Transcript V+A+I

Attenuating Agency in Russian
Dagmar Divjak
[email protected]
Laura Janda
[email protected]
News headline, July 2, 2009
Putinu xočetsja
počuvstvovat’
Obamu
[Putin-Dat, wants-REFL
feel-Inf Obama-Acc.]
Dative noun phrase +
finite impersonal
verb + infinitive
‘Putin feels like getting
a feel for Obama’
Where did the subject go?

Impersonal constructions that
attenuate or eliminate the role of
the agent
 Russian has a particularly rich
assortment of such patterns: are they
all created equal?


How do impersonal constructions
relate to prototypical personal &
transitive sentences?
How can this relation be accounted
for, theoretically?
Networks of constructions

(Radical) Construction Grammar
(Goldberg 1995 & 2006; Croft
2001)


Emphasis: organization of
constructions in larger networks of
related (personal and impersonal)
constructions


analyze the use of non-canonical
subjects in Russian
impersonal constructions as
peripheral members of the system
Case study: the role of the dative
case in impersonal [Vfin Vinf]
constructions


there are two such constructions
implications for the concepts of main
verb-hood and agentivity
Theoretical framework (i)?
A radical construction grammar approach
= non-reductionist
 the “primitive construct” = the construction, a
complex entity
 constructions contain categories and relations, and
these are defined by the constructions they appear in
= categories and relations are not theoretical primitives
 constructions are not derived from their parts
 the parts are derived from the constructions they
appear in
 the parts of a construction do not have an
independent existence outside of the whole
construction

Theoretical framework (ii)

Cognitive Grammar

Meaning: all linguistic units and
categories have meaning in all contexts


Options: construal


No distinction between grammatical and
semantic case
Different ways of arranging elements
Organization: radial category

Centre-periphery structure
The plan
Personal vs impersonal constructions in
Russian
•
Construction grammar
•
A network of constructions
•
Relationships among constructions
Impersonal constructions with Dative in
Russian
•
Radical construction grammar
•
What is an impersonal construction?
•
Ways to attenuate agentivity
Contrasting Russian with English
•
An experiment
Construction grammar


Goldberg 1995, 2006; Croft 2001;
Langacker 1991; Fillmore 1985
A construction is



a combination of form and meaning
a unit, not just an aggregate of
components
the unit that is most relevant for
linguistic analysis
What construction grammar has
not achieved

Construction grammar analyses




are typically based on English; an analysis of a
case language is likely to be very different
usually cover only small portions of a language
(e.g., the radial category of Subject-Auxiliary
Inversion in Goldberg 2006)
neglect transitions between related constructions
do not research relationships among the
meanings of neighboring constructions
Main points


Grammatical models, e.g. Construction
grammar, are seldom used to describe the
entire grammar of a language
Radial categories are seldom used to show the
relationships among syntactic constructions
 Construction grammar can describe a radial
network of constructions and the relationships
that hold among them
 Syntax is a continuum, the differences
between constructions are continuous and
neighboring constructions give (semantic)
support to each other
Case in Russian


Case is always obligatory in Russian
Russian has six cases, and the most
important ones for this study are:




Nominative = Subject
Accusative = Direct Object
Dative = Indirect Object, Experiencer
Instrumental = Instrument, Means
Network of constructions


The most important constructions and
their network
The transitions between constructions



The prototypical construction


addition or removal of a participant
transformation of a participant
N+V+A: personal, transitive
Peripheral constructions are


intransitive
impersonal
V+A+I
N+V+A+I
N+V+I
V+A
N+V+A
N+V+A+D
N+V+G
N+V+D
N+V+PP
N+V
D+V+A
D+V
V+A+I
N+V+A+I
N+V+I
V+A
N+V+A
N+V+A+D
N+V+G
N+V+D
N+V+PP
N+V
D+V+A
D+V
V+A+I
N+V+A+I
N+V+I
V+A
N+V+A
N+V+A+D
N+V+G
N+V+D
N+V+PP
N+V
D+V+A
D+V
V+A+I
N+V+A+I
N+V+I
V+A
N+V+A
D+V+A
N+V+A+D
D+V
N+V+G
N+V+D
N+V+PP
Transitive
constructions
N+V
Intransitive
constructions
V+A+I
N+V+A+I
N+V+I
V+A
N+V+A
N+V+A+D
N+V+G
N+V+D
N+V+PP
N+V
Impersonal constructions
D+V+A
D+V
Relationships between constructions

Transitions tend to be continuous



What neighboring constructions
contribute to a given construction


An example of addition/removal
An example of transformation
An example of a peculiar construction
Fully vs. mildly impersonal
constructions
Continuous transition with addition/removal
of a participant

See examples 1 (intransitive) and 9
(transitive) and examples 14-18
V+A+I
N+V+A+I
N+V+I
V+A
N+V+A
N+V+A+D
N+V+G
N+V+D
N+V+PP
N+V
D+V+A
D+V
N+V+A
Ženščina šila kostjum.
Transition between
The woman sewed a suit.
transitive and intransitive
N+V+A
Ženščina šila vsju noč’.
The woman sewed all night.
N+V+A
Ženščina spala vsju noč’.
The woman slept all night.
N+V
Ženščina šila.
The woman sewed.
N+V
Ženščina spala.
The woman slept.
Continuous transition with transformation of
a participant

In the N+V+D construction, Dative
can have five meanings:





recipient of something
experiencer of something good or bad
participant in a relationship of
equality/opposition
participant in a relationship of
submission
See examples 19-28
V+A+I
N+V+A+I
N+V+I
V+A
N+V+A
N+V+A+D
N+V+G
N+V+D
N+V+PP
N+V
D+V+A
D+V
What neighboring constructions
contribute to a given construction

A peculiar construction: V+A+I


A mystery: Why can’t the
instrumental express agent in this
construction?


See examples 29-32
See example 33
Neighboring constructions:


V+A (See examples 34-37)
N+V+A+I (See examples 38-41)
V+A+I
N+V+A+I
N+V+I
V+A
N+V+A
N+V+A+D
N+V+G
N+V+D
N+V+PP
N+V
D+V+A
D+V
Fully
impersonal
V+A+I
N+V+A+I
N+V+I
Mildly
impersonal
V+A
N+V+A
D+V+A
N+V+A+D
D+V
N+V+G
N+V+D
N+V+PP
N+V
Impersonal constructions
The remainder
of the talk will
focus on these
What neighboring constructions
contribute to V+A+I


V+A contributes a situation that
involves an unnamed force that has
negative results in an impersonal
construction
N+V+A+I contributes a situation
that involves use of the
Instrumental to express an
instrument or means (agent is
impossible because N fills this role)
Summing up personal and impersonal
constructions




Construction grammar can describe syntax
(or a large portion of it) as a radial category
The ordinary transitive construction (=
Langacker’s 1991: 285-6 “canonical event
model”) is prototypical; intransitive and
impersonal constructions are peripheral
Transitions between constructions involve
the addition/removal or transformation of a
participant and are continuous
Neighboring constructions contribute to a
given construction’s meaning
MIC: Mildly ill/impersonal constructions
Some examples

See examples (42) through (45) on
handout
Impersonal Constructions =




finite verb is “impersonal”
 “every verb without an acting person or
thing [canonically in the nominative] can
be considered impersonal” OR all “3rd (n)
sg verb forms and infinitives are
impersonal forms” (KG 1990: 283-284,
§285)
lack a subject with nominative case marking
accusative or dative required or possible
[with infinitive]
Disparity of Views in 3 Areas

Disagreement on the structure of
impersonal constructions and function
of their components



The construction as a whole: monopartite
or bipartite?
The status of the infinitive: grammatical
subject or not?
The function of the (accusative or)
dative: semantic subject or not?
Monopartite or Bipartite?

Sentence structure: subject predicate
Monopartite: infinitive does not
dominate the predicate, i.e. does not
initiate morphological subject-verb
agreement  1 part missing
 Bipartite: sentence consists of 2 parts,
yet …
Infinitive = “main member” of bipartite
sentence?

Infinitive = Grammatical Subject?

No: function grammatical subject
reserved for nouns and pronouns
e.g. RG (1970, 1980), KG (1990); Šachmatov
(1941), Zolotova (1973), Belošapkova (1978),
Šmeleva (1978)

Yes: infinitive can substitute
(pro)nominal grammatical subject
e.g. RG (1960, 1970); Protogenova (1955), Metlina
(1953), Ermakova (1974), Kokorina (1979), Barsov
(1981), Guiraud-Weber (1984)
! cannot express exactly the same
meaning (≠ forms) e.g. Peškovskij (1956), Bricyn
(1990)
Function of the Dative = ?

Semantic subject?

What is a “semantic subject”?



intuitive
no criteria given (cf. Zaiceva 1990, which
element qualifies as a semantic subject,
under which conditions etc.)
only reformulations (non-nominative or
oblique subject, subject of the
action/state, logical subject)
Radical construction grammar (i)
A radical construction grammar approach
= non-reductionist
 the “primitive construct” = the construction, a
complex entity
 constructions contain categories and relations, and
these are defined by the constructions they appear in
= categories and relations are not theoretical
primitives
 constructions are not derived from their parts
 the parts are derived from the constructions they
appear in
 the parts of a construction do not have an
independent existence outside of the whole
construction

Radical construction grammar (ii)

This is reminiscent of the Gestalt
paradigm: the whole is greater than
the sum of its parts
“difference of form entails
difference in categorization, identity
of form does not entail identity of
categorization” (Croft 2001: 76)
1.
2.
A family of constructions that share
the same form, but have distinct
properties
Elements filling up constructional
slots that do not always relate to
each other in the same way
A Radical Construction-based Proposal
Status of infinitive and dative depends on
the type of finite verb
= Bipartite structure possible
 Infinitive can be subject of construction
 Dative can take on subject-like function
 different ways of encoding signal different
sorts of relationships between the
participants
 analysis reconciles different insights put
forward in literature [Guiraud-Weber (1984) or

Bricyn (1990)]
A scale of agentivity
Nom > Dat > Instr > Acc > Gen > Loc
(Janda & Clancy 2002)
// Langacker’s (1991) role archetypes and
their agentivity
See examples on handout:
9. Nominative is prototypical subject
8. Dative is logical subject of further
action
5. Accusative is logical subject
Further examples along the scale
of agentivity

Instrumental is agent of passive:
Kostjum byl sšit ženščinoj.
[Suit-N was sewn woman-I.]
‘The suit was sewn by the woman.’

Genitive is logical agent of HAVE via BE:
U ženščiny est’ kostjum.
[By woman-G is suit-N.]
‘The woman has a suit.’

Locative has no agentive or subject-like
function
Two non-prototypical types of agents

Non-nominal entities occupying the
NOM slot



Ex 46 a/b/c: noun, that-clause,
infinitive
Ex 47 a/b: pronoun, infinitive
Absence of a NOM slot


Ex 48 vs 49
Ex 50
Assigning subject-status
Brown (1987: 166)
“those noun phrases with which the verb agrees in
person and number (in gender too, for some verb
forms). Then we observe that an infinitive
construction or a subordinate clause can play the
same role as a noun phrase and is mutually
exclusive with it; therefore we extend the term
“subject” to these infinitive constructions or
clauses, and mention in our description the
special verb-agreement which they are associated
with (3rd singular neuter)”.
Methodology


52 Russian verbs [dat + impersonal verb + inf]
Small n-design:



5 native speakers studied intensively (25-50 yr)
10 control judgments gathered
Elicitation test for argument structure based on
substitution with pro-nouns & pronoun
schemes (relevant in language acquisition)

does infinitive fit in an argument structure slot?
His parents decided to buy him a Mercedes.
What did his parents decide? To buy him a Mercedes.
To buy him a Mercedes, that’s what they decided.
Infinitive as Grammatical Subject
Morphologically complete verb
Rebenku
nadoelo
Kid
it bored
dat m sg
pf ind past 3sg n
Reading bored the kid.
Čto
rebenku
What
kid
Nom n sg
dat m sg
What bored the kid?
Čitat’.
Or
Impf inf
To read/Reading.
čitat’.
to read
impf inf
nadoelo?
it bored
pf ind past 3sg n
Čtenie/bljudo/moloko/selo.
Nom n sg
Reading/dish/milk/village
An Interpretation: Reification

Level of argument structure:



nominative slot

infinitive read fits in nominative slot occupied by
čto or what (neuter singular)

infinitive = subject, although not prototypical
subject, initiates finite verb event
dative slot: occupied by person = typical
experiencer
Level of event structure:

~ “reification” (Langacker 1987): infinitive event
reduced to, treated like any other “thing” [~ n sg
(cf. Smith 1994)] that can be subject of finite verb
event and bring it about
Infinitive ≠ Argument Structure
Morphologically defective verb-sense
Bol’nomu
Ill person
doctor
dat m sg
The ill person
ponadobitsja
it is necessary
pokazat’sja
vraču.
to show himself
pf ind fut 3sg n
pf inf
dat m sg
will need/have to go and see a doctor.
Čto
bol’nomu
ponadobitsja?
What
ill person
it is necessary
Nom n sg
dat m sg
pf ind fut 3sg
What will the ill person need?
*Pokazat’sja vraču.
pf ind fut 3sg
To go and see a doctor.
Vs
Lekarstvo.
nom n sg
Medicine.
Infinitive ≠ Argument Structure
Morphologically defective verb
Vam
nadležit
javit’sja v ukazannyj srok
You
it is required
to appear at time indicated
dat pl
impf ind pres 3sg
pf inf
You are to present yourself at the time indicated.
*Čto
What
Nom n sg
Čto
What
Nom n sg
What are you
vam
you
dat pl
nadležit?
it is required
impf ind pres 3sg
vam
you
dat pl
to do?
nadležit sdelat’?
it is required to do
impf ind pres 3sg
Interpretation: Complex Events

Level of argument structure:



infinitive go and see a doctor or appear at the time
indicated do not fit in nominative slot occupied by
čto or what (neuter singular) or in prepositional
slot
finite verb (defective paradigm!) ≠ construction
kernel [cf. Butler 1967]
Level of event structure:

no “reification” (Langacker 1987) = infinitive event
go and see a doctor or appear at the time indicated
cannot be reduced to, treated like any other “thing”
that can be subject of finite verb event need or
have to, infinitive event does not initiate finite verb
event
 infinitive = ?
Interpretation: Complex Events
Finite verb cannot pull infinitive into its
argument structure: infinitive is
stronger than usual, finite verb needs
infinitive to carry load of construction
~ auxiliary-like behavior
 Finite verb event modifies infinite verb
event
 together finite verb and infinitive form a
complex event
// status claimed for modal verbs in
general

Some Complex Event Examples

Modality-verbs


Volition: e.g. chočetsja, ne terpitsja, chvatit etc.
Suitability: e.g. (ne) goditsja, nadležit, polagaetsja
etc.


Necessity: e.g. trebuetsja, predstoit, ostalos’ etc.
Result-verbs:


Success only e.g. udalos’
Success + Associated (mis)fortune, e.g. (+)
povezlo, posčastlivilos’, (-) podfartilo etc.

Success + Reason for acting (circumstances,
chance, higher forces), e.g. dovelos’, slučilos’,
dostalos’ etc.
Zaliznjak & Levontina
(1996: 253)
Russian has
invested heavily
in lexical items
that express an
intermediate
degree of
subject
responsibility
for the final
result
Russian and Middle-English (i)
Allen 1997
“[t]he disappearance of the impersonal
constructions with a preposed nonnominative Experiencer (…) was largely
due to the decline of the case-marking
system of English, which often made the
preposed Experiencer ambiguous as to
case marking and liable to reanalysis as
the subject”
Russian and Middle-English (ii)
! some verbs, such as bihoven, began to be
used impersonally in Early Middle English,
i.e. by the time cases had disappeared
“this increase in the use of a nonnominative Experiencer was semantically
motivated. […] verbs of emotions so
frequently had non-nominative
Experiencers […] a useful way of showing
that the Experiencer was not in control of
the situation, i.e. not agentive. […] Thus
we get non-nominative subjects with
modal verbs (…) which talk about
necessity over which the human
argument had no control. ”
An Interpretation: De-agentivization

Finite verb does not open up nominative slot
Bol’nomu [dat] ponadobitsja [Vfin] pokazat’sja [Vinf] vraču
The ill person will need/have to go and see a doctor.

Infinitive event opens up nominative slot
Bol’noj
pokazalsja
vraču.
Nom m sg
pf past m sg dat m sg
The ill person went to see a doctor.

Morphologically defective finite verb that
modifies infinitive blocks nominative subject
 appears as dative = “discontinuous
(syntactic) subject”
An Interpretation: De-agentivization



Nominative: agent,
typical initiator of
the finite verb event
Dative: typical
experiencer of the
finite verb event
Dative in absence of
Nominative:
“experiencing
agent” or “agentive
experiencer”
A Cline of De-agentivization
True Agent
Bol’noj
pokazalsja
Ill person
showed himself
nom m sg
pf past m sg
The ill person went to see a doctor.
vraču.
doctor
dat m sg
Agentive experiencer
Bol’nomu
ponadobitsja
pokazat’sja
Ill person
it is necessary to show himself
dat m sg
pf ind fut 3sg pf inf
The ill person will have to go and see a doctor.
True experiencer
Rebenku
Kid
dat m sg
Reading bored
nadoelo
it bored
pf ind past 3sg n
the kid.
čitat’.
to read
impf inf
vraču.
doctor
dat m sg
Analysis of dative in infinitive
constructions

Dative in such constructions typically analyzed
as syntactic subject because the infinitive
action needs a subject to initiate it.
Mne

zakančivat’
Dat sg
impf inf
Me
finish
I need to finish this article.
Mne bylo/budet
past/pres 3sg
Me was/will be
I needed/will need to
stat’ju
acc f sg
article
zakančivat’
stat’ju
impf inf
acc f sg
finish
article
finish this article.
Inequality matters!
≠ constructional patterns, ≠ semantic possibly
≠ conceptual structure
 not all ‘impersonal’ verbs are equal: there
are finite verbs that function as
construction kernel and finite verbs that
merely modify the infinitive
 not all infinitives are equal: some fulfil the
syntactic subject or [prepositional] object
requirements, others act as (part of the)
construction kernel
 not all datives are equal: a certain group
has to be classified as discontinuous
syntactic subject, another as indirect object
A Constructional Conclusion



Analysis takes into account case
semantics and relationships among
constructions in assessing how agency
is assigned or avoided in Russian
impersonal constructions,
making it possible to tease apart the
differences between two impersonal
constructions that appear identical in
structure.
Everyone is ‘right’ … as long as they
select the ‘right’ finite verb!
Lining up semantics and concepts:
Is Russian a Non-Egotistical
Language?
A striking difference

Russian
 No modal verbs
(except мочь ‘be
able’)
 Many impersonal
constructions: мне
холодно/48 лет

English
 Lots of modal verbs
 Personal subjectheaded
constructions : I’m
cold/48 yrs old
Often Russian Dative Experiencer + Impersonal
verb constructions correspond to English
Nominative Agent + Personal verb constructions
e.g. Мне хочется спать = I feel like sleeping
How to be impersonal in English
Some theoretical background
•
Do these grammatical differences influence
thought?
• Thinking for speaking (Dan Slobin)
Is the organization of thought influenced by
specific organizational properties of an
individual language?
• Speaking a language requires paying
attention to those properties that are
grammaticalized in that language, e.g.
number, gender, tense, aspect ...
 Speakers of different languages might be
thinking differently to this extent.
•
A grammatical difference
Russian
DAT + Vfin + Vinf +
ACC expresses
enjoyment,
necessity,
opportunity:
Мне хотелось бы
порадовать моих
девчонок.

English
NOM + Vfin + Vinf +
Obj
for corresponding
expressions:
I’d like to make my
girls happy.

Do Russian and English speakers think differently
when speaking about these experiences? If so, in
which way(s)?
A typical interpretation

Wierzbicka (1988: 233): the unknown
Limitations of human reason/knowledge 
dependence on fate, destiny
 Uncontrollable passions govern lives of people
= some things are beyond human control


Israeli (1997: 21)

Some things come from outside the subject,
are imposed upon him/her
Verbs used in the study

Nominative slot: infinitive = grammatical
subject, dative = true experiencer:



(R) Грозило, идет, льстило, нравилось, опротивело,
не светит
(E) Be in danger of, look good, be flattered to, enjoy,
be sick of, be fated to
No nominative slot, dative takes on subject-like
function, i.e. Agentive experiencer:


(R) Осталось, повезло, полагалось, пришлось,
хотелось, удалось
(E) Have to, be lucky enough to, be supposed to,
have to, feel like, manage to
An experiment: discourse cohesion

Trigger: Мне хотелось бы порадовать моих девчонок

3 “Instigator” types:
чем-нибудь необычным, сказочным. / I’d like to make
my girls happy with something special, something
fantastic.



Subject …Я - хороший отец, люблю своих детей,
люблю доставлять им удовольствие./ I am a good
father, I love my children and I like giving them
pleasure.
Object …В школе они получили только пятерки и
заслуживали награду. / They got the best grades in
school and deserved a reward.
Circumstance …Новый год был близок, и надо было
отметить этот день./ New Year’s day was near and it
was necessary to mark that day.
Experimental design

36 questionnaires per language



(E) college age, non-linguists, non-slavists,
responded in class
(R) various ages, responded via email, most live in
US
6 benchmark sentences, 12 fillers, 6 triggers



Benchmark sentences: training, also test
participant reliability (3 at beginning, 3 at end)
Filler sentences: to prevent participants from
guessing what we were testing
Trigger sentences: contained the independent
variables
! fillers and triggers presented in randomized order in
every questionnaire to avoid order effects
Experimental design, cont’d.

Independent variables: 2 kinds



type of experiencer: 2 levels
 True vs. Agentive Experiencer
type of instigator: 3 levels
 Subject, Object, Circumstance
! 12 verbs, 3 different token sets per verb to
avoid lexical effects
Dependent variable: discourse coherence,
measured on 5-point Likert-scale (-2 to 2)
Statistical evaluation

Data:



36 judgments for every factor level combination
 every subject judged all 6 factor level combinations
once
 every subject got only one example from each token
set
Data analyzed using both the Means Model
(models means) and the Multinomial Model
(models proportions)
No statistically significant contrasts:

speakers of Russian do not significantly prefer
situations in which the circumstances are held
responsible for need, opportunity etc. to do sth.
Discussion

Is there a difference in
expectation pattern that
this design does not
show?



Different type of task?
Different type of
measure?
Is there no difference in
expectation pattern?
 evidence from a corpus
(BNC/RNC)
Many thanks to




Stef Grondelaers (K.U.Leuven, Belgium),
Christina Hellman (SU, Sweden) and Stefan
Gries (UCSB, USA) for discussing the
experimental set-up;
Masja Koptjevskaja (SU, Sweden) and Eleonora
Magomedova (UNC, USA) for scrutinizing the
experimental items;
Our 72 participants for filling out the
questionnaires;
Chris Wiesen (UNC, USA) for statistical analysis.