up, come
down, carry away...), the default interpretation appears to be the allative. Transitive directional
verbs in the past have both interpretations--allative and ablative--equally available (see (75),
(79)), while for intransitives, particularly in the non-past, the allative alone is possible (see (74),
(77)). These patterns applied to all three relativization strategies (with the active and passive
participial markers indexing non-past and past respectively). Context plays an important part in
the interpretation as allative or ablative, rather than any formal trait: a non-past directional verb,
where the action is either in progress or in the future, will quite logically focus on the goal of that
motion, rather than its source.
(74) go jo-ŋro-m
khom
1SG come.down-1SG.PST-NMLZ place
'the place that I came down to'
(75) go bʌlʌkpu kheʈ-pu-mim
del
1SG potato
bring.up-1SG-NMLZ village
'the village I am bringing potatoes up to'
(76) go bʌlʌkpu kheʈ-to-m
del
1SG potato bring.up-1SG>3SG.PST-NMLZ village
'the village I brought potatoes up to/from'
(77) go bik-pa
del
1SG come-ACT.PTCP village
'the village I am coming to'
(78) bik-ma
del
come-PASS.PTCP village
'the village I came to'
(79) bʌlʌkpu lʌk-ma
del
potato take-PASS.PTCP village
'the village I am taking potatoes to/from'
5.5 Genitive
For a genitive-marked argument, only relativization with -m ((80) and (81)) and -pa (82) is
available. Furthermore, there are additional constraints when the verb is transitive (83): only -m
14
can be used then, as the finite nature of the verb makes clear the argument structure; relativization
in -pa would lead to the head noun being interpreted as an agent, as in (84).
(80) basta
najime nɵ-ɖa-m
myʦy oram ŋa
yesterday wife be.ill-3SG.PST-NMLZ person this INT
'the man whose wife was ill yesterday is this one'
(81) wotsy si:ɖa-m
me:sem
husband die-3SG.PST-NMLZ woman
'the woman whose husband died'
(82) malo nem ʣhar-pa
myʦy
just house fall.apart-ACT.PTCP person
'the person whose house fell apart recently' (lit. the just fallen.apart-house person)
(83) go ʣam thʌk-saʈ-to-m
ŋopsɵ
1SG food hide-BEN-1SG>3SG.PST-NMLZ friend
'the friend whose food I hid'
(84) ʣam thʌk-pa
ŋopsɵ
food hide-ACT.PTCP friend
'the friend hiding food'
*the friend whose food is hidden
Table 1 sums up the possible relativization strategies for the non-core arguments discussed in
section 5.
Indirect
Object
Locative
-pa
No
(interp. as
A)
Yes
-ma
Yes
Yes
Instrumental
Comitative Ablative/allative Genitive
Yes
No
(interp. as A)
Yes
Yes
No
Yes
Yes
(only for
vi)
No
15
-m/mim
Yes
Yes
Yes
Yes
Yes
Yes
Table 1. Relativization strategies available for non-core arguments
Relativization with the standard nominalizer -m is clearly the strategy which is available to the
most non-core arguments. As mentioned in the individual subsections, one of the main factors
constraining the use of participial relativizers for non-core arguments is that they result in
interpretations which pick up a core argument: by virtue of their non-finiteness and the
unavailability of agreement markers, participial relativizers make it impossible to disambiguate
roles, and when there is a possibility of interpreting the head noun as the A, this strategy is not
available.
6. Typological issues
The data presented above is interesting to considering with respect to the Keenan and Comrie's
1977 accessibility hierarchy. They propose that arguments which are accessible to relativization
in a given language will respect the following hierarchy:
subject>direct object>indirect object>oblique>genitive>object of comparison
The primary relativization strategy of a language will start at the top of the hierarchy, and go
partway or all the way down, in a continuous segment, depending on the language. If we look at
the three strategies described for Thulung, they are available for relativization on different
arguments as per table 2:
S
A
DO
IO
Obl
Gen
Obj Comp
-pa
✓
✓
✓
✕
mostly no
✓ (with
(depends
restrictions)
on type)
x
-ma
✕
✕
✓
✓
mostly yes
(depends
on type)
✕
x
-m/-mim
✓
✕
✓
✓
✓
✓
x
Table 2. Overview of relativization strategies for all arguments
Table 2 raises two issues in terms of the applicability of the accessibility hierarchy to Thulung.
The first point to note is that of the two strategies that can relativize on S (-pa and -m)—and
16
according to Keenan and Comrie (1977), relativization on S is a requirement for determining the
primary relativization strategy of a language—neither has a continuous run down the hierarchy:
-pa does not cover IO or Obl (generally); -m does not cover A.
A great many linguists have mentioned the constraint on relativization in ergative
languages. Among them, Creissels (2006:215) notes that "dans les langues à alignement ergatif,
il peut arriver que le sujet des constructions transitives ne soit pas relativisable". In the same
vein, Croft (1990:110-11) points out that for ergative languages, the A is only accessible to
relativization if the verb is antipassivized, making A into S. In such languages, Subject in the
hierarchy is to be interpreted as S. From this point of view, the standard nominalizer -m can be
considered the primary relativizer for Thulung; it is the only one which can relativize on all other
available arguments.
One interesting issue with respect to strategies for marking A vs S to this is that Thulung is
a language with a split ergative system: ergative marking occurs on A's that are 2PL4 and 3
person, and the other persons are unmarked when in the A role (see Lahaussois 2003b).
Nonetheless, all head nouns, regardless of person/number, undergo the restriction on
relativization with -m when they are in an A role, whether or not they take ergative marking. It is
quite challenging to generate relative clauses with 1st or 2nd person pronouns as the head noun,
but still possible (see examples (28) and (29)): when these are produced, they block relativization
with any strategy other than -pa. As 1st and 2nd person (apart from 2PL) do not take ergative
marking when functioning as the A, it is expected that they should behave as do the A's of other
nominative-accusative languages and be available for relativization with the primary relativizer,
but this is not the case. It seems that the relativization patterns in languages with ergative splits
could be an interesting area of study to pursue. In Thulung, the primary relativization strategy
remains blocked for A arguments no matter what their position in the split ergative system.
A different issue which is raised by the Keenan and Comrie scale is that it is not the
arguments highest in the animacy hierarchy which have most constructions available for
relativization—this is the expected scenario--but rather arguments that are lower down in the
hierarchy: in Thulung, a few arguments use all three strategies—patient, locative, instrumental,
ablative/allative--and these are found lower down in the hierarchy. This situation suggests that
the notion of hierarchy may need to be revised in terms of its predictive power in the accessibility
of relativization strategies.
Conclusion
In conclusion, the three relativization strategies available for Thulung have different distributions,
depending on the argument of the head noun of the relative clause. Like other ergative
languages, and counter to the accessibility hierarchy, the A argument is less accessible to
relativization than other core arguments. In the case of Thulung, a single strategy--participial pa--was available for relativization on A. This was true whatever the person/number of the A
argument, despite that person-based split system.
Nonetheless, the primary relativization strategy for Thulung appears to be the standard
nominalizer -m: it can be used to relativize upon all arguments other than A, and additionally it
relativizes finite verbs, with person, number, tense indexed, whereas this information is lost in
relativization with participial markers; the loss in participial relatives then makes it impossible to
4It
seems likely that the split was originally between 1, 2 on the one hand and 3 on the other, but shifted "upwards" based on the
influence of Nepali and pressure to create, from the 2PL pronoun, a 2SG polite pronoun form, resulting in the creation of a new
plural marked 2PL pronoun.
17
disambiguate the role of the head noun in some cases, as agreement markers cannot be used to
recover argument structure.
Relativization with participial -pa is possible for all core arguments, but the fact that it is
limited to core arguments and some obliques makes it a bad candidate as the primary
relativization strategy in the language. Because of the non-finite nature of the clauses, the
interpretation of the argument of a head noun relativized upon with -pa often defaults to A or S.
Abbreviations
A
most agent-like argument of a
transitive scenario
ABL ablative
ACT.PASS active participle
ALL allative
AUX auxiliary
BEN benefactive
COM comitative
CONJ conjunction
CONT continuative
CONTR
contrastive
DAT dative
DE
dual exclusive
DEM demonstrative
DO
direct object
DU
dual
DU
dual
ERG ergative
FOC focus
GEN genitive
HS
hearsay
IMP imperative
INT intensifier
IO
indirect object
LOC locative
NEG negative
NMLZ nominalizer
Obj Comp
object of comparison
Obl oblique
ONOM
onomatopeion
P
most patient-like argument of a
transitive scenario
PASS.PTCP passive participle
PE
plural exclusive
PI
plural inclusive
PL
plural
POSS possessive
PST past
PURP purposive
REFL reflexive
S
sole argument of intransitive scenario
SG
singular
TEMP temporal
TOP topic
X>Y agent X acting on patient Y
References
Allen N.J. 1975. Sketch of Thulung Grammar, with three texts and a glossary. Ithaca: ChinaJapan Program.
Bickel, Balthasar. 1999. Nominalization and focus constructions in some Kiranti languages, In
Yogendra P. Yadava & Warren G. Glover [eds.], 1999. Topics in Nepalese Linguistics.
Kathmandu: Royal Nepal Academy, pp. 271 – 296
Creissels, Denis. 2006. Syntaxe générale, une introduction typologique, tome 2, la phrase. Paris:
Hermes Science Publications.
Croft, W. 1990. Typology and Universals. Cambridge: Cambridge University Press
18
DeLancey, Scott. 1989. Relativization and nominalization in Tibeto-Burman. Ms., University of
Oregon.
Dixon, RMW. 2009. Basic Linguistic Theory. Vol 2.: Grammatical Topics. Oxford: Oxford
University Press.
Genetti, Carol. 2011. Nominalization in Tibeto-Burman languages of the Himalayan area:
A typological perspective. In Yap et al (eds), Nominalization in Asian Languages.
Amsterdam/Philadelphia: John Benjamins. Pp. 163-194.
Genetti, Carol, Bartee, Ellen, Coupe, Alec, Hildebrandt, Kristine & Lin, You-Jing. 2008.
Syntactic aspects of nominalization in five Tibeto-Burman languages of the Himalayan area.
Linguistics of the Tibeto-Burman Area 31(2): 97–143.
Jacques, Guillaume. 2016. Subjects, objects and relativization in Japhug. Journal of Chinese
Linguistics 44(1). 1–28.
Keenan, Edward L. & Bernard Comrie. 1977. Noun Phrase Accessibility and Universal
Grammar. Linguistic Inquiry 8.1. 63–99.
Lahaussois, Aimée. 2003a. “Nominalization and its various uses in Thulung Rai”, Linguistics of
the Tibeto-Burman Area 26(1): 33-57.
Lahaussois, Aimée. 2003b, Ergativity in Thulung Rai: a shift in the position of pronominal
split, in D. Bradley, R. LaPolla, B. Michailovsky & G. Thurgood (eds), Language Variation:
Papers on variation and change in the Sinosphere and in the Indosphere in honour of James A.
Matisoff, Canberra, Australian National University (Pacific Linguistics), p. 101-112.
Lahaussois, Aimée. 2011. "The Thulung Rai verbal system: an account of verb stem
alternation" Cahiers de Linguistique Asie Orientale Vol. 40(2): 189-224.
Noonan, Michael P. 1997. Versatile nominalization. In: Bybee, Joan, John Haiman and Sandra A.
Thompson (eds.), Essays on language function and language type in honor of Talmy Givón, pp.
374 – 94. Amsterdam: Benjamins
Matisoff, James A. 1972. Lahu nominalization, relativization, and genitivization. In: Kimball,
John P. (ed.), Syntax and Semantics, Vol. 1, pp. 237 - 57. New York: Academic Press
Yap, Foong Ha, Grunow-Hårsta, Karen, Wrona, Janick (eds). 2011. Nominalization in Asian
Languages. Amsterdam/Philadelphia: John Benjamins.
19