Measures of Central Tendency: Mean, Median and Mode
The Evolution of Morphological Agreement
1. The Evolution of Morphological Agreement
Richard Littauer
Saarland University"
@Richlitt
2. OUTLINE
• What
I
mean
by
agreement
• The
evolu3on
of
morphology
• The
arguments
regarding
simultaneous
evolu3on
• Using
the
agreement
hierarchy
• Differences
in
protolanguage
communi3es
3. What is Agreement
• “The
term
agreement
commonly
refers
to
some
systema3c
covariance
between
a
seman3c
or
formal
property
of
one
element
and
a
formal
property
of
another.”
(Steele
1978:
610)
• “The
essen3al
no3on
is
the
covariance
or
matching
of
feature
specifica3ons
between
two
separate
elements.”
(CorbeM
1998:
191)
4. What is Agreement
• The
most
produc3ve
case
of
agreement
across
languages
appears
to
be
subject-‐verb
agreement.
Even
languages
with
liMle
or
no
agreement
elsewhere
in
their
grammars,
such
as
English,
may
exhibit
subject-‐verb
agreement,
however
residually.
– Hawkins
1994:
370
5. What is Agreement
• Controller:
the
element
that
determines
the
agreement.
(also
trigger,
source)
• Target:
The
element
whose
form
is
determined
by
agreement
.
• Domain:
The
syntac3c
environment
in
which
agreement
occurs.
• Features:
The
means
or
manner
in
which
agreement
operates.
(also
category)
• Condi2ons:
other
factors
which
have
an
effect
on
agreement
but
are
not
directly
reflected.
(CorbeM
1998:
191)
6. Simultaneous Evolution
• Where
did
morphology
come
from?
• Uses
for
agreement
• Varying
complexity
• The
case
of
pidgins,
creoles,
and
gramma3caliza3on
7. Whence Morphology?
• “The
conven3onal
historical
explana3on
for
morphology
traces
it
to
proto-‐syntax
and
phonology.”
– Carstairs-‐McCarthy
1994:
46
• There
are
clear
controversies
over
where
to
put
morphology:
– its
own
component
(Aranoff
1993)
– wherever
it
is
relevant
to
the
syntax
(Anderson
2004)
– out
of
access
of
the
syntax
en3rely
(Chomsky
1970)
– in
the
lexicon
(Jensen
2004:
237)
– as
a
cohesive
whole
with
syntax
(Bickerton
1990)
– par3ally
overlapping
with
syntax
(Sadock
2004)
8. Whence Morphology?
• There
is
also
a
common
view
of
morphology
as
independently
built
on
top
of
protolanguage,
at
the
same
3me
as
syntax.
• There
are
arguments
for
this:
– Agreement
markers
do
not
always
follow
syntac3c
order
(Comrie
1980)
– Rela3vely
free
word
order
of
some
languages
(like
La3n)
(Samson
2009:
4)
– Its
use
for
clause
combining
(Heine
&
Kuteva
2007:
349)
9. Whence Morphology?
• “Thus
we
might
think
of
phrasal
syntax
and
morpho-‐syntax
as
independently
evolved
systems,
each
built
on
top
of
the
system
of
protolanguage,
each
refining
communica3on
through
its
own
expressive
techniques.
In
a
similar
vein,
Casey
and
Kluender
(1995)
suggest
that
agreement
inflec2on
evolved
as
an
extra
system
to
provide
redundant
(and
hence
more
reliable)
informa3on
about
seman3c
rela3ons
of
arguments.
I
see
no
immediate
reason
to
assert
the
temporal
priority
of
one
of
these
systems
over
the
other
in
the
course
of
evolu2on.”
– Jackendoff
2002:
260
10. Whence Agreement?
• Six
gradual
stages
from
protolanguage
to
modern,
and
agreement
occurs
on
the
sixth.
– Heine
and
Kuteva
(2007)
• “Agreement
is
a
purely
morpho-‐syntac3c
phenomenon,
and
serves
the
purpose
of
marking
those
cons3tuents
that
are
bound
together
in
close
gramma3cal
rela3onships.
Such
close
gramma3cal
rela3onships
ofen
reflect
closeness
in
the
conceptual
representa3on,
but
clearly
in
the
mental
representa3on
itself
such
closeness
is
inherent
and
does
not
stand
in
need
of
marking.
Agreement
is
part
of
the
apparatus
for
mapping
pre-‐linguis2c
representa2ons
onto
strings.”
– Hurford
(2002:
332)
11. Whence Agreement?
• The
historical
sources
of
various
agreement
markings
in
modern
languages
are
ofen
used
diagnos3cally
to
suggest
late
evolu3on.
• However,
agreement
is
not
always
telis3c,
nor
affected
only
by
erosion
(CorbeM
2006:
273)
• “A
purely
historical
explana3on
for
why
morphology
exists
amounts
to
an
asser3on
that
all
morphological
phenomena
can
be
traced
back
to
ancestral
phenomena
that
were
en3rely
non-‐
morphological,
involving
only
syntax
or
phonology.”
– Carstairs-‐McCarthy,
2010:
46)
12. Why Agreement?
• "Given
that
the
evidence
for
each
of
the
proposed
func3ons
is
not
fully
convincing,
it
appears
unlikely
that
agreement
is
to
be
explained
in
terms
of
a
single
func3on.
Rather,
it
has
different
combina3ons
of
func3ons
in
different
languages."
– Carstairs-‐McCarthy
2010:
275
13. Why Agreement?
• Givón
(1976:
173)
gives
many
examples:
1. Pro-‐drop
(arguable)
2. In
Redundant,
predictable,
obligatory
verb-‐
subject
agreement
cases,
the
agreement
can
become
a
way
of
signaling
the
syntac3c
type.
3. Correct
case
marking
in
iden3cal
parsed
forms
can
be
iden3fied
due
to
mismatching
of
agreement
features.
4. Agreement
allows
a
synchronic
analysis
of
evolu3onarily
transi3onal
processes
5. Verb
agreement
marks
the
verb’s
syntac3c
type,
as
well
as
its
general
seman3c-‐selec3onal
typology.
14. Why Agreement?
• Pro
drop
(cont.)
• The
ability
to
use
pro-‐drop
correctly
ofen
develops
much
later
in
children
than
other
morphological
agreement
abili3es
(e.g.
Snyder,
Senghas
&
Inman,
2001).
However,
pro-‐drop
may
be
different
from
other
agreement
phenomena.
Pro-‐drop
can
be
viewed
an
interface
phenomenon
-‐
it
must
be
processed
on-‐line
by
combining
informa3on
from
the
syntac3c
and
pragma3c
domains
(Sorace,
2011)
15. Why Agreement?
• "Uninterpretable
features
are
the
mechanism
that
implements
the
displacement
property.”
– Chomsky
2000:
12;13-‐14
• Carstairs-‐McCarthy
disputes
this,
using
La3n
as
an
example.
16. Why Agreement?
• Some
other
proposed
uses
for
agreement:
– Syntac3c
agreement
may
be
a
way
of
marking
nodes
for
help
in
parsing.
(Hawkins
(1994)
Kirby
(1999)
followed
this
up
in
simula3ons.
– Help
with
reference
tracking.
(Levin
2001)
– Marking
cons3tuency.
(Levin
2001)
– Agreement
allows
expression
of
different
seman3c
perspec3ves
(the
commiMee
has/have
...)
(CorbeM
1999)
– Signals
thema3c
roles.
(Jackendoff
2002)
– Pronominal
effect,
which
allows
pro-‐drop.
(Anderson,
others)
– Agreement
markers
as
arguments,
in
Autolexical
syntax.
(Sadock
1991)
17. Modern Morphogenesis
• Pidgins?
– Almost
no
inflec3onal
morphology.
– The
closest
example
has
been
that
of
Palu’e,
an
Austronesian
language
from
Indonesia,
which
has
begun
to
cli3cize
its
first
person
pronoun
subject
to
the
front
end
of
the
verb.
(CorbeM
2006:
266)
18. Modern Morphogenesis
• Children?
– Children
figure
out
the
basic
proper3es
of
the
agreement
system
very
early
on,
at
the
same
3me
as
syntac3cally
significant
produc3on
(Cinque
&
Kayne
2005:
99)
– Children
learning
languages
with
complex
morphological
systems
learn
agreement
markers
faster.
(Atsos
2011)
19. Modern Morphogenesis
• Pathological
cases?
– “Broca's
and
Wernicke's
aphasics
both
seem
to
be
significantly
impaired
in
the
produc3on
of
gramma3cal
morphology
-‐
par3cularly
when
their
performance
is
compared
with
evidence
for
sparing
of
pragma3cs
and
word
order
in
the
same
transcripts.”
(Batalli
2004:
291)
20. Modern Morphogenesis
• Primate
cogni3ve
abili3es?
– AnBn
grammars:
“no
syntac3c
rules
implemen3ng
embedded
nonadjacent
dependencies
were
learned
in
these
experiments”
– “Distribu3onal
regulari3es
explain
the
data
beMer
than
grammar
learning.”
• Hochmann
et
al.
2008
21. Modern Morphogenesis
• ”Gramma3caliza3on
can
hardly
explain
fully
the
origin
of
morphology
as
a
paMern
of
gramma3cal
organiza3on
dis3nct
from
syntax.”
(Carstairs-‐McCarthy
2010:
50)
• Furthermore,
studies
like
Dunn,
Gray,
&
Greenhill
suggest
that
phylogeny
is
more
important
for
language
change
than
universals
or
UG.
Quick,
almost
a
priori
languages
such
as
pidgins
and
creoles
may
not
be
the
best
guide.
22. Agreement Hierarchy
• CorbeM
reduced
his
hierarchy
to
three
basic
principles,
which
fit
the
bill
for
what
proto-‐morphology
might
have
looked
like
(CorbeM
2006:
26-‐7):
I. Canonical
agreement
is
redundant
rather
than
informa3ve.
II. Canonical
agreement
is
syntac3cally
simple.
III. The
closer
the
expression
of
agreement
is
to
canonical
inflec3onal
morphology,
the
more
canonical
it
is
as
agreement.
23. Agreement Hierarchy
• What
is
canonicity?
– “’Canonical’
instances
of
agreement
[are
the]
“best,
clearest,
indisputable
(according
to
the
'canon');
such
cases
need
not
be
common.”
(CorbeM
2001:
109)
24. Agreement Hierarchy
• Some
examples:
– Controller
present
>
controller
absent
– Controller’s
part
of
speech
irrelevant
>
relevant
– Bound
>
free
– Inflec3on
marking
>
cli3c
>
free
word
– Obligatory
>
op3onal
25. Agreement Hierarchy
• Some
examples:
– Regular
>
supple3ve
– Allitera3ve
>
opaque
– Produc3ve
>
sporadic
– Doubling
>
independent
only
– Target’s
part
of
speech
irrelevant
>
relevant
– Local
>
non-‐local
26. Agreement Hierarchy
• A
quick
example
of
one
of
the
canonical
hierarchies
(allitera3ve
agreement):
ki-‐kapu
ki-‐kubwa
ki-‐moja
ki-‐lianguka
7-‐basket
7-‐one
7-‐fell
7-‐large
'one
large
basket
fell’
(CorbeM
2001:
116)
27. Agreement Hierarchy
• “It
is
not
good
enough
simply
to
define
a
structural
complexity
hierarchy
and
assume
it
directly
gives
rise
to
a
cross-‐linguis3c
hierarchy,
because
one
needs
to
explain
why
not
all
languages
opt
for
minimum
complexity.”
(Kirby
1999:
119)
• Complexity
may
arise
from
constraints
regarding
costs,
benefits,
and
func3onal
load.
Alterna3vely,
it
may
be
due
to
the
possible
nature
of
“language
universals
as
products
of
cultural
influence.”
(Sampson
2009:
15)
28. Varying Complexity
• Languages
differ
in
complexity.
(eg.
Sampson
2009)
See
Lupyan
and
Dale
(2010),
LiMle
(2011),
and
other
studies
on
community
size,
second
language
learners,
foreigner-‐directed
speech,
etc.
and
morphological
complexity.
• “Complexi3es
in
morphology
are
accompanied
by
complexi3es
in
syntax."
(Dahl
2009:
63)
29. Evolutionary Environment?
• Smaller
communi3es
=
more
agreement.
• As
Hurford
(2012)
states,
language
evolved
gradually
–
complexity
on
the
scale
of
modern
language
comes
into
it
later.
• Gramma3caliza3on
not
necessarily
a
good
theory
for
showing
early
language
change.
30. Future Work?
• Possible
future
work
would
include:
– Cross-‐linguis3c
first
language
agreement
acquisi3on
(specifically
across
families)
– More
studies
into
linguis3c
complexity
involving
speaking
community
size
– Experimental
studies
using
agreement
morphology
in
lieu
of
syntax
to
convey
meaning
– simula3ons
of
morphological
redundancy
(which,
computa3onally,
may
not
be
easy.)
31. Conclusion
“Nothing
in
biology
makes
sense
except
in
the
light
of
evolu3on.”
(Dobzhansky
1973)
Agreement
may
be
a
living
fossil
of
protolanguage.
32. THANKS!
(Refs on request)
www.replicatedtypo.com
@richlitt
Editor's Notes
Agreement phenomena evolved simultaneously, or at least not far after, syntax. Because: Overt sign of compositional structure on the combinatorial level redundancy for the hearer the many different proposed functions of agreement the inadequacy of historical analyses and syntactic models to discount morphology evolutionarily I’ll go on to say that the canonical hierarchy could be used as a diagnostic on early fossilization, before complexity was introduced too much I’m not going to be focusing in too much depth on the evolution of morphology – that has been covered much better elsewhere, including in this conference.
Agreement phenomena evolved simultaneously, or at least not far after, syntax. Because: Overt sign of compositional structure on the combinatorial level redundancy for the hearer the many different proposed functions of agreement the inadequacy of historical analyses and syntactic models to discount morphology evolutionarily I’ll go on to say that the canonical hierarchy could be used as a diagnostic on early fossilization, before complexity was introduced too much I’m not going to be focusing in too much depth on the evolution of morphology – that has been covered much better elsewhere, including in this conference.
Of course, the amount of agreement differs between languages.
This is partly why agreement hasn’t been considered much by evolutionary linguistics. Archi
Two contrasting views: last, or first? I should clarify here that when I say grammaticalization, I mean the full theory laid out by Heine and Kuteva. As far as an increase in productivity and generality against a decrease in compositionality – I don’t think that this is at ends with my stance. My problem is mainly that agreement is normally taken to happen at the very end of the evolutionary road, which I think doesn’t work as well evolutionarily. I agree with Hurford here that it is part of the mapping, but also that it is useful for the hearer, as well as the speaker.
Now, agreement lies on the interface between syntax and semantics So all of this has to be incorporated into a theory that unifies them, which means that describing a straightforward origin of a single marker will not help. Next: Why is agreement important?
So, what are the different functions of agreement that we see in the literature?
The ability to use pro-drop correctly often develops much later in children than other morphological agreement abilities (e.g. Snyder,Senghas & Inman, 2001). However, pro-drop may be different from other agreement phenomena. Antonella Sorace sees pro-drop as an interface phenomenon - it must be processed on-line by combining information from the syntactic and <i>pragmatic domains (see <a href=&quot;http://www.lel.ed.ac.uk/~antonell/Sorace-LAB2011.pdf&quot; rel=&quot;nofollow&quot;>Sorace, 2011</a>).</i> 2. This is apparently attested in Tok Pisin, where ‘him’ has grammaticalised into a marker for transitive verbs. (Givón 1976: 168). 4. This is useful when, for a short time, agreement differentiates between the stable subordinate clauses and the innovative environment of the main clause.
Siewierska (1998: 505-8) noted that freer word order leads to more agreement. You’ll notice that most of these work within the syntactic-semantic interface. We can see other functions outside of this interface in the agreement hierarchy laid out by Corbett (which I’ll get to in a minute), and in the neural processing that I already mentioned from Hurford. Now, the grammaticalisationists and pro-protosyntacticians would argue that pidgins, creoles, and the like show that agreement isn’t that important or doesn’t occur earlier diachronically, using modern examples.
This study was done on Greek. I wish it had been done on Archi, which has 1.5 million contrastive forms. over 1.5 million contrasting forms.) (Atsos 2011, Samson 2009) I think that this is, then, another case where ontogeny recapitulates phylogeny.
Grammaticalisation assumes syntax having a stronger role than morphology in protolanguage, which I don’t think is a given on an evoutionary scale. So what would early agreement look like? >>> canonicity in agreement?
It should be noted that the hierarchy specified here is not based on distribution of typological features, but rather on what is the clearest example of agreement. (as shown by next slide)
So what I’m saying is that the hierarchy, as the simplest and clearest, might be justifiably viewed as a baseline for what early morphology might have looked like. Teasing apart simplicity of the system from complexity in reality is, of course, a difficult task – and one I would like to do in experimental work in the future. But here are some examples of canonical features that will illustrate what I am trying to get at.
So, what do I hope you take from this? That canonical agreement, would be clearly beneficial to mapping onto strings in it’s simple make-up, and that it would provide clear, redundant information that could be used by the receiver. What’d I’d also like to point out is that it is possible that canonical agreement could give clear clues that could be used in place of syntax.
Here’s a good example. In such canonical, non-opaque alliterative noun adjective agreement, if regular, we could see how easily this would be used in place of syntax What’s clear in this example – and for all agreement phenomena – is that we see a physical, phonetic presence on the combinatorial level (a lá Kirby’s talk where he didn’t want to use the term duality of patterning ) for compositional structure.
Relevant to Luke McCrohon’s talk on the borrowing hierarchy, where we run into the same problem. I think the latter is true – due to the nature of cultural evolution, complexity gets introduced into the system, which causes the gap between the hierarchy and the typological distribution we see today. There’s been a of work, of course, on that distribution – Greenberg, or the Dunn, Gray, Greenhill group, or Daumé and Campbell 2010 – but what I want to focus on now is the forces that influence that complexity, and why it is possible that morphological agreement and morphology have been sidelined by syntacticians who work mainly with larger languages.
The essential notion is that there would be smaller community sizes and less pressure to keep the language simple, as we’ve seen can affect morphological complexity (and complexity as a whole – tones in Pirahã might be a good example of this.) When thinking about varying complexity, and agreement, we need to think about what would have been the situation for protolanguage communities. And this will be more beneficial in the long run for us than getting stuck in historical analyses.