Realis mood, focus, and existential closure in Tundra Yukaghir

Abstract

The nature and the typological validity of the categories ‘realis’ and ‘irrealis’ has been a matter of intensive debate. In this paper we analyse the realis/irrealis dichotomy in Tundra Yukaghir (isolate, north-eastern Siberia), and show that in this language realis is associated with a meaningful contribution, namely, existential quantification over events. This contribution must be expressed overtly by a combination of syntactic and prosodic means. Irrealis is the default category: the clause is interpreted as irrealis in the absence of the marker of realis. This implies that the relevant typological question may turn out to be the semantics of realis, rather than irrealis. We further argue that the Tundra Yukaghir realis is a hybrid category composed of elements from different domains (information structure, lexical semantics, and quantification) unified at the level of interpretation via pragmatic enrichment. The concept of notional mood must therefore be expanded to include moods which come about in interpretation and do not constitute a discrete denotation.

Full text

Realis
mood,
focus,
and
existential
closure
in
Tundra
Yukaghir
Dejan
Matic
´
a,
*,
Irina
Nikolaeva
b
a
Syntax,
Typology,
and
Information
Structure
Group,
Max-Planck-Institute
for
Psycholinguistics,
Wundtlaan
1,
6525
XD
Nijmegen,
The
Netherlands
b
Department
of
Linguistics,
School
of
Oriental
and
African
Studies,
University
of
London,
Thornhaugh
Street,
Russell
Square,
London
WC1H
0XG,
United
Kingdom
Received
3
June
2013;
received
in
revised
form
25
July
2014;
accepted
26
July
2014
Available
online
29
August
2014
Abstract
The
nature
and
the
typological
validity
of
the
categories
‘realis’
and
‘irrealis’
has
been
a
matter
of
intensive
debate.
In
this
paper
we
analyse
the
realis/irrealis
dichotomy
in
Tundra
Yukaghir
(isolate,
north-eastern
Siberia),
and
show
that
in
this
language
realis
is
associated
with
a
meaningful
contribution,
namely,
existential
quantification
over
events.
This
contribution
must
be
expressed
overtly
by
a
combination
of
syntactic
and
prosodic
means.
Irrealis
is
the
default
category:
the
clause
is
interpreted
as
irrealis
in
the
absence
of
the
marker
of
realis.
This
implies
that
the
relevant
typological
question
may
turn
out
to
be
the
semantics
of
realis,
rather
than
irrealis.
We
further
argue
that
the
Tundra
Yukaghir
realis
is
a
hybrid
category
composed
of
elements
from
different
domains
(information
structure,
lexical
semantics,
and
quantification)
unified
at
the
level
of
interpretation
via
pragmatic
enrichment.
The
concept
of
notional
mood
must
therefore
be
expanded
to
include
moods
which
come
about
in
interpretation
and
do
not
constitute
a
discrete
denotation.
©
2014
Elsevier
B.V.
All
rights
reserved.
Keywords:
Notional
mood;
Realis;
Irrealis;
Focus;
Information
structure;
Event
semantics;
Tundra
Yukaghir
1.
Introduction
The
terms
‘realis’
and
‘irrealis’
are
often
used
as
semantic
categories:
they
refer
to
two
opposite
values
of
the
functional
domain
loosely
characterised
in
terms
of
‘reality
status’
or
‘actualization’.
This
may
be
realised
differently
in
different
languages,
either
by
verbal
inflection
(‘grammatical
mood’)
or
non-inflectional
means
such
as
free-standing
particles
and
adverbs,
modal
verbs
or
verb-like
items,
constituent
order,
etc.
A
convenient
label
for
this
broader
notion
of
mood
that
has
gained
popularity
in
the
past
decade
is
notional
mood.
Notional
mood
describes
a
wide
range
of
grammatical
patterns
associated
with
mood-like
meanings,
the
use
of
which
can
be
explained
on
the
basis
of
the
theory
that
is
used
to
explain
grammatical
moods
(Portner,
1999,
2011).
This
is
how
we
will
be
using
the
term
‘mood’
throughout
this
paper.
www.elsevier.com/locate/lingua
Available
online
at
www.sciencedirect.com
ScienceDirect
Lingua
150
(2014)
202--231
Abbreviations:
ACC,
accusative;
AF,
transitive
subject
focus;
ATTR,
attributive;
AUG,
augmentative;
CJT,
conjectural;
COM,
comitative;
COND,
conditional;
DAT,
dative;
DES,
desiderative;
DS,
different
subject;
EV,
evidential;
EX,
existential;
NP/FCI,
negative
polarity/free
choice
item;
FOC,
focus;
FUT,
future;
IMP,
imperative;
INCH,
inchoative;
INS,
instrumental;
INTERR,
interrogative;
INTJ,
interjection;
INTR,
intransitive;
IPF,
imperfective;
HAB,
habitual;
HORT,
hortative;
HYP,
hypothetical;
LOC,
locative;
NEC,
necessitative;
NEG,
negative;
NEUT,
neutral;
NLZR,
nominaliser;
O,
object;
OF,
object
focus;
0,
epenthetic
vowel;
PF,
perfective;
PL,
plural;
POSS,
possessive;
POT,
potential;
PRP,
proprietive;
PTC,
participle;
PTL,
particle;
S,
subject;
SF,
intransitive
subject
focus;
SG,
singular;
SS,
same
subject;
STAT,
stative;
TEMP,
temporal;
TR,
transitive;
V,
verb.
*
Corresponding
author.
Tel.:
+31
24
3521
187;
fax:
+31
24
3521
213.
E-mail
addresses:
dejan.matic@mpi.nl,
dejanmatic@hotmail.com
(D.
Matic
´),
in3@soas.ac.uk
(I.
Nikolaeva).
http://dx.doi.org/10.1016/j.lingua.2014.07.016
0024-3841/©
2014
Elsevier
B.V.
All
rights
reserved.

The
question,
however,
is
whether
this
notional
understanding
of
the
opposition
between
realis
and
irrealis
is
universally
applicable.
Some
authors
speak
of
the
reality
status
as
a
cross-linguistic
grammatical
category
in
its
own
right
(e.g.
Foley
and
Van
Valin,
1984;
Van
Valin
and
La
Polla,
1997:
40ff;
Elliott,
2000:
80).
Its
two
values
are
characterised
in
terms
of
actualisation
vs.
non-actualisation
of
a
given
state
of
affairs.
A
proposition
is
said
to
be
realis
when
it
asserts
that
a
state
of
affairs
is
an
‘‘actualised
and
certain
fact
of
reality’’
(Elliott,
2000:
66).
In
contrast,
irrealis
implies
that
a
state
of
affairs
‘‘belongs
to
the
realm
of
the
imagined
or
hypothetical,
and
as
such
it
constitutes
a
potential
or
possible
event
but
it
is
not
an
observable
fact
of
reality’’
(Elliott,
2000:
67).
This
notion
typically
includes
potential
actualisations
(epistemic,
deontic
and
desiderative-
intentional
meanings)
and
non-actualisations
(counterfactual
and
negative
constructions,
i.e.
the
realm
of
unrealised).
What
remains
unclear
is
whether
there
is
one
notional
feature
that
lies
in
the
centre
of
this
distinction.
Cross-
linguistically,
there
is
no
one-to-one
correspondence
between
the
status
of
an
eventuality
and
the
use
of
realis
or
irrealis.
Both
categories
are
known
to
convey
a
wide
array
of
meanings
which
differ
greatly
across
languages,
so
that
proposition
types
marked
as
realis
in
one
language
may
be
marked
as
irrealis
in
another.
The
fact
that
the
contents
of
these
categories
are
not
comparable
from
language
to
language
and
that
the
same
constructions
seem
to
be
able
to
express
both
realised
and
unrealised
states
of
affairs
(at
least
in
the
first
approximation)
has
led
some
linguists
to
doubt
the
typological
validity
of
the
realis/irrealis
distinction
(see
de
Haan,
2012
and
Mauri
and
Sansò,
2012a
for
recent
overviews).
For
instance,
Bybee
et
al.
(1994:
236--40)
and
Bybee
(1998)
suggest
that
grammars
do
not
make
a
consistent
binary
distinction
related
to
reality
status.
Cristofaro
(2012)
argues
that,
although
the
notion
of
‘unrealised
state
of
affairs’
may
be
relevant
in
the
sense
that
that
it
plays
a
role
in
determining
the
patterns
of
polyfunctionality
for
some
linguistic
constructions,
there
is
no
evidence
that
it
forms
a
feature
that
is
part
of
a
speaker’s
linguistic
knowledge
at
the
synchronic
level.
On
the
other
hand,
a
number
of
researchers
working
on
individual
languages
have
argued
that
the
realis/irrealis
distinction
does
make
a
stable
interpretive
contribution,
even
though
it
may
differ
from
one
language
to
another
and
require
a
separate
semantic
definition
for
each.
For
instance,
Verstraete
(2005)
suggests
that
the
basis
of
the
irrealis
category
in
several
non-Pama-Nyungan
languages
of
Australia
is
the
expression
of
potentiality.
This
is
a
robust
core
meaning
which
may
trigger
the
implicature
that
the
described
event
did
not
take
place;
that
is,
under
certain
conditions
the
basic
semantics
of
potential
actualisation
is
extended
into
the
domain
of
non-actualisation
(counterfactuality).
McGregor
and
Wagner
(2006)
propose
that
in
Nyulnyulan,
too,
the
irrealis
encodes
a
single
core
meaning,
namely,
the
construal
by
the
speaker
of
a
situation
as
unreal,
either
in
the
actual
world
or
some
possible
world.
The
potentiality
meaning
is
not
a
semantic
invariant
associated
with
the
category
itself,
but
is
accounted
for
by
pragmatic
inferences
arising
from
the
relevant
speech
act.
It
may
then
turn
out
to
be
impossible
to
provide
a
universal
cross-linguistic
definition
of
the
realis/irrealis
dichotomy.
However,
it
can
probably
be
conceptualised
as
a
spectrum
encompassing
language-particular
binary
categories
which
stand
in
the
family
resemblance
relation
to
one
another.
This
seems
to
be
the
position
shared
by
some
typologically
oriented
studies
that
argue
for
the
theoretical
relevance
of
these
notions
(Givo
´n,
1994;
Mithun,
1995;
Palmer,
2001).
In
this
paper
we
provide
an
analysis
of
the
realis/irrealis
dichotomy
in
Tundra
Yukaghir
(TY
henceforth),
a
member
of
a
small
language
family
in
north-eastern
Siberia.
This
family
additionally
comprises
Kolyma
Yukaghir,
at
present
nearly
extinct
(see
Nikolaeva,
2000:
89
ff.,
2006
on
other
extinct
varieties),
and
may
be
distantly
related
to
Uralic
(Fortescue,
1998;
Nikolaeva,
2000:
25ff.,
2006,
among
others).
TY
is
spoken
in
the
villages
of
Andryushkino
and
Kolymskoe
(the
Lower
Kolyma
District
of
the
Sakha
Republic,
Russia)
by
63
speakers
(according
to
the
survey
made
by
Dejan
Matic
´and
Cecilia
Odé
in
2010--11).
The
data
used
in
this
paper
come
from
natural
discourse,
namely
the
texts
collected
by
Dejan
Matic
´in
2008--2012
and
the
texts
published
in
Maslova
(2001)
and
Kurilov
(2005),
as
well
as
from
elicitation.
We
argue
that
the
contrast
between
moods
in
TY
has
to
do
with
existential
quantification
over
events
and
is
derived
from
different
types
of
denotations
via
pragmatic
enrichment.
At
the
first
glance,
our
analysis
follows
the
non-universalist
line
of
thought
outlined
above.
It
suggests
that
the
opposition
between
the
two
categories
is
real,
even
though
the
dividing
line
is
defined
on
a
language-particular
basis.
That
is,
the
TY
realis
does
not
mean
exactly
the
same
as,
say,
the
realis
in
Nyulnyulan,
but
it
is
still
a
notional
mood
expressed
by
a
mixture
of
morphological,
prosodic
and
syntactic
means.
However,
we
go
further
than
this:
we
argue
that
semantically,
the
TY
realis
is
a
hybrid
category,
composed
of
elements
from
different
domains
(information
structure,
lexical
semantics,
and
quantification),
but
unified
by
a
higher-level
notion
of
existential
closure
of
events
in
Davidson’s
(1967)
sense.
In
order
to
describe
the
realis
mood
in
TY,
we
will
use
the
notional
apparatus
of
event
semantics.
Events
are
spatio-temporal
particulars
that
occupy
a
certain
space
at
a
certain
time.
We
take
it
that
predicates
have
an
event
argument,
a
variable
representing
the
event
they
describe.
1
To
guarantee
that
the
sentence
denotes
a
proposition,
some
kind
of
quantification
over
events
is
necessary.
This
is
achieved
by
means
of
an
D.
Matic
´,
I.
Nikolaeva
/
Lingua
150
(2014)
202--231
203
1
We
bypass
some
controversial
issues
of
event
semantics
that
are
irrelevant
for
our
purposes.
In
particular,
we
are
agnostic
as
to
the
syntactic
status
of
the
event
argument
(see
Beck
and
von
Stechow,
2006
for
a
discussion).
The
exact
nature
of
the
relationship
between
this
argument
and
the
predicate
is
immaterial
for
our
purpose,
too:
our
analysis
is
compatible
both
with
the
classical
Davidsonian
account,
according
to
which
the
event
argument
is
merely
one
of
the
arguments
of
the
predicate,
and
with
Neo-Davidsonian
semantics,
in
which
the
event
argument
is
the
only
proper
argument
(a
good
overview
of
this
issue
is
found
in
Maienborn,
2011).

existential
quantifier
which
scopes
over
the
event
variable.
The
existential
quantification
over
the
event
introduces
existential
closure
over
the
whole
proposition
and
unselectively
binds
all
variables
within
it.
This
is
illustrated
in
(1).
(1)
Peter
reads
a
newspaper
in
the
café.
9e
[READ(Peter,
newspaper,
e)
&
IN
(e,
the
café)]
We
will
argue
that
the
existential
closure
of
the
proposition
is
one
of
the
necessary
conditions
for
a
sentence
to
be
interpreted
as
realis
in
TY.
In
particular,
the
unselective
nature
of
the
closure
accounts
for
the
fact
that
those
expressions
which
cannot
introduce
discourse
referents
due
to
their
referential
weakness
or
for
some
other
reason
cannot
occur
in
realis
clauses.
However,
existential
closure
is
not
a
sufficient
condition
for
a
realis
interpretation.
Two
important
aspects
of
events
are
unaccounted
for
by
the
formula
in
(1).
First,
events
occur
in
a
particular
world.
This
obviously
relates
the
realis
meaning
in
TY
to
verbal
mood.
In
this
paper,
we
adopt
the
standard
approach
to
modal
semantics
as
developed
by
Kratzer
(1991,
2012).
There
are
two
important
notions
in
this
approach:
the
modal
base,
i.e.
the
set
of
possible
worlds
relevant
for
the
interpretation
of
the
sentence,
and
the
ordering
source,
which
ranks
the
worlds
according
to
some
parameter
so
as
to
establish
an
ordering
relevant
for
the
truth
of
the
proposition.
In
order
to
account
for
some
quirks
of
modal
marking
in
TY,
we
will
enrich
this
model
with
the
notion
of
the
speaker’s
contextual
commitment
to
the
modal
base/ordering
source,
which
implies
the
readiness
of
the
speaker
to
argue
rationally
for
the
optimal
status
of
the
event
given
the
ordering
source
(Portner
and
Rubinstein,
2012).
A
realis
clause
in
TY
is
specified
with
respect
to
all
three
of
these
parameters.
The
world
in
which
the
event
denoted
by
the
realis
clause
takes
place
is
the
actual
world;
if
possible
worlds
are
involved,
what
counts
as
the
actual
world
has
to
belong
to
the
set
of
best
worlds
according
to
the
ordering
source,
and
the
speaker
has
to
be
contextually
committed
to
this
world.
Second,
events
exist
in
time
and
place,
i.e.
in
specific
sections
of
the
world.
In
order
for
an
event
to
be
existentially
quantified,
a
reference
to
the
specific
spatio-temporal
section
of
the
world
in
which
it
unfolds,
that
is,
the
situation
in
the
sense
of
Austin
(1950),
must
be
established
(Krifka,
1989:
90).
We
take
it
that
situations
are
presupposed
and
temporally
delimited.
The
temporal
notions
we
use
to
capture
this
idea
stem
from
Klein
(1994).
In
particular,
we
make
use
of
the
distinction
between
topic
time
(TT)
and
event
time
(ET).
The
latter
refers
to
the
temporal
extension
of
an
event,
the
former
to
the
segment
of
time
about
which
the
speaker
makes
a
claim.
In
TY,
realis
events
must
take
place
within
the
topic
time,
i.e.
ET
<
TT.
The
notion
of
topic
time
is
thus
crucial
in
delimiting
the
situation
within
which
the
event
takes
place.
This
is
because
the
speaker
can
commit
herself
only
to
the
claim
that
the
event
exists
in
the
segment
of
time
about
which
the
assertion
is
made,
i.e.
within
TT.
Anything
beyond
TT
is
beyond
the
scope
of
the
claim
with
realis
interpretation.
The
full
specification
of
a
realis
clause
in
TY,
including
existential
binding
and
the
world
and
time/situation
restrictions,
is
given
in
(2).
(2)
A
clause
in
TY
is
realis
if
its
logical
form
is
9e
[P(e)
&
TIME(e)

t’
&
e
<
w]
and
it
is
evaluated
with
respect
to
topic
time
TT
and
world
w-actual,
such
that
(a)
t’
≤
TT
(b)
w
is
the
actual
world
or
highest
ranking
world
according
to
the
modal
base
and
ordering
source
(c)
the
speaker
is
contextually
committed
to
the
modal
base
and
ordering
source.
This
reads
as
follows:
a
clause
in
TY
is
realis
if
its
logical
form
is
such
that
it
denotes
an
event
that
corresponds
to
the
description
given
by
the
proposition
and
this
event
extends
over
a
subinterval
of
topic
time
and
takes
place
in
the
actual
world
or
in
the
preferred
world
to
which
the
speaker
is
committed.
Existential
closure
(and
its
additional
semantic
specifications)
can
have
different
sources
in
different
languages,
which
range
from
aspectual
and
temporal
operators,
verb
meaning
itself,
valence-increasing
morphology
and
covert
operators
to
overt
existential
quantifiers
(Zimmermann,
2007).
In
other
words,
a
sentence
that
satisfies
the
conditions
contained
in
(2)
can
do
so
by
virtue
of
its
being
perfective,
or
through
the
lexical
meaning
of
the
predicate,
or
it
can
contain
an
overt
existential
quantifier.
We
will
argue
that
in
TY
(2)
is
a
type
of
interpretation
which
can
be
reached
from
(at
least)
two
different
sources.
The
first
source
is
an
explicit
existential
quantifier,
the
particle
mə(r)=.
This
particle
presents
a
long-
standing
descriptive
problem
in
TY
syntax:
it
has
been
analysed
alternately
as
a
focus,
declarative
or
polarity
marker.
We
propose
that
the
use
of
the
existential
particle
mə(r)=
is
sensitive
to
possible
world
semantics
and
aspect,
but
also
to
the
commitment
of
the
speaker
to
the
truth
of
the
proposition.
The
second
possible
source
of
existential
closure
in
TY
is
information
structure,
whose
effects
come
about
both
through
explicit
bipartition
of
the
proposition
into
focus
and
background
and
through
constructions
with
inherently
focused
elements
(the
latter
include
adverbial
modifiers
and
loosely
incorporated
objects
in
TY).
In
order
to
capture
the
information-structural
effects
of
realis
interpretation,
we
will
use
a
simplified
version
of
Rooth’s
Alternative
Semantics
D.
Matic
´,
I.
Nikolaeva
/
Lingua
150
(2014)
202--231204

(Rooth,
1992)
and
treat
focus
as
the
element
of
the
sentence
that
carries
the
main
assertion
via
evocation
of
alternatives.
Our
approach
to
focus
differs
from
that
of
Rooth
(1999)
in
that
we
assume
that
focus
is
a
soft
presupposition
trigger
in
the
sense
of
Abusch
(2010),
i.e.
the
background
to
focus
is
routinely
presupposed
unless
there
is
evidence
to
the
contrary
(cf.
also
Geurts
and
van
der
Sandt,
2004).
This
default
presuppositional
interpretation
of
the
event
within
which
one
of
the
elements
is
treated
as
focused
leads
to
the
default
existential
closure
of
the
proposition
and
has
the
same
effect
as
the
overt
existential
quantifier.
As
a
result,
the
sentence
is
understood
as
fulfilling
the
conditions
given
in
(2).
This
analysis
explains
the
complementary
distribution
of
focus
marking
and
the
existential
quantifier
mə(r)=,
as
well
as
their
similar
properties
in
relation
to
mood.
Since
mə(r)=
and
different
types
of
non-verbal
foci
are
mutually
exclusive,
they
might
appear
to
form
a
kind
of
paradigm
whose
denotation
is
something
close
to
(2).
However,
we
show
that
this
is
not
the
case.
The
only
feature
that
the
diverse
structures
that
make
up
the
realis
mood
in
TY
share
is
that
they,
starting
off
from
different
source
meanings,
entail
the
existence
of
an
event
within
the
topic
time.
They
are
neither
unified
at
the
level
of
structure
nor
at
the
level
of
encoded
meaning:
the
existential
quantifier
encodes
existence
by
virtue
of
its
denotation,
while
focus
entails
it
via
inference.
The
modal
meaning
then
comes
about
through
interpretation
associated
with
more
fundamental
semantic
operations.
The
corollary
of
this
analysis
is
that
a
modal
category
need
not
be
based
on
Platonic
essences:
the
TY
realis
is
a
set
of
interpretive
effects
which
arise
from
different
sources.
If
the
notional
mood
is
something
that
can
be
explained
on
the
basis
of
the
theory
that
is
used
to
explain
grammatical
moods,
as
Portner
suggests,
the
catalogue
of
notional
moods
must
be
expanded
to
include
this
specific
type
of
expression:
moods
that
come
about
in
interpretation.
The
paper
is
structured
as
follows.
In
Section
2
we
introduce
the
realis/irrealis
distinction
in
TY
and
describe
its
most
important
formal
characteristics.
Section
3
analyses
the
formal
and
semantic
properties
of
foci
and
related
expressions,
while
Section
4
is
devoted
to
the
form
and
meaning
of
the
particle
mə(r)=;
this
leads
to
a
more
precise
definition
of
the
TY
realis
and
contributes
to
a
better
understanding
of
the
TY
focus
system.
In
Section
5
we
discuss
the
relation
between
the
realis
mood,
existential
quantification
and
focus.
In
the
final
section
we
summarise
our
findings
and
discuss
the
consequences
of
our
analysis
for
the
general
theory
of
linguistic
categorisation.
2.
Realis
and
irrealis
in
Tundra
Yukaghir
TY
is
a
typical
SOV
language,
with
a
fairly
rigid
verb-final
structure
in
main/independent
clauses;
the
same
rigid
head-
final
pattern
is
evident
in
noun
phrases,
postpositional
phrases
and
non-finite
dependent
clauses.
The
language
is
known
in
the
typological
literature
for
its
intricate
system
of
morphological
focus
marking
2
;
however,
the
structural
opposition
between
realis
and
irrealis
described
in
this
section
has
never
been
discussed
in
the
earlier
work.
2.1.
Mood
and
basic
clause
structure
TY
has
a
number
of
morphologically
marked
moods,
which
we
label
imperative
(3a),
potential
(the
suffix
-mori,
possible
future
situations)
(3b),
necessitative
(the
complex
suffix
-morawn
´ə,
deontic
modality)
(3c),
conditional
(the
proclitic
ət=,
possible
or
desirable
situations)
(3d),
desiderative
(the
suffix
-lbun
´,
desired
situations)
(3e),
and
conjectural
(the
complex
suffix
-l’əltə,
probable
situations)
(3f).
There
is
no
dedicated
morphological
marker
of
realis.
The
unmarked
verbal
form
can
have
quite
different
readings
with
respect
to
the
reality
status
of
the
proposition,
from
the
indicative-like
‘realis’
reading
to
various
directive
(4a)
or
hypothetical
meanings
(4b).
In
other
words,
the
lack
of
modal
inflections
on
the
verb
is
not
a
marker
of
the
‘realis
mood’
either.
(3)
a.
ann
´ə-k!
<speak-IMP.2>
‘Speak!’
b.
kewej-mori-jəŋ
<leave-POT-NEUT.INTR.1SG>
‘I
would/could
leave.’
c.
kuder-o:l-morawn
´ə-j
<put-STAT-NEC-NEUT.INTR.3>
‘He
must
put
(it).’
d.
ət=jaba-j
<COND-die-NEUT.INTR.3>
‘He
would
die.’
e.
law-lbun
´-i
<drink-DES-NEUT.INTR.3>
‘He
wants
to
drink.’
f.
mira:-l’əltə-jli
<walk-CJT-NEUT.INTR.1PL>
‘We
might
walk.’
(4)
a.
jo:γačəli
<stop.NEUT.INTR.1PL>
‘Let’s
stop!’
b.
mid’ek
med’i-m
<HYP
take-NEUT.TR.3>
‘He
might
take
(it).’
D.
Matic
´,
I.
Nikolaeva
/
Lingua
150
(2014)
202--231
205
2
The
primary
description
of
this
system
can
be
found
in
Krejnovič
(1958,
1982);
for
more
details
see
Comrie
(1992),
Fortescue
(1996),
and
Maslova
(2003,
2005,
2006).

On
the
basis
of
these
data
it
appears
that
realis
is
not
a
category
of
TY
grammar.
However,
it
is
a
relevant
notion
in
TY.
As
indicated
above,
realis
assertions
share
one
notional
property:
they
entail
the
existence
of
an
event
in
the
topic
time.
The
distinction
between
realis
and
irrealis
is
reflected
in
the
clause
structure.
In
a
sense,
realis
clauses
are
marked,
as
they
impose
an
additional
formal
constraint:
they
must
contain
a
syntactic
element
that
signals
the
realis
interpretation.
The
minimal
structure
of
a
realis
clause
is
as
follows:
(5)
Realis
clause:
X
+
V
where
X
is
(i)
a
non-verbal
focus
element,
or
(ii)
a
verb
modifier,
or
(iii)
the
verbal
particle
mə(r)=
Some
types
of
irrealis
clauses
can
contain
non-verbal
foci
and
verb
modifiers,
but
this
happens
rather
infrequently,
and
is
not
required
for
the
grammaticality
of
the
clause
(see
Section
5.2).
The
particle
mə(r)=
never
occurs
in
irrealis.
The
items
making
up
class
X
above
are
rather
heterogeneous.
The
first
subclass
includes
nominal
foci,
which
come
in
three
basic
forms.
When
the
focus
falls
on
S
or
O,
it
is
indicated
with
the
marker
-lə(ŋ)/-(ə)k.
3
The
verb
only
agrees
with
the
subject
(S
or
A),
but
the
actual
exponence
of
agreement
depends
on
which
argument
is
focused,
S
or
O.
This
is
glossed
as
OF
and
SF
for
‘object
focus’
and
‘intransitive
subject
focus’,
respectively.
(6)
peldude:
qad’ir
mət-in
a:ri:-lə
tadi:-mələ
old.man
PTL
I-DAT
gun-FOC
give3-OF.3
‘The
old
man
gave
me
a
GUN.’
(7)
mət-ek
werwə-l
I-FOC
be.strong-SF
‘I
am
the
one
who
is
strong.’
Focus
on
A
is
not
morphologically
marked,
and
the
verb
shows
no
agreement
in
this
case
(Ø,
glossed
as
AF
for
‘transitive
subject
focus’).
(8)
tuŋ
a:ri:
mət
nu:-Ø
this
gun
I
find-AF
‘It
is
me
who
found
this
gun.’
The
focus
on
oblique
arguments/adjuncts
receives
no
special
morphological
marking
either.
The
verb
agrees
with
the
subject
(S
or
A)
and
stands
in
the
so-called
‘neutral’
agreement
form
(NEUT
in
the
glosses).
(9)
tuŋ
qul’arqa:
ma:rqə-n
sa:l-γa
saγanə-j,
əl=lukunburəbə-γa
this
seagull
one-ATTR
tree-LOC
sit-NEUT.INTR.3SG
NEG=earth-LOC
‘This
seagull
is
sitting
on
a
tree,
not
on
the
earth.’
The
second
subclass
is
comprised
of
what
we
call
verb
modifiers.
We
borrow
this
term
from
studies
of
Hungarian
clause
structure,
where
it
refers
to
the
class
of
expressions
including
aspectual
prefixes,
bare
nominals
and
directives,
which
compete
with
the
focus
phrase
for
the
immediately
preverbal
position
(Koopman
and
Szabolcsi,
2000:
19ff;
Wedgwood,
2005:
81ff).
In
TY,
verb
modifiers
include
most
manner
adverbials,
such
as
e.g.
amutnəŋ
‘well’
and
amdur
‘quickly’,
and
non-referential
nouns,
occurring
both
in
idiomatic
verb-noun
complexes,
e.g.
jaγun
moŋorn
´ə-
‘snore’
(lit.
‘nose-sound’),
čuŋdə(lə)
gudiči:-
‘think’
(lit.
‘thought-put’),
nonγələ
law-
‘smoke’
(lit.
‘tobacco-drink’),
and
in
more
compositional
verb-noun
combinations,
e.g.
lawjələ
men-
‘fetch
water’
and
ča:jlə
law-
‘drink
tea’.
(10)
e:ru:če
amdur
mira:-nu-j
hunter
quickly
walk-IPF-NEUT.INTR.3SG
‘The
hunter
was
walking
quickly.’
D.
Matic
´,
I.
Nikolaeva
/
Lingua
150
(2014)
202--231206
3
The
choice
between
the
two
variants
of
the
focus
marker,
-lə(ŋ)
or
-(ə)k,
depends
on
the
internal
structure
of
the
noun
phrase
and
will
not
be
relevant
here.
Note
also
that
focus
marking
is
absent
on
3rd
person
pronouns,
proper
names
and
possessive
NPs,
but
focus
agreement
on
the
verb
is
still
obligatory
(for
details
see
Maslova,
2003).

(11)
k
ö
de
mə=law-nu-m,
nonγə-lə
law-nu-m
person
EX=drink-IPF-NEUT.TR.3SG
tobacco-ACC
drink-IPF-NEUT.TR.3SG
‘the
man
is
drinking,
he’s
drinking
tobacco
(=smoking).’
The
third
member
of
class
X,
mə(r)=,
is
a
proclitic,
phonologically
bound
to
the
verb
and
represented
by
two
variants,
mər=
before
a
vowel
and
mə=
before
a
consonant.
(12)
lawjə
moj-l-γa
mə=pugeč
water
hold-1/2-DS
EX=be.warm.NEUT.INTR.3SG
‘When
I
touch
the
water,
it’s
warm.’
The
function
of
mə(r)=
has
been
a
matter
of
some
debate,
partly
similar
to
the
debate
concerning
the
semantics
of
the
Somali
verbal
particle
waa,
which
has
been
analysed
by
various
authors
as
a
declarative,
predicative,
or
focus
marker
(Saeed,
1984:
160ff,
1999:
239ff.;
Lecarme,
1999).
Krejnovič
(1958),
who
provided
the
first
description
of
mə(r)=,
suggested
that
it
indicates
positive
polarity;
Fortescue
(1996)
and
Kurilov
(2006)
analyse
it
as
some
kind
of
declarative
marker;
Maslova
(2003:
24)
refers
to
it
as
an
affirmative
prefix
that
signals
narrow
focus
on
the
verb,
whereas
in
our
own
earlier
work
(Matic
´and
Nikolaeva,
2008)
we
suggested
that
the
function
of
mə(r)=
is
to
indicate
that
the
verb
is
within
the
focus
domain.
We
will
argue
in
Section
4
that
there
is
a
certain
truth
in
all
these
characterizations,
but
none
of
them
exhaustively
describes
the
meaning
of
mə(r)=.
As
mentioned
in
the
Introduction,
the
basic
conclusion
we
will
make
is
that
mə(r)=
is
an
existential
quantifier,
and
accordingly
we
gloss
it
as
EX=
throughout
the
paper.
What
needs
to
be
emphasised
at
this
stage
is
that
mə(r)=
is,
with
some
minor
exceptions,
mutually
exclusive
with
regard
to
other
elements
of
class
X:
(70)
*mət-ek
mə=werwə-l
I-FOC
EX=be.strong-SF
(90)
*qul’arqa:
ma:rqə-n
sa:l-γa
mə=saγanə-j,
əl=lukunburəbə-γa
seagull
one-ATTR
tree-LOC
EX=sit-NEUT.INTR.3SG
NEG=earth-LOC
(110)
??
k
ö
de
nonγə-lə
mə=law-nu-m
person
tobacco-ACC
EX=drink-IPF-NEUT.TR.3SG
The
ungrammatical
examples
(70),
(90)
and
(110)
show
that
mə(r)=
cannot
co-occur
with
either
a
non-verbal
focus
element
((70)
and
(90)),
or
a
verb
modifier
(110).
We
will
return
to
this
point
in
Section
5.
2.2.
Realis
and
irrealis
clauses
Before
proceeding
to
investigate
the
formal
properties
of
class
X,
we
will
illustrate
how
the
system
works
by
providing
a
first
semantic
and
formal
characterisation
of
the
realis
and
irrealis
in
TY.
In
our
first
approximation,
the
realis
meaning
comprises
non-hypothetical
positive
assertions
(see
Sections
4
and
5
for
a
revised
version
of
this
claim).
When
this
interpretation
is
intended,
verb-only
clauses
are
strictly
ungrammatical.
(13)
*keweč
<leave.NEUT.INTR.3SG>
intended
meaning:
‘He
left.’
(14)
*law-m
<drink-NEUT.TR.3SG>
intended
meaning:
‘He
drank
(it).’
Similarly,
realis
clauses
that
only
include
non-focused
non-verbal
elements
and
the
verb
without
mə(r)=,
but
contain
no
X,
were
judged
as
bad
or
at
best
very
marginally
acceptable
by
our
consultants.
(15)
*/?
la:mə-lə
pa:j-m
dog-ACC
hit-NEUT.TR.3SG
intended
meaning:
‘He
hit
a
dog.’
(16)
*n
´ime-γa
la:mə
segu-j
house-LOC
dog
enter-NEUT.INTR.3SG
intended
meaning:
‘A
dog
entered
the
house.’
D.
Matic
´,
I.
Nikolaeva
/
Lingua
150
(2014)
202--231
207

Crucially,
if
any
member
of
class
X
is
present,
such
clauses
become
fully
grammatical.
That
is,
the
examples
above
can
be
‘repaired’
by
inserting
a
focused
element
(150),
a
manner
adverb
(130),
a
non-referential
noun
(140),
or
the
particle
mə(r)=
(1300).
(150)
la:mə-ləŋ
pa:j-mələ
dog-FOC
hit-OF.3
‘He
hit
a
dog.’
(130)
amdur
keweč
quickly
leave.NEUT.INTR.3SG
‘He
left
quickly.’
(140)
nonγə-lə
law-m
tobacco-ACC
drink-NEUT.TR.3SG
‘He
smoked.’
(1300)
mə=keweč
EX=leave.NEUT.INTR.3SG
‘He
left.’
(Maslova,
2001:
57)
This
stands
in
contrast
to
the
irrealis
contexts.
Irrealis
in
TY
notionally
includes
at
least
the
following
categories:
negation,
various
types
of
directives
(whether
or
not
the
force
is
encoded
by
dedicated
verbal
morphology)
and
hypothetical
situations,
including
all
the
modal
meanings
that
have
been
mentioned
in
Section
2.1.
In
the
clauses
that
convey
these
meanings
the
element
X
is
not
obligatory
for
syntactic
well-formedness.
This
is
demonstrated
below.
Example
(17)
shows
that
negated
verbs
do
not
require
the
presence
of
X.
(17)
əl=ann
´ə-ŋu
NEG=speak-3PL(NEG)
‘They
were
not
talking.’
The
structure
of
directive
clauses
is
illustrated
in
(18)
and
(19):
(18)
exemplifies
a
morphological
imperative,
while
(19)
contains
a
verb
in
the
indicative
mood
used
to
convey
the
directive
force.
Both
are
well-formed
without
any
element
X.
(18)
mət-u-l
we:tə-k!
I-0-ACC
untie-IMP.2SG
‘Untie
me!’
(Kurilov,
2005:
366)
(19)
u:-jəli!
go-NEUT.INTR.1PL
‘Let’s
go!’
As
with
directives,
we
understand
‘hypothetical’
as
a
notional
feature
which
needs
not
be
encoded
by
specialised
verbal
morphology.
Thus,
in
(20)
the
verb
stands
in
the
inflectional
desiderative
mood;
(21)
is
an
example
of
the
potential
mood;
while
in
(22)
the
verb
is
in
the
indicative,
but
the
hypothetical
meaning
is
contributed
by
the
free-standing
particle.
(20)
lewdə-lbun
´-i
eat-DES-NEUT.INTR.3SG
‘He
wants
to
eat.’
(21)
law-mori-mək
drink-POT-NEUT.TR.2SG
‘You
should/could
drink.’
(22)
qa:laγaji:!
mid’ek
m
ö
ri-m!
INTJ
HYP
hear-NEUT.TR.3SG
‘Awful!
What
if
he
finds
out!’
In
all
these
irrealis
examples
the
verb
is
either
the
only
element
in
the
clause
or
it
combines
with
a
non-verbal
element
that
does
not
belong
to
class
X,
such
as,
for
example,
a
non-focused
object
in
(18)
or
the
particle
mid’ek
in
(22).
D.
Matic
´,
I.
Nikolaeva
/
Lingua
150
(2014)
202--231208

2.3.
Additional
evidence
for
realis:
negative
polarity/free
choice
items
There
is
independent
evidence
that
the
obligatory
presence
of
X
has
to
do
with
the
realis/irrealis
distinction
which
comes
from
the
distribution
of
negative
polarity/free
choice
items
(NP/FCIs;
TY
belongs
to
the
languages
that
collapse
these
two
categories,
see
Giannakidou,
2011).
In
TY
NP/FCIs
are
derived
from
wh-words
with
the
suffixes
-(ŋ)oll’əlk
(on
subjects)
or
-(ŋ)olləŋ
(on
objects).
Importantly,
they
are
only
licit
in
the
clauses
that
are
acceptable
without
X,
as
in
(23):
with
negation,
in
directives,
and
in
clauses
encoding
hypothetical
situations.
If
X
is
required,
i.e.
in
positive
non-
hypothetical
assertions,
their
use
is
ungrammatical,
as
in
(24),
where
mə(r)=
and
focus
agreement
block
the
use
of
NP/FCIs
(see
Louie,
2008:
108ff.
and
Bar-el
and
Denzer-King,
2008:
11
for
a
similar
distribution
in
Blackfoot).
(23)
a.
kin-oll’əlk
əl=kelu
who-NP/FCI.S
NEG=come(NEG.3SG)
‘Nobody
came.’
b.
neme-ŋolləŋ
mon-k!
what-NP/FCI.O
say-IMP.2SG
‘Say
anything!’
(i.e.
don’t
be
silent)
c.
mət-in
neme-ŋolləŋ
ət=saqsərə-mək!
I-DAT
what-NP/FCI.O
COND=pour-NEUT.TR.2SG
‘If
you
would
be
so
kind
to
pour
me
a
little
bit
(sc.
in
the
glass)!’
(24)
a.
*kin-oll’əlk
mə=kelu-j
who-NP/FCI.S
EX=come-NEUT.INTR.3SG
b.
*neme-ŋolləŋ
men
´-məŋ
what-NP/FCI.O
take-OF.1SG
Examples
(23)
and
(24)
show
that
NP/FCIs
cannot
co-occur
with
the
elements
of
class
X.
We
do
not
have
enough
data
for
all
relevant
contexts
to
judge
with
confidence
whether
this
is
due
to
their
sensitivity
to
veridicality,
i.e.
we
are
not
sure
whether
NP/FCIs
in
TY
can
only
occur
in
those
contexts
which
do
not
entail
the
truth
of
the
proposition,
as
has
been
claimed
for
NP/FCIs
in
many
other
languages
(Giannakidou,
1998,
2011
and
other
work).
What
does
seem
to
be
obvious
is
that
the
distribution
of
NP/FCIs
closely
follows
our
realis/irrealis
distinction.
The
explanation
at
hand
is
that
NP/FCIs
in
TY
are
referentially
deficient
in
such
a
way
that
they
cannot
introduce
discourse
referents:
in
Matthewson’s
(1998)
words,
NP/FCIs
‘fail
to
positively
assert
the
existence
of
an
entity’.
Since
they
cannot
refer,
they
are
licenced
only
if
they
are
in
the
scope
of
an
operator
that
allows
for
non-existential
readings.
Negation
and
the
illocutionary
and
modal
operators
introduced
above
are
thus
the
only
contexts
in
which
NP/FCIs
are
possible
because
they
do
not
entail
existence
(Giannakidou,
2011).
This
indicates
that
the
presence
of
X,
which
blocks
NP/FCIs,
must
somehow
entail
the
existence
of
the
referents
included
in
the
event
denoted
by
the
proposition.
At
this
stage
we
remain
vague
as
to
how
exactly
the
existential
import
of
X
comes
about.
In
Sections
4
and
5
we
will
suggest
that
this
is
achieved
by
unselective
existential
closure,
which
constitutes
the
essence
of
the
TY
realis.
Let
us
summarise
the
basic
distributional
data.
TY
makes
a
distinction
between
realis
and
irrealis
realised
in
the
following
formal
condition:
realis
clauses
must
contain
at
least
one
element
selected
from
the
heterogeneous
class
which
we
labelled
X,
for
want
of
a
better
term.
This
class
comprises
non-verbal
foci,
verb
modifiers
and
the
proclitic
mə(r)=.
The
latter
is
incompatible
with
the
other
two
members
of
the
class.
Notionally,
realis
encompasses
non-hypothetical
positive
assertions.
On
the
other
hand,
negative,
non-declarative
and
hypothetical
situations
are
treated
as
irrealis.
They
can
or
must
be
encoded
without
X.
Foci
and
verb
modifiers
are
optional
in
some
of
these
cases
and
impossible
in
others,
while
mə(r)=
is
completely
banned
from
all
irrealis
clauses.
The
distribution
of
NP/FCIs
corroborates
the
realis/irrealis
dichotomy
and
confirms
that
this
distinction
is
related
to
the
entailment
of
existence
in
TY.
In
the
following
two
sections
we
will
analyse
the
formal
and
semantic
properties
of
the
individual
members
of
the
class
X
in
order
to
explain
this
peculiar
configuration
and
offer
a
more
precise
definition
of
the
meaning
of
realis
in
TY.
We
demonstrate
that
the
elements
X
do
not
form
a
coherent
syntactic
or
semantic
class,
but
are
unified
at
a
higher
level,
that
of
entailing
the
existence
of
the
event
that
matches
the
proposition.
3.
Non-verbal
foci
and
verb
modifiers
In
this
section
we
show
that
two
members
of
class
X,
non-verbal
foci
and
verb
modifiers,
display
a
number
of
formal
similarities,
and
offer
a
semantic
explanation
for
this.
The
particle
mə(r)=
differs
from
them
in
a
number
of
important
ways
and
will
be
dealt
with
separately
in
Section
4.
D.
Matic
´,
I.
Nikolaeva
/
Lingua
150
(2014)
202--231
209

3.1.
Shared
properties
of
non-verbal
foci
and
verb
modifiers
This
section
discusses
the
formal
properties
of
foci
and
verb
modifiers.
Its
purpose
is
not
to
provide
an
elaborated
analysis
of
TY
sentence
structure,
but
rather
to
show
that
these
two
apparently
unrelated
classes
of
expressions
constitute
a
natural
syntactic
class.
3.1.1.
Prosody
Non-verbal
foci
and
verb
modifiers
carry
the
same
pitch
pattern,
a
prominent
high-low
pitch
contour
followed
by
a
deaccented
verb
(see
Matic
´and
Odé,
in
press,
for
details).
This
is
illustrated
by
the
following
examples,
in
which
the
high-
low
contour
appears
on
a
focus-marked
object
(25)
and
a
non-referential
nominal
(26)
(Figs.
1
and
2):
(25)
Ivan
kin-ek
j
ö
:-mələ?
tude-l
apanala:-lə
ŋ
j
ö
:-mələ
Ivan
who-FOC
see-OF.3
he-NOM
old.woman-FOC
see-OF.3
‘Who
did
Ivan
see?’--‘He
saw
the
old
woman.’
(26)
taŋ
ma:rqə-n
pajpə
lawjə-lə
law-nu-m
this
one-ATTR
woman
water-ACC
drink-IPF-NEUT.INTR.3SG
‘This
woman
is
drinking
water.’
The
phrases
apanala:ləŋ
j
ö
:mələ
(‘old.woman
saw’)
in
(25)
and
lawjələ
lawnum
(‘water
drank’)
in
(26)
display
a
rise-fall
contour
on
the
accented
syllable
of
the
first
word
and
a
flat
low
tone
on
the
verb;
this
is
the
regular
contour
with
all
non-
verbal
foci
and
verb
modifiers.
Importantly,
the
prominent
rise-fall
sequence
is
restricted
only
to
non-verbal
foci
and
verb
modifiers;
other
types
of
elements
that
happen
to
appear
in
front
of
the
verb
in
other
types
of
sentences
are
never
associated
with
this
prosodic
contour.
3.1.2.
Constituent
order
There
are
indications
that
non-verbal
foci
and
verb
modifiers
are
located
in
the
same
syntactic
position,